Decision aids for people facing health treatment or screening decisions

  • Review
  • Intervention

Authors


Abstract

Background

Decision aids are intended to help people participate in decisions that involve weighing the benefits and harms of treatment options often with scientific uncertainty.

Objectives

To assess the effects of decision aids for people facing treatment or screening decisions.

Search methods

For this update, we searched from 2009 to June 2012 in MEDLINE; CENTRAL; EMBASE; PsycINFO; and grey literature. Cumulatively, we have searched each database since its start date including CINAHL (to September 2008).

Selection criteria

We included published randomized controlled trials of decision aids, which are interventions designed to support patients' decision making by making explicit the decision, providing information about treatment or screening options and their associated outcomes, compared to usual care and/or alternative interventions. We excluded studies of participants making hypothetical decisions.

Data collection and analysis

Two review authors independently screened citations for inclusion, extracted data, and assessed risk of bias. The primary outcomes, based on the International Patient Decision Aid Standards (IPDAS), were:

A) 'choice made' attributes;

B) 'decision-making process' attributes.

Secondary outcomes were behavioral, health, and health-system effects. We pooled results using mean differences (MD) and relative risks (RR), applying a random-effects model.

Main results

This update includes 33 new studies for a total of 115 studies involving 34,444 participants. For risk of bias, selective outcome reporting and blinding of participants and personnel were mostly rated as unclear due to inadequate reporting. Based on 7 items, 8 of 115 studies had high risk of bias for 1 or 2 items each.

Of 115 included studies, 88 (76.5%) used at least one of the IPDAS effectiveness criteria: A) 'choice made' attributes criteria: knowledge scores (76 studies); accurate risk perceptions (25 studies); and informed value-based choice (20 studies); and B) 'decision-making process' attributes criteria: feeling informed (34 studies) and feeling clear about values (29 studies).

A) Criteria involving 'choice made' attributes:

Compared to usual care, decision aids increased knowledge (MD 13.34 out of 100; 95% confidence interval (CI) 11.17 to 15.51; n = 42). When more detailed decision aids were compared to simple decision aids, the relative improvement in knowledge was significant (MD 5.52 out of 100; 95% CI 3.90 to 7.15; n = 19). Exposure to a decision aid with expressed probabilities resulted in a higher proportion of people with accurate risk perceptions (RR 1.82; 95% CI 1.52 to 2.16; n = 19). Exposure to a decision aid with explicit values clarification resulted in a higher proportion of patients choosing an option congruent with their values (RR 1.51; 95% CI 1.17 to 1.96; n = 13).

B) Criteria involving 'decision-making process' attributes:

Decision aids compared to usual care interventions resulted in:

a) lower decisional conflict related to feeling uninformed (MD -7.26 of 100; 95% CI -9.73 to -4.78; n = 22) and feeling unclear about personal values (MD -6.09; 95% CI -8.50 to -3.67; n = 18);

b) reduced proportions of people who were passive in decision making (RR 0.66; 95% CI 0.53 to 0.81; n = 14); and

c) reduced proportions of people who remained undecided post-intervention (RR 0.59; 95% CI 0.47 to 0.72; n = 18).

Decision aids appeared to have a positive effect on patient-practitioner communication in all nine studies that measured this outcome. For satisfaction with the decision (n = 20), decision-making process (n = 17), and/or preparation for decision making (n = 3), those exposed to a decision aid were either more satisfied, or there was no difference between the decision aid versus comparison interventions. No studies evaluated decision-making process attributes for helping patients to recognize that a decision needs to be made, or understanding that values affect the choice.

C) Secondary outcomes

Exposure to decision aids compared to usual care reduced the number of people of choosing major elective invasive surgery in favour of more conservative options (RR 0.79; 95% CI 0.68 to 0.93; n = 15). Exposure to decision aids compared to usual care reduced the number of people choosing to have prostate-specific antigen screening (RR 0.87; 95% CI 0.77 to 0.98; n = 9). When detailed compared to simple decision aids were used, fewer people chose menopausal hormone therapy (RR 0.73; 95% CI 0.55 to 0.98; n = 3). For other decisions, the effect on choices was variable.

The effect of decision aids on length of consultation varied from 8 minutes shorter to 23 minutes longer (median 2.55 minutes longer) with 2 studies indicating statistically-significantly longer, 1 study shorter, and 6 studies reporting no difference in consultation length. Groups of patients receiving decision aids do not appear to differ from comparison groups in terms of anxiety (n = 30), general health outcomes (n = 11), and condition-specific health outcomes (n = 11). The effects of decision aids on other outcomes (adherence to the decision, costs/resource use) were inconclusive.

Authors' conclusions

There is high-quality evidence that decision aids compared to usual care improve people's knowledge regarding options, and reduce their decisional conflict related to feeling uninformed and unclear about their personal values. There is moderate-quality evidence that decision aids compared to usual care stimulate people to take a more active role in decision making, and improve accurate risk perceptions when probabilities are included in decision aids, compared to not being included. There is low-quality evidence that decision aids improve congruence between the chosen option and the patient's values.

New for this updated review is further evidence indicating more informed, values-based choices, and improved patient-practitioner communication. There is a variable effect of decision aids on length of consultation. Consistent with findings from the previous review, decision aids have a variable effect on choices. They reduce the number of people choosing discretionary surgery and have no apparent adverse effects on health outcomes or satisfaction. The effects on adherence with the chosen option, cost-effectiveness, use with lower literacy populations, and level of detail needed in decision aids need further evaluation. Little is known about the degree of detail that decision aids need in order to have a positive effect on attributes of the choice made, or the decision-making process.

摘要

面臨健康醫療或篩選選擇時,民眾的決策輔助

背景

決策輔助讓民眾準備好參與下列決策,包含評估益處、危害與科學的不確定性。

目的

想要評估面臨治療或篩選決策時,決策輔助對於民眾的有效性。

搜尋策略

為了此更新,我們搜尋了2006年1月到2009年12月 的MEDLINE (Ovid);考科藍對照試驗中央註冊(CENTRAL,考科藍圖書館2009年第4期); CINAHL (Ovid) (僅至2008年9月); EMBASE (Ovid); PsycINFO (Ovid); 與非正式發表之文獻。累計而言,我們自每一資料庫開始日期搜尋。

選擇標準

我們納入了採用隨機對照試驗(RCTs)來進行之已公開發表的決策輔助,其中介入是設計用於支援病患決策制定,藉由提供治療或篩選選項以及它們的相關成果有關的資訊,並與一般照護及/或替代選擇的介入比較。我們排除了參與者沒有進行積極治療或篩選決策的研究。

資料收集與分析

兩位審閱作者獨立篩選摘要以利採用、摘錄數據並評估可能的偏誤風險。根據於國際病患決策輔助標準的重要成果為:

A) 決策屬性;

B) 決策制定程序屬性。

次要成果為行為、建康與醫療體系影響。我們使用平均數差異(MD)與相對風險(RR)集合RCTs結果,採用隨機效果模型。

主要結果

具有34,316獨特引用,86個研究涵蓋20,209名參與者,符合合格準則並被採用。這些研究中有31個是此次更新中新加入。29個試驗持續進行。所有研究中可能的偏誤風險具變化性。最有問題的兩個準則為(1)缺乏盲目方式與(2) 缺乏選擇成果報告的潛力。也因此造成大多數較早的試驗沒有註冊。

其中86個所含研究, 63 (73%)個研究至少採用一個測量,建構於IPDAS有效性準則上:

A) 準則涉及決策特性: 知識分數(51個研究); 準確的風險認知(16個研究); 與經告知的價值導向選擇 (12個研究); 與B) 準則涉及決策程序特性: 感覺受告知(30個研究) 與感覺價值相關為明確 (18個研究)。

A) 準則涉及決策特性:

藉由增加知識,決策輔助執行的較一般照護佳 (MD 100個中13.77; 95%信賴區間(CI) 11.40 to 16.15; n = 26)。當更多詳細決策輔助與較簡化的決策輔助比較時,知識上相對的改善為顯著 (MD 100個中4.97; 95% CI 3.22 to 6.72; n = 15)。得到給予明確的機率之決策輔助會導致民眾有精確風險知覺的比例較高(RR 1.74; 95% CI 1.46 to 2.08; n = 14)。機率以數字表示時(RR 1.93; 95% CI 1.58 to 2.37; n = 11) 會比文字表示之效果更好(RR 1.27; 95% CI 1.09 to 1.48; n = 3)。 得到明確數值之決策輔助,與無明確數值者相比,病患比較會做出與被告知之訊息一致之決策(RR 1.25; 95% CI 1.03 to 1.52; n = 8)。

B) 準則涉及決策程序特性:決策輔助與一般照護介入相比:

a) 降低未被告知之感覺有所相關的決策衝突(MD -6.43 of 100; 95% CI -9.16 to -3.70; n = 17);

b) 降低對個人價值感覺不明確有所相關的決策衝突(MD -4.81; 95% CI -7.23 to -2.40; n = 14);

c) 降低被動的決策制定 (RR 0.61; 95% CI 0.49 to 0.77; n = 11); 與d) 降低介入後仍舊無法做出決策的比例 (RR 0.57; 95% CI 0.44 to 0.74; n = 9)。

決策輔助在四個有測量此成果的研究中,似乎在病患-醫師之溝通上有正面效果。就決策滿意度(n = 12) 及/或決策制定程序而言 (n = 12),與沒有得到決策輔助者相比,這些得到決策輔助者,可能有更高之滿意度,也可能沒有差異。目前沒有研究有在評估協助病患認知必須制定的決策或是了解影響選擇的價值上相關的決策程序特性。

C) 次要成果

得到決策輔助者與一般照護比較,顯示出降低了選擇重大侵入性手術,偏好採用保守性選項(RR 0.80; 95% CI 0.64 to 1.00; n = 11)。得到決策輔助者與一般照護比較,也降低去選擇採用前列腺癌篩選(RR 0.85; 95% CI 0.74 to 0.98; n = 7)。詳細的決策輔助與的簡易決策輔助比較時,選擇採用更年期賀爾蒙治療的病患減少(RR 0.73; 95% CI 0.55 to 0.98; n = 3)。就其它決策而言,選擇相關的效果各有不同。

決策輔助在諮詢的時間有從減少8分鐘到增加23分鐘 (中位數2.5分鐘)各有不同。焦慮程度 (n = 20)、一般健康成果(n = 7)、與疾病特定健康成果 (n = 9)這三方面,兩者皆沒有差異。決策輔助在其他成果 (依從決策、成本/資源使用) 上的效果尚無法做出結論。

作者結論

就此更新的審閱而言,為新的證據有: (1)得到給予明確的數值之決策輔助,可以改善經告知的價值導向之決策; (2)決策輔助似乎在病患-醫師溝通上有正面效果; (3)決策輔助在諮詢時間上有不同效果。

與先前審閱(包含2006年之前的研究) 的發現一致的有: (1)決策輔助增加民眾涉入並改善知識與實際成果知覺; (2)然而,影響規模在每個研究中各異。決策輔助在選擇上有不同影響。它們降低有彈性選擇之手術,即使如此在健康成果或滿意度上也沒有明顯之不良效果。奉行所選選擇、病患-醫師溝通、成本效益以及文化水平不高人口之使用等以上四種情況的影響,需要進一步之評估。對於決策輔助之需求,用於決策特性或決策制定程序上時,會產生何種正面效果的細節之程度,目前所知甚少。

Plain language summary

Decision aids to help people who are facing health treatment or screening decisions

Identifying and making a decision about the best health treatment or screening option can be difficult for patients. Decision aids can be used when there is more than one reasonable option, when no option has a clear advantage in terms of health outcomes, and when each option has benefits and harms that patients may value differently. Decision aids may be pamphlets, videos, or web-based tools. They make the decision explicit, describe the options available, and help people to understand these options as well as their possible benefits and harms. This helps patients to consider the options from a personal view (e.g., how important the possible benefits and harms are to them) and helps them to participate with their health practitioner in making a decision.

The updated review, with searches updated in June 2012, includes 115 studies involving 34,444 participants. Findings show that when patients use decision aids they: a) improve their knowledge of the options (high-quality evidence); b) feel more informed and more clear about what matters most to them (high-quality evidence); c) have more accurate expectations of possible benefits and harms of their options (moderate-quality evidence); and d) participate more in decision making (moderate-quality evidence). Patients who used decision aids that included an exercise to help them clarify what matters most to them, were more likely to reach decisions that were consistent with their values. However, the quality of the evidence was moderate for this outcome, meaning that further research may change these findings. Decision aids reduce the number of patients choosing prostate specific antigen testing and elective surgery when patients consider other options. They have a variable effect on most other actual choices. Decision aids improve communication between patients and their health practitioner. More detailed decision aids are better than simple decision aids for improving people's knowledge and lowering decisional conflict related to feeling uninformed and unclear about their personal values. Decision aids do not worsen health outcomes and people using them are not less satisfied. More research is needed to evaluate adherence with the chosen option, the associated costs, use with patients who have more limited reading skills, and the level of detail needed in a decision aid.

淺顯易懂的口語結論

決策輔助以協助面臨健康治療或篩選決策的民眾:

找出最佳健康治療或篩選選項的決策,對病患而言可能有困難。下列情況可採用決策輔助:(1)在有一個以上選項時,(2)就件康成果而言,當沒有選項有明確優點時,(3)每一選項都有利弊時,這些利弊對於不同的病患,所重視的程度的可能不同。決策輔助可以是小冊子、視頻或是網路導向工具。這些決策輔助說明可行之選項,並協助民眾了解這些選項以及可能的利弊。這讓病患能以個人觀點考慮選項(例如:可能的利弊對他們而言有多重要),並讓病患準備好制定決策。

86個研究的更新審閱發現當病患利用決策輔助時,他們: a) 提升對選項的知識; b) 接受協助而對可能利弊有更精確的預期; c) 完成與他們得知的價值較一致的選擇; 與d) 更加參與決策制定。決策輔助在實際選擇上有不同影響,但當病患考慮其他選擇時,它們降低選擇性手術的選擇。病患利用決策輔助時,與他們的醫師溝通似乎有正面效果,而需要此諮商的時間各有不同效果。雖然與較簡易的決策輔助相比,利用更詳細的決策輔助有所改善,但跟決策輔助與一般照護相比時,改善的程度較小。健康成果或滿意度上沒有明顯不良效果。需要更多研究評估所選選項的奉行程度、病患-醫師間的溝通與相關成本。

譯註

翻譯: East Asian Cochrane Alliance
翻譯補助: 台灣衛生福利部/台北醫學大學實證醫學研究中心

Резюме на простом языке

Поддержка принятия решений для помощи людям, которым предстоит лечение или принятие решений по скринингу

Процесс принятия решений о выборе лучшего лечения или варианта скрининга может быть трудным для пациентов. Способы поддержки принятия решений могут быть использованы, когда существует более одного разумного варианта, и когда ни один из вариантов не имеет явного преимущества с точки зрения исходов для здоровья, и когда каждый вариант имеет преимущества и недостатки, которые пациенты могут оценивать по-разному. С целью поддержки принятия решения могут быть использованы брошюры, видео- и web-материалы. Они помогают сделать решение не оставляющим сомнений, описывают имеющиеся варианты, помогают людям понимать эти варианты, так же как и их возможную пользу и вред. Это помогает пациентам рассмотреть эти варианты с их личной точки зрения (например, насколько важны потенциальные польза и вред от вмешательства для них) и помогает им совместно с их доктором принять решение.

Обновленный обзор (с обновлением поисков в июне 2012 года) включает 115 исследований с участием 34 444 пациентов. Результаты обзора показывают, что при использовании средств поддержки принятия решений пациенты: a) улучшают свои знания о вариантах (доказательства высокого качества); б) чувствуют себя более информированными и имеют более четкое представление о том, что для них самое главное (доказательства высокого качества); в) имеют более точные ожидания потенциальной пользы и вреда от своих вариантов вмешательства (доказательства среднего качества); г) более активно участвуют в принятии решения (доказательства среднего качества). Пациенты, которые использовали средства поддержки принятия решений, включающие упражнения, которые помогают им прояснить, что является наиболее важным для них, были более склонны принимать решения, согласующиеся с их ценностями. Однако, качество доказательств для этого исхода было умеренным (средним), что означает, что эти результаты могут быть изменены при дальнейших исследованиях. Средства поддержки принятия решений сокращают число пациентов, выбирающих тестирование на предмет простатического специфического антигена и элективную хирургию, когда пациенты рассматривают другие варианты. Они оказывают различное влияние на большинство других конкретных выборов. Средства поддержки принятия решений улучшают взаимодействие между пациентами и их врачами. Более детализированные (подробные) средства поддержки принятия решений лучше, чем простые улучшают знания людей и снижают конфликт решений, связанный с ощущением неосведомленности и неясности о своих личных ценностях. Эти средства не ухудшают исходы здоровья или степень удовлетворенности пациентов их использованием. Необходимы дополнительные исследования, чтобы оценить приверженность к выбранному варианту лечения, связанные с ним затраты, применение пациентами, которые имеют ограниченные навыки чтения, и уровень детализации, необходимой в средстве поддержки принятия решений.

Заметки по переводу

Перевод: Кораблева Анна Александровна. Редактирование: Зиганшина Лилия Евгеньевна. Координация проекта по переводу на русский язык: Казанский федеральный университет. По вопросам, связанным с этим переводом, пожалуйста, свяжитесь с нами по адресу: lezign@gmail.com

Laički sažetak

Alati za pomoć pri donošenju odluke namijenjeni osobama koje trebaju odlučiti o vrsti liječenja ili metodi ranog otkrivanja bolesti

Pacijentima može biti teško donijeti odluku o tome koja je vrsta liječenja ili metoda probira (ranog otkrivanja bolesti) za njih najbolja. Postoje „alati za pomoć pri donošenju odluke“ (prema engl. decision aids) koji se mogu koristiti u situacijama kada postoji više od jedne razumne mogućnosti, kada niti jedna mogućnost nema jasnu prednost što se tiče ishoda, i kada svaka metoda ima i korisne i štetne učinke koje pacijenti mogu vrednovati na različite načine. Alati za pomoć pri donošenju odluke mogu biti brošure, video zapisi ili web alati. One čine odluku eksplicitnom, opisuju dostupne mogućnosti, te pomažu ljudima u njihovom razumijevanju kao i u razumijevanju mogućih korisnih i štetnih posljedica. To omogućuje pacijentima sagledavanje različitih mogućnosti iz osobnog kuta (npr. koliko su potencijalni korisni učinci i štetne posljedice važne pacijentu osobno) i pomaže im da u donošenju odluke sudjeluju zajedno s liječnikom.

Ovo je obnovljeni Cochrane sustavni pregled koji prikazuje podatke iz literature dostupne do lipnja 2012. godine, a obuhvaća 115 studija s 34.444 ispitanika. Rezultati pokazuju da pacijenti koji koriste alate za pomoć pri odlučivanju: a) poboljšavaju svoje znanje o mogućnostima (dokazi visoke kvalitete); b) osjećaju se bolje informiranima i bolje razaznaju što je njima samima najvažnije (dokazi visoke kvalitete); c) imaju točnija očekivanja što se tiče mogućih korisnih i štetnih posljedica odabira (dokazi umjerene kvalitete); i d) više sudjeluju u donošenju odluke (dokazi umjerene kvalitete). Pacijenti koji su koristili alate za pomoć pri odlučivanju, a koji omogućuju razjasniti što je njima osobno važno, imali su veću vjerojatnost donošenja odluke koja je u skladu s njihovim vrijednostima. Ipak, kvaliteta dokaza za ovu tvrdnju je umjerena što znači da daljnja istraživanja mogu promijeniti te rezultate. Korištenje smjernica smanjuje broj pacijenata koji biraju pretragu pomoću PSA (prostatični specifični antigen) za otkrivanje karcinoma prostate i elektivne operacije nakon što uzmu u obzir i ostale mogućnosti. Oni imaju promjenjivi učinak na većinu drugih stvarnih izbora. Takvi alati poboljšavaju komunikaciju između pacijenata i njihovih liječnika obiteljske medicine. Detaljniji alati za pomoć pri odlučivanju bolji su od jednostavnih jer povećavaju znanje pacijenata i smanjuju pojavu konflikta pri odlučivanju, a koji je povezan s osjećajem neinformiranosti i nerazumijevanja osobnih vrijednosti. Korištenje alata za pomoć pri donošenju odluke ne pogoršava zdravstvene ishode i oni koji ih koriste nisu manje zadovoljni. Potrebna su daljnja istraživanja na ovu temu kako bi se procijenila dosljednost nakon donošenja odluke, povezani troškovi, korisnost tih alata u pacijenata s ograničenom vještinom čitanja, i količina detalja koja treba biti zastupljena u takvim alatima.

Bilješke prijevoda

Cochrane Hrvatska
Prevela: Dubravka Komić
Ovaj sažetak preveden je u okviru volonterskog projekta prevođenja Cochrane sažetaka. Uključite se u projekt i pomozite nam u prevođenju brojnih preostalih Cochrane sažetaka koji su još uvijek dostupni samo na engleskom jeziku. Kontakt: cochrane_croatia@mefst.hr

Laienverständliche Zusammenfassung

Entscheidungshilfen für Patienten, die Entscheidungen hinsichtlich medizinischer Behandlung oder Screening treffen müssen

Die Auswahl und Entscheidung für eine bestimmte medizinische Behandlung oder ein Screening kann für Patienten schwierig sein. Entscheidungshilfen können angewendet werden, wenn es mehr als eine relevante Option gibt, wenn keine der Optionen einen klaren Vorteil hinsichtlich Behandlungsergebnisse aufweist, oder wenn jede der Optionen Nutzen und Schäden birgt, die von verschiedenen Patienten unterschiedlich bewertet werden. Informationsbroschüren, Videos oder auf dem Internet basierende Hilfen können als Entscheidungshilfen dienen. Diese machen die Entscheidung explizit, beschreiben die verfügbaren Optionen und helfen Patienten, diese Optionen und deren potentielle Nutzen und Schäden zu verstehen. Sie unterstützen den Patienten dabei, alle Optionen aus persönlicher Sicht (z.B. wie wichtig die möglichen Nutzen und Schäden für den Einzelnen sind) abzuwägen und eine Entscheidung gemeinsam mit ihrem behandelnden Arzt zu treffen.

Der aktuelle Review, dessen Suchen im Juni 2012 aktualisiert wurden, umfasst 115 Studien mit 34.444 Teilnehmern. Die Ergebnisse zeigen, dass wenn Patienten Entscheidungshilfen nutzen sie a) ihr Wissen über die Alternativen verbessern (hohe Qualität der Evidenz); b) sich informierter über ihre eigenen Prioritäten fühlen (hohe Qualität der Evidenz); c) präzisere Erwartungen an die möglichen Nutzen und Schäden der Alternativen haben (moderate Qualität der Evidenz); und d) stärker an der Entscheidungsfindung teilnehmen (moderate Qualität der Evidenz). Patienten, die Entscheidungshilfen nutzen, in denen eine Übung zu ihren Prioritäten war, trafen eher Entscheidungen, die im Einklang mit ihren Wertvorstellungen waren. Jedoch war die Qualität der Evidenz für diesen Endpunkt moderat, was bedeutet, dass zukünftige Forschung dieses Ergebnis möglicherweise ändert. Entscheidungshilfen verminderten die Anzahl der Patienten, die ein Prostata-spezifisches Antigen Test (PSA-Test) und elektive Operationen wählten, wenn sie andere Optionen in Erwägung zogen. Auf die meisten anderen konkreten Entscheidungen haben Entscheidungshilfen eine unterschiedliche Wirkung. Entscheidungshilfen verbessern die Kommunikation zwischen Patienten und Ärzten. Detailliertere Entscheidungshilfen sind besser als einfache Entscheidungshilfen, wenn es darum geht das Wissen der Patienten zu verbessern und den Entscheidungskonflikt bezüglich des Gefühls, uninformiert und sich nicht im klaren über die eigenen Wertvorstellungen zu sein, zu vermindern. Entscheidungshilfen verschlechtern weder die Gesundheit noch die Patientenzufriedenheit. Mehr Forschung ist nötig, um die Einhaltung der gewählten Optionen, die damit zusammenhängenden Kosten, die Verwendung durch Patienten mit Leseschwäche und den benötigten Detaillierungsgrad zu untersuchen.

Anmerkungen zur Übersetzung

A. Lieder, Koordination durch Cochrane Schweiz

Summary of findings(Explanation)

Summary of findings for the main comparison. 
  1. 1. The vast majority of studies measuring this outcome were not at high risk of bias.

    2.The GRADE rating was downgraded given the lack of precision.

    3.The GRADE rating was downgraded given the lack of consistency.

    4.The GRADE rating was downgraded given the lack of directness.

Patient decision aids compared with usual care for adults considering treatment or screening decisions

Patient or population: adults considering treatment or screening decisions

Settings: all settings

Intervention: patient decision aid

Comparison: usual care

OutcomesIllustrative comparative benefits* (95% CI)Relative effect
(95% CI)
No of Participants
(studies)
Quality of the evidence
(GRADE)
Comments
Assumed benefitCorresponding benefit
Usual carePatient decision aid

Knowledge: decision aid versus usual care - all studies

standardized on score from 0 (no knowledge) to 100 (perfect knowledge) [soon after exposure to the decision aid]

The mean knowledge score was 56.9% ranged across control groups from 31% to 85.2%The mean knowledge score in the intervention groups was 13.34 higher (11.17 to 15.51 higher) 10,842
(42 studies)
⊕⊕⊕⊕
high 1
Higher scores indicate better knowledge. 41 out of 42 studies showed an improvement in knowledge.

Accurate risk perceptions - all studies

[soon after exposure to the decision aid]

296 patients per 1000 542 patients per 1000 RR 1.82 (95% CI: 1.52 to 2.16)5868
(19 studies)
⊕⊕⊕⊝
moderate 1,2
 

Congruence between the chosen option and their values - all studies

[soon after exposure to the decision aid]

316 patients per 1000 498 patients per 1000 RR 1.51 (95% CI: 1.17 to 1.97)

4670

(13 studies)

⊕⊕⊝⊝
low 1,2,3,4
 

Decisional conflict: decision aid versus usual care - all studies - Uninformed sub-scale

standardized on score from 0 (not uninformed) to 100 (uninformed) [soon after exposure to the decision aid]

The mean feeling uninformed ranged across control groups from 12.75 to 49.1. Scores of 25 or lower are associated with follow-through with decisions; whereas scores that exceed 38 are associated with delay in decision makingThe mean feeling uninformed in the intervention groups was 7.26 lower (9.73 to 4.78 lower) 

4343

(22 studies)

⊕⊕⊕⊕
high 1
Lower scores indicate feeling more informed.

Decisional conflict: decision aid versus usual care - all studies - Unclear values sub-scale

standardized on score from 0 (not unclear) to 100 (unclear) [soon after exposure to the decision aid]

The mean feeling unclear values ranged across control groups from 15.5 to 51.29. Scores of 25 or lower are associated with follow-through with decisions; whereas scores that exceed 38 are associated with delay in decision makingThe mean feeling unclear values in the intervention groups was 6.09 lower (8.50 to 3.67 lower) 

3704

(18 studies)

⊕⊕⊕⊕
high 1
Lower scores indicate feeling more clear about values

Participation in decision making: decision aid versus usual care - all studies - Practitioner controlled decision making

[soon after consultation with practitioner]

174 patients per 1000 103 patients per 1000 RR 0.66 (95%CI: 0.53 to 0.81)3234
(14 studies)
⊕⊕⊕⊝
moderate 1,3
Patient decision aids aim to increase patient involvement in making decisions. Lower proportion of practitioner controlled decision making is better.
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; RR: Risk Ratio
GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

Background

Many health treatment and screening decisions have no single 'best' choice. These types of decisions are considered 'preference-sensitive' because there is insufficient evidence about outcomes or there is a need to trade off known benefits and harms. Clinical Evidence classified 3000 treatments as: 50% having insufficient evidence, 24% likely to be beneficial, 7% requiring 'trade-offs between benefits and harms', 5% unlikely to be beneficial, 3% likely to be ineffective or harmful, and only 11% being clearly beneficial (Clinical Evidence 2013). Not only does one have to take into account the strength of the evidence, but even for 11% that show beneficial effects for populations, translating the probabilistic nature of the evidence for patients is necessary to reach an informed value-based decision. Patient decision aids are an intervention that can be used to present evidence (Brouwers 2010).

According to the International Patient Decision Aids Standards (IPDAS) Collaboration (Elwyn 2006; IPDAS 2005a; Joseph-Williams 2013), decision aids are evidence-based tools designed to help patients to participate in making specific and deliberated choices among healthcare options. Patient decision aids supplement (rather than replace) clinicians' counselling about options. The specific aims of decision aids and the type of decision support they provide may vary slightly, but in general they:

  1. Explicitly state the decision that needs to be considered;

  2. Provide evidence-based information about a health condition, the options, associated benefits, harms, probabilities, and scientific uncertainties;

  3. Help patients to recognize the values-sensitive nature of the decision and to clarify, either implicitly or explicitly, the value they place on the benefits, harms, and scientific uncertainties. (To accomplish this, patient decision aids may describe the options in enough detail that clients can imagine what it is like to experience the physical, emotional, and social effects and/or guide clients to consider which benefits and harms are most important to them).

Decision aids differ from usual health education materials because decision aids make explicit the decision being considered, and provide detailed, specific, and personalized focus on options and outcomes for the purpose of preparing people for decision making. In contrast, health education materials help people to understand their diagnosis, treatment, and management in general terms, but given their broader perspective these materials do not necessarily help them to participate in decision making. Many decision aids are based on a conceptual model or theoretical framework (Durand 2008; Mulley 1995; O'Connor 1998b; Rothert 1987).

Decision aids can be used before, during, or after a clinical encounter to enable patients to become active, informed participants. Decision aids can also facilitate shared decision making. Shared decision making is defined as a process by which a healthcare choice is made by practitioners together with the patient (Charles 1997; Makoul 2006) and is said to be the crux of patient-centred care (Weston 2001). However, the way information is provided by the practitioner may strongly affect how patients construct preferences (Hibbard 1997); thereby suggesting the need for standardized information such as patient decision aids. Patients who are more active in making decisions about their health have better health outcomes and healthcare experiences (Hibbard 2013; Kiesler 2006).

Decision aids have been developed primarily in Australia, Europe, North America, and the United Kingdom. Since 1999, there has been a rapid proliferation of patient decision aids. For example, decision aids from large-scale producers were accessed over 8 million times in 2006 (O'Connor 2007). In response to concerns about variability in the quality of patient decision aids, the IPDAS Collaboration reached agreement on criteria for judging their quality (Elwyn 2006). More than 100 researchers, practitioners, patients, and policy makers from 14 countries participated. Participants addressed three domains of quality: clinical content, development process, and evaluation of a patient decision aid's effectiveness. Subsequently, an international team of researchers reached consensus on a shorter set of qualifying and certifying criteria (Joseph-Williams 2013). The background papers informing the original IPDAS criteria were updated in 2012 (IPDAS 2013).

The ultimate goal of patient decision aids is to improve decision making in order to reach a high-quality decision. Over the past decade, there has been considerable debate about the definition of a 'good decision', when there is no single 'best' therapeutic action and choices depend on how patients value benefits versus harms (Briss 2004; O'Connor 2003a; O'Connor 1997b; Sepucha 2004). IPDAS reached agreement on criteria for judging "the things that you would need to observe in order to say that after using a patient decision aid, the way the decision was made was good and that the choice that was made was good" (Elwyn 2006; IPDAS 2005b; Sepucha 2013). The criteria were as follows:

  • choice made: the patient decision aid improves the match between the chosen option and the features that matter most to the informed patient;

  • decision-making process: the patient decision aid helps patients to: recognize that a decision needs to be made; know the options and their features; understand that values affect the decision; be clear about the option-features that matter most; discuss values with their practitioner; and become involved in their preferred ways.

Several individual studies examining the efficacy of decision aids have been published. There are annotated bibliographies, reports, and general reviews of decision aids (Bekker 1999; Bekker 2003; RTI 1997 Estabrooks 2000; Molenaar 2000; O'Connor 1997a; O'Connor 1999c; RTI 1997; Whelan 2002). We published the first systematic review of 17 randomized trials of decision aids in 1999 (O'Connor 1999b; O'Connor 2001a), followed by updates in 2003 with a total of 35 studies (O'Connor 2003b), in 2009 with a total of 55 studies (O'Connor 2009) and in 2011 with a total of 86 studies (Stacey 2011). Findings from this review have been used to inform clinical practice guidelines such as Patient Experience in Adult NHS Services (NCGC/NICE 2012) and Decision Support for Adults Living with Chronic Kidney Disease (RNAO 2009).

Several systematic reviews focus on the use of patient decision aids as one type of intervention to facilitate shared decision making in clinical practice (Coyne 2013; Duncan 2010; Elwyn 2013; Legare 2010).

Objectives

To assess the effects of decision aids for people facing treatment or screening decisions.

Methods

Criteria for considering studies for this review

Types of studies

We included all published studies using a randomized controlled trial (RCT) design comparing decision aids to no intervention, usual care, alternative interventions, or a combination.

Types of participants

We included studies involving people who were making decisions about screening or treatment options for themselves, for a child, or for an incapacitated significant other. We excluded studies in which participants were making hypothetical choices.

Types of interventions

We included studies that evaluated a patient decision aid as part of the intervention. Decision aids were defined as interventions designed to help people make specific and deliberative choices among options (including the status quo), by making the decision explicit and by providing (at the minimum) a) information on the options and outcomes relevant to a person's health status and b) implicit methods to clarify values. The aid also may have included: information on the disease/condition; costs associated with options; probabilities of outcomes tailored to personal health risk factors; an explicit values clarification exercise; information on others' opinions; a personalized recommendation on the basis of clinical characteristics and expressed preferences; and guidance or coaching in the steps of making and communicating decisions with others.

We excluded studies if interventions focused on: decisions about lifestyle changes, clinical trial entry, or general advance directives (e.g., do not resuscitate); education programs not geared to a specific decision; and interventions designed to promote adherence to or to elicit informed consent regarding a recommended option. We also excluded studies on decision aids that were not available to us, and so we could not determine that they provided the minimum criteria to qualify as a patient decision aid and their characteristics.

Types of outcome measures

To ascertain whether the decision aids achieved their objectives, we examined a broad range of positive or negative effects. Although the decision aids focused on diverse clinical decisions, many had similar objectives such as improving knowledge, accurate risk perceptions and participation in decision making. Many of these evaluation criteria mapped onto the International Patient Decision Aids Standards (IPDAS) criteria for evaluating the effectiveness of decision aids (Elwyn 2006; IPDAS 2005b; Sepucha 2013). A total list of outcomes, specified in advance of the review, included:

Primary outcomes
Evaluation criteria that map onto the IPDAS criteria
  • Attributes of the choice made: Does the patient decision aid improve the match between the chosen option and the features that matter most to the informed patient (demonstrated by outcomes such as knowledge, accurate risk perceptions, and chosen option congruent with their values)?

  • Attributes of the decision-making process: Does the patient decision aid help patients to: recognize that a decision needs to be made; know the options and their features; understand that values affect the decision; be clear about the option features that matter most; discuss values with their practitioner; and become involved in preferred ways?

Other decision-making process variables
  • Decisional conflict

  • Patient-practitioner communication

  • Participation in decision making

  • Proportion undecided

  • Satisfaction

Secondary outcomes
Behaviour
  • Choice (actual choice implemented; if not reported, the preferred option was used as a surrogate measure)

  • Adherence to chosen option

Health outcomes
  • Health status and quality of life (generic and condition-specific)

  • Anxiety, depression, emotional distress, regret, confidence

Healthcare system
  • Costs, cost effectiveness

  • Consultation length

  • Litigation rates

Search methods for identification of studies

Our search strategy for the review included:

  1. searching electronic medical and social science databases; and

  2. searching trial registries (World Health Organization, National Institutes of Health, Clinicaltrial.gov), the Internet, reference lists of included trials, and the Decision Aid Library Inventory (http://decisionaid.ohri.ca).

For this update, the search strategy (January 2009 to June 2012) was revised with the Trials Search Coordinator at the Cochrane Consumers and Communication Review Group. To validate the new search strategy, we conducted duplicate searches for 2009 and 2010 using the old and the new search strategies.

Therefore, the cumulative search of electronic databases is as follows:

  • MEDLINE (Ovid) (1966 to June 2012);

  • Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library; Issue 6 of 12, June 2012);

  • EMBASE (Ovid) (1980 to June 2012);

  • PsycINFO (Ovid) (1806 to June 2012); and

  • CINAHL (Ovid) (1982 to September 2008 (then in Ebsco when no longer indexed by Ovid; 2009 to June 2012)).

We present the search strategies in Appendix 1 and Appendix 2.

Data collection and analysis

In the 2009 Cochrane review (O'Connor 2009), the update used the International Patient Decision Aid Standards (IPDAS) criteria; this was a different strategy from that used in the earlier reviews (O'Connor 2001b; O'Connor 2003a). For the 2013 update reported here, we continued to use the IPDAS criteria and we focused only on new publications that had appeared since the previous publication (Stacey 2011).

Two review authors (CB, DS, RT, MB, MHR, NC, KE, BV, DR) screened all reports of randomized controlled trials (RCTs). Two of four research assistants extracted data independently and appraised studies using the Cochrane tool for judging risk of bias (SB, CC, RW, JW) (Higgins 2011, Chapter 8). No review author screened or extracted data from any of his or her own studies in this update or in any other versions of this review. We resolved inconsistencies by discussion with the principal investigator (DS).

First, we described study characteristics individually. According to the original protocol, the planned comparisons between groups were: a) usual care versus decision aids; and b) simple versus detailed decision aids. For studies in which there was more than one intervention group and one control group, we extracted data from the two groups that used a patient decision aid and provided the strongest contrast. For example, the group that used the most detailed decision aid was compared with the group who used the least detailed decision aid (detailed versus simple) or received usual care (decision aid versus control).

We pooled results across studies in cases where: a) similar outcome measures were used; and b) the effects were expected to be independent of the type of decision studied. For example, decision aids were expected to improve knowledge of options, benefits, and harms; to create accurate perceptions of benefits/harms; to reduce decisional conflict; and to enhance active participation in decision making. Therefore, we pooled data from the RCTs for these outcomes, if comparable measures were used. To facilitate pooling of data for some outcomes (e.g., knowledge, decisional conflict), the scores were standardized to range from 0 to 100 points. When analysing the effects of decision aids on choices, we pooled outcomes on more homogenous subgroups of decisions (choice of major surgery versus conservative options; testing or not; menopausal hormone therapy or not; etc.). In addition, we analyzed studies comparing usual care to decision aids separately from studies comparing simple to more detailed decision aids. For this update, we conducted a subgroup analysis of studies comparing decision aids for treatment decisions to usual care separately from studies comparing decision aids for screening decisions to usual care.

We used Review Manager 5.2.6 software (RevMan 2013) to estimate a weighted treatment effect (with 95% confidence intervals). For continuous measures, we used mean differences (MD); for dichotomous outcomes, we calculated pooled relative risks (RR). We analyzed all data with a random-effects model because of the diverse nature of the studies being combined. New for this update, we summarized all of the findings for the primary outcomes and rated the strength of evidence using GRADE (Andrews 2013), presenting these in a 'Summary of findings' table (Higgins 2011). Funnel plots were used to assess publication bias.

We performed post-hoc sensitivity analyses to examine the effect of excluding trials of lower methodological quality. The first analysis excluded trials that had 'high' risk of bias for any of the seven categories in the 'Risk of bias' assessment (Higgins 2011). The second sensitivity analysis excluded trials that had been rated as 'high' or 'unclear' in three or more of the seven categories.

In the 2009 update, the post-hoc analysis included the IPDAS effectiveness criteria to explore heterogeneity according to the following factors: the type of decision (treatment versus screening), the type of media of the decision aid (video/computer versus audio booklet/pamphlet), and the possibility of a ceiling effect based on usual-care scores (resulting in the removal of studies with lower knowledge and accurate perception scores, and the removal of studies with higher decisional conflict scores for the sub-scales measuring feeling uninformed and unclear values). We analyzed the effect of removing the biggest outlier(s) (defined by visual inspection of forest plots). Given that the post-hoc analysis did not alter the findings in the 2009 update (O'Connor 2009), the post-hoc analysis for the IPDAS effectiveness criteria was not re-conducted.

Results

Description of studies

The current version of our review updates our 2011 version (Stacey 2011, which included 86 studies) with 33 new studies: Allen 2010; Arterburn 2011; Berry 2013; Bjorklund 2012; Chambers 2012; de Achaval 2012; Evans 2010; Fagerlin 2011; Hanson 2011; Hess 2012; Jibaja-Weiss 2011; Labrecque 2010; Langston 2010; Legare 2011; Leighl 2011; Lewis 2010; Mann D 2010; Mann E 2010; Marteau 2010; Mathieu 2010; McCaffery 2010; Miller 2011; Montori 2011; Myers 2011; Raynes-Greenow 2010; Rubel 2010; Schroy 2011; Schwalm 2012; Sheridan 2011; Smith 2010; Solberg 2010; Steckelberg 2011; van Peperstraten 2010. We re-assessed four previously-included studies as 'excluded' due to quasi-experimental design (Dunn 1998; Herrera 1983; Phillips 1995) or the same intervention in both arms but delivered using different formats (Frosch 2003).

Results of the search

In total, we identified 38,069 unique citations from the electronic database searches and 247 citations from other sources. Of these, only 2072 citations focused on people's decision making (see Figure 1).

Figure 1.

Study flow diagram.

Of the 2072 citations identified, 358 appeared to be evaluations of patient decision aids. We excluded 186 of these upon close perusal of the paper (see Characteristics of excluded studies). The reasons for exclusion were: a) the study was not a randomized controlled trial; b) the decision was hypothetical, with participants not actually at a point of decision making; c) the intervention was not focused on making a choice; d) the intervention offered no decision support in the form of a decision aid or did not provide enough information about the decision aid; e) no comparison outcome data were provided; f) the study did not evaluate the decision aid; g) the study was a protocol; h) the decision aid was about clinical trial entry, lifestyle choice, or advanced care planning; or i) the study involved testing the presentation of decision aid, but with no difference in the content of the decision aid between study groups.

We also identified 30 ongoing RCTs through trial registration databases, personal contact, and published protocols in the electronic database searches (see references to Ongoing studies and table Characteristics of ongoing studies).

Using the old and new search strategies for 2009 and 2010, there was no difference in the included articles identified despite that the newer search strategy yielded fewer citations.

Included studies

The remaining 142 citations provided data on 115 studies that met our inclusion criteria, 33 of which are new for this update. The 115 RCTs, involving 34,444 participants, presented results from 9 countries (Australia (n = 15), Canada (n = 21), China (n = 1), Finland (n = 2), Germany (n = 5), Netherlands (n = 2), Sweden (n = 1), the United Kingdom (n = 14), the United States (n = 53), and Australia plus Canada (n = 1)). Study details are presented below and in the table Characteristics of included studies.

Unit of randomization

One hundred studies randomized individual patients and 15 randomized clusters. Allen 2010 randomized 12 company work sites; Goel 2001 randomized 57 surgeons; Hamann 2006 randomized 12 inpatient psychiatric units; Legare 2003 randomized 40 family physicians; Legare 2011 randomized 4 family medicine group practices; Lewis 2010 randomized 32 family medicine group practices; Loh 2007 randomized 30 general practitioners; McAlister 2005 randomized 102 primary care practices; Mullan 2009 randomized 40 clinicians; Nagle 2008 randomized 60 general practitioners; Solberg 2010 randomized 8 gynaecologist group practices; family-wise randomization was used for Wakefield 2008, Wakefield 2008a, and Wakefield 2008b; and Whelan 2004 randomized 27 surgeons. For 11 studies (Allen 2010; Goel 2001; Legare 2011; Loh 2007; Mullan 2009; Nagle 2008; Solberg 2010; Wakefield 2008; Wakefield 2008a; Wakefield 2008b; Whelan 2004), the cluster effect was taken into account in the published outcome data and the meta-analysis used published results. Although Hamann 2006 did not account for the cluster effect in the published outcome data, the way this study was reported did not allow us to include it in the meta-analysis, and, as such, we did not re-analyze the data and the study is reported separately. For McAlister 2005, meta-analysis was done applying the design effect (based on the published intra-cluster correlation coefficient (ICC)). For Legare 2003, the authors stated that for the Decisional Conflict Scale's results, “clustering had no impact on individual scores of women and, therefore, we present the results without adjustment”. The analysis in Lewis 2010 did not account for clustering.

Decision aids and comparisons

The 115 included studies evaluating decision aids focused on 46 different decisions (Table 1). The most common decisions were about prostate cancer screening (n = 15), colon cancer screening (n = 10), menopausal hormone therapy for menopausal women (n = 10), breast cancer genetic testing (n = 7), prenatal screening (n = 6), medication for atrial fibrillation (n = 3), and surgery (mastectomy breast cancer (n = 5), hysterectomy (n = 4), prostatectomy (n = 4)). New decisions included contraception (n = 2), medications for acute respiratory infection (n = 1), bariatric surgery (n = 1), long-term feeding tube placement (n = 1), labour analgesia (n = 1), embryo transplant (n = 1), influenza immunization (n = 1), access site for coronary angiography (n = 1), screening for diabetes (n = 2) or cervical cancer (n = 1), and stress test for chest pain (n = 1).

Table 1. Decision aids evaluated in the trials
StudyTopicAvailabilitySourceContact Information
Allen 2010Prostate cancer screeningNoAllen,Center for Community-Based Research, Dana-Farber Cancer Institute, Boston, MA, US, 2010requested access
Arterburn 2011Bariatric surgeryYesInformed Medical Decisions Foundation, MA,US, 2010informedmedicaldecisions.org/imdf_decision_aid/making-decisions-about-weight-loss-surgery/
Auvinen 2004Prostate cancer treatmentYesAuvinen, Helsinki, Finland, 1993included in publication
Barry 1997Benign prostate disease treatmentYesInformed Medical Decisions Foundation, MA,US, 2001informedmedicaldecisions.org/imdf_decision_aid/treatment-options-for-benign-prostatic-hyperplasia/
Bekker 2004Prenatal screeningYesBekker, Leeds, UK, 2003included in publication
Bernstein 1998Ischaemic heart disease treatmentYesInformed Medical Decisions Foundation, MA,US, 2002informedmedicaldecisions.org/imdf_decision_aid/treatment-choices-for-carotid-artery-disease/
Berry 2013Prostate cancer treatmentNoBerry, Phyllis F. Cantor Center, MA, USA, 2011donna_berry@dfci.harvard.edu 
Bjorklund 2012Antenatal Down syndrome screeningYesSödersjukhuset, Department of Obstetrics and Gynecology, Stockholm, Swedenvimeo.com/34600615/
Chambers 2012Healthcare personnel’s influenza immunizationYesA McCarthy.  Ottawa Influenza Decision Aid Planning Group, CA, 2008decisionaid.ohri.ca/decaids.html#oida
Clancy 1988Hepatitis B VaccineNoClancy, Richmond VA, US, 1983 
Davison 1997Prostate cancer treatmentNoDavison, Manitoba CA, 1992-1996 
de Achaval 2012Total
knee arthroplasty treatment
YesInformed Medical Decisions Foundation, MA,USinformedmedicaldecisions.org/imdf_decision_aid/treatment-choices-for-knee-osteoarthritis/
Deschamps 2004Hormone replacement therapyNoO'Connor, Ottawa, CA, 1996decisionaid.ohri.ca/decaids-archive.html
Deyo 2000Back surgeryYesInformed Medical Decisions Foundation, MA,US, 2001informedmedicaldecisions.org/imdf_decision_aid/managing-chronic-low-back-pain/
Dodin 2001Hormone replacement therapyNoO'Connor, Ottawa, CA, 1996decisionaid.ohri.ca/decaids-archive.html
Dolan 2002Colon cancer screeningNoDolan, Rochester NY, US, 1999 
Evans 2010Prostate cancer screeningYesElwyn, Cardiff, UKwww.prosdex.com
Fagerlin 2011Breast cancer preventionYesFagerlin, Ann Arbor, MI, US 
Fraenkel 2007Osteoarthritis knee treatmentNoFraenkel, New Haven CT, USauthor said DA never fully developed, all info in paper
Frosch 2008Prostate cancer screeningNoFrosch, Los Angeles, USScreenshots from author
Gattellari 2003Prostate cancer screeningYesGatellari , Sydney, AU, 2003included in publication
Gattellari 2005Prostate cancer screeningYesGatellari , Sydney, AU, 2003included in publication
Goel 2001Breast cancer surgeryNoGoel/Sawka, Toronto CAN, 2001 
Green 2001aBreast cancer genetic testingYesGreen, Hershey PA, US, 20001-800-757-4868 dwc@mavc.com
Green 2004Breast cancer genetic testingYesGreen, Hershey PA, US, 20001-800-757-4868 dwc@mavc.com
Hamann 2006Schizophrenia treatmentYesHamann, Munich, GERemailed by author (in German)
Hanson 2011Feeding options in advanced
dementia
YesMitchell, Tetroe, O'Connor; 2001 (updated 2008)decisionaid.ohri.ca/decaids.html#feedingtube
Heller 2008Breast reconstructionYesUniversity of Texas M.D. Anderson Cancer Center, Houston TX, US, 2003Disc mailed
Hess 2012Stress testing for chest painYesHess, Rochester, MN, US, 2012Included in publication
Hunter 2005Prenatal screeningNoHunter, Ottawa, CA, 2000decisionaid.ohri.ca/decaids-archive.html
Jibaja-Weiss 2011Breast cancer treatmentYesJibaja-Weiss, Baylor College of Medicine, 2010www.bcm.edu/patchworkoflife
Johnson 2006Endodontic treatmentYesJohnson, Chicago, US, 2004Included in publication
Kasper 2008Multiple SclerosisNoJürgen Kasper 
Kennedy 2002Abnormal uterine bleeding treatmentNoKennedy/Coulter, London UK, 1996 
Krist 2007Prostate cancer screeningYesKrist, Fairfax VA, USwww.familymedicine.vcu.edu/research/misc/psa/index.html
Kuppermann 2009Prenatal screeningNoKuppermann, San Francisco CA, USComputerized tool
Labrecque 2010VasectomyYesLabrecque, Quebec City, CA, 2010www.vasectomie.net (in French)
Lalonde 2006Cardiovascular health treatmentNoLalonde, Ottawa, CA, 2002decisionaid.ohri.ca/decaids-archive.html
Langston 2010Contraceptive method choiceYesWorld Health Organization, 2005www.who.int/reproductivehealth/publications/family_planning/9241593229index/en/index.html
Laupacis 2006Pre-operative autologous blood donationNoLaupacis, Ottawa, CA, 2001decisionaid.ohri.ca/decaids-archive.html
Legare 2003Hormone replacement therapyNoO'Connor, Ottawa, CA, 1996decisionaid.ohri.ca/decaids-archive.html
Legare 2008aNatural health productsNoLegare, Quebec City, CA, 2006 
Legare 2011Use of antibiotics for acute
respiratory infections
YesLegare, Quebec City, CA, 2007www.decision.chaire.fmed.ulaval.ca/index.php?id=192&L=2
Leighl 2011Advanced colorectal cancer chemotherapyYesPrincess Margaret Hospital, Toronto, 2011Natasha.Leighl@uhn.on.ca.
Lerman 1997Breast cancer genetic testingNoLerman/Schwartz, Washington DC, US, 1997 
Leung 2004Prenatal screeningNoLeung, Hong Kong, China, 2001 
Lewis 2010Colorectal cancer
screening
YesLewis, University of North Carolina, Chapel Hill, NC, USA, 2010decisionsupport.unc.edu/CHOICE6/
Loh 2007Depression treatmentYesLoh, Freiburg, GER(emailed to us by author - in German)
Man-Son-Hing 1999Atrial fibrillation treatmentNoMcAlister/Laupacis, Ottawa CA, 2000decisionaid.ohri.ca/decaids-archive.html
Mann D 2010Diabetes treatment - statinsYesMontori, Rochester MN, USmayoresearch.mayo.edu/mayo/research/ker_unit/form.cfm
Mann E 2010Diabetes
screening
YesMarteau, King's College London, London, England, 2010Additional file 2 of publication
Marteau 2010Diabetes
screening
YesMarteau, King's College London, London, England, 2010Provided by author, same DA as Mann E 2010
Mathieu 2007MammographyYesMathieu, Sydney, AU,DA emailed by author 
Mathieu 2010MammographyYesMathieu, University of Sydney, AUS, 2010http://www.psych.usyd.edu.au/cemped/com_decision_aids.shtml
McAlister 2005Atrial fibrillation treatmentNoMcAlister/Laupacis, Ottawa CAN, 2000decisionaid.ohri.ca/decaids-archive.html
McBride 2002Hormone replacement therapyYes, update in progressSigler/Bastien, Durham NC, US, 1998basti001@mc.duke.edu
McCaffery 2010Screening after mildly abnormal pap smearYesScreening & test evaluation program, School of public health, University of Sydney 2007kirstenm@health.usyd.edu.au
Miller 2005BRCA1/BRCA2 gene testingNoMiller, Fox Chase PA, US 
Miller 2011Colorectal
cancer screening
YesUniversity of North Carolina, Chapel Hill, NC, USA, 2007intmedweb.wakehealth.edu/choice/choice.html (no longer available)
Montgomery 2003Hypertension treatmentNoMontgomery, UK, 2000 
Montgomery 2007Birthing options after caesareanYesMontgomery, Bristol, UK, last update 2004www.computing.dundee.ac.uk/acstaff/cjones/diamond/Information.html
Montori 2011Osteoporosis treatmentYesMontori, Mayo Foundation for Medical Education and Research, 2007shareddecisions.mayoclinic.org/decision-aids-for-diabetes/other-decision-aids/
Morgan 2000Ischaemic heart disease treatmentYesInformed Medical Decisions Foundation, MA,US, 2002informedmedicaldecisions.org/imdf_decision_aid/treatment-choices-for-carotid-artery-disease/
Mullan 2009Diabetes treatmentYesMontori or Mayo Foundation?, Rochester MN, US,included in publication
Murray 2001aBenign prostate disease treatmentYesInformed Medical Decisions Foundation, MA,US, 2001informedmedicaldecisions.org/imdf_decision_aid/treatment-options-for-benign-prostatic-hyperplasia/
Murray 2001bHormone replacement therapyNo, update in progressInformed Medical Decisions Foundation, MA,USinformedmedicaldecisions.org/imdf_decision_aid/treatment-choices-for-managing-menopause/
Myers 2005aProstate cancer screeningNoMyers, Philadelphia PA, US, 1999 
Myers 2011Prostate cancer screeningYesMyers, Philadelphia PA, 1999 
Nagle 2008Prenatal screeningYesNagle, Victoria, AUwww.mcri.edu.au/Downloads/PrenatalTestingDecisionAid.pdf
Nassar 2007Birth breech presentationYesNassar, West Perth WA, AUsydney.edu.au/medicine/public-health/shdg/resources/decision_aids.php
O'Connor 1998aHormone replacement therapyNoO'Connor, Ottawa CA, 1996decisionaid.ohri.ca/decaids-archive.html
O'Connor 1999aHormone replacement therapyNoO'Connor, Ottawa CA, 1996decisionaid.ohri.ca/decaids-archive.html
Oakley 2006Osteoporosis treatmentNoCranney, Ottawa CA, 2002decisionaid.ohri.ca/decaids-archive.html
Ozanne 2007Breast cancer preventionNoOzanne, Boston MA, US, 
Partin 2004Prostate cancer screeningYesInformed Medical Decisions Foundation, MA,US, 2001informedmedicaldecisions.org/imdf_decision_aid/deciding-if-the-psa-test-is-right-for-you/
Pignone 2000Colon cancer screeningYesPignone, Chapel Hill NC, US, 1999www.med.unc.edu/medicine/edusrc/colon.htm
Protheroe 2007Menorrhagia treatmentNoProtheroe, Manchester, UKcomputerized decision aid, Clinical Guidance Tree - no longer in existence, author sent chapter in thesis
Raynes-Greenow 2010Labour
analgesia
YesRaynes-Greenow, Sydney,Australia, 2004http://www.psych.usyd.edu.au/cemped/com_decision_aids.shtml
Rostom 2002Hormone replacement therapyNoO'Connor, Ottawa CA, 1996decisionaid.ohri.ca/decaids-archive.html
Rothert 1997Hormone replacement therapyNo, update in progressRothert, East Lansing MI, US, 1999 
Rubel 2010Prostate cancer screeningNoCenters for Disease Control and Prevention (CDC), US, 2010[No longer available]
Ruffin 2007Colorectal cancer screeningYesRegents of the University of Michigan (copyright info), Ann Arbor MI, US, 2006colorectalweb.org
Schapira 2000Prostate cancer screeningYesSchapira, Milwaukee WI, US, 1995mschap@mcw.edu
Schapira 2007Hormone replacement therapyYesSchapira, Milwaukee WI, UScomputer-based DA
Schroy 2011Colorectal
cancer screening
YesSchroy III, Boston, USAPaul.schroy@bmc.org
Schwalm 2012Coronary angiogram access siteYesSchwalm, Hamilton, ON, Canada, 2009http://www.phri.ca/workfiles/studies/presentations/PtDA%20Vascular%20Access%2023-May-2012.pdf
Schwartz 2001Breast cancer genetic testingNoSchwartz/Lerman, Washington DC, US, 1997 
Schwartz 2009BRCA mutation prophylactic surgeryNoSchwartz, Washington DC, US 
Sheridan 2006Cardiovascular preventionYesSheridan, Chapel Hill, NC, UShttp://www.med-decisions.com/cvtool/
Sheridan 2011Coronary heart
disease prevention
YesSheridan, University of North Carolina at Chapel Hill, Division of General Internal Medicine, North Carolina, US, 2011http://www.med-decisions.com/h2hv3/
Shorten 2005Birthing options after previous caesareanYes (updated 2006)Shorten, Wollongong, AU, 2000ashorten@uow.edu.au or www.capersbookstore.com.au/product.asp?id=301
Smith 2010Bowel
cancer screening
YesSmith, Sydney, AU 2008sydney.edu.au/medicine/public-health/shdg/resources/decision_aids.php
Solberg 2010Uterine fibroid treatmentYesInformed Medical Decisions Foundation, MA,US, 2006informedmedicaldecisions.org/imdf_decision_aid/treatment-choices-for-uterine-fibroids/
Steckelberg 2011Colorectal cancer screeningYesSteckelberg, Hamburg, Germany 
Street 1995Breast cancer surgeryNoStreet, College Station TX, US, 1995 
Thomson 2007Atrial fibrillation treatmentYesThomson, Newcastle Upon Thyne, UKdisc sent by mail
Tiller 2006Ovarian cancer risk managementNoTiller, Randwick NSW, AU 
Trevena 2008Colorectal cancer screenYesTrevena, Sydney, AUsydney.edu.au/medicine/public-health/shdg/resources/decision_aids.php
van Peperstraten 2010Embryos transplantYesRadboud University Nijmegen Medical Centre; 2006www.umcn.nl/ivfda-en
Vandemheen 2009Cystic Fibrosis referral transplantYesAaron, Ottawa ON, CA, 2009 (last update 2011)decisionaid.ohri.ca/decaids.html#cfda
van Roosmalen 2004BRCA1/2 mutation: prophylactic surgeryYesvanRoosmalen, Netherlands, 1999see publication
Vodermaier 2009Breast cancer surgeryYesVodermaier, Vancouver BC, CAreceived by email (in German)
Volk 1999Prostate cancer screeningYesInformed Medical Decisions Foundation, MA,US, 1999informedmedicaldecisions.org/imdf_decision_aid/deciding-if-the-psa-test-is-right-for-you/
Volk 2008Prostate cancer screeningNoVolk, Houston TX, US 
Vuorma 2003Menorrhagia treatmentNoVuorma, Helsinki Finland, 1996 
Wakefield 2008Colorectal cancer screeningYesWakefield, Sydney, AU,www.genetics.edu.au/Information/PublicationsBrochuresandPamphlets/Understanding%20Genetic%20Tests%20for%20Lynch%20Syndrome
Wakefield 2008aBreast cancer genetic testingYesWakefield, Sydney, AU, 
Wakefield 2008bBreast cancer genetic testingYesWakefield, Sydney, AU, 
Watson 2006Prostate cancer screeningYesOxford, UKincluded in publication
Weymiller 2007Diabetes mellitus type 2 treatmentYesMontori, Rochester MN, USmayoresearch.mayo.edu/mayo/research/ker_unit/form.cfm 
Whelan 2003Breast cancer chemotherapyYesWhelan, Hamilton CA, 1995included in publication
Whelan 2004Breast cancer surgeryYesWhelan, Hamilton CA, 1997included in publication
Wolf 1996Prostate cancer screeningYesWolf, Charlottesville VA, US, 1996Script in publication
Wolf 2000Colon cancer screeningYesWolf, Charlottesville VA, US, 2000Script in publication
Wong 2006Pregnancy terminationNoBekker, Leeds, UK, 2002 

The decision aids used a variety of formats and were compared to a variety of control interventions (e.g., usual care, no intervention, guideline, placebo intervention). We noted the nature of usual care when reported (see table Characteristics of included studies). For this review, we have grouped control interventions and refer to them as usual care unless the intervention meets the definition of a patient decision aid.

According to the definition of a patient decision aid, all of the studies evaluated patient decision aids that included information about the options and outcomes, and provided at least implicit values clarification. Most patient decision aids included information on the clinical problem (91.3%) as well as outcome probabilities (87.8%). Fewer patient decision aids provided guidance in the steps of decision making (62.6%), explicit methods to clarify values (59.1%), and/or examples of others' experiences (50.4%) (see table Characteristics of included studies).

The comparison interventions ranged from no intervention through to usual care, and general information through to simple decision aids that varied in their number of elements. However, a simple decision aid had to meet the minimum definition of a decision aid (see table Characteristics of included studies).

Risk of bias in included studies

Details on the ratings and rationale for risk of bias are in the Characteristics of included studies table and displayed in Figure 2 and Figure 3. The risk of bias was summarized in Table 2 based on the primary outcomes.

Figure 2.

Risk of bias summary as percentages across all included studies.

Figure 3.

Risk of bias summary for each included study.

Table 2. Risk of bias by primary outcome
OutcomeKnowledgeAccurate risk perceptionValue-choice agreementUninformedUnclear valuesParticipation - practitioner controlled
Total studiesn = 42n = 19n = 13n = 22n = 18n = 14

Random sequence

generation

low35 (83.3%)8 (42.1%)7 (53.8%)19 (86.4%)17 (94.4%)12 (87.7%)
unclear7 (16.7%)11 (57.9%)6 (46.2%)3 (13.6%)1 (5.6%)2 (14.3%)
high000000

Allocation

concealment

low30 (71.4%)12 (63.2%)11 (84.6%)20 (90.9%)17 (94.4%)10 (71.4%)
unclear12 (28.6%)7 (36.8%)2 (15.4%)2 (9.1%)1 (5.6%)4 (28.6%)
high000000

Incomplete

outcome data

low26 (61.9%)11 (57.9%)11 (84.6%)15 (68.2%)13 (72.2%)10 (71.4%)
unclear16 (38.1%)8 (42.1%)2 (15.4%)7 (31.8%)5 (27.8%)4 (28.6%)
high000000

Selective

reporting

low15 (35.7%)7 (36.8%)6 (46.2%)9 (40.9%)8 (44.4%)4 (28.6%)
unclear27 (64.3%)12 (63.2%)7 (53.8%)13 (59.1%)10 (55.6%)10 (71.4%)
high000000
Other biaslow34 (81.0%)14 (73.7%)11 (84.6%)19 (86.4%)17 (94.4%)11 (78.6%)
unclear7 (16.7%)5 (26.3%)2 (15.4%)3 (13.6%)1 (5.6%)3 ( 21.4%)
high1 (2.4%)00000

Blinding of participants

and personnel

low9 (21.4%)1 (5.3%)2 (15.4%)3 (13.6%)2 (11.1%)2 (14.3%)
unclear31 (73.8%)17 (89.5%)11 (84.6%)18 (81.8%)15 (83.3%)9 (64.3%)
high2 (4.8%)1 (5.3%)01 (4.5%)1 (5.6%)3 (21.4%)

Blinding of outcome

assessment

low41 (97.6%)19 (100%)13 (100%)22 (100%)18 (100%)13 (92.9%)
unclear1 (2.4%)00001 (7.1%)
high000000

Allocation

For assessing risk of selection bias, random sequence generation was rated as being at low risk of bias in 81 of 115 studies (70.4%) and unclear risk of bias in 34 studies (29.6%). Allocation concealment was rated as being at low risk of bias in 72 of 115 studies (62.6%) and unclear risk of bias in 43 studies (37.4%).

Blinding

Blinding of participants and personnel was rated as being at low risk of bias in 21 studies (18.3%), unclear risk of bias in 91 studies (79.1%), and high risk of bias in 3 studies (2.6%). High risk of bias was due to lack of blinding of physicians who were involved with patients randomized to both the patient decision aid and alternative interventions (Auvinen 2004; Krist 2007; Man-Son-Hing 1999).

Blinding of outcome assessment was low risk of bias in 109 studies (94.8%) and unclear risk of bias in 6 studies (5.2%).

Incomplete outcome data

Incomplete outcome data which could lead to attrition bias were adequately described in 78 studies (67.8%), inadequately described to judge risk of bias in 36 studies (31.3%), and for 1 study (0.9%) there was high risk of bias (Chambers 2012). In Chambers 2012, few participants in the intervention arm compared to usual care completed the study (65% versus 77%).

Selective reporting

Of 115 studies, 33 (28.7%) were rated as low risk of bias because the protocol was registered publicly and 82 (71.3%) were rated as being at unclear risk of bias for this domain.

Other potential sources of bias

Of 115 studies, 90 (78.3%) did not indicate any other potential sources of bias, 21 (18.3%) did not provide an adequate description to judge other potential sources of bias, and 4 (3.4%) discussed other potential risks of bias. Clancy 1988 describes a potential for selection bias, given that non-randomized medical residents were added to the decision analysis group and that there was a low response rate among those offered decision analysis. In a study focused on the decision about menopausal hormone therapy for menopausal women. Rostom 2002 reported that there was a potential for bias, given that there was an uneven balance of pre-menopausal women who were not appropriate for hormone therapy with more women in the detailed decision aid group. Hamann 2006 and Lewis 2010 did not account for clustering in the analysis.

Effects of interventions

See: Summary of findings for the main comparison

In addition to Summary of findings for the main comparison, see the Data and analyses figures for pooled data and Additional tables 3 to 22 for outcome data that were not pooled.

A) Attributes of the choice made:

Does the patient decision aid improve the match between the chosen option and the features that matter most to the informed patient?

The randomized controlled trials (RCTs) used three measures that correspond to this definition: knowledge, accuracy of risk perceptions, and chosen option congruent with their values.

Knowledge

Seventy-six of the 115 studies (66.1%) assessed the effects of decision aids on knowledge; 56 of these compared decision aids to usual care (74%) and 20 compared detailed decision aids to simple decision aids (26%). The studies' knowledge tests were based on information contained in the decision aid. The proportion of accurate responses was transformed to a percentage scale ranging from 0% (no correct responses) to 100% (perfectly accurate responses).

For patient decision aids compared to usual care (n = 42): people exposed to decision aids had higher average knowledge scores (MD 13.34%; 95% CI 11.17 to 15.51; Analysis 1.1). Fourteen additional studies that compared decision aids to usual care presented knowledge data that could not be included in the pooled outcome (see Table 3). Six of these studies reported statistically-significantly higher knowledge for those exposed to the decision aid compared to usual care (Evans 2010; Hamann 2006; Nagle 2008; Partin 2004; Trevena 2008; Watson 2006).

Table 3. Knowledge
  1. CI: confidence interval; DA: decision aid; SD: standard deviation

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA versus usual care
Evans 201012 true or false questions; scores ranging from -12 to +12immediately post894.91032.17P < 0.001
Hamann 20067-item multiple choice knowledge test (unable to standardize results)on discharge (˜ 1 month)4915 (4.4 SD)5810.9 (5.4 SD)P = 0.01
Heller 200812-item multiple choicepre-operatively6614%*678%*

*mean increase from baseline

P = 0.02

Legare 2008a10-item yes/no/unsure general knowledge test about natural health products (not specific to outcomes of options)change scores from baseline to 2 weeks43

0.86 ± 1.77

P = 0.002

410.51 ± 1.47 P = 0.031No difference between groups (P = 0.162)
Mann D 201014 items surveyimmediately post    No difference in level of knowledge between groups
Mathieu 20079 item - 4 concept questions and 5 numeric questions 351 357 Significantly higher mean increase for the intervention group (2.62 ) compared to control group (0.68) from baseline, P < 0.001
Miller 20058 items survey2-week, 2-month, and 6-month follow-ups    Intervention type had no impact on general or specific knowledge
Nagle 2008Good level knowledge was scored higher than the mid point of the knowledge scale (greater than 4)     88% (147/167) in DA group compared to 72% (123/171) pamphlet group. Odds ratio (3.43 95%CI 1.79 to 6.58)
Ozanne 2007Change in knowledge from baselinepost-test1548% to 64%1545% to 57%change in knowledge score was significant for decision aid (P = 0.01) but not control (P = 0.13)
Partin 200410-item knowledge index score2 weeks3087.442906.9P = 0.001
Rubel 201024-items adapted from existing prostate cancer knowledge measuresimmediately post100 100 the total mean standardized knowledge score was 84.38 (SD 12.38).
Trevena 2008Adequate knowledge (positive score: understanding benefits/harms)1 month13428/1341378/137P = 0.0001
Watson 200612-item true/false/don't knowpost-test46875% (range 0 to 100)52225% (range 0 to 100)P < 0.0001
Weymiller 200714-item - 9 addressed by decision aid; 5 were notimmediately post52 46 Mean difference between groups 2.4 (95% CI 1.5 to 3.3) P < 0.05 (when decision aid administered during the consultation only - not if prior to the consultation)
Detailed versus simple DA
Volk 2008 2 weeks233 223 Significant improvement in knowledge with no difference between groups (entertainment decision aid or audio-booklet)

One study (Weymiller 2007) reported a higher mean difference when the decision aid was administered during the consultation, but not if administered before the consultation. Mann D 2010 and Miller 2005 reported no difference between groups. Four other studies (Heller 2008; Legare 2008a; Mathieu 2007; Ozanne 2007) reported a change in knowledge from baseline: two found a statistically-significant improvement in the decision aid group (Heller 2008; Mathieu 2007); Ozanne 2007 reported a statistically-significant improvement in the decision aid group (P = 0.01) but not in the control group (P = 0.13); and Legare 2008a reported a statistically-significant improvement in both the decision aid group (P = 0.002) and the control group (P = 0.031) but no difference between groups. Rubel 2010 reported knowledge scores with no comparisons. The funnel plot for knowledge as an outcome in studies comparing decision aid to usual care shows low risk for publication bias (Figure 4).

Figure 4.

Funnel plot of comparison: 1 Knowledge, outcome: 1.1 Knowledge: DA vs usual care - all studies.

For detailed compared to simple decision aids (n = 19): people exposed to detailed decision aids had higher average knowledge scores but this effect was smaller (MD 5.52%; 95% CI 3.90 to 7.15; Analysis 1.4). One additional study that compared a detailed to simple decision aid (Volk 2008) reported significant improvement in all groups from baseline but no significant differences between groups (see Table 3).

Accurate risk perceptions (i.e. perceived probabilities of outcomes)

Of 115 studies, 25 (21.7%) examined the effects of including probabilities in decision aids on the accuracy of patients' perceived probabilities of outcomes (see Analysis 2.1; Table 4). Of these 25 studies, 15 measured perceived probabilities as percentages (see Analysis 2.4), 4 gauged probabilities in words (see Analysis 2.5), and 6 were not able to be pooled (Table 4). Perceived outcome probabilities were classified according to the percentage of individuals whose judgments corresponded to the scientific evidence about the chances of an outcome for similar people. For studies that elicited risk perceptions using multiple items, the proportion of accurate risk perceptions was averaged.

Table 4. Accurate risk perceptions
  1. CI: confidence interval; DA: decision aid; OR: odds ratio; SD: standard deviation

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA versus usual care
Hanson 2011Expectation of benefit index 11 items score from 1 to 4 with lower score indicating better knowledgepost (after reviewing DA)1272.31292.6P = 0.001
Mann E 20103 of 8 multiple choice items in the knowledge test (question 4, 5, 7)2 weeks post    total knowledge reported only
Mathieu 20105 item numerical questions (max = 5)post1133.021892.45P < 0.001
Miller 2005 2-week, 2-month, and 6-month follow-ups    Intervention type had no impact on risk perceptions
Smith 20108 numerical questions (max = 8) 3572.93 (SD 2.91)1730.58 (SD1.28)P < 0.001
Weymiller 2007 immediately52 46 

Difference between group

OR 22.4 (95% CI 5.9 to 85.8) when decision aid administered during the consultation only (not if prior to)

OR 6.7 (95% CI 2.2 to 19.7) when the decision aid administered prior to or during the consultation

People who received a patient decision aid with descriptions of outcome probabilities were more likely to have accurate risk perceptions than those who did not receive this information; the pooled relative risk (RR) of having accurate risk perceptions was 1.82 (95% CI 1.52 to 2.16, n = 19; Analysis 2.1). The pooled RR for probabilities measured as numbers was 2.00 (95% CI 1.65 to 2.43, n = 15; Analysis 2.4) and the pooled RR for probabilities gauged in words was 1.31 (95% CI 1.13 to 1.52, n = 4; Analysis 2.5). Six studies reported results that could not be pooled (see Table 4). Hanson 2011; Mathieu 2010; and Smith 2010 reported a statistically-significant improvement in accurate perceptions of outcomes for the decision aid group compared to usual care, and Miller 2005 reported no impact on risk perception. In another study, Weymiller 2007 reported a statistically-significant difference in the accurate perception of baseline risks in the group receiving a decision aid with probabilities compared to the usual care group, when the decision aid was administered during the consultation but not when it was administered before the consultation. The difference in accurate estimations of the potential absolute risk reduction with statin drugs was also statistically significant between the decision aid and usual care groups, and this difference remained significant regardless of the timing of delivery. Although three of eight knowledge test items measured accurate risk perceptions (Mann E 2010), results were presented for total knowledge and not individual items.The funnel plot for accurate risk perception as an outcome in studies comparing decision aid to usual care shows low risk for publication bias (Figure 5).

Figure 5.

Funnel plot of comparison: 2 Accurate risk perceptions: Decision aid with outcome probabilities vs no outcome probability information, outcome: 2.1 Accurate risk perceptions - all studies.

Chosen option congruent with values

Of 115 studies, 20 (17.4%) measured congruence between with the chosen option and their values; however, 7 did not present quantitative data to permit pooling across studies (Arterburn 2011; Frosch 2008; Legare 2008a; Lerman 1997; Rothert 1997; Solberg 2010; Vandemheen 2009; see Table 5).

Table 5. Values congruent with chosen option
  1. DA: decision aid; SD: standard deviation

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA versus usual care
Arterburn 2011Percent match procedures described by Sepucha et al (2007; 2008). For values items were most predictive and used to specify logistic models to estimate predicted probability of selecting surgery > 0.5.post intervention75 77 The intervention group experienced a more rapid early improvement in value concordance immediately after the intervention compared to control, see Figure 2.
Frosch 2008Concordance between patient's preferences and values for potential outcomes related to the decision and the choice madewithin weeks155 151 Men assigned to the decision aid who chose not to have a PSA test rated their concern about prostate cancer lower than did men who requested a PSA test. Men assigned to usual care provided similar ratings of concern about prostate cancer regardless of their PSA decision. There was no statistically significant difference between groups.
Legare 2008a      Women valuing of non chemical aspect of nature health products was positively associated with their choice of nature health products, P = 0.006.
Lerman 1997Association between values and choice------------------------------No difference; between group differences were not reported
Vandemheen 2009Congruence between personal values and decision3 weeks70 70 Patient choices were consistent with their values across both randomised groups
Detailed versus simple DA
Rothert 1997Correlation between expected utilities and their likelihood of taking hormones------------------------------Simple DA showed lower correlations between expected value of hormones and likelihood of taking hormones than did more detailed DA
Solberg 2010My decision was consistent with my personal values. (Likert Scale, ranged from 1-5)4-5 weeks after intervention10387.5 (SD 20)11280 (SD 22.5)P < 0.01
multi-nomial logistic regression analysis     No significant difference between groups

Nine of these studies used the Multi-Dimensional Measure of Informed Choice (Bjorklund 2012; Mathieu 2007; Mathieu 2010; Nagle 2008; Smith 2010; Trevena 2008; Wakefield 2008; Wakefield 2008a; Wakefield 2008b), which assesses the extent to which the choice is based on relevant knowledge, is consistent with a person's values/attitudes, and is behaviorally implemented (Michie 2002). These studies operationalized the measure in terms of knowledge test scores higher than the mid-point, attitude scale scores higher than the mid-point, and choice being congruent with attitude.

People who received a patient decision aid with an explicit values clarification exercise were more likely to achieve a chosen option congruent with their values: the pooled RR was 1.51 (95% CI 1.17 to 1.96, n = 13; Analysis 3.1). A sub-analysis of studies using the Multi-Dimensional Measure of Informed Choice revealed a pooled RR of 1.35 (95% CI 1.12 to 1.61, n = 9). Of the seven studies that were not pooled, Arterburn 2011 reported that, compared to the control group, those exposed to the decision aid experienced a more rapid early improvement of value concordance immediately after exposure. Legare 2008a reported that women's valuing of the non-chemical aspect of natural health products was positively associated with their choice of natural health products in managing menopausal symptoms (P = 0.006). Rothert 1997 reported higher correlations between the expected value of hormones and the likelihood of taking hormones in women exposed to the detailed decision aid compared those exposed to the simple decision aid. No differences between groups were reported in the other studies (Frosch 2008; Lerman 1997; Solberg 2010; Vandemheen 2009; see Table 5). However, Frosch 2008 observed that men exposed to the decision aid who chose not to have a prostate-specific antigen (PSA) test rated their concern about prostate cancer lower than men who requested a PSA test, while men assigned to the usual care group provided similar ratings of concern regardless of their PSA choice. The funnel plot for congruence between the chosen option and their values as an outcome in studies comparing decision aid to usual care shows low risk for publication bias (Figure 6).

Figure 6.

Funnel plot of comparison: 3 Values congruent with chosen option, outcome: 3.1 Values congruent with chosen option - all studies.

B) Attributes of the decision process:

Does the patient decision aid help patients to: recognize that a decision needs to be made; know the options and their features; understand that values affect the decision; be clear about the option features that matter most; discuss values with their practitioner; and become involved in their preferred ways?

In relation to the International Patient Decision Aids Standards (IPDAS) decision process criteria, no studies evaluated the extent to which patient decision aids helped patients to recognize that a decision needs to be made or understand that values affect the decision.

Some studies measured patients' self-reports about feeling informed and clear about personal values. The measures used to evaluate these two criteria were two sub-scales of the previously validated Decisional Conflict Scale (DCS) (O'Connor 1995).

Decisional conflict

Of 115 studies, 58 (50.4%) evaluated overall decisional conflict using the DCS (O'Connor 1995). The DCS is reliable, discriminates between those who make or delay decisions, is sensitive to change, and discriminates between different decision support interventions (Morgan 2000; O'Connor 1995; O'Connor 1998a). The scale measures the constructs of overall decisional conflict and the particular factors contributing to uncertainty (e.g., feeling uncertain, uninformed, unclear about values, and unsupported in decision making). A final sub-scale measures perceived effective decision making. The scores were standardized to range from 0 (no decisional conflict) to 100 points (extreme decisional conflict). Scores of 25 or lower are associated with follow-through with decisions, whereas scores that exceed 38 are associated with delay in decision making (O'Connor 1998a). When decision aids are compared to usual care, a negative score indicates a reduction in decisional conflict, which is in favour of the decision aid.

Analysis 4.1.6 summarizes the decisional conflict results for the 28 studies that compared decision aids to usual care, and Analysis 4.4.6 summarizes the results for the 17 studies that compared detailed to simple decision aids. Fifteen studies that were not able to be pooled are reported in Table 6 and Table 7.

Table 6. Decisional Conflict Score
  1. CI: confidence interval; DA: decision aid; DCS: decisional conflict scale; IVF: in vitro fertilisation; SD: standard deviation

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA versus usual care
Arterburn 2011Total Decisional Conflict- change from baseline (standardised values)immediately post75mean (-20) SD (19.44)77mean (-11.8) SD (22.83)P = 0.03
Berry 2013Decisional conflict scaleuncertainty -3.61 units  P = 0.04
uninformed    No significant difference
unclear values -3.57 units  P = 0.002
unsupported    No significant difference
Ineffective decision    No significant difference
total -1.75 units  P = 0.07
Fagerlin 2011Decisional conflict scaleimmediately post    DCS was higher in the intervention group compared to control, P < 0.001.
Frosch 2008Decisional conflict - sub-scales onlyFeeling uninformed15523.3715129.68P < 0.05
Feeling unclear values15532.2515137.93P < 0.05
Feeling supported15530.5115135.21P < 0.05
Feeling uncertain155 151 No difference
Effective decisions155 151 No difference
Krist 2007Decisional conflictimmediately after office visit1961.54751.58No difference
Leighl 2011

Decisional conflict scale

median (range)

1-2 weeks post intervention10726 (range 0-79)10026 (range 0-67)No difference
Mathieu 2010Based on approaches suggested by Marteau et al. (informed choice)immediately after intervention9171%11064%P = 0.24
Ozanne 2007Decisional conflictpost consultation15 15 Both groups showed lower decisional conflict post-consultation (P < 0.001) but no difference between groups
Rubel 2010Decisional conflictimmediately post    The total mean score was 24.5 with a SD of 15.25 (n=200)
Schwartz 2009aDecisional conflict12 of 16 items of the original scale    Significant longitudinal impact of the decision aid was moderated by baseline decision status; decision aid led to significant decreases in decisional conflict for those who were undecided at the time of randomisation
Thomson 2007Decisional conflictpost consultation53 56 Difference between decision aid and control group were -0.18 (95% CI -0.34 to -0.01). P = 0.036
3-months post51 55 Difference between decision aid and control group were -0.15 (95% CI -0.37 to 0.06), no significant difference.
van Peperstraten 201015 item questionnaire (1-5) - satisfaction-uncertaintypost intervention, pre IVF12472.512875P = 0.76
15 item questionnaire (1-5) - informed (includes some items from DCS).post intervention, pre IVF12477.512887.5P = 0.001
Weymiller 2007Decisional conflictimmediately post52 46 

Mean difference indicates statistically significantly lower decisional conflict for decision aid compared to usual care.

Total DCS -10.6 (-15.4 to -5.9)

Uncertain -12.8 (-18.4 to -7.3)

Informed -17.3 (-22.6 to -12.0) if administered during consult

-6.6 (-14.3 to -1.1) if administered prior to consult

Values clarity -8.5 (-15.7 to -1.3)

Support -9.4 (-14.8 to -3.9)

Effective decision -10.0 (-15.0 to -5.0)

Table 7. Decisional Conflict Score - low literacy version
  1. DA: decision aid; DCS: decisional conflict scale; SD: standard deviation

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA versus usual care
Smith 2010Total DCS2 week follow-up35713.63 (SD 20.55)17314.91(SD 18.34)P = 0.02
Detailed versus simple DA
Volk 2008Uncertainty2 weeks395.8 (SD 18.0)486.8 (SD 18.0)P = 0.80
Informed2 weeks399.1 (SD 26.0)4618.8 (SD 26.1)P = 0.09
Values2 weeks4017.4 (SD 36.8)4834.9 (SD 36.6)P = 0.03
Social Support2 weeks3917.8 (SD 29.6)4827.6 (SD 29.5)P = 0.12
Total DCS2 weeks3812.0 (SD 21.9)4621.7 (SD 21.8)P = 0.04

The overall MD was -6.22 out of 100 points for decision aid compared to usual care (95% CI -8.00 to -4.44; see Analysis 4.1.6) and -1.77 for detailed compared to simple decision aid (95% CI -2.64 to -0.91; see Analysis 4.4.6). Three studies that could not be pooled (Table 6) reported statistically-significantly less total decisional conflict (Arterburn 2011; Schwartz 2009; Weymiller 2007), three no difference (Krist 2007; Leighl 2011; Ozanne 2007), and one higher decisional conflict (Fagerlin 2011). Smith 2010 used the low literacy version and reported statistically-significant improvement in total decisional conflict in the decision aid group, compared to usual care (Table 7). Rubel 2010 did not report results by group.

The 'feeling uninformed' sub-scale of the DCS was reported in 32 studies. Because this DCS sub-scale measures self-reported comfort with knowledge and not actual knowledge, we elected to consider it a process measure and to reserve the gold standard of objective knowledge tests for assessing decision quality. The MD for 'feeling uninformed' about options, benefits, and harms was -7.26 (95% CI -9.73 to -4.78) in the 22 studies that compared patient decision aids to usual care (see Analysis 4.1.2). The 10 studies that compared detailed with simple patient decision aids had a MD for 'feeling uninformed' of -2.39 (95% CI -4.39 to -0.39; Analysis 4.4.2). For the studies that could not be pooled (Table 6), compared to usual care, those exposed to the decision aid felt more informed in three studies (Frosch 2008; Mathieu 2010; Weymiller 2007) but were no different in one study (Berry 2013).The funnel plot for feeling uninformed as an outcome in studies comparing decision aid to usual care shows low risk for publication bias (Figure 7).

Figure 7.

Funnel plot of comparison: 4.1 Decisional conflict: DA vs usual care - all studies, outcome: 4.1.2 Uninformed sub-scale

The 'feeling unclear about values' sub-scale of the DCS was reported in 18 studies comparing patient decision aids to usual care (MD -6.09; 95% CI -8.50 to -3.67; see Analysis 4.1.3). In the 10 studies that compared detailed to simple decision aids, the MD for 'feeling unclear about values' was -2.31 (95% CI -4.67 to -0.05; see Analysis 4.4.3) Compared to usual care, those exposed to the decision aid in all three studies that could not be pooled (Table 6) felt more clear about their values (Berry 2013; Frosch 2008; Weymiller 2007).The funnel plot for feeling unclear about values as an outcome in studies comparing decision aid to usual care shows low risk for publication bias (Figure 8).

Figure 8.

Funnel plot of comparison: 4.1 Decisional conflict: DA vs usual care - all studies, outcome: 4.1.3 Unclear sub-scale

Volk 2008 compared detailed to simple decision aids and showed improvements in decisional conflict only in lower literacy sub-groups of participants. For example, low literacy study participants used the low literacy version of the DCS and their results were reported separately from participants at the higher literacy site. The lower literacy study participants exposed to the more detailed edutainment decision aid reported significantly lower levels of overall decisional conflict and higher levels of 'feeling clear about values', compared to the lower literacy study participants exposed to the simpler audio-booklet decision aid (see Table 7).

Patient-practitioner communication

Of 115 studies, 9 (7.8%) measured the effect of decision aids on patient-practitioner communication. Four studies (Hess 2012; Montori 2011; Mullan 2009; Weymiller 2007) compared the effect of a decision aid used within the clinical encounter (or, in one study, half the decision aid participants were exposed just prior to the encounter) to usual care, and evaluated the extent of shared decision making by analysing the audio-recordings using the OPTION scale. All four studies reported statistically-higher mean OPTION scores when patients were exposed to the decision aid, and this effect was greater when the decision aid was used within the clinical encounter (see Table 8).

Table 8. Patient-practitioner communication
  1. CHD: coronary heart disease; CI: confidence interval; DA: decision aid; DCS: decisional conflict scale; ICC: intraclass correlation coefficient; OPTION scale: observing patient involvement scale; RR: risk ratio; SD: standard deviation

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA versus usual care
Hanson 2011Discussed feeding with physician, nurse practitioner, or physician's assistant3 months12646%12733%P = 0.04
Discussed feeding with other nursing home staff3 months12664%12771%P = 0.42
Hess 2012OPTION scaleanalysis of the consultation using video-recorded consultations101Mean of 26.6 (95% CI 24.9 to 8.2)103Mean of 7% (95%CI 5.9 to 8.1)Significantly greater in the intervention arm
Legare 2011DCS / Dolan's Provider DCSimmediately post    Difference 0.26 (95%CI -0.06 to 0.53, P = 0.06)
Montori 2011OPTION 100 point Scaleanalysis of the consultation using video-recorded consultations3849.83227.3P < 0.001
Mullan 2009OPTION Scaleanalysis of the consultation using video-recorded consultations48 used decision aid within consultation49.7% (SD 17.74)37 usual care27.7% (SD 11.75MD 21.8 (95% CI 13.0, 30.5) for decision aid vs usual care. All but 2 of the 12 items significantly
favoured the decision aid
Sheridan 2006Discussed CHD with doctorpatient reported immediately post16/41 decision aid pre-consult with summary report to bring to consult 8/34 usual care absolute difference 16%; 95% CI
-4% to 37%
Plan to reduce CHD risk & discussed with doctorpatient reported immediately post15/41 decision aid pre-consult with summary report to bring to consult 8/34 usual care absolute difference 13%; 95% CI -7% to 34%).
Plan to reduce CHD risk & not discussed with doctorpatient reported immediately post37/41 decision aid pre-consult with summary report to bring to consult 25/34 usual care absolute difference 16%; 95% CI
-1% to 33%
Weymiller 2007OPTION Scaleanalysis of the consultation using video-recorded consultations1/2 used decision aid prior to consult and 1/2 used it during consult usual care Greater patient participation (MD 4.4; 95% CI 2.9 to 6.0) in decision aid compared to usual care
Detailed versus simple DA
Legare 2003Agreement between women’s and physicians’ decisional conflict scoresimmediately post87ICC = 0.44 (95% CI 0.25 to 0.59)80ICC = 0.28 (95% CI 0.06 to 0.47)Agreement measure was higher for the DA group.
DCS / Dolan's Provider Decision Process Assessment Instrumentimmediately post97 detailed decision aid pre consultICC 0.44 (0.9 SD)87 simple decision aid pre consultICC 0.28 (1.0 SD)Agreement measure was higher for the DA group (ICC 0.44; 95% CI 0.25 to 0.59) than for the pamphlet group (ICC 0.28; 95% CI 0.06 to 0.47)
Myers 2011Informed decision makinganalysis of the physician–patient encounter using audio-recordings 3.0 items 2.4 items

RR 1.30 (CI 1.03 to 1.64)

P = 0.029

Myers 2011 analyzed audio-recorded encounters using the Informed Decision Making observer instrument (Braddock III 1997; Braddock III 1999; Price 2012). Findings reported significantly higher scores in the detailed decision aid group compared to simple decision aid (P = 0.029).

For agreement between physicians and women on decisional conflict scores as an indicator of communication about the decision within the consultation, Legare 2003 reported higher agreement for the decision aid group than for the usual care group, but Legare 2011 reported no statistically-significant difference between groups (see Table 8).

Sheridan 2006 and Hanson 2011 found that, of those exposed to the decision aid, a higher proportion compared to usual care reported having discussed the decision with their practitioner (see Table 8)

Participation in decision making

Of 115 studies, 22 (19.1%) measured the effect of decision aids on patient participation in decision making: of these, 20 compared the effects of decision aids to usual care (Analysis 5.1; Table 9) and 2 (Deschamps 2004; Raynes-Greenow 2010) compared a detailed decision aid to a simple one (Analysis 5.4). The Davison 1997 paper used the Control Preferences Scale (Degner 1992). This scale measures the role in decision making using five response statements: two represent an active or patient-controlled role, one a shared or collaborative role, and two response statements represent a passive or practitioner-controlled role. Most other studies used comparable response statements that could be classified within each of the three groupings of the Control Preferences Scale, except for Hamann 2006 which used the COMRADE instrument to measure patient perception of involvement, and two others that used other measures of perceived involvement (Hanson 2011; Loh 2007) (see Table 9).

Table 9. Participation in decision making
  1. DA: decision aid; SD: standard deviation

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA versus usual care
Allen 2010control preferences - patients choosing active/ collaborative decision makingpost intervention29195%33492%No difference
control preferences did not changepost intervention29192%33487%No difference
control preferences changed to passivepost intervention2913%3345%No difference
control preferences changed to active/ collaborativepost intervention2913%3347%No difference
Hamann 2006COMRADE used to measure patients' perceived involvement in decisionspost-consultation49

79.5 (SD 18.6)

76.8 (SD 20.9)

58

69.7 (SD 20.0)

73.5 (SD 19.3)

increased patient involvement in decision aid group post intervention compared to usual care at baseline. At discharge there was no difference between groups.
Hanson 2011surrogates feeling somewhat or very involved in decision makingpost intervention 83% 77%P = 0.18
Leighl 2011achieved decision involvementpost intervention 32% 35%No difference
Loh 2007patients' perceived involvement in decision makingpost-consultation19126.3 pre 28.0 post96

24.5 pre

25.5 post

Improved patient participation from baseline to post exposure to the decision aid (P = 0.010) and in comparison to the usual care group (P = 0.003) but there was no change in the control group for the pre-post comparison
Rubel 2010adapted from the Control Preferences Scalepost-intervention    the total mean scores were: 2.74±1.25 (n=99) pre and 2.83±1.16 (n=199) post, no statistically significant difference.
van Peperstraten 2010Decision Evaluation scale (15 item questionnaire) Decision control subscalepost-consultation1248512887.5P = 0.33

For patients assuming an active (patient-controlled) role in decision making, the pooled RR for 12 studies compared decision aid to usual care was 1.28 (95% CI 1.02 to 1.60; Analysis 5.1.1). The proportion adopting a shared decision-making role in 12 studies showed no difference between decision aid and usual care (decision aid versus usual care pooled RR 0.96; 95% CI 0.82 to 1.13; Analysis 5.1.2). Given that patient decision aids are hypothesized to increase patient participation in decision making, there was a reduction in practitioner-controlled decision making; the pooled RR based on 14 studies comparing decision aids to usual care was 0.66 (95% CI 0.53 to 0.81; Analysis 5.1.3). For studies that could not be pooled, Allen 2010, Leighl 2011, Rubel 2010, and van Peperstraten 2010 reported no difference in these roles between groups.

For studies that could not be pooled in which a decision aid was compared to usual care, Loh 2007 and Hamann 2006 reported that a statistically-significant proportion of patients exposed to the decision aid described feeling involved in decision making. However, Hamann 2006 did not analyze findings accounting for cluster. Hanson 2011 reported that a higher proportion described feeling involved (83% vs 77%) but that the difference between groups was not statistically significant (Table 9).

There was no statistically-significant difference in patient participation in decision making for the two studies that compared a detailed decision aid to a simple one (Deschamps 2004; Raynes-Greenow 2010) (see Analysis 5.4)

Proportion undecided

Of 115 studies, 21 (18.3%) measured the proportion remaining undecided: of these, 18 pooled studies compared decision aids to usual care, 3 pooled studies compared detailed to simple decision aids, and 1 not able to be pooled compared decision aid to usual care. For 18 studies comparing decision aids to usual care, a statistically-significantly lower proportion of people remained undecided after exposure to a decision aid (RR 0.59; 95% CI 0.47 to 0.72; Analysis 6.1 ).

None of the studies (Deschamps 2004; Labrecque 2010; Leung 2004) comparing detailed to simple decision aids showed a statistically-significant difference between groups (pooled RR 0.98; 95% CI 0.69 to 1.37; Analysis 6.4).

Kasper 2008 measured progress in decision making using a single item ranging from '0 = completely undecided' to '100 = made my decision'. Given the different measure used, these findings were not included in the meta-analysis. In this study, both the patients exposed to a decision aid and the usual care group progressed in their decision making, with no difference between groups (Table 10).

Table 10. Proportion undecided
  1. DA: decision aid

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA versus usual care
Kasper 2008single item - ranging from '0 = completely undecided' to '100 = made my decision'     No difference
Satisfaction

Satisfaction was measured as it relates to satisfaction with the choice, satisfaction with the process of decision making, and preparation for decision making. Satisfaction with preparation for decision making was measured in three studies using the Preparation for Decision Making Scale (Bennett 2010). When possible, the scores were standardized to a 0-to-100 point scale, with higher scores reflecting greater satisfaction.

Of 115 studies, 20 (17.4%) measured satisfaction with the choice: 15 compared decision aids to usual care and 5 compared detailed to simple decision aids. Of these 15 studies, 3 (Heller 2008; Laupacis 2006; Montgomery 2007) reported that people exposed to the decision aid had higher satisfaction with their choice compared to usual care, and the other 12 reported no statistically-significant difference (see Analysis 7.1 and Table 11). Of the five studies that compared detailed to simple decision aids, four found no across-group differences in satisfaction with the choice (Deschamps 2004; Raynes-Greenow 2010; Rothert 1997; Schapira 2007), and one reported higher satisfaction with the choice after using the detailed decision aid (Solberg 2010) (Analysis 7.4 and Table 11).

Table 11. Satisfaction with the choice
  1. DA: decision aid

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA versus usual care
Heller 20081-item; pleased with treatment choice1 month post-surgery62/66 55/67 P = 0.03
Leighl 2011

satisfaction with decision scale:

median (range)

1 month post intervention10722(13-25)10021(15-25)No difference
Marteau 20107-point scale: ranging from 1-74 weeks 91.17 (14) 91.33(14.50)No difference
Schwartz 2009a6-item1, 6, 12 months100 114 Overall, no difference between groups; decision aid led to significantly increased satisfaction compared to US among those who were undecided at randomisation but not among those who had made a decision before randomisation; (only graph in paper with no raw data)
Trevena 2008satisfaction with the decisionimmediately post134 137 No difference (P = 0.56)
Detailed versus simple DA
Rothert 19976-item scale (measured on 1 to 5)1 day834.0 (0.56)893.8 (0.66)No difference
6 months633.8 (0.63)753.8 (0.67)No difference
12 months623.9 (0.62)743.9 (0.67)No difference
Schapira 20076-item scale3 months    No difference

Of 115 studies, 17 (14.8%) measured satisfaction with the decision-making process: 14 compared decision aids to usual care and 3 compared detailed to simple decision aids. Of 14 comparing decision aids to usual care, 10 measured satisfaction with the decision-making process (see Analysis 7.6; Hess 2012; Kennedy 2002; Montori 2011; Vodermaier 2009 in Table 12), 3 measured satisfaction with information received (Laupacis 2006; Miller 2005; Oakley 2006) and 1 (Green 2004) measured satisfaction with genetic counselling. Of the 14 studies, 5 showed statistically-significant improvement in satisfaction with the decision-making process (Barry 1997; Hess 2012; Kennedy 2002; Laupacis 2006; Schroy 2011) and with information provided (Laupacis 2006) when patient decision aids were used compared to usual care, and 9 showed no difference (Bernstein 1998; Green 2004; Jibaja-Weiss 2011; Man-Son-Hing 1999; Miller 2005; Montori 2011; Morgan 2000; Oakley 2006; Vodermaier 2009) (see Analysis 7.6; Table 12). No studies reported that those exposed to patient decision aids were statistically-significantly less satisfied compared to usual care. Although there was no difference in satisfaction with the information between patients in the Montori 2011 study, clinicians had higher satisfaction.

Table 12. Satisfaction with the decision-making process
  1. DA: decision aid; SD: standard deviation

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA versus usual care
Green 2004Effectiveness of consultation - patient assessment. Single item 1 (not at all effective) to 7 (extremely effective) 1066.61056.6No difference
Effectiveness of consultation - counsellor assessment. Single item 1 to 7  5.9 5.8No difference
Hess 2012Satisfaction with decision process (0 for strongly agree to 5 for strongly disagree) 101 103 Patients in DA group reported greater satisfaction with the DM process (strongly agree, 61% DA vs 40% usual care)
Kennedy 2002Measured satisfaction with opportunities to participate in decision making using a single item     Compared to usual care, women who received the decision aid followed by nurse coaching were statistically significantly more satisfied with the opportunities to participate in decision making (OR 1.5; 95% CI 1.1 to 2.0).
Laupacis 2006Satisfaction with information received sub-scale 4-item (0 to 100; low to high)average 10 days5476 (15.5 SD)5659 (23.3 SD)P = 0.001
Satisfaction with practitioner treatment during decision process sub-scale 4-item (0 to 100; low to high)average 10 days5469 (25.3 SD)5654 (26.7 SD)P = 0.004
Miller 2005Satisfaction with cancer information service 1-item (1 to 5; low to high)2 weeks 4.37 (0.84 SD) 4.38 (0.86 SD)No difference
6 months 4.51 (0.75 SD) 4.51 (0.64 SD)No difference
Montori 2011

(7 point scales)

Participants' satisfaction with knowledge transfer

-amount of information

-clarity of information

-helpfulness of the information

-would want other decisions

-recommend to others

post intervention49

6.6

6

6

6.1

6.4

46

6.3

6

5.8

5.8

6.2

P = 0.798

P = 0.296

P = 0.624

P = 0.248

P = 0.435

Clinicians' satisfaction with knowledge transfer

-helpfulness of the information

-would want other decisions

-recommend to others

post intervention39

5.8

6.1

5.9

33

5.2

4.9

4.8

P = 0.006

P < 0.001

P < 0.001

Oakley 2006Satisfaction with information about medicines4 months post1610.4 (SD 2.9)1710.1 (SD 2.2)No difference
Vodermaier 2009

- physician helped me understand

- physician understood important to me

- physician answered questions

- satisfied with involvement

- satisfied with physician's involvement

- satisfied with process

1 week follow-up53

49 (92.5%)

47

47

44

36

42

56

53 (94.6%)

50

51

45

36

50

High satisfaction with no difference by group
Detailed versus simple DA
Deyo 2000Satisfaction with decision making process 7-item scale (5-point response)3 months171separate responses provided with no total172separate responses provided with no totalNo difference except DA more likely to report they had as much information as they wanted and less likely to report having relied too much on physician's opinion
Hunter 2005Satisfaction with genetic counselling 11-item short form (range 4 to 44; low to high)immediately post11637.27 (5.74 SD)12640.48 (4.26 SD)P < 0.001 higher satisfaction with individual counselling compared to decision aid
Kuppermann 2009Satisfaction with involvement in decision making (3 questions)26 to 30 weeks gestation244

44.8

44.3

72.6

252

49.2

48.1

79.9

P = 0.40

P = 0.45

P = 0.10

Of three studies comparing detailed and simple decision aids, Deyo 2000 measured satisfaction with the decision-making process, Kuppermann 2009 measured satisfaction with involvement in decision making, and Hunter 2005 measured satisfaction with genetic counselling (see Table 12). Hunter 2005 reported higher satisfaction among those exposed to genetic counselling compared to decision aid alone, and the other two studies reported no difference between groups.

Of 115 studies, 3 (2.6%) measured preparation for decision making (Table 13). Compared to usual care, two studies reported significant improvements in people's satisfaction with their preparation for making decisions after using decision aids about management of knee osteoarthritis (Fraenkel 2007) or referral to a lung transplant centre (Vandemheen 2009) (see Table 13). The third study (Deschamps 2004) found no statistically-significant difference between those exposed to the detailed or simple decision aid.

Table 13. Preparation for decision making
  1. DA: decision aid; SD: standard deviation

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA versus usual care 
Fraenkel 2007Preparation for Decision Making ScalePre-consultation4335 (median)4020.5 (median)P = 0.0001
Vandemheen 2009Preparation for Decision Making Scale3 weeks7065.1 (24.9 SD)7953.9 (27.1 SD)P = 0.009
Detailed versus simple DA
Deschamps 2004Preparation for Decision Making ScalePost-physician consultation4828 (6.1 SD)4227(5.5 SD)No difference

Behaviour

Choice

Choice was defined as the actual choice implemented. However, when the actual choice was not reported, the preferred option was used as a surrogate measure. Ninety-three studies (80.9%) assessed the effects of decision aids on the participants' actual choice implemented (n = 57), their preferred option (n = 33), or used both (n = 3) (Table 14). Actual choice or preferences were reported as the percentage of individuals actually implementing or stating a preference for the most intensive or most invasive option.

Table 14. Choice
  1. DA: decision aid; OR: odds ratio

StudyType of comparison N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
Other elective surgery - uptake
Hanson 2011DA versus usual care12711293No difference
Wong 2006DA versus usual care    No difference
Other elective surgery - preference
Labrecque 2010Detailed versus simple DA32133114No difference
Screening - Breast cancer genetic testing - uptake
Wakefield 2008aDetailed versus simple DA    No difference
Wakefield 2008bDetailed versus simple DA    No difference
Screening - Breast cancer genetic testing - preference
Miller 2005DA versus usual care    The intervention decreased intention to obtain genetic testing among women at average risk, but increased in women at high risk
Screening - Cardiac stress testing - uptake
Hess 2012DA versus usual care10158%10077%P < 0.0001
Screening - Colorectal cancer genetic testing - uptake
Wakefield 2008Detailed versus simple DA    No difference
Screening - Breast screening - uptake
Mathieu 2007DA versus usual care    No difference
Mathieu 2010DA versus usual care11782%20961%P < 0.001
Screening - Diabetes - uptake
Marteau 2010DA versus usual care633353639368P = 0.51
Screening - Diabetes - preference
Mann E 2010DA versus usual care273 134 No difference
Screening - Prenatal - uptake
Bekker 2004DA versus usual care    No difference
Bjorklund 2012DA versus usual care18450%20653.8%No difference
Nagle 2008DA versus usual care    No difference
Screening - PSA - uptake
Frosch 2008DA versus usual care    The experimental interventions led to significant reductions in requests for prostate-specific antigen tests ( ˜2 times greater decline).
Medication - Antibiotics for upper respiratory infections - uptake
Legare 2011DA versus usual care81337049P = 0.08
Medication - Cardiovascular disease - preference
Sheridan 2011DA versus usual care7963%7842%P < 0.01
Medication - Breast cancer prevention - uptake
Fagerlin 2011DA versus usual care3820.5%1000%No difference
Medication - Chemotherapy for advanced cancer
Leighl 2011DA versus usual care10777%10071%No difference
Medication - Hormone replacement therapy - uptake
Murray 2001bDA versus usual care    8% decrease in DA group, not statistically significant
Schapira 2007Detailed versus simple DA    No difference
Medication - Natural heath products - preference
Legare 2008aDA versus usual care 41% 41%No difference
Medication - Anti-thrombosis - uptake
Man-Son-Hing 1999DA versus usual care    25% decrease in DA group, not statistically significant
McAlister 2005DA versus usual care    No difference
Thomson 2007DA versus usual care 93.8% 25%risk ratio 0.27 (95% CI 0.11 to 0.63)
Medication - Hypertension - uptake
Montgomery 2003DA versus usual care    No difference
Medication - Chemotherapy for breast cancer - preference
Whelan 2003DA versus usual care    No difference
Medication - Osteoporosis - uptake
Montori 2011DA versus usual care5244%4840%No difference
Medication - Immunotherapy - uptake
Kasper 2008DA versus usual care    No difference
Medication - Schizophrenia treatment - uptake
Hamann 2006 - prescriptionsDA versus usual care    No difference
Hamann 2006 - psycho-educationDA versus usual care    Higher uptake in DA group (P = 0.003)
Obstetrics - Birth control method - preference
Langston 2010DA versus usual care114 108 No difference in the methods chosen between groups, participants in the intervention group were not more likely to initiate the requested method immediately compared to those in
the usual care group (OR 0.65, 95% CI 0.31 to 1.34)
Obstetric - Childbirth procedure - uptake
Montgomery 2007DA versus usual care    No difference
Nassar 2007DA versus usual care    No difference
Obstetric - Childbirth procedure - preference
Shorten 2005DA versus usual care    No difference
Obstetric - Embryo transplant - uptake
van Peperstraten 2010 - single embryo transferDA versus usual care15243%15632%P = 0.05
Obstetric - Pain relief in labour - uptake
Raynes-Greenow 2010Detailed versus simple DA308 146 No difference
Other- Lung transplant referral
Vandemheen 2009DA versus usual care    No difference
Other - Pre-operative blood transfusion - uptake
Laupacis 2006DA versus usual care    No difference
Vaccine - Flu shot - uptake
Chambers 2012DA versus usual care4846%5927%No difference
Vaccine - Hepatitis B - uptake
Clancy 1988DA versus usual care    Significant increase of 76% in the DA group
Choice for surgery
Major elective surgery

Eighteen studies (15.3%) focused on choices regarding a more major elective surgery. Fifteen (Arterburn 2011; Auvinen 2004; Barry 1997; Berry 2013; Bernstein 1998; Morgan 2000; Murray 2001a; Jibaja-Weiss 2011; Kennedy 2002; Protheroe 2007; Schwartz 2009; Solberg 2010; Whelan 2004; Vodermaier 2009; Vuorma 2003) compared decision aids to usual care (Analysis 8.1), and three (Deyo 2000; Street 1995; Tiller 2006) compared detailed to simple decision aids (Analysis 8.2).

Using intention-to-treat analysis, there was a reduction in the number of patients choosing major elective surgery in the group receiving the decision aid compared to usual care (RR 0.79; 95% CI 0.68 to 0.93, n = 15; Analysis 8.1.2). Schwartz 2009 reported a statistically-significant uptake of prophylactic mastectomy for women who are BRCA1/2 gene carriers (114%). Only three other studies showed statistically-significant changes in surgery rates: -29% for cardiac revascularization (Morgan 2000), -74% for mastectomy (Whelan 2004), and -33% for orchiectomy (Auvinen 2004). Eight other studies (Arterburn 2011; Barry 1997; Bernstein 1998; Berry 2013; Jibaja-Weiss 2011; Kennedy 2002; Solberg 2010; Vodermaier 2009) showed reductions in uptake of the more intensive surgical treatment by 14% to 58%, but the results were not statistically significant. Two studies (Protheroe 2007; Vuorma 2003) showed non-significant higher rates of hysterectomy in the decision aid group compared to usual care. Another study (Murray 2001a) reported a non-significant five-fold increase in uptake of prostatectomy.

Using intention-to-treat analysis, there was a non-statistically-significant reduction in the number of patients choosing major elective surgery in the group receiving the detailed compared to simple decision aids (RR 0.82; 95% CI 0.63 to 1.08; Analysis 8.2.2). None of the three studies comparing detailed to simple decision aids reported a statistically-significant difference in surgery rates for mastectomy in women with breast cancer (Street 1995), back surgery for people with herniated disc or spinal stenosis (Deyo 2000), prophylactic oophorectomies for women with a family history of breast or ovarian cancer, or non-polyposis colon cancer (Tiller 2006).

Other elective surgery

Three studies evaluated the effect of decision aids versus usual care on other elective surgical decisions. Decision aids did not significantly influence surgical abortion rates (Wong 2006), feeding tube insertions (Hanson 2011), or preference for vasectomy (Labrecque 2010).

Choice for screening
Prostate-specific antigen screening

The effects of decision aids on prostate-specific antigen (PSA) screening decisions were variable in 13 studies (11.3%): 10 that compared decision aids to usual care and 3 that compared detailed to simple decision aids. The pooled RR for nine studies was 0.87 (95% CI 0.77 to 0.98; Analysis 8.3.1); Frosch 2008 reported a reduction in screening rates but we were not able to pool the data. Of 10 studies that compared decision aids with usual care, 3 showed significant reductions in screening by 9% to 42% (Frosch 2008; Volk 1999; Wolf 1996). The results of the other seven studies (Allen 2010; Evans 2010; Gattellari 2003; Gattellari 2005; Partin 2004; Krist 2007; Watson 2006) were not statistically significant.

Three studies compared a detailed and a simple decision aid and the pooled RR was 0.98 (95% CI 0.82 to 1.17; Analysis 8.3.2). There were non-significant reductions of 2% and 11% in PSA screening in two studies (Myers 2011; Schapira 2000). One study reported a non-significant increase in screening of 89% (Myers 2005a).

Colon cancer screening

Of 10 studies (8.7%) of colon cancer screening, 3 reported statistically-significant changes and 7 showed no difference. Two studies reported that the decision aid, when compared to usual care, significantly increased the uptake of screening by 64% and 70%, respectively (Pignone 2000; Ruffin 2007), and the other study reported a statistically-significant reduction of 21% for screening (Smith 2010). There was an increase in uptake of screening in five studies, by 6% to 39%, but the difference was not statistically significant (Lewis 2010; Miller 2011; Schroy 2011; Steckelberg 2011; Wolf 2000). In two studies (Dolan 2002; Trevena 2008), there was a 73% and 4% decrease in screening rates that was not statistically significant. The pooled RR was 1.12 (95% CI 0.95 to 1.31, n = 10; Analysis 8.3.3).

Cancer genetic screening

Preferences or uptake of cancer genetic screening were reported in 8 studies (7.0%): seven focused on breast cancer and one focused on colorectal cancer genetic testing. Preferences for breast cancer genetic screening were not statistically-significantly affected when a decision aid was compared to usual care. The pooled RR was 1.01 (95% CI 0.83 to 1.22, n = 4; Analysis 8.3.4). One study reported an increased uptake of screening by 14% (Lerman 1997), a second study reported an increase of 18% (Green 2001a), a third study reported a decrease in uptake by 29% (Schwartz 2001), and the other study reported no difference (Green 2004). Miller 2005 reported that women exposed to the decision aid who were at higher risk of breast cancer increased their intention to obtain genetic testing, while those at average risk decreased their intention.

When detailed decision aids were compared to simple ones, there was no difference in uptake of genetic testing for breast or colorectal cancer (Wakefield 2008; Wakefield 2008a; Wakefield 2008b).

Breast screening

There was higher uptake of mammography screening among women aged 38 to 45 years of age (Mathieu 2010) but no difference in women aged 70 or older (Mathieu 2007) who were exposed to a decision aid versus usual care.

Prenatal screening

The uptake of prenatal testing was not affected by a decision aid compared to usual care (Bekker 2004; Bjorklund 2012; Nagle 2008), nor by a more detailed decision aid compared to a simple decision aid (Hunter 2005; Leung 2004; pooled RR 0.96, 95% CI 0.90 to 1.03; Analysis 8.3.5).

Stress test for chest pain

Compared to usual care, adults presenting with chest pain in the emergency department who received the decision aid had significantly less stress testing done (58% versus 77%) (Hess 2012).

Screening for diabetes

There was no difference in uptake (Marteau 2010) or preference (Mann E 2010) to be screened for diabetes in adults exposed to a decision aid compared to usual care.

Choice for medication
Antibiotics for upper respiratory infection

There was a decrease in prescriptions for antibiotics for upper respiratory infections when a decision aid was used in the consultation compared to usual care, but this difference was not statistically significant (Legare 2011).

Cardiovascular disease prevention

There was an increase in patient preferences for medication to lower cardiovascular disease risk when a decision aid was used compared to usual care (63% vs 42%) (Sheridan 2011). Three studies evaluated decision aids with people with diabetes considering cardiovascular disease prevention medications and the RR was 1.84 (95% CI 0.77 to 4.39). Compared to usual care, those exposed to the decision aid had increased uptake of statins therapy (Mann D 2010; Weymiller 2007), but the findings were not statistically significant (Analysis 8.4). Mullan 2009 reported that a higher proportion of people with type II diabetes started medications after exposure to the decision aid (33%), compared to usual care (22%).

Breast cancer prevention medication

There was no difference in uptake of medications for women at risk of breast cancer who were exposed to the decision aid versus usual care (Fagerlin 2011).

Chemotherapy for advanced cancer

There was no statistically-significant difference in the uptake of chemotherapy for adults with advanced colorectal cancer (77% versus 71%) (Leighl 2011).

Menopausal hormone therapy

Preferences regarding hormone therapy for menopausal women were affected when a detailed decision aid was compared to a simple decision aid in three studies, with a statistically-significant decrease of 36% (Dodin 2001), a non-statistically-significant decrease of 25% (Deschamps 2004) and non-statistically-significant increase of 12% (O'Connor 1998a) respectively. There was a statistically-significant reduction of 27% in the uptake of hormone therapy when these studies were pooled (RR 0.73; 95% CI 0.55 to 0.98; Analysis 8.5). Schapira 2007 reported no difference in the use of hormone therapy between those exposed to the detailed or simple decision aid. In a single study comparing a decision aid to usual care (Murray 2001b), there was a decrease of 8%, which was not statistically significant.

Natural health products

Preferences for natural health products in women experiencing menopausal symptoms were no different for women exposed to the decision aid compared to women exposed to the usual education materials (Legare 2008a).

Anti-thrombosis medication

Three studies evaluated the effect of a decision aid on the use of anti-thrombotic therapy for atrial fibrillation versus usual care. One study demonstrated a non-significant reduction of uptake of warfarin of 25% (Man-Son-Hing 1999). The second study evaluated the proportions of patients choosing the option that was appropriate relative to their level of risk, and found no significant difference between the groups (McAlister 2005). Thomson 2007 reported that patients in the usual care group (guided by practice recommendations) were much more likely to start warfarin (15/16; 93.8%) compared to the decision aid group (4/16; 25%; RR 0.27; 95% CI: 011 to 0.63).

Hypertension medication

Montgomery 2003 found no significant effect of decision aids over usual care on the initiation of medication for hypertension.

Breast cancer medication

Whelan 2003 also found no significant effect on preferences for adjuvant chemotherapy versus no chemotherapy for breast cancer.

Immunotherapy

Kasper 2008 reported no difference in the uptake of immunotherapy in people with multiple sclerosis who were exposed to a decision aid, compared to usual care based on practice guidelines.

Osteoporosis treatment

Montori 2011 found no significant effect of decision aids over usual care on the uptake of medication for osteoporosis treatment.

Schizophrenia treatment

Although Hamann 2006 found no difference in prescriptions for antipsychotic medications but a statistically-significant increase in the uptake in psycho-education (P = 0.003) in people with schizophrenia exposed to the decision aid compared to usual care.

Influenza (flu) vaccine

Compared to usual care, there was a non-statistically-significant increase in intentions to get the flu vaccine in those exposed to the decision aid (46% versus 27%) (Chambers 2012).

Hepatitis B vaccine

Compared to usual care, there was a statistically-significant increase in uptake of Hepatitis B vaccination with decision aids (Clancy 1988).

Blood transfusions

There was no difference in the uptake of pre-operative autologous blood donation when a decision aid was compared to usual care (Laupacis 2006).

Obstetrical choices

Childbirth procedures

Three studies focused on childbirth issues, using a decision aid compared to usual care. There was no difference in preference for (Shorten 2005) or actual vaginal mode of delivery (Montgomery 2007) following previous cesarean section. Another study found no difference in actual choice to undergo external cephalic version for women with breech presentation (Nassar 2007). Raynes-Greenow 2010 reported that there was no difference in uptake of pain relief in labour for those exposed to a detailed versus simple decision aid.

Birth control approaches

There was no difference in the birth control methods chosen for those in the decision aid versus usual care groups (Langston 2010).

Embryo transplantation

Compared to usual care, those in the decision aid group were statistically significantly more likely to choose a single embryo transplant (43% versus 32%) (van Peperstraten 2010).

Other choices
Lung transplant referral

There was no difference in referral rates for consideration of lung transplant in people with advanced cystic fibrosis exposed to a decision aid versus usual care (Vandemheen 2009).

Summary: choice

In summary, patient decision aids decrease the number of patients choosing elective surgical procedures, PSA testing, and use of hormone therapy in multiple studies. Single studies showed that decision aids increased the number of people choosing: hepatitis B vaccination, psycho-educational therapies for schizophrenia, and medication for cardiovascular disease prevention; and decreased cardiac stress testing and the number of embryos being transplanted. The effect on patients' choice in other situations was more variable. There were mixed results for the choice of colon cancer screening, genetic testing, prenatal testing, anti-thrombosis therapy, breast screening, and diabetes medications. There was no difference between groups for choices about natural health products, hypertension therapy, breast cancer chemotherapy, schizophrenia medication, immunotherapy for multiple sclerosis, flu vaccine, diabetes screening, birth control, osteoporosis treatment, chemotherapy for advanced cancer, chemopreventive medications, antibiotics for upper respiratory tract infections, use of blood transfusions, and childbirth procedures.

Adherence (continuance/compliance) with chosen option

Of 115 studies, 13 (11.3%) measured adherence with the chosen option: 10 compared a decision aid to usual care, and 3 compared detailed to simple decision aids (Table 15). Of the 10 that compared a decision aid to usual care, 3 studies showed a statistically-significant difference between groups, with adherence rates reported by Mullan 2009 favouring usual care (97.5% decision aid compared to 100% usual care at 6 months), and with adherence rates reported by Montori 2011 and Sheridan 2011 favouring the decision aid. Montori 2011 reported that 100% of the participants in the decision aid group versus 74% in the usual care group at 6 months had taken their medication on more than 80% of the days for which it was prescribed, based on pharmacy records. Sheridan 2011 found higher adherence in the decision aid group compared to the usual care group for any therapy described in the decision aid, any therapy whether or not it was described in the decision aid, and aspirin (P < 0.02). Although trends appeared positive for decision aids, the Sheridan 2011 study was underpowered to determine if observed differences between decision aid and usual care were statistically significant for adherence to cholesterol medication, blood pressure medication, or smoking cessation. The other seven studies found no difference in adherence to medication for atrial fibrillation (warfarin versus aspirin) at six months (Man-Son-Hing 1999), oral bisphosphonate medication for osteoporosis at four months (Oakley 2006), blood pressure medication at three years (Montgomery 2003), anti-depressant medication at two months (Loh 2007), statins for high cholesterol at three or six months (Mann D 2010; Weymiller 2007), or use of effective contraceptive method (Langston 2010).

Table 15. Adherence with chosen option
  1. DA: decision aid; OR: odds ratio

ReferenceScale usedN Decision aidMean (SD) Decision aidN ComparisonMean (SD) ComparisonNotes
DA versus usual care
Langston 20103 months - Using a contraceptive method that was in the same effectiveness group as the method requested at enrolment, 'very effective', as chosen option - eg. if chose sterilization and ended up using an IUD counted as adhering4885%5277%P = 0.28
3 months - Using a contraceptive method that was in the same effectiveness group, 'effective', as chosen option4168%3168%P = 0.96
Loh 20076 to 8 weeks - Patient reported - 5-point Likert scale on steadiness of following the treatment plan: 1-very bad to 5-very good1914.3 (0.9)963.9 (1.0)P = 0.073
6 to 8 weeks - Physician reported - 5-point Likert scale steadiness of following the treatment plan: 1-very bad to 5-very good1914.8 (0.6)964.3 (1.1)P = 0.56
Mann D 20103 months - telephone administration of the 8-item Morisky adherence (7 yes/no items and 1 item with 5 point Likert scale to elicit behaviours such as skipping medicines when they have no symptoms)    70% of participants reported good adherence to statins with no difference between groups
6 months - telephone administration of the 8-item Morisky adherence (7 yes/no items and 1 item with 5 point Likert scale to elicit behaviours such as skipping medicines when they have no symptoms)    80% of participants reported good adherence to statins with no difference between groups
Man-Son-Hing 19996 months - Self reported – Measured % of patients taking therapy initially chosen12995.35%13493.28%P = 0.44
Montgomery 2003˜ 3 years - Self reported – 6 item Adherence Questionnaire: from "I take all my tablets at the same time of day" to "I take hardly any of my tablets“    No difference
Montori 20116 months - Percentage of participants that self-reported currently taking medication who have not missed one dose within last week1765%1963%P = 0.92
6 months - Percentage of participants who opted to take biophosphonates who took their medication on more than 80% of the days for which it was prescribed, based on pharmacy records23100%1974%P = 0.009
Mullan 20096 months - Pharmacy records - days covered (range)4897.5% (range 0 to 100)37100 (range 73.9 to 100)

AMD −8.88 (−13.6% to −4.14%)

Positive AMD favours decision aid arm. This finding is statistically significant

6 months - Self reported by telephone call – did not miss a dose in last week4176%3181%

OR 0.74

(95% CI 0.24-2.32)

Oakley 20064 months - Extent to which the patients' behaviour in taking medications coincides with the clinical prescription1610.4% (32) [improvement from baseline]172% (26) [improvement from baseline]Not significant
Sheridan 20113 month - adherence to initial choice post intervention
Any therapy promoted in decision aid7645 (59%)7325 (34%)P < 0.01
any therapy promoted in decision aid + others (eg. diet or physical activity)7764 (83%)7752 (68%)P = 0.02
aspirin3230 (94%)1911 (58%)P < 0.01
cholesterol medicine1412 (86%)65 (83%)The intervention had little effect blood pressure or cholesterol medication,
however, the sample sizes for these estimates were
small and underpowered
blood pressure medicine99 (100%)1211 (92%)
stop smoking 825%520%No effect on smoking, although subgroups were small
and underpowered
Weymiller 2007

3 months - Self reported – mailed surveys & telephone call to non-respondents

on adherence to statin use: missed 1 dose or more within the last week.

3393.94%2979.31%No difference in adherence when analysis adjusted by sex, cardiovascular disease, and number of medications
Detailed versus simple DA
Deschamps 2004

12 months - Self reported – Telephone call to patients to ask estimated days missed per week and reasons

Response categories: 1) taking medication as prescribed (omitting no more than one day/week) , 2) missing doses occasionally and randomly, 3) systematically deviating from the prescribed directions

16˜72%20˜72%No difference
Rothert 199712 months - Self reported – daily adherence recorded on a calendar62˜89%74˜89%No difference
Trevena 20081 month - faecal occult blood test uptake1345.2%1376.6%P = 0.64

Three studies that compared a detailed to a simple decision aid reported no difference in adherence to hormone therapy at 12 months (Deschamps 2004; Rothert 1997), or in colorectal cancer screening rates at 1 month (Trevena 2008).

Health outcomes

General health outcomes

Ten studies (8.7%) compared a decision aid to usual care and one study compared detailed to simple decision aids in terms of general health outcomes. Eight of these (Barry 1997; Bernstein 1998; Kennedy 2002; Legare 2011; McCaffery 2010; Morgan 2000; Murray 2001a; Murray 2001b) used the previously validated Medical Outcomes Study 36-item Short-Form Health Survey (SF-36) or the 12-item Short-Form Health Survey (SF-12) (Stewart 1992), and one study (Vuorma 2003) used the RAND-36 (Hays 1993). As shown in Table 16, there were no significant differences for mental health function or social function in any of the seven studies. In one study (Barry 1997), general health and physical function outcome scores were significantly better in the decision aid group compared to usual care for men considering treatments for benign prostatic disease. Of the two studies evaluating the effect of a decision aid for women considering treatment for abnormal uterine bleeding, Kennedy 2002 found a statistically-significant improvement in role physical function, and Vuorma 2003 found a statistically-significant improvement in emotional role functioning for women.

Table 16. General health outcomes
  1. DA: decision aid; SF-36: Medical Outcomes Study 36-item Short-Form Health Survey; SF-12: 12-item Short-Form Health Survey;

    RAND-36: the 36-item short form survey from the RAND Medical Outcomes Study

    FACT-G: Functional Assessment of Cancer Therapy-General

ReferenceTimingN Decision aidMean Decision aid (SD)Change from baselineN ComparisonMean Comparison (SD)Change from BaselineNotes
General health - DA versus usual care
Barry 1997 (SF-36)Baseline10467.2 (19.0) 12371.1 (17.6) P = 0.02
3 months  -0.96 (1.41)  -3.59 (1.57)
6 months  -1.46 (1.41)  -4.93 (1.45)
12 months  0.61 (1.58)  -4.99 (1.44)
Legare 2011 (percentage of people who felt they had a stable and better health, (SF-12))2 weeks postnot reported94+7not reported85-6P = 0.08
Morgan 2000 (SF-36)6 months post7262 (23)+4.08865 (20)+7.0No difference
Kennedy 2002 (SF-36)2 years176  157  No difference
Vuorma 2003 (RAND-36)1 year156 2.2159 2.8No difference
Physical function - DA versus usual care
Barry 1997 (SF-36)Baseline10481.9 (20.0) 12383.0 (18.9) P = 0.02
3 months  -0.34 (1.61)  -1.81 (1.07)
6 months  0.10 (1.28)  -3.26 (1.37)
12 months  0.15 (1.40)  -3.74 (1.18)
Morgan 2000 (SF-36)6 months post7267 (29)+7.08871 (24)+10.0No difference
Kennedy 2002 (SF-36)2 years176  157  No difference
Vuorma 2003 (RAND-36)1 year156 2.4159 2.2No difference
Physical function - Detailed versus simple DA
Bernstein 1998 (SF-12)3 months post6138 (12.1)+0.64837.6 (10.6)+3.8No difference
Social function - DA versus usual care
Barry 1997 (SF-36)Baseline10490.6 (15.5) 12391.7 (15.7) P = 0.17
3 months  0.34 (1.58)  -2.26 (1.36)
6 months  -0.05 (1.92)  -2.46 (1.45)
12 months  -1.46 (1.85)  -3.52 (1.71)
Kennedy 2002 (SF-36)2 years176  157  No difference
McCaffery 2010 (SF-36)2 weeks7784.7 7182.1 P = 0.39
Vuorma 2003 (RAND-36)1 year156 5.2159 7.1No difference
Mental function - DA versus usual care
McCaffery 2010 (SF-36)2 weeks7771.3 7171.6 P = 0.46
Kennedy 2002 (SF-36)2 years176  157  No difference
Vuorma 2003 (RAND-36)1 year156 4.7159 5.3No difference
Mental function - Detailed versus simple DA
Bernstein 1998 (SF-12)3 months post6149.1 (11.4)0.04848.9 (10.8)+0.9No difference
Role function - DA versus usual care
Morgan 2000 (SF-36)6 months post7262 (44)+20.08858 (43)+15.0No difference
Kennedy 2002 (SF-36)2 years176  157  P = 0.04
Vuorma 2003 (RAND-36)1 year  9.2  6.3No difference
Bodily pain - DA versus usual care
Morgan 2000 (SF-36)6 months post7281 (22)+6.08877 (24)+5.0No difference
Kennedy 2002 (SF-36)2 years176  157  No difference
Vuorma 2003 (RAND-36)1 year156 6.5159 6.2No difference
Role emotional - DA versus usual care
Kennedy 2002 (SF-36)2 years176  157  No difference
McCaffery 2010 (SF-36)2 weeks7780.3 7177.4 P = 0.61
Vuorma 2003 (RAND-36)1 year156 12.6159 1.9P = 0.01
Energy/vitality - DA versus usual care
Kennedy 2002 (SF-36)2 years176  157  No difference
McCaffery 2010 (SF-36)2 weeks7755.2 7154.1 P = 0.09
Vuorma 2003 (RAND-36)1 year156 8.9159 8.8No difference
SF-36 all dimensions - DA versus usual care
McCaffery 2010 (SF-36)2 weeks7747 7146.3 P = 0.35
Murray 2001b (SF-36)9 months93  94  No difference
Murray 2001a (SP-36)9 months54  48  No difference
Functional status - DA versus usual care
Deyo 2000 (Roland Disability Questionnaire)1 year17120.4+5.417320.9+5.7No difference
Leighl 2011 (FACT-G) median (range)1 month post7417 (6-28) 6817.5 (7-28) P = 0.02
Health utilities - DA versus usual care
Murray 2001a (Euroqol EQ-5D)       No difference
Murray 2001b (Euroqol EQ-5D)       No difference

Deyo 2000, using the previously validated Roland Disability Questionnaire (Roland 1983) to measure functional status in patients with back pain, found no difference between the detailed decision aid and simple decision aid groups.

In two studies measuring health utilities using the Euroqol EQ-5D (Murray 2001a; Murray 2001b), there was no difference between the decision aid and usual care groups.

Condition-specific health outcomes

Twelve studies (10.4%) used various measures to assess condition-specific health outcomes (see Table 17). Ten of these compared decision aids to usual care (Barry 1997; Bernstein 1998; Leighl 2011; Morgan 2000; Murray 2001a; Murray 2001b; Protheroe 2007; Thomson 2007; van Peperstraten 2010; Vuorma 2003), and two compared a detailed decision aid to a simple decision aid (Deyo 2000; Raynes-Greenow 2010). Outcomes included urinary symptoms (Barry 1997; Murray 2001a), angina (Bernstein 1998; Morgan 2000), back pain (Deyo 2000), menopausal symptoms (Murray 2001b), menstrual symptoms (Protheroe 2007; Vuorma 2003), stroke or bleed (Thomson 2007), pregnancies and twin pregnancies (van Peperstraten 2010), and newborn Apgar score and birth weight (Raynes-Greenow 2010). Nine of the 12 studies (Bernstein 1998; Leighl 2011; Morgan 2000; Murray 2001a; Murray 2001b; Raynes-Greenow 2010; Thomson 2007; van Peperstraten 2010; Vuorma 2003) found no significant effects on condition-specific health outcomes. Protheroe 2007 reported statistically-significantly higher menorrhagia-related quality of life in women exposed to the decision aid compared to usual care. Deyo 2000 found no significant differences according to most measures, except for back pain severity -- for which improvement was shown, one year later, in the decision aid group. Barry 1997 showed an improvement in urinary symptoms in favour of the decision aid group, but it was not statistically significant.

Table 17. Condition-specific health outcomes
  1. AUA: American Urological Association; CCVA: Canadian Cardiovascular Angina; BPH: benign prostatic hyperplasia; DA: decision aid; SAQ: Seattle Angina Questionnaire;

StudyOutcomeScale usedTimingN Decision aidDecision aid mean changeN ComparisonComparison mean changeNotes
DA versus usual care
Barry 1997Urinary symptomsAUA Symptom Index (0 to 100)3 months104-4.80% (1.74)117-1.40% (1.37)No difference; trend toward DA
Urinary symptomsAUA6 months104-3.66% (2.06)117-3.17% (1.77)No difference
Urinary symptomsAUA12 months104-2.51% (2.11)117-4.14% (1.66)No difference; trend toward control
Impact of symptomsBPH Impact Index (0 to 100)3 months104-6.58% (1.10)117-3.00% (1.05)No difference; trend toward DA
Impact of symptomsBPH6 months104-4.37% (1.32)117-3.89% (1.16)No difference; trend toward DA
Impact of symptomsBPH12 months104-5.53% (1.32)117-2.63% (1.32)No difference; trend toward DA
Bernstein 1998SatisfactionSAQ (0 to 100)3 months61+6.2%48+10.5%Control significantly more satisfied
Angina stabilitySAQ3 months61+17.2%48+28.3%No difference
Angina frequencySAQ3 months61+5.5%48+15.3%No difference
Disease PerceptionSAQ3 months61+14.1%48+18.8%No difference
Physical CapacitySAQ3 months61-0.5%48+7.1%No difference

Leighl 2011

(FACT-G) median (range)

Physical function at 1 month post7421 (0-28) 6820 (4-28) No difference
Role emotional at 1 month post7417 (0-20) 6817(7-20) No difference
Morgan 2000No AnginaCCVA6 months72+49%88+48%No difference
Class I AnginaCCVA6 months72-1%88+6%No difference
Class II AnginaCCVA6 months72-23%88-26%No difference
Class III AnginaCCVA6 months72-26%88-28%No difference
Class IV AnginaCCVA6 months720%880%No difference
Murray 2001aUrinary symptomsAUA symptom Index (0 to100)     No difference
Murray 2001bMenopausal symptomsMenQol     No difference
Protheroe 2007Menorrhagia specific utility scale(0 to 100)6 months6059.3 (30.0)5650.9 (25.1)P = 0.03 higher menorrhagia quality of life favouring DA group
Thomson 2007Strokes or bleeds requiring admission 3 months51 55 No strokes and no bleeds requiring admission. 1 bleed and 1 transient stroke both in control group that required GP consultation
van Peperstraten 2010Ongoing pregnancies (> 12 weeks gestation) after 1st IVF cycle152 156 32% of participants in the intervention group and 38% of participants in the control group had ongoing pregnancies, P = 0.25
Twin pregnancies (> 12 weeks gestation) after 1st IVF cycle152 156 4% of participants in intervention group and 6% of participants in control group had twin pregnancies, P = 0.33
Vuorma 2003Inconvenience due to menstrual bleeding(5 to 25)1 year15610.415910.5No difference
Menstrual pain(0 to 12)1 year1564.71594.6No difference
Detailed versus simple DA
Deyo 2000% working 1 year171+17.3%173+18.3%No difference
% missed 1+ day work within past month 1 year171-38.4%173-35.2%No difference
Back pain severity 1 year171-22.4%173-22%1 year scores: DA 27.6% significantly better than control 37.2%
Leg pain severity 1 year171-42.1%173-43.9%No difference
Seeking compensation 1 year171-2.9%173-5.9%No difference
Satisfied with symptoms 1 year171+32.1%173+32.4%No difference
Raynes-Greenow 2010Apgar scorescores > 71 minute after birth395221 (82%)201149 (75%)P = 0.12
scores > 75 minutes after birth395235 (90%)201167 (84%)P = 0.68
Birth weightin gramsmean (SD)3953445 (451)2013412 (450)P = 0.11
Preference-linked health outcomes

None of the 115 studies measured preference-linked health outcomes—that is, whether the patients experienced the outcomes they preferred and avoided the outcomes they wanted to avoid.

Anxiety

Of 115 studies, 30 (26.1%) measured anxiety, with 19 using the previously validated 20-item State Trait Anxiety Inventory (Spielberger 1970) and 1 using a single question on a 7-point Likert scale (Johnson 2006) (see Table 18). Twenty-four of these studies involved decision aid/usual care comparisons, and five (Goel 2001; Hunter 2005; Raynes-Greenow 2010; Tiller 2006; van Roosmalen 2004) involved detailed/simple decision aid comparisons (see Table 18). Of 23 studies that measured anxiety within 1 month, 2 (8.7%) reported that the decision aid group had statistically-significantly lower anxiety scores for people considering birthing options after previous caesarean (Montgomery 2007) and for women considering options for treatment of menorrhagia (Protheroe 2007). Green 2004 reported a greater reduction in anxiety for high-risk women considering genetic testing in the control group and a greater reduction in anxiety for low-risk women considering genetic testing in the decision aid group.

Table 18. Anxiety
  1. DA: decision aid; HPV: human papilloma virus; PSA: prostate-specific antigen

Study Timing

N

Decision aid

Mean

Decision aid (SD)

Change from baseline

N

Comparison

Mean Comparison (SD) Change from Baseline Notes
State Anxiety Inventory: < 30 days post-intervention - DA versus usual care
Bekker 2004; prenatal screeningImmediately post5058.9 (16.6) 5661.2 (13.7) No difference
Evans 2010; PSA screeningimmediately post DA894.98 1034.88 P = 0.98
Green 2004; breast cancer screening (low risk group)Immediately post5629-46130-3P = 0.04 (for difference in change score)
Green 2004; breast cancer screening (high risk group)Immediately post5030-34433-5P = 0.04 (for difference in change score)
Leighl 2011post consult, 1-2 weeks and 4 weeks post      No difference; see Figure 3
Mathieu 2007; mammography screeningimmediately after32129.61 31529.34 No difference
McCaffery 2010; HPV screening (state trait anxiety inventory)2 weeks7710.5 7110.6 P = 0.25
Montgomery 2003; hypertensionimmediately post DA4435.45 (10.52) 5037.67 (13.92) No difference
Montgomery 2007; previous cesarean section37 weeks gestation19638.7 (12.2) 19542.1 (12.2) P = 0.016
Nassar 2007; breech presentation1 week9841.4 (12.5) 9044.4 (13.9) No difference
Protheroe 2007; menorrhagia2 weeks59

11.6

(3.7)

 6112.2 (3.7) P = 0.016
Rubel 2010; PSA screeningimmediately after      

20 items adapted from state portion of State-Trait Anxiety Inventory Scale STAI - Form Y;

total mean score= 1.66±0.59 (n=200) for patients in both groups

Smith 2010; bowel cancer screening2 week follow-up35713.67 17314.05 P= 0.80
Thomson 2007; anti-thrombotic treatment for atrial fibrillationimmediately after53  56  Significant fall in anxiety (-4.57) but no difference between groups (P = 0.98)
Trevena 2008 colorectal cancer screeningimmediately after134  137  No difference (P = 0.59)
van Peperstraten 2010; number of embryos transferredimmediately after15227.33% 15624.5% P = 0.14
Whelan 2004; breast cancer surgery7 days post DA9442.3 (1.3) 10741.9 (1.3) No difference
Whelan 2003; breast chemotherapy7 days post DA8245.6+2.29347.4+0.8No difference
Wong 2006; pregnancy terminationImmediately post15454 (15.8) 15954 (16.1) No difference
State Anxiety Inventory: < 30 days post-intervention - Detailed versus simple DA
Goel 2001; breast cancer surgery1 to 3 days post DA7451.2 (14.2)-0.74350.7 (14.8)-0.1No difference
Hunter 2005; prenatal screeningImmediately post11645.50 (9.69)-1.1712647.98 (10.14)-0.37No difference
Raynes-Greenow 2010; labour analgesia37 weeks gestation395

33.3

(9.3)

-0.6201

34.3

(11.0)

0P = 0.32
Tiller 2006; prophylactic ovarian cancer treatment2 weeks5838.2 (13.4) 6038.0 (15.2) No difference
State Anxiety Inventory: 1 month post-intervention - DA versus usual care
Bekker 2004; prenatal screening1 month post DA2935.3 (12.5) 3934.7(14.8) No difference
Davison 1997; prostate cancer treatment5 to 6 weeks post DA3035.5-9.03034.5-2.5No difference
State Anxiety Inventory: 1 month post-intervention - Detailed versus simple DA
van Roosmalen 20041 month post DA4335.4 (11.7) 4337.4 (10.7) No difference
State Anxiety Inventory: 3 months post-intervention - DA vs usual care
Murray 2001a; benign prostatic hypertrophy3 months post DA5536.36 (14.99)+2.44832.08 (9.836)+0.7No difference
Murray 2001b; hormone replacement therapy3 months post DA9338.42 (10.83)-0.59540.53 (12.96)+1.8No difference
Nagle 2008; prenatal screening˜1 to 12 weeks post DA16737.2 (12.1) 17137.36 (12.6) No difference
Nassar 2007; breech presentation3 months post DA8629.2 (9.9) 8430.8 (10.5) No difference
Vuorma 2003; menorrhagia treatment3 months post DA18437.1+1.017935.9-1.0No difference
Whelan 2003; breast chemotherapy3 months post DA8236.0 9337.8 No difference
State Anxiety Inventory: 6 months post-intervention - DA versus usual care
Protheroe 2007; menorrhagia6 months post DA4711.2 (4.2) 5213.3 (4.9) P = 0.067
Whelan 2004; breast cancer surgery6 months post DA9439.3 (1.3) 10738.9 (1.6) No difference
Whelan 2003; breast chemotherapy6 months post DA8238.2 9338.2 No difference
State Anxiety Inventory: 6 months post-intervention - Detailed versus simple DA
Goel 2001; breast cancer surgery6 months post DA5936.6 (12.9)-15.33934.3 (11.6)-16.5No difference
Tiller 2006; prophylactic ovarian cancer treatment6 months post DA5335.7 (9.0) 5536.2 (13.6) No difference
State Anxiety Inventory: 12 months post-intervention - DA versus usual care
Whelan 2004; breast cancer surgery12 months post DA9437.5 (1.4) 10736.6 (1.5) No difference
Whelan 2003; breast chemotherapy12 months post DA8239.2 9340.2 No difference
Other - DA versus usual care
Johnson 2006; endodontic treatmentImmediately post - single question 7-point Likert scale.323.2 (1.7) 353.8 (2.1) P = 0.27
Lewis 2010; colorectal cancer screeningintrusive thoughts - 3 items; 4 point scale - not at all13966.2% 15768.0% P = 0.92
intrusive thoughts - 3 items; 4 point scale - sometimes6631.4% 6929.9% 
intrusive thoughts - 3 items; 4 point scale - often52.4% 52.2% 
McCaffery 2010intrusive thoughts - measured using 1 item from the impact of events scale7743% 7132% No difference
Smith 2010Worry about developing bowel cancer - quite or very3576% 1738% P = 0.78
Worry about developing bowel cancer - none or a bit35794% 17392% 

None of the studies demonstrated significant differences in effects on people's state anxiety at one month (n = 3 studies), at three months (n = 6 studies), at six months (n = 5 studies), or at one year (n = 2 studies).

Depression

Of 115 studies, 9 (7.8%) measured the effect of decision aids on depression using various instruments (Table 19). None of the studies reported a statistically-significant difference between groups for decisions about cancer treatment (Davison 1997; Whelan 2004), depression (Loh 2007), prenatal genetic testing (Nagle 2008), people at higher risk of cancer who were considering risk management (Tiller 2006), women considering number of embryos to transplant (van Peperstraten 2010), or genetic testing (Wakefield 2008; Wakefield 2008a; Wakefield 2008b).

Table 19. Depression
  1. CES-D: Centre for Epidemiology Studies Depresion Scale; DA: decision aid; IVF: in vitro fertilisation

Study Timing

N

Decision aid

Mean

Decision aid (SD)

Change from Baseline

N

Comparison

MeanComparison (SD) Change from Baseline Notes
DA versus usual care
Davison 1997 (20-item CES-D)5 to 6 weeks3029.8-0.63029.5+1.3No difference
Loh 2007 (Brief Patient Health Questionnaire-D)6 to 8 weeks19129.8 (2.7) 9627.0 (3.6) P = 0.236
Nagle 2008 (Edinburgh Postnatal Depression Scale)˜1 to 12 weeks post DA16719 (11.6) 17119 (11.2) No difference
Tiller 2006 (Hospital Anxiety and Depression Scale)2 weeks post DA5810.9 (5.6) 6110.7 (6.4) P = 0.03
6 mos post DA5010.1(4.7) 5610.8 (6.4) P = 0.12
van Peperstraten 2010 (Beck Depression Inventory)after multifaceted intervention/ before IVF12616 (13%) 1365 (4%) P = 0.01
at uptake of IVF14716 (11%) 151113 (9%) No difference
Whelan 2004 (20-item CES-D)1 week post DA9413.8 (1.0) 10713.4 (1.1) No difference
6 months post DA9415.1 (1.1) 10714.2 (1.2) No difference
12 months post DA9413.2 (1.3) 10712.8 (1.2) No difference
Detailed versus simple DA
Wakefield 2008 (Hospital Anxiety and Depression Scale)1 week post48  61  No difference
Wakefield 2008a (Hospital Anxiety and Depression Scale)1 week post56  63  No difference
Wakefield 2008b (Hospital Anxiety and Depression Scale)immediately55  55  No difference
Regret

Of 115 studies, 7 (6.1%) measured the effect of decision aids on decision regret, using the 5-item Decisional Regret scale (Brehaut 2003) (see Table 20). Of the seven studies, two compared decision aids to usual care and five studies compared detailed to simple decision aids. There was no statistically-significant difference for any of the seven studies.

Table 20. Decisional regret
  1. DA: decision aid

AuthorItem

N

Decision aid

Proportion or

Mean (SD)

N

Control

Proportion or

Mean (SD)

Notes
DA vs usual care
Hanson 20115-item Decisional Regret Index12611.912714.3P = 0.14
Legare 2011Proportion of patients with decisional regret 7% 9%P=0.91
Detailed vs simple DA
Goel 2001Right decision6358 (92.06%)4442 (95.45%)No difference
Regret choice638 (12.70%)445 (11.36%)No difference
Would make same choice6354 (85.71%)4440 (90.91%)No difference
Choice did me harm637 (11.11%)443 (6.82%)No difference
Decision was wise6354 (85.71%)4441 (93.18%)No difference
Kuppermann 2009

Decisional Regret - 3 items

at 26 to 30 weeks gestation

2449.625212.8P = 0.28
Wakefield 2008Decision Regret Scale at 6 months41 54 No difference
Wakefield 2008aDecision Regret Scale˜577.04 (12.12)˜636.39 (13.68)No difference
Wakefield 2008bDecision Regret Scale˜569.78 (14.49)˜495.13 (10.16)No difference
Confidence

Of 115 studies, 9 (7.8%) measured the effect of decision aids on confidence levels: 8 compared decision aids to usual care and Rothert 1997 compared detailed to simple decision aids (see Table 21). Four of these studies used the Decisional Self-efficacy Scale (Allen 2010; Arterburn 2011; Fraenkel 2007; Smith 2010). Of these eight studies, four reported a statistically-significant improvement in confidence or self-efficacy with decision making in the decision aid compared to the usual care groups (Chambers 2012; Fraenkel 2007; Gattellari 2003; McBride 2002) and the other studies reported no difference between groups. The other study comparing detailed to simple decision aids found no difference in confidence scores immediately post decision aid or at 12 months (Rothert 1997).

Table 21. Confidence
  1. CI: confidence interval; DA: decision aid; SD: standard deviation

Study Scale used Timing N Decision aid Decision aid - mean N Comparison Comparison - mean Notes
DA vs usual care
Allen 201011-item self-efficacy scalepost intervention291

83%

(40.26% SD)

334

79%

(33.08% SD)

No difference
Arterburn 2011Decisional self efficacychanges from baseline75+ 3.0 (95% CI 0.6 to 5.4)77+ 2.8 (95%CI 0.9 to 4.8)P = 0.78
Chambers 2012Mean confidence with decision: scale from 1 (low confidence) to 5 (high confidence)post intervention484593.6P = 0.02
Fraenkel 2007Decisional self-efficacy scalepre-consultation4332 (median)4027 (median)P = 0.001
Gattellari 2003Perceived ability to make an informed choice 1-item; 5-point Likert scale3 days post106 108 P = 0.008; DA group more likely to agree that they could make an informed choice about PSA screening
Gattellari 2005Perceived ability to make an informed choice 1-item; 5-point Likert scaleImmediately post131 136 No difference
McBride 2002Confidence with ability to understand outcomes of hormone replacement therapy, make a decision, engage in discussion with practitioner 3-items (0 to 10; low to high confidence)1 month post27378% (18% SD)28470% (19% SD)P < 0.0001
9 months post26180% (17%SD)27875% (20% SD)P = 0.0004
Smith 20103 items adapted from the Decisional self-efficacy scale2 week follow-up3574.67 (0.54 SD)1734.61(0.62 SD)P = 0.26
Detailed versus simple DA
Rothert 19978-items (1 to 10; low to high confidence)post DA8378% (16% SD)8980% (19% SD)No difference
12 months post6378% (15% SD)7480% (19% SD)No difference

Healthcare system effects

Consultation length

Of 115 studies, 10 (8.7%) evaluated the effect of a decision aid compared to usual care (n = 9) or simple decision aid (n = 1) on consultation length, with a range from 8 minutes shorter to 23 minutes longer (median 2.5 minutes longer) (see Table 22). Four studies evaluated decision aids in the form of decision boards used primarily within the consultation (Loh 2007; Vodermaier 2009; Weymiller 2007; Whelan 2003). Of 9 studies, Bekker 2004 reported consultations about prenatal diagnostic testing were 6 minutes longer for women who prepared for the consultation using a decision aid, Thomson 2007 reported consultations about treatment for atrial fibrillation were 23 minutes longer when using a computerized decision aid with standard gamble method within the consultation compared to guideline driven consultation, and Green 2004 reported that consultations about breast cancer genetic testing were shorter by 8 minutes when women prepared using a decision aid. The other six studies that evaluated decision aids compared to usual care and one study that evaluated detailed compared to simple decision aids reported no statistically-significant difference in consultation length (see Table 22). Results were not pooled given the variability in the way length of time was reported, including many studies that did not include standard deviations.

Table 22. Healthcare system effects
  1. CI: confidence interval; DA: decision aid; SD: standard deviation

Study Scale used N Decision aid Decision aid - mean N Comparison Comparison - meanDifference between groups Notes
Consultation length - DA versus usual care
Bekker 2004Consultation length using DA in consult (minutes)5032.2 (13.0 SD)5626.3 (11.5 SD)+5.9 minutesP = 0.01 (longer with decision aid)
Green 2004Consultation length with practitioner post DA (minutes)1068210590-8 minutesP = 0.03 (shorter with decision aid)
Krist 2007Time spent discussing prostate cancer with practitioner post DA (minutes) -patient reported1965.3755.2+0.1 minutesNo difference between groups
Time spent discussing prostate cancer with practitioner post DA (minutes) - physician reported1963.8754.2-0.4 minutesNo difference between groups but physicians thought they spent less time than patients (P < 0.001)
Loh 2007Consultation length using DA in consult (minutes)19129.2 (10.7)9626.7 (12.5)+2.5 minutesP = 0.681
Ozanne 2007Consultation length using DA in consult (minutes)15241521+3 minutesP = 0.42
Thomson 2007Consultation length using DA in consult (minutes)844 (39 to 55)1021 (19 to 26)+23 minutes

P = 0.001

Compared computerized decision aid with standard gamble within the consultation to guideline driven consultation

Vodermaier 2009Consultation length with practitioner post DA
5 to 10 min536 (11.3%)545 (9.3%) P = 0.91
10 to 15 min17 (32.1%)19 (35.2%) 
15 to 25 min15 (28.3%)14 (25.9%) 
25 to 35 min7 (13.2%)5 (9.3%) 
Above 35 min8 (15.1%)11 (20.4%) 
Whelan 2003Consultation length using DA in consult (minutes)5068.35065.7+2.6 minutesP = 0.53
Weymiller 2007Consultation length using DA in consult (minutes)52 46 +3.8 minutes

Not statistically significant

3.8 min longer in DA group (95%CI -2.9 to 10.5)

Consultation length - Detailed versus simple DA
Myers 2011Encounter length with practitioner post DA (minutes)     Median 16 minutes for both groups (range 6 to 44)
Cost and resource use - DA versus usual care
Hollinghurst 2010; Montgomery 2007Cost-consequences analysis235£2019 (SD £741)238£2033 (SD £677) no difference
Kennedy 2002Cost effectiveness204$1556 USD215$2751 USD Mean difference $1184 (95%CI $684 to $2110)
Murray 2001aTotal costs excluding intervention57£310.3 (SD £602.0)48£188.8 (SD £300.4) Mean difference £121.5 (95% CI £ -58.9 to £302.0)
Total costs including intervention57£594.10 (SD £602)48£188.8 (SD £300.4) Mean difference £405.4 (95% CI £224.9 to £585.8) P < 0.001
Murray 2001bTotal costs excluding intervention85£90.584£90.9 (SD £39.2) No difference
Total costs including intervention85£306.5 (SD £42.8)84£90.9 (SD £39.2) Mean difference £215.5 (95% CI £203.1 to £228.0) P < 0.001
Thomson 2007GP consultations post intervention39/51 32/54  P = 0.35
Hospital appointments post intervention29/51 10/54  P = 0.06
van Peperstraten 2010Mean total savings per couple     Mean total saving per couple in the intervention group were €169.75 ($219.12 USD)
Vuorma 2003Cost and productivity losses184€2760 Euro179€3094 Euro 

P = 0.1

No difference between intervention and control when treatment cost and productivity losses were analysed.

Cost and resource use - Detailed versus simple DA
Deyo 2000Healthcare use at 1 year171 172  No difference in most services; DA less surgery for herniated disk
Cost and resource use

Eight studies (7.0%) evaluated the impact of decision aids compared to usual care on costs (Kennedy 2002;Montgomery 2007; Murray 2001a; Murray 2001b; van Peperstraten 2010; Vuorma 2003) or resource use only (Deyo 2000; Thomson 2007) (see Table 22).

Both studies by Murray involved a cost-minimization economic analysis from the perspective of the healthcare system decision-maker, with less than 4% of resource use items being replaced by conditional means due to missing data. There was no significant difference between the groups in terms of health service resource use. There was a difference in costs, when the additional costs of interactive videodisc equipment was considered in the analysis.

The cost-effectiveness analysis in the Kennedy 2002 study was also conducted from the healthcare system perspective, using 1999 to 2000 US dollars and calculated over two years. The decision aid with nurse coaching demonstrated the lowest mean cost ($1566) compared to decision aid alone ($2026) or usual care ($2751).

In the Vuorma 2003 study, despite the statistically-insignificant trend for fewer diagnostic procedures (55 versus 89; P = 0.07) and lower rates for uterine-preserving surgery procedures (16 versus 26; P = 0.08) in the intervention group, there was no difference between the intervention and control group when treatment cost and productivity losses were analyzed at one year follow-up.

van Peperstraten 2010 evaluated the costs from a healthcare perspective and determined the difference in total costs per couple between groups. The mean total savings in the decision aid compared to usual care group was €169.75 per couple.

Montgomery 2007 evaluated the costs from the United Kingdom National Health Service perspective and reported that there was no difference in mean costs per patient between groups.

For healthcare resource use, there was no difference in most services for Deyo 2000 (except fewer surgeries for herniated disc in the detailed versus simple decision aid group) and no difference in general practitioner consultations was reported by Thomson 2007.

Litigation rates

None of the 115 studies examined the effect of decision aids on litigation.

Post-hoc analysis

Effects of study quality

To examine the potential bias arising from including trials of low methodological quality, eight trials with a high risk of bias for any of the seven risk of bias criteria were excluded from the analysis (Auvinen 2004; Chambers 2012; Clancy 1988; Hamann 2006; Krist 2007; Lewis 2010; Man-Son-Hing 1999; Rostom 2002). Overall, the results remained the same (Table 23). For a more conservative post-hoc analysis, we also excluded 64 trials with a 'high' risk of bias or 'unclear' risk of bias for at least 3 of the 7 criteria (Figure 3). Overall, the results of this more conservative analysis remained the same (Table 23).

Table 23. Sub-analysis using higher quality trials
  1. CI: confidence interval

OutcomeOverall mean effect (95% CI)Without trials having high risk of bias on at least 1 of 7 criteriaWithout trials having high or unclear risk of bias for at least 3 of 7 criteria
Knowledge - decision aid versus usual care13.29 (11.32 to 15.25) n = 4213.67 (11.60 to 15.74) n = 3914.97 (11.84 to 18.10) n = 21
Knowledge - detailed versus simple decision aid5.52 (3.9 to 7.15) n = 195.48 (3.78 to 7.18) n = 186.97 (2.39 to 11.55) n = 3
Accurate risk perceptions - with probabilities versus no probabilities1.82 (1.52 to 2.16) n = 191.76 (1.48 to 2.10) n = 182.28 (1.78 to 2.92) n = 7
Values congruent with chosen option1.51 (1.17 to 1.96) n = 131.52 (1.17 to 1.97) n = 132.15 (1.35 to 3.44) n = 6
Uninformed sub-scale of Decisional Conflict Scale - decision aid versus usual care-7.26 (-9.73 to -4.78) n = 22-7.40 (-10.06 to -4.74) n = 21-8.26 (-11.74 to -4.78) n = 15
Uninformed sub-scale of Decisional Conflict Scale - detailed versus simple decision aid-2.39 (-4.39 to -0.39) n = 10-2.39 (-4.39 to -0.39) n = 10too few to assess, n = 1
Unclear values sub-scale of Decisional Conflict Scale - decision aid versus usual care-6.09 (-8.5 to -3.67) n = 18-6.40 (-9.02 to -3.79) n = 17-7.02 (-10.00 to -4.04) n = 14
Unclear values sub-scale of Decisional Conflict Scale - detailed versus simple decision aid-2.31 (-4.67 to 0.05) n = 10-2.31 (-4.67 to 0.05) n = 10too few to assess, n = 1

We applied a fixed-effect model for the primary outcomes and compared it to the random-effects model used in the analysis reported earlier. The results were similar. For example, knowledge results were 13.34 (95% CI 11.17 to 15.51) using a random-effects model compared to 13.61 (95% CI 12.83 to 14.38) using a fixed-effect model. Therefore, there is little concern about the impact of small studies being included that could potentially have shown more beneficial effects (Sterne 2011).

Heterogeneity

When patient decision aids were compared to usual care, there was statistically-significant heterogeneity in five of six of the IPDAS effectiveness criteria: knowledge; accurate risk perceptions; values congruence with choice; feeling uninformed; and feeling unclear regarding personal values. There was no statistically-significant heterogeneity for participation in decision making. It should be noted that the heterogeneity of the effect was not manifested in its direction but only in its size. For the 2009 update (O'Connor 2009), we explored the potential factors contributing to heterogeneity (Table 24). Overall, scores for outcomes were similar to the overall effect regardless of sub-analysis conducted as indicated by overlapping confidence intervals.

Table 24. Heterogeneity (based on 55 trials in search to 2006)
OutcomeOverall effectTreatment decisionScreening decisionVideo/computer Decision aidAudio/pamphlet Decision aidBase risk controlRemoval of Outliers*
Knowledge - decision aid versus usual care15.2 (11.7 to 18.7)16.5 (11.9 to 21.2)13.1 (7.7 to 18.5)21.3 (16.3 to 26.2)11.9 (8.3 to 15.6)15.5 (11.3 to 19.8)17.3 (13.6 to 20.9) (*Bekker 2004, Gattellari 2003, Johnson 2006)
Accurate risk perceptions - probabilities versus no probabilities1.6 (1.4 to 1.9)1.6 (1.4 to 1.9)1.6 (1.1 to 2.3)No data1.6 (1.4 to 1.9)1.3 (1.2 to 1.5) (P = 0.3)1.5 (1.3 to 1.7) (*Gattellari 2003)
Uninformed sub-scale of the Decisional Conflict Scale - decision aid versus usual care-8.4 (-11.9 to -4.8)-9.4 (-13.3 to -5.5)-3.5 (-12.9 to 5.8)-12.6 (-19.5 to -5.8)-4.9 (-7.6 to -2.3) (P = 0.06)-5.4 (-7.7 to -3.2) (P = 0.11)-6.2 (-8.4 to -4.1) (P = 0.06) (*Montgomery 2003)
Unclear values sub-scale of the Decisional Conflict Scale - decision aid versus usual care-6.3 (-10.0 to -2.7)-6.0 (-9.8 to -2.3)Insufficient data-8.0 (-15.1 to -1.0)-4.5 (-8.4 to -0.6)-3.6 (-6.8 to -0.5)-4.0 (-6.7 to -1.3) (*Montgomery 2003)

Discussion

Summary of main results

The addition of 33 studies in this updated review confirms many of the observations reported in the previous versions of our reviews (O'Connor 2003b; O'Connor 2009; Stacey 2011). Based on the GRADE assessment (Summary of findings for the main comparison; Table 2), there is high-quality evidence that decision aids compared to usual care improve people's knowledge regarding options and reduce their decisional conflict related to feeling uninformed and unclear about their personal values. There is moderate-quality evidence that decision aids compared to usual care stimulate people to take a more active role in decision making and that decision aids with probabilities compared to interventions without probabilities increase accurate risk perception. There is low-quality evidence that decision aids improve congruence between the chosen option and their values. For secondary outcomes, there is a decrease in the proportion of people remaining undecided. Compared to simple versions, detailed decision aids improved knowledge only marginally.

The impact of decision aids on increasing or decreasing the numbers of patients choosing particular options continues to be variable. New in this update is pooled evidence indicating a non-significant trend for a 12% increase in colorectal cancer screening in those exposed to a decision aid compared to usual care. The numbers of patients choosing to have major elective surgery continues to be decreased in favour of more conservative options except when base rates are low (e.g., surgery for benign prostate hyperplasia, prophylactic mastectomy for women who are carriers of the BRCA gene). The numbers of patients choosing hormone replacement therapy and prostate-specific antigen (PSA) testing were decreased with exposure to decision aids.

Decision aids do no better than alternative interventions in terms of their effects on people’s satisfaction with decision making, anxiety, or health outcomes such as general quality of life or condition-specific quality of life. However, no studies measured preference-linked health outcomes. There continue to be too few studies to determine the effects of decision aids on costs/resource use. Although there may be additional costs of delivering decision aids, an independent review of decision aid trials with economic outcomes concluded that "this was likely to be small relative to the benefit to patients in terms of improved decision quality when effective decision aids are used" (NCGC/NICE 2012). Although several studies have measured adherence, the variability in the measurement makes it difficult to determine the effect of patient decision aids on adherence to the chosen option.

New for this update, we analyzed the pooled data for screening decision aids separately from treatment decision aids, and found that the results were similar.

Quality of the evidence

Risk of bias ratings show the variability in risk of bias across studies. The two criteria for which studies scored the worst were selective outcome reporting, and blinding of participants and personnel. When a post-hoc analysis was conducted that involved removing studies at high risk of bias, there was no effect on the results. The conclusions of this review are limited by: a) inadequate power to detect important differences in effectiveness in subgroups; and b) the wide variability in the decision contexts, the elements within the patient decision aids, the type of comparison interventions, the targeted outcomes, and the evaluation procedures. The small number of studies for most outcomes did not allow for analysis of publication bias due to failure to publish negative studies. Moreover, there may have been publication bias due to failure to report all negative findings in a published study. Several of the outcomes demonstrated statistically-significant heterogeneity. For the outcome of knowledge, for example, heterogeneity would be expected, given that the knowledge tests themselves were not standardized. Furthermore, the heterogeneity found in the various outcomes reflects differences across clinically-diverse studies; therefore, the pooled effect size and confidence intervals should be interpreted as a range across conditions, which may not be applicable to a specific condition.

Main effects of decision aids  

The largest and most consistent benefits of decision aids, relative to usual care, are better knowledge of options and outcomes, more accurate perceptions of outcome probabilities, and congruence between the chosen option and the person's values. These observations are clinically important because the usual care group's knowledge and understanding of probable outcomes were less than the intervention group; both knowledge and understanding of probable risk are important for ensuring informed decision making. These effects on knowledge and risk perceptions suggest that current 'usual care' may not be good enough when informing people about these complex, values-sensitive decisions. People need to comprehend the options and probable outcomes in order to consider and communicate to their practitioners the personal value they place on the benefits versus the harms. As well, in this update there is a significant increase in values-based choice when decision aids with explicit values clarification exercises were compared to a simple decision aid without explicit values clarification or usual care.

Decision aids, when compared to usual care, also help people feel more comfortable with their choices. This is revealed by the reduced scores for the decisional conflict sub-scales. People who use decision aids generally feel more informed about options and clearer regarding their personal values.

Compared to usual care strategies, decision aids improve individuals' involvement in decision making. This observation suggests that the International Patient Decision Aids Standards criterion of helping patients participate ‘in ways that they prefer’ needs to be assessed after a patient has adequate information about what involvement means. People may have a mistaken preference for passivity because they believe that the best choice relies on the expertise of the clinician (which option is medically reasonable?) rather than the opinions of the person who will experience the outcomes (which outcomes matter most to me?).

Evidence continues to build that decision aids have a positive effect on the patient-practitioner consultation in all nine studies that measured it, and have a variable effect on the length on these consultations. Of the studies that measured patient-practitioner communication, five involved using decision aids within the consultation and in three decision aids were used in preparation for the consultation. Interestingly, most studies of consultation length were conducted on decision aids intended to be used within the consultation, and fewer studies were focused on consultation length when decision aids were used in preparation for the consultation. Of the 10 studies that measured consultation length, 2 that also measured patient-practitioner communication reported that there was increased patient participation in the consultation for those exposed to the patient decision aid within the consultation compared to usual care (Weymiller 2007) but there was no difference when the detailed compared to simple decision aid was used in preparation for the consultation (Myers 2011). However, few studies have evaluated the impact of patient decision aids in clinical practice and further research is underway to better evaluate outcomes when the trial was conducted within clinical practice versus as another research study.

Variable effects of decision aids 

There may be several reasons for the variable impact of decision aids on choices. First, most studies were under-powered to detect important differences in choices. Second, in the five studies reporting choices at baseline and post decision aid, some options may have been under-used and others over-used, relative to the choices individuals would make if they were more fully informed. Under these circumstances, one could expect to observe directional effects on choices once people become better informed and more involved in decision making. Examples of relatively under-used options at baseline were colon cancer screening and hepatitis B vaccination. Another illustration lies in the non-significant five-fold increase in the number of people choosing prostate surgery in the UK study and uptake in prophylactic mastectomy in women who were carriers of the BRCA gene. For the prostate example, there was a shortage of urologists and low referral rates for benign prostatic hyperplasia; whereas the breast example reflects the growing number of women who test gene positive who are aware of their options for preventing breast cancer. This situation may have resulted in under-use of a chosen option, which was corrected with exposure to a decision aid. In contrast, the other surgical decision aid studies had higher numbers of people choosing surgery in the control group. The procedure may have been chosen due to people’s inflated perceptions of the probabilities of benefits, lack of appreciation of the probabilities of harms, and lack of awareness of alternatives. Exposure to the decision aid reduced the number of people choosing surgery in favour of more conservative alternatives.   

Limited effects of decision aids

The limited effects of decision aids on reported satisfaction with the decision-making process and with the actual choice made may indicate that decision aids have a limited effect on satisfaction. The null effects may also be due to measurement insensitivity. This is especially likely when satisfaction with usual care is already quite high (e.g., ceiling effects) and when choices are inherently difficult to make because of competing benefits and harms. Furthermore, once the decision is made, people may find it psychologically more comforting to say that they are satisfied rather than entertain doubts about what they have chosen (Gruppen 1994). Interestingly, in the two studies that used the Preparation for Decision Making Scale to compare the decision aid to usual care, both reported positive outcomes for satisfaction in the patient decision aid group.

The small differences in knowledge and decisional conflict scores between detailed and simple versions of decision aids are likely due to the overlapping information presented in the two interventions. This raises questions about the minimum information needed for the decision aid to be effective. For example, in the study by Goel 2001, a simple pamphlet describing options and outcomes of mastectomy versus lumpectomy was comparable to a detailed audio-workbook for women newly diagnosed with breast cancer. However, a post-hoc analysis revealed that women who were uncertain about their choice at baseline or were leaning toward mastectomy, appeared to benefit more from the detailed aid. These observations indicate a need to establish the 'essential ingredients' in decision aids and to identify the people who are most likely to benefit from more detailed versions. As the body of available research grows, it will become easier and more important to assess the usefulness of different components of decision support for different clinical contexts, decision problems, and groups of people. The IPDAS Collaboration is trying to establish the evidence for the various components in decision aids in the recent set of reviews underlying the IPDAS checklist (IPDAS 2013). At a minimum, the IPDAS collaboration proposes criteria for defining the intervention as a patient decision aid and certifying criteria (Joseph-Williams 2013).

It is not surprising that decision aids had limited effects on health outcomes. One reason for using a decision aid is that there is often no option with a clear health outcome advantage. For example, when men with localized prostate cancer consider active treatment options, their health outcomes can be different, depending on whether they choose surgery with higher risks of longer term urinary incontinence, or radiation therapy with higher risks of longer term bowel irritation. Therefore, if health outcomes are used in future investigations of decision aids in situations in which there is clearly no health outcome advantage, the key question to pose is: do patients experience the health outcomes they prefer and avoid the outcomes to which they are averse?

More recently, decision aids are being used in situations in which there may be a longer-term health advantage; for example, in preventive decisions about the management of type II diabetes (Mann D 2010; Mullan 2009; Weymiller 2007) and/or hypertension (Montgomery 2003), when the longer-term health outcome maybe to avoid may be stroke. Interestingly, of these studies one reported a statistically-significant difference in medication initiation when exposed to the decision aid compared to alternative interventions (Mullan 2009), thereby highlighting tensions between outcome and process measures (Bekker 2010; Thomson 2005).

Unknown effects of decision aids

The effect of patient decision aids on adherence is an area of uncertainty. The adherence results are difficult to interpret due to incomplete data, varying length of follow-up (4 to 36 months), and small sample size (n = 33 in one study). Moreover, studies such as Man-Son-Hing 1999 had very little variation in choice (over 90% of long-term aspirin users decided to stay on aspirin). When examining adherence, it would be important to do so: a) in the early phase, when presumably the issue is actually decisional in nature (e.g., filling the prescription, picking up the prescription, refilling the prescription) rather than involving the management of side effects; and b) in a manner that separates those choosing to change versus those remaining with the status quo.

Despite the positive effects of decision aids on patient-practitioner communication, some authors are concerned about the potential negative influence that decision aids may have on the relational aspects of the decision-making process; this concern highlights the need for further evaluation when decision aids are implemented as part of the routine process of care (Charles 2010; LeBlanc 2010).

Cost-effectiveness, health utilities, and preference-linked health outcomes are other secondary outcomes about which little is known and further evaluation is required. It is unlikely that we will observe the effect of decision aids on litigation rates in trials of decision aids, given the time delay to litigation and the rareness of this type of event. In fact, a mock trial that used a patient decision aid for prostate-specific antigen testing found that the majority of jurors (94%) would indicate that the standard of care had been met (Barry 2008).

Limitations

This systematic review is limited by inadequate power to detect important differences in effectiveness in subgroups and the wide variability in the decision contexts, the elements within the patient decision aid, the type of comparison interventions, the targeted outcomes, and the evaluation procedures. Several of the outcomes demonstrated statistically-significant heterogeneity. This reflects differences across clinically-diverse studies; therefore, the pooled effect size and confidence intervals should be interpreted as a range across conditions, which may not be applicable to a specific condition. A sub-analysis to explore three potential sources of heterogeneity (e.g. type of control intervention, decision aid IPDAS quality score, patients' baseline accurate risk perception) found that patients' baseline accurate risk perception was an important variable for explaining heterogeneity (Gentles 2013). When patients' baseline scores for accurate risk perception are lower, there is great improvement observed. Furthermore, we limited the study data to only two comparison groups (e.g. most intensive and least intensive).

Authors' conclusions

Implications for practice

The positive effects of decision aids on improving people's knowledge of risks and benefits, feeling informed and feeling clear about their values provides sufficient evidence for using them in clinical practice. As well they can facilitate accurate risk perception and active participation in decision making. However, several conditions may be necessary for successful implementation: a) good quality decision aids to meet the needs of the population; b) practitioners willing to use decision aids in their practice; c) effective systems for delivering decision support; and d) practitioners and healthcare consumers who are skilled in shared decision making. Although some strides have been made in achieving these conditions (O'Connor 2007), the use of patient decision aids will not occur without adequate attention to the barriers to implementation and consideration of effective interventions for implementing them as part of routine clinical practice (Gravel 2006; Legare 2008b; Legare 2010).

Implications for research

Studies are needed to deepen our understanding of: interactions between patient decision aid use and the patterns of patient-practitioner communication; the effect of decision aids on lower health literacy and low numeracy populations; various cultural groups; format issues such as web-based delivery of patient decision aids; timing issues regarding most effective use of decision aids before or during a consultation; and downstream effects on cost, resource use, and adherence.

With the addition of more studies in the systematic review, it may be possible to tease out the reasons for heterogeneity of results, including variability in: a) study quality; b) comparison intervention; c) elements within patient decision aids; d) decision type; and e) format of decision aid (e.g., video, Internet, booklet). The degree of detail in patient decision aids that is required for positive effects on IPDAS criteria should also be explored. In particular, evaluation is needed to determine the effect of those that meet the minimal IPDAS criteria for certification versus those that also meet the IPDAS quality criteria (Joseph-Williams 2013).

Acknowledgements

The Cochrane Consumers and Communication Review Group (editors, academic and consumer referees) provided peer review and advice regarding the protocol and review and checked extracted data for newly-included studies. We thank Jessie McGowan, a librarian who assisted with designing the original search strategy and John Kis-Rigo at La Trobe University who revised the search strategy used in this current update. Robert Volk and David Rovner helped with screening studies for inclusion. Joan Peterson, Sarah Mullan, Amanda Dresch, Robert Wu, and Christine Courtemanche assisted with the screening, data extraction, and 'Risk of bias' assessment. Anton Saarimaki set up and managed a web-based title and abstract screening application that facilitated independent screening of citations by the review authors and verified references for patient decision aids in Table 1. Alain Mayhew provided guidance in the interpretation of 'Risk of bias' findings and the 'Summary of findings' table. George Wells and Dean Fergusson provided consultation on the statistical analysis.

Data and analyses

Download statistical data

Comparison 1. Knowledge
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Knowledge: DA vs usual care - all studies4210842Mean Difference (IV, Random, 95% CI)13.34 [11.17, 15.51]
2 Knowledge: DA vs usual care - treatment only233977Mean Difference (IV, Random, 95% CI)13.75 [11.08, 16.43]
3 Knowledge: DA vs usual care - screening only196865Mean Difference (IV, Random, 95% CI)12.76 [9.66, 15.86]
4 Knowledge: Detailed vs simple decision aids - all studies193531Mean Difference (IV, Random, 95% CI)5.52 [3.90, 7.15]
5 Knowledge: Detailed vs simple decision aids - treatment only121930Mean Difference (IV, Random, 95% CI)4.98 [2.64, 7.33]
6 Knowledge: Detailed vs simple decision aids - screening only71601Mean Difference (IV, Random, 95% CI)6.33 [4.49, 8.17]
Analysis 1.1.

Comparison 1 Knowledge, Outcome 1 Knowledge: DA vs usual care - all studies.

Analysis 1.2.

Comparison 1 Knowledge, Outcome 2 Knowledge: DA vs usual care - treatment only.

Analysis 1.3.

Comparison 1 Knowledge, Outcome 3 Knowledge: DA vs usual care - screening only.

Analysis 1.4.

Comparison 1 Knowledge, Outcome 4 Knowledge: Detailed vs simple decision aids - all studies.

Analysis 1.5.

Comparison 1 Knowledge, Outcome 5 Knowledge: Detailed vs simple decision aids - treatment only.

Analysis 1.6.

Comparison 1 Knowledge, Outcome 6 Knowledge: Detailed vs simple decision aids - screening only.

Comparison 2. Accurate risk perceptions: decision aid with outcome probabilities vs no outcome probability information
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Accurate risk perceptions - all studies195868Risk Ratio (M-H, Random, 95% CI)1.82 [1.52, 2.16]
2 Accurate risk perceptions - treatments only122435Risk Ratio (M-H, Random, 95% CI)1.72 [1.47, 2.01]
3 Accurate risk perceptions - screening only73433Risk Ratio (M-H, Random, 95% CI)2.03 [1.40, 2.93]
4 Accurate risk perceptions - numbers154776Risk Ratio (M-H, Random, 95% CI)2.00 [1.65, 2.43]
5 Accurate risk perceptions - words41092Risk Ratio (M-H, Random, 95% CI)1.31 [1.13, 1.52]
Analysis 2.1.

Comparison 2 Accurate risk perceptions: decision aid with outcome probabilities vs no outcome probability information, Outcome 1 Accurate risk perceptions - all studies.

Analysis 2.2.

Comparison 2 Accurate risk perceptions: decision aid with outcome probabilities vs no outcome probability information, Outcome 2 Accurate risk perceptions - treatments only.

Analysis 2.3.

Comparison 2 Accurate risk perceptions: decision aid with outcome probabilities vs no outcome probability information, Outcome 3 Accurate risk perceptions - screening only.

Analysis 2.4.

Comparison 2 Accurate risk perceptions: decision aid with outcome probabilities vs no outcome probability information, Outcome 4 Accurate risk perceptions - numbers.

Analysis 2.5.

Comparison 2 Accurate risk perceptions: decision aid with outcome probabilities vs no outcome probability information, Outcome 5 Accurate risk perceptions - words.

Comparison 3. Values congruent with chosen option
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Values congruent with chosen option - all studies134670Risk Ratio (M-H, Random, 95% CI)1.51 [1.17, 1.96]
2 Values congruent with chosen option - treatment only3452Risk Ratio (M-H, Random, 95% CI)1.35 [0.80, 2.30]
3 Values congruent with chosen option - screening only104321Risk Ratio (M-H, Random, 95% CI)1.56 [1.16, 2.11]
Analysis 3.1.

Comparison 3 Values congruent with chosen option, Outcome 1 Values congruent with chosen option - all studies.

Analysis 3.2.

Comparison 3 Values congruent with chosen option, Outcome 2 Values congruent with chosen option - treatment only.

Analysis 3.3.

Comparison 3 Values congruent with chosen option, Outcome 3 Values congruent with chosen option - screening only.

Comparison 4. Decisional conflict
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Decisional conflict: DA vs usual care - all studies32 Mean Difference (IV, Random, 95% CI)Subtotals only
1.1 Uncertainty sub-scale234837Mean Difference (IV, Random, 95% CI)-2.47 [-4.28, -0.66]
1.2 Uninformed sub-scale224343Mean Difference (IV, Random, 95% CI)-7.26 [-9.73, -4.78]
1.3 Unclear values sub-scale183704Mean Difference (IV, Random, 95% CI)-6.09 [-8.50, -3.67]
1.4 Unsupported sub-scale193851Mean Difference (IV, Random, 95% CI)-4.77 [-6.86, -2.69]
1.5 Ineffective choice sub-scale193878Mean Difference (IV, Random, 95% CI)-4.86 [-7.04, -2.68]
1.6 Total decisional conflict score285830Mean Difference (IV, Random, 95% CI)-6.22 [-8.00, -4.44]
2 Decisional conflict: DA vs usual care - treatment only23 Mean Difference (IV, Random, 95% CI)Subtotals only
2.1 Uncertainty sub-scale163020Mean Difference (IV, Random, 95% CI)-3.06 [-5.33, -0.79]
2.2 Uninformed sub-scale173007Mean Difference (IV, Random, 95% CI)-8.06 [-10.52, -5.60]
2.3 Unclear values sub-scale142474Mean Difference (IV, Random, 95% CI)-6.31 [-9.01, -3.61]
2.4 Unsupported sub-scale152621Mean Difference (IV, Random, 95% CI)-5.28 [-7.74, -2.82]
2.5 Ineffective choice sub-scale152746Mean Difference (IV, Random, 95% CI)-6.07 [-8.41, -3.72]
2.6 Total decisional conflict score223783Mean Difference (IV, Random, 95% CI)-6.14 [-7.78, -4.50]
3 Decisional conflict: DA vs usual care - screening only9 Mean Difference (IV, Random, 95% CI)Subtotals only
3.1 Uncertainty sub-scale71817Mean Difference (IV, Random, 95% CI)-1.32 [-4.47, 1.83]
3.2 Uninformed sub-scale51336Mean Difference (IV, Random, 95% CI)-4.67 [-10.61, 1.27]
3.3 Unclear values sub-scale41230Mean Difference (IV, Random, 95% CI)-5.94 [-13.44, 1.56]
3.4 Unsupported sub-scale41230Mean Difference (IV, Random, 95% CI)-2.94 [-6.90, 1.02]
3.5 Ineffective choice sub-scale41132Mean Difference (IV, Random, 95% CI)-0.17 [-1.88, 1.55]
3.6 Total decisional conflict score62047Mean Difference (IV, Random, 95% CI)-6.83 [-12.64, -1.03]
4 Decisional conflict: Detailed vs simple decision aid - all studies19 Mean Difference (IV, Random, 95% CI)Subtotals only
4.1 Uncertainty sub-scale142130Mean Difference (IV, Random, 95% CI)-2.15 [-4.42, 0.12]
4.2 Uninformed sub-scale101264Mean Difference (IV, Random, 95% CI)-2.39 [-4.39, -0.39]
4.3 Unclear values sub-scale101260Mean Difference (IV, Random, 95% CI)-2.31 [-4.67, 0.05]
4.4 Unsupported sub-scale101268Mean Difference (IV, Random, 95% CI)-2.05 [-5.37, 1.27]
4.5 Ineffective choice sub-scale91541Mean Difference (IV, Random, 95% CI)-1.06 [-2.83, 0.71]
4.6 Total decisional conflict score173277Mean Difference (IV, Random, 95% CI)-1.77 [-2.64, -0.91]
5 Decisional conflict: Detailed vs simple decision aid - treatment only12 Mean Difference (IV, Random, 95% CI)Subtotals only
5.1 Uncertainty sub-scale91101Mean Difference (IV, Random, 95% CI)-2.02 [-6.65, 2.61]
5.2 Uninformed sub-scale6672Mean Difference (IV, Random, 95% CI)-1.16 [-4.40, 2.09]
5.3 Unclear values sub-scale6669Mean Difference (IV, Random, 95% CI)-0.46 [-3.72, 2.80]
5.4 Unsupported sub-scale6674Mean Difference (IV, Random, 95% CI)-0.65 [-3.09, 1.79]
5.5 Ineffective choice sub-scale7849Mean Difference (IV, Random, 95% CI)-0.27 [-2.97, 2.44]
5.6 Total decisional conflict score101732Mean Difference (IV, Random, 95% CI)-0.96 [-2.30, 0.38]
6 Decisional conflict: Detailed vs simple decision aid - screening only7 Mean Difference (IV, Random, 95% CI)Subtotals only
6.1 Uncertainty sub-scale51029Mean Difference (IV, Random, 95% CI)-2.16 [-4.20, -0.12]
6.2 Uninformed sub-scale4592Mean Difference (IV, Random, 95% CI)-3.42 [-5.81, -1.02]
6.3 Unclear values sub-scale4591Mean Difference (IV, Random, 95% CI)-4.54 [-6.77, -2.32]
6.4 Unsupported sub-scale4594Mean Difference (IV, Random, 95% CI)-3.65 [-9.74, 2.44]
6.5 Ineffective choice sub-scale2692Mean Difference (IV, Random, 95% CI)-2.18 [-3.60, -0.75]
6.6 Total decisional conflict score71545Mean Difference (IV, Random, 95% CI)-2.26 [-3.33, -1.19]
Analysis 4.1.

Comparison 4 Decisional conflict, Outcome 1 Decisional conflict: DA vs usual care - all studies.

Analysis 4.2.

Comparison 4 Decisional conflict, Outcome 2 Decisional conflict: DA vs usual care - treatment only.

Analysis 4.3.

Comparison 4 Decisional conflict, Outcome 3 Decisional conflict: DA vs usual care - screening only.

Analysis 4.4.

Comparison 4 Decisional conflict, Outcome 4 Decisional conflict: Detailed vs simple decision aid - all studies.

Analysis 4.5.

Comparison 4 Decisional conflict, Outcome 5 Decisional conflict: Detailed vs simple decision aid - treatment only.

Analysis 4.6.

Comparison 4 Decisional conflict, Outcome 6 Decisional conflict: Detailed vs simple decision aid - screening only.

Comparison 5. Participation in decision making
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Participation in decision making: DA vs usual care - all studies14 Risk Ratio (M-H, Random, 95% CI)Subtotals only
1.1 Patient controlled decision making122438Risk Ratio (M-H, Random, 95% CI)1.28 [1.02, 1.60]
1.2 Shared decision making122402Risk Ratio (M-H, Random, 95% CI)0.96 [0.82, 1.13]
1.3 Practitioner controlled decision making143234Risk Ratio (M-H, Random, 95% CI)0.66 [0.53, 0.81]
2 Participation in decision making: DA vs usual care - treatment only11 Risk Ratio (M-H, Random, 95% CI)Subtotals only
2.1 Patient controlled decision making102147Risk Ratio (M-H, Random, 95% CI)1.33 [1.05, 1.68]
2.2 Shared decision making102111Risk Ratio (M-H, Random, 95% CI)0.95 [0.78, 1.15]
2.3 Practitioner controlled decision making112318Risk Ratio (M-H, Random, 95% CI)0.70 [0.54, 0.90]
3 Participation in decision making: DA vs usual care - screening only4 Risk Ratio (M-H, Random, 95% CI)Subtotals only
3.1 Patient controlled decision making3887Risk Ratio (M-H, Random, 95% CI)0.99 [0.82, 1.20]
3.2 Shared decision making3887Risk Ratio (M-H, Random, 95% CI)1.13 [0.89, 1.45]
3.3 Practitioner controlled decision making41512Risk Ratio (M-H, Random, 95% CI)0.67 [0.44, 1.01]
4 Participation in decision making: Detailed vs simple decision aid - all studies2 Risk Ratio (M-H, Random, 95% CI)Subtotals only
4.1 Patient controlled decision making2687Risk Ratio (M-H, Random, 95% CI)1.06 [0.68, 1.64]
4.2 Shared decision making2687Risk Ratio (M-H, Random, 95% CI)1.07 [0.63, 1.81]
4.3 Practitioner controlled decision making2687Risk Ratio (M-H, Random, 95% CI)1.12 [0.56, 2.23]
5 Participation in decision making: Detailed vs simple decision aid - treatment only2 Risk Ratio (M-H, Random, 95% CI)Subtotals only
5.1 Patient controlled decision making2687Risk Ratio (M-H, Random, 95% CI)1.06 [0.68, 1.64]
5.2 Shared decision making2687Risk Ratio (M-H, Random, 95% CI)1.07 [0.63, 1.81]
5.3 Practitioner controlled decision making2687Risk Ratio (M-H, Random, 95% CI)1.12 [0.56, 2.23]
Analysis 5.1.

Comparison 5 Participation in decision making, Outcome 1 Participation in decision making: DA vs usual care - all studies.

Analysis 5.2.

Comparison 5 Participation in decision making, Outcome 2 Participation in decision making: DA vs usual care - treatment only.

Analysis 5.3.

Comparison 5 Participation in decision making, Outcome 3 Participation in decision making: DA vs usual care - screening only.

Analysis 5.4.

Comparison 5 Participation in decision making, Outcome 4 Participation in decision making: Detailed vs simple decision aid - all studies.

Analysis 5.5.

Comparison 5 Participation in decision making, Outcome 5 Participation in decision making: Detailed vs simple decision aid - treatment only.

Comparison 6. Proportion undecided
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Proportion undecided: DA vs usual care - all studies184753Risk Ratio (M-H, Random, 95% CI)0.59 [0.47, 0.72]
2 Proportion undecided: DA vs usual care - treatment only142830Risk Ratio (M-H, Random, 95% CI)0.63 [0.51, 0.78]
3 Proportion undecided: DA vs usual care - screening only41923Risk Ratio (M-H, Random, 95% CI)0.52 [0.30, 0.90]
4 Proportion undecided: Detailed vs simple decision aids - all studies3352Risk Ratio (M-H, Random, 95% CI)0.98 [0.69, 1.37]
5 Proportion undecided: Detailed vs simple decision aids - treatment only2151Risk Ratio (M-H, Random, 95% CI)1.02 [0.71, 1.47]
6 Proportion undecided: Detailed vs simple decision aids - screening only1201Risk Ratio (M-H, Random, 95% CI)0.63 [0.21, 1.86]
Analysis 6.1.

Comparison 6 Proportion undecided, Outcome 1 Proportion undecided: DA vs usual care - all studies.

Analysis 6.2.

Comparison 6 Proportion undecided, Outcome 2 Proportion undecided: DA vs usual care - treatment only.

Analysis 6.3.

Comparison 6 Proportion undecided, Outcome 3 Proportion undecided: DA vs usual care - screening only.

Analysis 6.4.

Comparison 6 Proportion undecided, Outcome 4 Proportion undecided: Detailed vs simple decision aids - all studies.

Analysis 6.5.

Comparison 6 Proportion undecided, Outcome 5 Proportion undecided: Detailed vs simple decision aids - treatment only.

Analysis 6.6.

Comparison 6 Proportion undecided, Outcome 6 Proportion undecided: Detailed vs simple decision aids - screening only.

Comparison 7. Satisfaction
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Satisfaction with the choice: DA vs usual care - all studies10 Mean Difference (IV, Random, 95% CI)Totals not selected
2 Satisfaction with the choice: DA vs usual care - treatment only9 Mean Difference (IV, Random, 95% CI)Totals not selected
3 Satisfaction with the choice: DA vs usual care - screening only1 Mean Difference (IV, Random, 95% CI)Totals not selected
4 Satisfaction with the choice: Detailed vs simple DA - all studies3 Mean Difference (IV, Random, 95% CI)Totals not selected
5 Satisfaction with the choice: Detailed vs simple DA - treatment only3 Mean Difference (IV, Random, 95% CI)Totals not selected
6 Satisfaction with the decision making process: DA vs usual care - all studies6 Mean Difference (IV, Random, 95% CI)Totals not selected
7 Satisfaction with the decision making process: DA vs usual care - treatment only5 Mean Difference (IV, Random, 95% CI)Totals not selected
8 Satisfaction with the decision making process: DA vs usual care - screening only1 Mean Difference (IV, Random, 95% CI)Totals not selected
Analysis 7.1.

Comparison 7 Satisfaction, Outcome 1 Satisfaction with the choice: DA vs usual care - all studies.

Analysis 7.2.

Comparison 7 Satisfaction, Outcome 2 Satisfaction with the choice: DA vs usual care - treatment only.

Analysis 7.3.

Comparison 7 Satisfaction, Outcome 3 Satisfaction with the choice: DA vs usual care - screening only.

Analysis 7.4.

Comparison 7 Satisfaction, Outcome 4 Satisfaction with the choice: Detailed vs simple DA - all studies.

Analysis 7.5.

Comparison 7 Satisfaction, Outcome 5 Satisfaction with the choice: Detailed vs simple DA - treatment only.

Analysis 7.6.

Comparison 7 Satisfaction, Outcome 6 Satisfaction with the decision making process: DA vs usual care - all studies.

Analysis 7.7.

Comparison 7 Satisfaction, Outcome 7 Satisfaction with the decision making process: DA vs usual care - treatment only.

Analysis 7.8.

Comparison 7 Satisfaction, Outcome 8 Satisfaction with the decision making process: DA vs usual care - screening only.

Comparison 8. Choice
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Choice: Surgery over conservative option: DA vs usual care15 Risk Ratio (M-H, Random, 95% CI)Subtotals only
1.1 As treated analysis152915Risk Ratio (M-H, Random, 95% CI)0.80 [0.67, 0.95]
1.2 Intention to treat analysis153553Risk Ratio (M-H, Random, 95% CI)0.79 [0.68, 0.93]
2 Choice: Surgery over conservative option: Detailed vs simple decision aid3 Risk Ratio (M-H, Random, 95% CI)Subtotals only
2.1 As treated analysis3513Risk Ratio (M-H, Random, 95% CI)0.82 [0.63, 1.08]
2.2 Intention to treat analysis3584Risk Ratio (M-H, Random, 95% CI)0.82 [0.63, 1.08]
3 Choice for screening28 Risk Ratio (M-H, Random, 95% CI)Subtotals only
3.1 PSA screening: DA vs usual care93565Risk Ratio (M-H, Random, 95% CI)0.87 [0.77, 0.98]
3.2 PSA screening: detailed DA vs simple decision aid3782Risk Ratio (M-H, Random, 95% CI)0.98 [0.82, 1.17]
3.3 Colorectal cancer screening: DA vs usual care104529Risk Ratio (M-H, Random, 95% CI)1.12 [0.95, 1.31]
3.4 Breast cancer genetic testing: DA vs usual care4949Risk Ratio (M-H, Random, 95% CI)1.01 [0.83, 1.22]
3.5 Prenatal diagnostic testing: Detailed vs simple decision aid2443Risk Ratio (M-H, Random, 95% CI)0.96 [0.90, 1.03]
4 Choice: Diabetes medication (uptake new medication): DA vs usual care3277Risk Ratio (M-H, Random, 95% CI)1.84 [0.77, 4.39]
5 Choice: Menopausal hormone therapy: Detailed vs simple decision aid3357Risk Ratio (M-H, Random, 95% CI)0.73 [0.55, 0.98]
Analysis 8.1.

Comparison 8 Choice, Outcome 1 Choice: Surgery over conservative option: DA vs usual care.

Analysis 8.2.

Comparison 8 Choice, Outcome 2 Choice: Surgery over conservative option: Detailed vs simple decision aid.

Analysis 8.3.

Comparison 8 Choice, Outcome 3 Choice for screening.

Analysis 8.4.

Comparison 8 Choice, Outcome 4 Choice: Diabetes medication (uptake new medication): DA vs usual care.

Analysis 8.5.

Comparison 8 Choice, Outcome 5 Choice: Menopausal hormone therapy: Detailed vs simple decision aid.

Appendices

Appendix 1. Revised Search Strategies January 2009 to June 2012

MEDLINE (Ovid)

1. decision support techniques/

2. decision support systems clinical/

3. decision trees/

4. (decision making or choice behavior).mp. and informed consent.sh.

5. ((decision* or decid*) adj4 (support* or aid* or tool* or instrument* or technolog* or technique* or system* or program* or algorithm* or process* or method* or intervention* or material*)).tw.

6. (decision adj (board* or guide* or counseling)).tw.

7. ((risk communication or risk assessment or risk information) adj4 (tool* or method*)).tw.

8. decision-making computer assisted/

9. (computer* adj2 decision making).tw.

10. interactive health communication*.tw.

11. (interactive adj (internet or online or graphic* or booklet*)).tw.

12. (interacti* adj4 tool*).tw.

13. ((interactiv* or evidence based) adj3 (risk information or risk communication or risk presentation or risk graphic*)).tw.

14. shared decision making.tw.

15. (informed adj (choice* or decision*)).tw.

16. adaptive conjoint analys#s.tw.

17. or/1-16

18. randomized controlled trial.pt.

19. controlled clinical trial.pt.

20. randomized.ab.

21. placebo.ab.

22. clinical trials as topic.sh.

23. randomly.ab.

24. trial.ti.

25. or/18-24

26. exp animals/ not humans.sh.

27. 25 not 26

28. 17 and 27

29. limit 28 to yr="2009 -Current"

CENTRAL (The Cochrane Library)

  1. (decision-support or decision-aid):kw in Trials

  2. decision-tree:kw in Trials

  3. patient-decision-making:kw

  4. (decision-making or choice-behavior):ti,ab,kw and (informed-consent:kw,ti or (patient or parent* or carer or caregiver or care-giver):ti,ab,kw) in Trials

  5. ((decision or decid*) near/4 (support* or aid* or tool or instrument or technolog* or technique or system or program* or algorithm or process or method or intervention or material)):ti,ab,kw

  6. (decision next (board or guide or counseling)):ti,ab,kw

  7. ((risk-communication or risk-assessment or risk-information) near/4 (tool or method)):ti,ab,kw

  8. (computer* near/2 decision-making):ti,ab,kw

  9. (interactive-health-communication or (interacti* near/4 tool)):ti,ab,kw

  10. (interactive next (internet or online or graphic* or booklet)):ti,ab,kw

  11. ((interactiv* or evidence-based) near/3 (risk-information or risk-communication or risk-presentation or risk-graphic*)):ti,ab,kw

  12. shared-decision-making:ti,ab,kw

  13. (informed next (choice or decision)):ti,ab,kw

  14. adaptive-conjoint-analysis:ti,ab,kw

  15. (#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12 OR #13 OR #14), from 2009 to 2012

(Last line restricted to “Trials”, and to date range 2009 to 2012)

EMBASE (Ovid)

1. decision support system/

2. patient decision making/

3. decision aid/

4. "decision tree"/

5. decision making.hw,kw,tw. and informed consent.hw,kw.

6. ((decision* or decid*) adj4 (support* or aid* or tool* or instrument* or technolog* or technique* or system* or program* or algorithm* or process* or method* or intervention* or material*)).tw,kw.

7. (decision adj (board* or guide* or counseling)).tw,kw.

8. ((risk communication or risk assessment or risk information) adj4 (tool* or method*)).tw,kw.

9. (computer* adj2 decision making).tw,kw.

10. interactive health communication*.tw,kw.

11. (interactive adj (internet or online or graphic* or booklet*)).tw,kw.

12. (interacti* adj4 tool*).tw,kw.

13. ((interactiv* or evidence based) adj3 (risk information or risk communication or risk presentation or risk graphic*)).tw,kw.

14. shared decision making.tw,kw.

15. (informed adj (choice* or decision*)).tw,kw.

16. adaptive conjoint analys#s.tw,kw.

17. or/1-16

18. randomized controlled trial/

19. controlled clinical trial/

20. single blind procedure/ or double blind procedure/

21. crossover procedure/

22. random*.tw.

23. placebo*.tw.

24. ((singl* or doubl*) adj (blind* or mask*)).tw.

25. (crossover or cross over or factorial* or latin square).tw.

26. (assign* or allocat* or volunteer*).tw.

27. or/18-26

28. nonhuman/ not (human/ and nonhuman/)

29. 27 not 28

30. 17 and 29

31. 30 and 2009:2012.(sa_year).

32. limit 31 to exclude medline journals

  

PsycINFO (Ovid)

1. decision support systems/

2. (decision making or choice behavior).mp. and (informed consent.sh. or (patient* or parent* or carer* or caregiver* or care giver*).mp.)

3. ((decision* or decid*) adj4 (support* or aid* or tool* or instrument* or technolog* or technique* or system* or program* or algorithm* or process* or method* or intervention* or material*)).ti,ab,id.

4. (decision adj (board* or guide* or counseling)).ti,ab,id.

5. ((risk communication or risk assessment or risk information) adj4 (tool* or method*)).ti,ab,id.

6. computer assisted therapy/

7. (computer* adj2 decision making).ti,ab,id.

8. interactive health communication*.ti,ab,id.

9. (interactive adj (internet or online or graphic* or booklet*)).ti,ab,id.

10. (interacti* adj4 tool*).ti,ab,id.

11. ((interactiv* or evidence based) adj3 (risk information or risk communication or risk presentation or risk graphic*)).ti,ab,id.

12. shared decision making.ti,ab,id.

13. (informed adj (choice* or decision*)).ti,ab,id.

14. adaptive conjoint analys#s.ti,ab,id.

15. or/1-14

16. random*.ti,ab,hw,id.

17. intervention.ti,ab,hw,id.

18. trial.ti,ab,hw,id.

19. placebo*.ti,ab,hw,id.

20. ((singl* or doubl* or trebl* or tripl*) and (blind* or mask*)).ti,ab,hw,id.

21. (cross over or crossover).ti,ab,hw,id.

22. latin square.ti,ab,hw,id.

23. (assign* or allocat* or volunteer*).ti,ab,hw,id.

24. treatment effectiveness evaluation/

25. mental health program evaluation/

26. exp experimental design/

27. or/16-26

28. 15 and 27

29. limit 28 to yr="2009 -Current"

CINAHL (EBSCO)

#QueryLimiters/Expanders
S31S30 Limiters - Exclude MEDLINE records
Search modes - Boolean/Phrase
S30S28 and S29Search modes - Boolean/Phrase
S29EM 2009-Search modes - Boolean/Phrase
S28S17 and S27Search modes - Boolean/Phrase
S27S18 or S19 or S20 or S21 or S22 or S23 or S24 or S25 or S26Search modes - Boolean/Phrase
S26TI (singl* or doubl* or tripl* or trebl*) and TI (blind* or mask*)Search modes - Boolean/Phrase
S25AB (singl* or doubl* or tripl* or trebl*) and AB (blind* or mask*)Search modes - Boolean/Phrase
S24AB (random* or trial or placebo*) or TI (random* or trial or placebo*)Search modes - Boolean/Phrase
S23MH Quantitative StudiesSearch modes - Boolean/Phrase
S22MH PlacebosSearch modes - Boolean/Phrase
S21MH Random AssignmentSearch modes - Boolean/Phrase
S20MH Clinical Trials+Search modes - Boolean/Phrase
S19PT Clinical TrialSearch modes - Boolean/Phrase
S18PT "randomi?ed controlled trial"Search modes - Boolean/Phrase
S17S1 or S2 or S3 or S4 or S5 or S6 or S7 or S8 or S9 or S10 or S11 or S12 or S13 or S14 or S15 or S16Search modes - Boolean/Phrase
S16"informed choice*" or "informed decision*"Search modes - Boolean/Phrase
S15"shared decision making"Search modes - Boolean/Phrase
S14"adaptive conjoint analys?s"Search modes - Boolean/Phrase
S13(interactive N2 "risk information") or (interactive N2 "risk communication") or (interactive N2 "risk presentation") or (interactive N2 "risk graphic*")Search modes - Boolean/Phrase
S12"interactive internet" or "interactive online" or "interactive graphic*" or "interactive booklet*" or (interacti* N3 tool*)Search modes - Boolean/Phrase
S11"interactive health communication*"Search modes - Boolean/Phrase
S10computer* N1 "decision making"Search modes - Boolean/Phrase
S9("risk communication" N3 tool*) or ("risk communication" N3 method*) or ("risk information" N3 tool*) or ("risk information" N3 method*) or ("risk assessment" N3 tool*) or ("risk assessment" N3 method*)Search modes - Boolean/Phrase
S8"evidence based risk communication" or "evidence based risk information"Search modes - Boolean/Phrase
S7"decision board*" or "decision guide*" or "decision counseling"Search modes - Boolean/Phrase
S6(decision* N3 support*) or (decision* N3 aid*) or (decision* N3 tool*) or (decision* N3 instrument*) or (decision* N3 technolog*) or (decision* N3 technique*) or (decision* N3 system*) or (decision* N3 program*) or (decision* N3 algorithm*) or (decision* N3 process*) or (decision* N3 method*) or (decision* N3 intervention*) or (decision* N3 material*)Search modes - Boolean/Phrase
S5("decision making" or "choice behavior") and MH consentSearch modes - Boolean/Phrase
S4MH decision making, computer assistedSearch modes - Boolean/Phrase
S3MH decision making, patientSearch modes - Boolean/Phrase
S2MH decision support systems, clinicalSearch modes - Boolean/Phrase
S1MH decision support techniques+Search modes - Boolean/Phrase

 

 

Appendix 2. Search strategies to 2009

MEDLINE,1966 to December 2009 (Ovid)

1. choice behavior/

2. decision making/

3. exp decision support techniques/

4. Educational Technology/

5. decision$.tw.

6. (choic$ or preference$).tw.

7. communication package.tw.

8. or/1-7

9. exp health education/

10. Health Knowledge, Attitudes, Practice/

11. informed consent.tw,hw.

12. patient.tw,hw.

13. consumer.tw,hw.

14. or/9-13

15. 8 and 14

16. ((patient$ or consumer$) adj1 (decision$ or choice or preference or participation)).tw.

17. ((women or men) adj1 (decision$ or choice or preference or participation)).tw.

18. (parent$ adj1 (decision$ or choice or preferenc$ or participat$)).tw.

19. ((personal or interpersonal or individual) adj (decision$ or choice or preference$ or participat$)).tw.

20. shared decision making.tw.

21. decision aid$.tw.

22. informed choice.tw.

23. or/16-22

24. 15 or 23

25. clinical trial.pt.

26. randomized controlled trial.pt.

27. random$.tw.

28. (double adj blind$).tw.

29. double-blind method/

30. or/25-29

31. 24 and 30

CENTRAL

CENTRAL, The Cochrane Library was searched using the MEDLINE search above in Ovid to the end of 2006; for the 2011 update, the CENTRAL search was conducted at www.thecochranelibrary.com to the end of 2009 using the following search strategy:

1. decision.tw,hw.

2. patient.tw,hw.

3. consumer.tw,sh.

4. 1 and (2 or 3)

5. shared decision making.tw.

6. decision aid$.tw.

7. informed choice.tw.

8. or/4-7

9. clinical trial.pt.

10. randomized controlled trial.pt.

11. random$.tw.

12. or/9-11

13. 8 and 12

CINAHL, 1982 to September 2008 (Ovid)

1. exp Decision Making/

2. information seeking behavior/

3. Help Seeking Behavior/

4. (choic$ or preference$).tw.

5. decision$.tw.

6. Educational Technology/

7. or/1-6

8. exp Health Behavior/

9. consumer participation/

10. exp Health Education/

11. health knowledge/ or exp professional knowledge/

12. exp Consent/

13. informed consent.tw.

14. patient.tw,hw.

15. consumer.tw,sh.

16. or/8-15

17. 7 and 16

18. ((patient$ or consumer$) adj1 (decision$ or choice or preference or participation)).tw.

19. ((women or men) adj1 (decision$ or choice or preference or participation)).tw.

20. (parent$ adj1 (decision$ or choice or preferenc$ or participat$)).tw.

21. ((personal or interpersonal or individual) adj (decision$ or choice or preference$ or participat$)).tw.

22. shared decision making.tw.

23. decision aid$.tw.

24. informed choice.tw.

25. or/18-24

26. 17 or 25

27. exp clinical trials/

28. Clinical trial.pt.

29. (clinic$ adj trial$1).tw.

30. random$.tw.

31. Random assignment/

32. placebo$.tw,sh.

33. Quantitative studies/

34. Allocat$ random$.tw.

35. ((singl$ or doubl$ or trebl$ or tripl$) adj (blind$3 or mask$3)).tw.

36. or/27-35

37. 26 and 36

EMBASE,1980 to December 2009 (Ovid)

1. decision making/

2. decision theory/

3. decision$.tw.

4. Educational Technology/

5. or/1-4

6. exp health behavior/

7. exp Patient Attitude/

8. exp health education/

9. informed consent.tw,sh.

10. patient.tw,sh.

11. consumer.tw,sh.

12. or/6-11

13. 5 and 12

14. ((patient$ or consumer$) adj1 (decision$ or choice or preference or participation)).tw.

15. ((women or men) adj1 (decision$ or choice or preference or participation)).tw.

16. (parent$ adj1 (decision$ or choice or preferenc$ or participat$)).tw.

17. ((personal or interpersonal or individual) adj (decision$ or choice or preference$ or participat$)).tw.

18. shared decision making.tw.

19. decision aid$.tw.

20. informed choice.tw.

21. or/14-20

22. 13 or 21

23. Controlled Study/

24. Randomized Controlled Trial/

25. Clinical Study/

26. Clinical Trial/

27. Major Clinical Study/

28. Prospective Study/

29. Multicenter Study/

30. Randomization/

31. Double Blind Procedure/

32. Single Blind Procedure/

33. Crossover Procedure/

34. Placebo.tw,sh.

35. random$.tw.

36. (double adj blind$).tw.

37. or/23-36

38. 22 and 37

PsycINFO, 1806 to December 2009 (Ovid)

1. decision$.tw.

2. (choic$ or preference$).tw.

3. exp decision making/

4. computer assisted instruction/

5. or/1-4

6. exp health education/

7. exp health personnel attitudes/

8. informed consent.tw,sh.

9. patient.tw,hw.

10. consumer.tw,hw.

11. exp health behavior/

12. or/6-11

13. 5 and 12

14. ((patient$ or consumer$) adj1 (decision$ or choice or preference or participation)).tw.

15. ((women or men) adj1 (decision$ or choice or preference or participation)).tw.

16. (parent$ adj1 (decision$ or choice or preferenc$ or participat$)).tw.

17. ((personal or interpersonal or individual) adj (decision$ or choice or preference$ or participat$)).tw.

18. shared decision making.tw.

19. decision aid$.tw.

20. informed choice.tw.

21. or/14-20

22. 13 or 21

23. random$.tw.

24. (double adj blind$).tw.

25. placebo$.tw,hw.

26. or/23-25

27. 22 and 26

What's new

DateEventDescription
29 January 2014AmendedMinor typographical corrections.

History

Protocol first published: Issue 1, 1999
Review first published: Issue 3, 2001

DateEventDescription
5 December 2013New citation required and conclusions have changed

This update added 33 new studies for a total of 115 studies involving 34,444 participants. GRADE was used to summarize the quality of the evidence, and findings were reported using a 'Summary of findings' table. We excluded three previously-included trials on the basis of their quasi-randomized controlled trial (q-RCT) design identified using the more rigorous 'Risk of bias' assessment tool, as well as one other study that used the same decision aid content for both groups but varied the format used.

Overall, the results are similar to the previous update, but this update indicates the quality of the evidence to support the reported outcomes (high-quality evidence that decision aids compared to usual care improve people’s knowledge and reduce their decisional conflict related to feeling uninformed and unclear about their personal values; moderate-quality evidence that decision aids compared to usual care stimulate people to take a more active role in decision making and improve accurate risk perceptions when probabilities are included; and low-quality evidence that decision aids improve the congruence between the chosen option and their values).

We added two new authors to the review, LT in Sydney and JW in Ottawa who helped coordinate this update.

30 June 2012New search has been performedSearch strategies were updated and new searches run in June 2012.
18 January 2012AmendedMinor change to wording, Plain Language Summary.
5 September 2011New search has been performedAn update of this review was conducted in 2010 and published on issue 10 2011 of The Cochrane Library. Citations were searched from 2006 to December 2009.
5 September 2011New citation required but conclusions have not changed

This update added 31 new studies, and all 86 included studies were assessed for risk of bias. Overall the results were consistent with the previous update.

New in this update is the meta-analysis of informed values-based choices for decision aids including explicit values-clarification compared to those with no explicit values-clarification. We have also conducted a post-hoc analysis to evaluate the effect of risk of bias assessment ratings on outcomes.

29 April 2009New search has been performedSee the 'History' items dated 29 April 2009 and 28 July 2006.
29 April 2009New citation required and conclusions have changed

A substantially updated version of this review was published on issue 1 2009 of The Cochrane Library. The changes are outlined in the 'History' (date 28 July 2006). The updated review ought to have had a new citation to reflect the new authorship and substantial changes to the review and its conclusions; however because of a technical error this new citation was not given to the updated review.

The new citation for this review for issue 3 2009 (O'Connor 2009) reflects the updated review contents as actually published from issue 1 2009 onwards.

28 April 2009AmendedCorrected mislabelled table 'Summary of pooled outcomes'.
17 July 2008AmendedConverted to new review format.
28 July 2006New search has been performed

Changes for the 2006 update (first published on issue 1 2009 of The Cochrane Library):

  • Outcomes focus on the new effectiveness criteria of the International Patient Decision Aids Standards (IPDAS) Collaboration.

  • There are now 55 randomized controlled trials evaluating decision aids in the review. Twenty-five new randomized controlled trials have been added for this update. Four trials that were previously included were excluded from this review as the decision support intervention was not available to determine whether it met the inclusion criteria - a requirement for this update in light of the new IPDAS standards. There are an additional 15 trials in progress.

  • The number of included countries has doubled from the last update. We now have results from 7 countries (AU, CA, China, Finland, Netherlands, US, UK).

Findings from the 2006 update (*new to this update):

  • * Thirty-eight trials used at least one measure that mapped onto an IPDAS effectiveness criterion. No trials evaluated the extent to which patient decision aids achieve the IPDAS decision process criteria: helped patients to recognize that a decision needs to be made, understand that values affect the decision, or discuss values with their practitioner.

  • * Exposure to a decision aid with probabilities resulted in a higher proportion of people with accurate risk perceptions; the effect was stronger when probabilities were measure quantitatively rather than qualitatively.

  • Compared to usual care, exposure to decision aids improved knowledge, decreased decisional conflict, reduced the proportion of people who were passive in decision making, reduced the proportion who remained undecided, and reduced rates of elective invasive surgery.

  • Detailed decision aids (compared to simpler decision aids) improved knowledge and reduced the uptake of hormone replacement therapy.

  • * Compared to usual care, exposure to decision aids reduced prostate-specific antigen (PSA) screening.

  • There are too few studies to comment on the effects of decision aids on length of the consult, patient-practitioner communication, persistence with chosen option, costs, and resource use.

21 February 2003New search has been performed

For the 2002 update (O'Connor 2003b), the following changes were made:

  • There are now 221 decision aids (increased from 87) that have been identified for the inventory with 131 available and up-to-date: many of which are available on the Internet. However few have undergone any form of evaluation for impact on decision making.

  • There are now 35 randomized controlled trials evaluating decision aids in the review. Eleven new randomized controlled trials have been added for this update including 1 large scale trial that evaluated a suite of 8 decision aids in a number of health services.

  • There are an additional 6 trials pending publication and 24 trials in progress.

  • In conjunction with the benefits reported in the earlier reports, there is now evidence that decision aids compared to usual care also help with making actual choices and there is a statistically-significant reduction in major elective surgery by a quarter. Detailed compared to simple decision aids also show an improved agreement between values and actual choice.

  • There continues to be too few studies to comment on the effects of decision aids on persistence with chosen therapy, costs, resource use, or efficacy of dissemination. 

Contributions of authors

1999 Review (O'Connor 1999b):
AO, AR, VF, JT, VE, HLT, MHR, VF, MB, JJ contributed to the design of the protocol, the interpretation of results, and the revision and final approval of the final paper.
AO led the team, JT coordinated the project.
AO, MH-R, AR, VF, and JT pilot tested the data extraction forms.
AR, VF, JT screened studies and extracted data.
AR, JT, and AO analyzed the results.

2001 Review (O'Connor 2001b):
AO, DS, DR, MHR, HLT, VE, MB, JT, VF, AR contributed to the interpretation of results, and the revision and final approval of the final paper.
AO lead the team and DS coordinated the update.
AO, DR, MHR, HLT, JT, DS, JP screened studies and extracted data.
DS, JP evaluated decision aids using the CREDIBLE criteria.
AO and DS analyzed the results.

2002 Review (O'Connor 2003b):
AO, DS, DR, MHR, HLT, VE, MB, JT, VF contributed to the interpretation of results, and the revision and final approval of the paper.
AO lead the team and DS coordinated the update.
DS, JP, VT, JT screened studies and extracted data.
DS, JP, VT, SK evaluated decision aids using the CREDIBLE criteria.
AO and DS analyzed the results.

2006 Review (O'Connor 2009):
AO, CB, DS, MB, NC, KE, VE, VF, MHR, SK, HLT, DR, contributed to the interpretation of results, and the revision and final approval of the paper.
AO led the team and CB coordinated the update.
CB, SK, DS, AO, VF screened studies and extracted data.
AO and CB analyzed the results.

2009 Review (Stacey 2011):

DS, CB, MB, NC, KE, FL, AL, MHR, HLT, RT contributed to the interpretation of results, and the revision and final approval of the paper.
DS led the team and CB coordinated the update.
CB, DS screened studies; SM, AD extracted data; CB entered the data; DS verified the data entered.
DS and CB analyzed the results.

2013 (current) Review

DS, CB, MB, NC, KE, FL, AL, MHR, HLT, RT, and LT contributed to the interpretation of results, and the revision and final approval of the paper.
DS led the team with help coordinating the update from SB and JW.
CB, DS, RT, MB, MHR, NC, KE, BV, DR, AS screened studies; SB, RW, JW, and CC extracted data; SB and JW entered the data; DS verified the data entered.
DS and JW analyzed the results.

Declarations of interest

Several of the investigators have developed patient decision aids (DS, FL, HL, MHR, MB, NC, KE, RT, LT), but none reviewed their own studies.

Within the last five years, two investigators (HL, MB) have received financial support from the not-for-profit Informed Medical Decisions Foundation (IMDF). MB serves on the Board of and receives salary support as President of the Foundation. Several investigators (DS, FL, HL, MHR, MB, KE, RT, LT), who were involved in a special issue in BMC Medical Informatics and Decision Making that included a series of 14 papers focused on the theoretical and empirical evidence underlying the International Patient Decision Aid Standards (IPDAS), received partial funding from the Foundation to cover publishing costs. The Foundation has a licensing agreement with Health Dialog (a commercial firm) that distributes and promotes patient decision aids.

NC is the founder of Shared Decision Making Resources, an organization devoted to the development and dissemination of interactive patient decision aids; served as an adviser to Emmi Solutions LLC, Janssen Scientific Affairs, LLC, Expert Medical Navigation Inc, University of Chicago, Miami University and BlueCross/Blue Shield; and has received travel and/or speakers fees/honoraria from various organizations that have sponsored conferences addressing Shared Decision Making (including the World Congress Leadership Summit on Shared Decision Making).

Sources of support

Internal sources

  • University of Ottawa, Canada.

    University Research Chair in Knowledge Translation to Patients

  • Ottawa Hospital Research Institute, Canada.

    Director, Patient Decision Aids Research Group

External sources

  • No sources of support supplied

Differences between protocol and review

There are three main differences between the original protocol and the review. The 2009 update (O'Connor 2009) was re-structured to organize the long list of outcomes into primary and secondary outcomes based on the new effectiveness criteria of the International Patient Decision Aid (IPDAS) Collaboration (Elwyn 2006). For the 2011 update (Stacey 2011), study quality assessment was changed to the 'Risk of bias' assessment (Higgins 2011). For the 2013 (current) update, GRADE was used to summarize the quality of the evidence and findings were reported using a 'Summary of findings' table.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Allen 2010

MethodsRandomised to decision aid vs usual care
Participants398 + 414 men considering prostate cancer screening in the USA
Interventions

DA: computer tailored program on clinical problem, outcome probabilities, explicit values clarification, others' opinion and guidance (step by step process for making the decision; interactive computer program: inherently guided the patient through the decision aid and decision making process), tailored print out given to patients to promote discussion with others (practitioner, significant others)

COMPARE: received no intervention

Outcomesdecisional status* (pre, post DA), knowledge* (pre, post DA), *decision self-efficacy(pre, post DA), decisional consistency* (pre, post DA), desire for involvement in decision making (pre, post DA), decisional conflict(pre, post DA), preferred options.
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg.2173 (Setting): "Sites were blocked on size and percent of male employees and randomly assigned by computer-generated random numbers to condition within blocks."
Allocation concealment (selection bias)Unclear riskThe study does not address this criterion.
Incomplete outcome data (attrition bias)
All outcomes
Low riskNo missing outcome data.
Selective reporting (reporting bias)Unclear riskNo mention of protocol.
Other biasLow riskpg.2175 (Intervention delivery): mention of money incentive to complete paperwork, but was judged to have no effect on outcomes measured
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskThe study does not address this criterion.
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes measured were not subjective to interpretation

Arterburn 2011

MethodsRandomised to decision aid vs usual care
Participants75 + 77 participants considering bariatric surgery in the USA
Interventions

DA: booklet + video on options' outcomes, clinical problem, outcome probabilities, others' opinion, guidance (list of questions to discuss with clinician)

COMPARE: usual care (general information pamphlets on clinical problem)

Outcomesknowledge* (pre, immediately post and 3 month follow-up), values (pre, immediately post and 3 month follow-up), values concordance* (pre, immediately post and 3 month follow-up), treatment preference (pre and immediately post), decisional conflict (pre and immediately post), decisional self efficacy (pre and immediately post), proportion undecided.
Notes*primary outcomes
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg.1670 (Participants and randomization): "used computer-assisted, block randomisation process to ensure balanced allocation of participants"
Allocation concealment (selection bias)Unclear riskNo mention of allocation concealment and no mention of impact on study
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskpg.1671 (measures): mentioned 4 choices for treatment preference (surgery, drug therapy, diet and/or exercise program and unsure) but only reported on surgery and unsure options
Selective reporting (reporting bias)Unclear riskNo mention of study protocol or trial registration; all pre-specified outcomes included
Other biasLow riskThe study appears to be free of other sources of bias
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskpg.1670 (Participants and randomization): "study was not blinded"; no mention of impact on study
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured and not subject to interpretation

Auvinen 2004

MethodsRandomised to decision aid vs usual care
Participants103 + 100 men newly diagnosed with prostate cancer in Finland
InterventionsDA: pamphlet patient decision aid created for study on options' outcomes, outcome probability, guidance
COMPARE: usual care by clinical guideline
OutcomesUptake of options*, participation in decision making
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low risk

Auvinen, 2001, BJU International: pg. 2 "randomized centrally, using software based on a random number generator" No blocking used

Auvinen, 2004, BJU International (Primary Study): pg. 1 "randomized using a computer algorithm based on random numbers"

Allocation concealment (selection bias)Unclear risk

Auvinen, 2001, BJU International: pg. 2 - Patients and Methods randomized centrally at the Finnish Cancer Registry

Auvinen, 2004, BJU International (Primary Study): pg. 1 - randomized centrally

I think central allocation is considered as low risk

Incomplete outcome data (attrition bias)
All outcomes
Low risk

Auvinen, 2001, BJU International: pg. 3 flow-chart; pg. 4 "imbalance in the numbers of patients between the arms within two hospitals. Not expected to affect the results in any way" "some participants refused to give informed consent, health deterioration, not seen by urologist"

Auvinen, 2004, BJU International (Primary Study): pg. 2 - flow diagram & results; Baseline data not included.

Selective reporting (reporting bias)Unclear riskNo indication that trial registered in central trials registry. Auvinen, 2001, BJU International: Protocol mentioned pg. 2 "The study protocol was approved by an ethical committee in each participating hospital"; Auvinen, 2004, BJU International (Primary Study): pg. 1 "The study protocol was approved by the institutional review board at each participating hospital"
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
High riskAuvinen, 2001, BJU International: pg. 3 "recognized carry-over effect because same physician in charge for intervention and control groups, diminish contrast between groups, as these physicians were more motivated to inform patients than those physicians not participating"; Auvinen, 2004, BJU International (Primary Study): No blinding but primary outcome is choice of treatment for prostate, objectively recorded. But unsure how physicians may have influenced decisions
Blinding of outcome assessment (detection bias)
All outcomes
Low riskNo blinding but primary outcome is choice of treatment for prostate, objectively recorded.

Barry 1997

MethodsRandomised to decision aid vs usual care
Participants104 + 123 patients considering benign prostatic hyperplasia treatment in the USA
InterventionsDA: Health Dialog interactive videodisc on options' outcomes, clinical problem, outcome probability, others' opinion
COMPARE: usual care using general information on the clinical problem
Outcomesuptake of option; knowledge*; satisfaction with DM process; satisfaction with decision; interest in DM; general health outcomes; condition specific health outcomes
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 2 "Stratified by study site in concealed blocks of 10"
Allocation concealment (selection bias)Low riskpg. 2 - study coordinator opening serially numbered, opaque, sealed envelopes
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 3 - patient accrual and follow-up; Reasons for withdrawal mentioned; pg. 4 Baseline characteristics included.
Selective reporting (reporting bias)Unclear riskNo indication that trial registered in central trials registry
Other biasLow riskAppears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Low riskNo blinding but phase 1 eliminated risk of contamination
Blinding of outcome assessment (detection bias)
All outcomes
Low riskNo blinding but phase 1 eliminated risk of outcome assessor interfering with decision

Bekker 2004

MethodsRandomised to detailed vs routine consultation
Participants59 + 58 pregnant women who have received a maternal serum screening positive test result for Down syndrome in the UK
InterventionsDA: decision analysis plus routine consultation on options' outcomes, clinical problem, outcome probability, values clarification, guidance/coaching
COMPARE: routine consultation on options' outcomes, outcome probability
Outcomesuptake of option; knowledge; decisional conflict; anxiety*; informed decision making; satisfaction with consultation; consultation length
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear risk

Bekker, 2003, Pt Ed & Counseling: pg. 2 - section 2.3 Sample and Procedure "randomly allocated... using previously numbered... envelopes"

Bekker, 2004, Prenat Diagn (Primary Study): pg. 3 "Participants were randomly allocated by previously numbered envelopes"; Does not mention how sequence was generated

Allocation concealment (selection bias)Low risk

Bekker, 2003, Pt Ed & Counseling: pg. 2 - section 2.3 Sample and Procedure "using previously numbered, sealed, opaque envelopes"

Bekker, 2004, Prenat Diagn (Primary Study): pg. 3 - previously numbered, sealed, opaque envelopes

Incomplete outcome data (attrition bias)
All outcomes
Unclear riskBekker, 2003, Pt Ed & Counseling; Bekker, 2004, Prenat Diagn (Primary Study): pg. 4 - results/flow diagram; Baseline characteristics not included.
Selective reporting (reporting bias)Unclear risk

Bekker, 2003, Pt Ed & Counseling: The coding frame was developed from literature. Does not mention protocol.

Bekker, 2004, Prenat Diagn (Primary Study): no information provided about central trials registry

Other biasUnclear riskBekker, 2003, Pt Ed & Counseling: does not directly address baseline characteristics of participants; Bekker, 2004, Prenat Diagn (Primary Study): appears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Low riskParticipants blinded, personnel not blinded. Same personnel did control & intervention. Tape recorded sessions to ensure no bias
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured

Bernstein 1998

MethodsRandomised to decision aid vs usual care
Participants65 + 53 patients with coronary artery disease considering revascularization surgery in the USA
InterventionsDA: Health Dialog video on options' outcomes, clinical problem, outcome probability, others' opinion
COMPARE: usual care (no information provided)
Outcomesuptake of option, knowledge, satisfaction with care, satisfaction with decision and decision making process*, general health outcomes, condition specific health outcomes
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 3 "Randomization was stratified by study site in blocks of 10"
Allocation concealment (selection bias)Low riskpg. 3 "randomization performed by a study coordinator opening opaque, sealed envelopes at study headquarters"
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 3 - flow diagram; Baseline data comparison included.
Selective reporting (reporting bias)Unclear riskNo information provided indicating trial was included in central trials registry
Other biasLow riskAppears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskNeither subjects nor study staff were blinded to treatment assignment - could lead to different satisfaction ratings based on knowing the treatment received
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured

Berry 2013

MethodsRandomised to decision aid vs usual care
Participants266 + 228 men considering prostate cancer treatment in the USA
Interventions

DA: interactive web based video on options' outcomes, clinical problem, outcome probabilities, others' opinion, guidance (list of questions to ask doctor and automated summary)

COMPARE: usual care

Outcomesdecisional conflict*, preferred/actual treatment choice (pre and post DA), proportion undecided
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg.3 (Methods section- second paragraph) "Participants were randomized automatically by the P3P application to study groups (1:1 using a simple randomization scheme with no blocking")
Allocation concealment (selection bias)Low riskpg.3 (Methods section) "Participants were randomized automatically by the P3P application to study groups"
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg.4: used ITT analysis and low dropout
Selective reporting (reporting bias)Low riskProtocol made available
Other biasUnclear riskWas a multicentre trial which could have lead to contamination, protocol violation and biased questionnaire completion
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskParticipants were not blinded and study does not address the effect on the results
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear whether outcome assessors are blinded, but outcomes are not subject to interpretation

Bjorklund 2012

MethodsRandomised to decision aid vs usual care
Participants236 + 247 women less than 11 weeks pregnant considering Down syndrome screening in Sweden
Interventions

DA: linear video on options' outcomes, clinical problem, outcome probabilities, others' opinion, and guidance (step by step process for making the decision)

COMPARE: usual care using pamphlet

Outcomesknowledge* (post DA), values congruent with chosen option (post DA), attitude* (post DA), uptake of CUB* (post DA)
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskpg. 391: "The midwife allocated the participants randomly by sealed envelopes..." but does not state the actual sequence generation method
Allocation concealment (selection bias)Low riskpg. 391: used sealed envelopes, "prepared, sequentially coded and distributed to the maternity units by the research group".
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskNo mention of why some participants' data were excluded in Tables 2, 3 and 4
Selective reporting (reporting bias)Unclear riskNo mention of study protocol
Other biasLow riskAppears to be free of other sources of bias
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskp. 395: 'It was not possible to blind neither [sic] the midwives nor the participants due to the characteristics of the intervention.'
Blinding of outcome assessment (detection bias)
All outcomes
Low riskNo blinding but outcomes were objectively measured and not subjective to interpretation

Chambers 2012

MethodsRandomised to DA vs usual care
Participants74 + 77 healthcare workers who did not receive the influenza vaccine considering receiving the vaccine in Canada
Interventions

DA: web based DA on options' outcomes, clinical problem, outcome probabilities, explicit values clarification and guidance

COMPARE: usual care using pamphlet

Outcomesconfidence in decision* (post DA), impact on immunization intent (post DA), proportion undecided.
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 199 - "The randomization list was generated using the randomization function in Excel 2002 (version 10.6856.6856 SP3)."
Allocation concealment (selection bias)Low riskpg. 199 - "The list was imported from Excel into a Microsoft SQL Server
database. The online application would sequentially assign a random identification number and their decision aid status (seeing the decision aid or not) from the randomization list when users logged into the survey."
Incomplete outcome data (attrition bias)
All outcomes
High risk65% completion rate in intervention arm and 77% completion rate in control arm: attrition could be different where the respondents and non-respondents are different
Selective reporting (reporting bias)Low riskprotocol available
Other biasUnclear riskFigure 1 numbers for exclusion are not logical
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskNot reported whether or not they were blinded during the course of the intervention
Blinding of outcome assessment (detection bias)
All outcomes
Low riskquestionnaire scores are objective and not subject to interpretation

Clancy 1988

MethodsRandomised to decision aid vs usual care
Participants753 + 263 Health physicians considering Hep B vaccine in the USA
InterventionsDA: pamphlet on options' outcomes, clinical problem, outcome probability, explicit values clarification (personal decision analysis), guidance/coaching
COMPARE: usual care (no information provided)
Outcomesuptake of option*
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 2 - random numbers table; all incoming residents were assigned to Group 2 (non-randomised residents identified as Subgroup)
Allocation concealment (selection bias)Unclear riskNo information provided
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskpg. 2 "35 physicians excluded because already received vaccine" Flow chart not included. Baseline characteristics not included.
Selective reporting (reporting bias)Unclear riskNo information provided
Other biasHigh riskpg. 287 - potential selection bias - non-randomised residents were added to group 2 and therefore potential unbalanced distribution; Plus low response rate among those offered decision analysis.
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskNo blinding of participants or personnel. Not clear how this may affect their decision
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but decisions for screening were retrieved from health records

Davison 1997

MethodsRandomised to decision aid + audio-taped consultation vs usual care
Participants30 + 30 men with prostate cancer considering treatment in Canada
InterventionsDA: written + audiotape consultation of options' outcomes, clinical problem, outcome probability, others' opinion
COMPARE: usual care (general information pamphlets on clinical problem)
Outcomesrole in decision making*, anxiety, depression
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 5 - Data Collection "The group to which subjects were assigned was predetermined by a block randomization procedure. This ensured there were an equal number of subjects in both groups for each physician."
Allocation concealment (selection bias)Unclear riskNot mentioned; pg. 5 - group assignment predetermined by block randomisation procedure
Incomplete outcome data (attrition bias)
All outcomes
Low riskno flow diagram; pg. 12 explains why certain men did not listen to audiotape. Baseline characteristics included. All men approached by study investigator agreed to participate, only one man refused to complete the second set of questionnaires.
Selective reporting (reporting bias)Unclear riskProtocol not mentioned
Other biasLow riskAppears to be free of other sources of bias. Similar baseline characteristics,
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskNo blinding. Unclear how participant's willingness to participate was affected by knowing they received the intervention
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskUnclear blinding, and whether outcomes could be affected by not blinding the assessor

de Achaval 2012

MethodsRandomised to detailed vs simple vs usual care
Participants70 + 70 + 71 patients diagnosed with knee OA considering OA treatment in the USA
Interventions

COMPLEX DA: videobooklet + interactive conjoint analysis on options' outcomes, clinical problem, outcome probabilities, explicit values clarification, others' opinion and guidance (list of questions)

COMPARE DA: videobooklet on options' outcomes, clinical problem, outcome probabilities, others' opinion and guidance (list of questions)

COMPARE: usual care receiving generic booklet

Outcomesdecisional conflict* (pre and post DA)
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 231: computer generated list with uneven blocks
Allocation concealment (selection bias)Low riskpg. 231 (procedure): numbered, sealed and opaque envelopes
Incomplete outcome data (attrition bias)
All outcomes
Low risk3 dropouts; missing data effect size unlikely to have significant impact on study outcome
Selective reporting (reporting bias)Unclear riskprotocol not available
Other biasLow riskappears to be free of other sources of bias
Blinding of participants and personnel (performance bias)
All outcomes
Low riskpg. 231 (procedures): likely not blinded, but low threat to causality in study
Blinding of outcome assessment (detection bias)
All outcomes
Low riskpg. 231: patients were not blinded but outcome was objectively measured

Deschamps 2004

MethodsRandomised to detailed decision aid vs pharmacist consultation
Participants67 + 61 women considering hormone replacement therapy in Canada
InterventionsDA: audiotape booklet on options' outcomes, clinical problem, outcome probabilities, explicit values clarification, others' opinions, guidance/coaching (Ottawa Decision Support Framework)
COMPARE: 40-minute pharmacist consultation on options' outcomes, outcome probability
Outcomespreferred option*, decisional conflict*, role in decision making*, satisfaction with preparation for decision making, satisfaction with decision*, adherence*
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskno information provided
Allocation concealment (selection bias)Unclear riskno information provided
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 4 - flow diagram; pg.3 reasons for attrition mentioned. Baseline characteristics included.
Selective reporting (reporting bias)Unclear riskno information provided
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskUnclear blinding
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured

Deyo 2000

MethodsRandomised to detailed vs simple decision aid
Participants190 + 203 adults with herniated disc or spinal stenosis considering back surgery in the USA
InterventionsDA: Health Dialog interactive videodisc on options' outcomes, clinical problem, outcome probability, other's opinions
COMPARE: simple DA pamphlet with clinical problem, options outcomes.
Outcomesuptake of option*, satisfaction with DM process, satisfaction with care, condition specific health outcomes
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 3 "computer generated simple randomization sequence"; Phelan - pg. 2 computer generated
Allocation concealment (selection bias)Low riskpg. 3 "series of numbered opaque envelopes"; Phelan - pg. 2 - concealed in serially marked, opaque envelopes
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 5 - flow diagram; Reasons for attrition mentioned and participants balanced across study groups. Baseline data not included; Phelan - Flow of participants not included. Baseline characteristics included.
Selective reporting (reporting bias)Unclear riskNo indication that the trial was registered in a central trials registry
Other biasLow riskpg. 4 - There were no significant group differences; appears to be free of other potential biases; Phelan - appears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskUnclear blinding
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured

Dodin 2001

MethodsRandomised to detailed vs simple decision aid
Participants52 + 49 women considering hormone replacement therapy in Canada
InterventionsDA: audiotape booklet on options' outcomes, clinical problem, outcome probability, explicit values clarification, others' opinions, guidance/coaching (Ottawa Decision Support Framework)
COMPARE: simple decision aid pamphlet with options' outcomes, clinical problem
Outcomespreferred option, knowledge, decisional conflict*, accurate risk perceptions, congruence between values and choice
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskpg. 2 - eligible women randomly assigned - no information on how sequence was generated
Allocation concealment (selection bias)Unclear riskno information provided
Incomplete outcome data (attrition bias)
All outcomes
Low riskBaseline characteristics included. pg. 3 "Toutes les 101 femmes recrutées ont complété l’étude" [all 101 women recruited completed the study]
Selective reporting (reporting bias)Unclear riskno information provided
Other biasUnclear riskwomen of 50-59 years and married were significantly more numerous in the experimental group
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskpg. 2 - a research assistant met the women during a debriefing of 20 minutes in small groups of 4-5 women assigned to the same intervention to avoid inter-group contact, thus ensuring blinding. Not sure if the physicians were blinded
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured

Dolan 2002

MethodsRandomised to decision aid vs usual care
Participants50 + 47 average risk for colorectal cancer considering screening in the USA
InterventionsDA: computer with analytic hierarchy process on options' outcomes, clinical problem, outcome probability, explicit values clarification, guidance/coaching
COMPARE: usual care with information on options, clinical problem.
Outcomesuptake of option*, decisional conflict*, role in decision making
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 2 (Study Interventions) "randomization schedules were created using a computer random number generator"
Allocation concealment (selection bias)Low riskpg. 2 (Study Interventions) - computer-based
Incomplete outcome data (attrition bias)
All outcomes
Low riskSee flow diagram and description in results
Selective reporting (reporting bias)Unclear riskNothing specifically mentioned re: study protocol
Other biasLow riskAppears to be free of other sources of bias
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskUnclear blinding of participants. All patient interviews in both the experimental and control groups were done by the same investigator, unclear on how this could contribute to risk of bias
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured and not subjective to interpretation

Evans 2010

MethodsRandomised to online decision aid vs paper decision aid vs questionnaire vs usual care
Participants129 + 126 + 127 + 132 men considering PSA screening in Wales
Interventions

DA: online program on options' outcomes, clinical problem, outcome probabilities, explicit values clarification, others' opinion, guidance (interactive computer program; summary)

COMPARE: paper version of online DA on options' outcomes, clinical problem, outcome probabilities, explicit values clarification, others' opinion, guidance (interactive computer program; summary)

COMPARE: received a questionnaire

COMPARE: received nothing

OutcomesKnowledge* (post DA), attitude (post DA), intention to undergo PSA testing (post DA), anxiety (post DA), uptake of PSA test (post DA), total decisional conflict
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskpg.4 (recruitment process): "a random sample of 100 men was selected from the list." "The process ensured individual level randomization"
Allocation concealment (selection bias)Low riskpg.4 (recruitment process): "affirmative consent forms from each practice were transferred to the research officer who allocated each participant with a number provided remotely by the trial statistician to ensure concealment"
Incomplete outcome data (attrition bias)
All outcomes
Low risksee Figure 1 for flow diagram and Table 1 for baseline characteristics of participants
Selective reporting (reporting bias)Low riskregistered as a trial
Other biasLow riskthe study appears free of other sources of bias
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskthe study does not address this outcome
Blinding of outcome assessment (detection bias)
All outcomes
Low riskunclear blinding but outcomes were objectively measured and not subjective to interpretation

Fagerlin 2011

MethodsDecision aid vs delayed intervention vs control
Participants382 + 159 + 100 women with an elevated five year risk of breast cancer considering breast cancer prevention medication in the USA
Interventions

DA: tailored DA on options' outcomes, clinical problem, outcome probabilities, and explicit values clarification

COMPARE: given DA after 3-month follow-up

COMPARE: given DA after all outcome measures were taken

Outcomesdecisional conflict (post DA), behavioural intent (post DA), actual behaviour (post DA), proportion undecided, perception of benefits (post DA), perception of risk (post DA).
Notesprimary outcome was not specified
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskrandom sequence generation was provided by the author
Allocation concealment (selection bias)Low riskcentral and web-based allocation
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskdoes not report exclusions
Selective reporting (reporting bias)Unclear riskno mention of study protocol
Other biasLow riskappears to be free of other sources of bias
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskunclear blinding - study does not address this outcome
Blinding of outcome assessment (detection bias)
All outcomes
Low riskunclear blinding but outcomes were objectively measured and not subjective to interpretation

Fraenkel 2007

MethodsRandomised to decision aid vs usual care
Participants47 + 40 patients with knee pain considering treatment options in the USA
Interventions

DA: interactive computer tool options' outcomes, outcome probability, explicit values clarification

COMPARE: usual care using the Arthritis Foundation information pamphlet

OutcomesDecisional self-efficacy, preparation for decision making
Notesprimary outcome was not specified
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 2 - computer-generated randomisation sequence
Allocation concealment (selection bias)Unclear riskno information provided; computer generated
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 3 - results; baseline characteristics included and balanced
Selective reporting (reporting bias)Unclear riskno information provided; no indication of trial was registered centrally
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskNo blinding but unclear if it has impact on the outcomes measured
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured and not subjective to interpretation

Frosch 2008

MethodsRandomised to decision aid vs. decision aid + chronic disease trajectory vs chronic disease trajectory vs usual care (Internet information)
Participants155 + 152 + 153 + 151 men considering prostate cancer screening
Interventions

DA:information on options' outcomes, clinical problem, outcome probabilities, others' opinions

COMPARE: information on options' outcomes, clinical problem, outcome probabilities, others' opinions, explicit values clarification (utilities for outcomes associated with prostate cancer)

COMPARE: explicit values clarification (utilities for outcomes associated with prostate cancer)

COMPARE: usual care using public information on prostate cancer screening on American Cancer Society and Centers for Disease Control and Prevention websites 2005-2006

Outcomesknowledge*, actual option*, decisional conflict*, concern about prostate cancer, treatment preference if prostate cancer diagnosed
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskcomputer algorithm randomly assigned participants to the 4 study groups
Allocation concealment (selection bias)Low riskrevealed after signed consent and completed baseline measures
Incomplete outcome data (attrition bias)
All outcomes
Low riskused intention to treat analysis; imputed missing data for participants who did not complete follow-up assessments
Selective reporting (reporting bias)Unclear riskno indication of published protocol
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskaccessed a secure Internet site that hosted all study materials; participants had unlimited access to assigned intervention, unclear blinding of personnel
Blinding of outcome assessment (detection bias)
All outcomes
Low riskunclear blinding but outcomes were measured via questionnaires and not subjective to interpretation

Gattellari 2003

MethodsRandomised to decision aid vs usual care
Participants126 + 122 men considering PSA testing in Australia
InterventionsDA: pamphlet on options' outcomes, clinical problem, outcome probability, explicit values clarification
COMPARE: usual care using brief information on screening test and chances of false-positive results
Outcomespreferred option, knowledge, decisional conflict, accurate risk perceptions, perceived ability to make an informed choice
Notesprimary outcome was not specified
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskpg. 1 - pre-randomised code - no further information
Allocation concealment (selection bias)Low riskpg. 1 - pre-randomised code unobtrusively marked on envelopes
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskpre-test characteristics included. Flow chart not included and reasons for attrition not mentioned.
Selective reporting (reporting bias)Unclear riskno information provided
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskconsenting men were blinded to allocation, but unclear if personnel were blinded
Blinding of outcome assessment (detection bias)
All outcomes
Low riskunclear blinding but outcomes were objectively measured and not subjective to interpretation

Gattellari 2005

MethodsRandomised to decision aid booklet vs decision aid video vs usual care
Participants140 + 141 + 140 men considering PSA testing in Australia
InterventionsDA: pamphlet on options' outcomes, clinical problem, outcome probability, explicit values clarification
COMPARE: video on clinical problem, outcome probability, others' opinion
COMPARE: usual care using brief information on screening test and chances of false-positive results
Outcomespreferred option, knowledge, decisional conflict, perceived ability to make an informed choice
Notesprimary outcome was not specified
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 2 - 2.3.1. Unique identification codes assigned to participants according to date and time enrolled into the interventional component of the study. Block randomisation of identification codes then performed via computer software
Allocation concealment (selection bias)Low riskpg. 2 - 2.3.1. "Allocation concealment was ensured as the interviewers, responsible for enrolling participants onto the trial, were blinded to the randomised study design while one of the
authors (MG) was responsible for randomisation. Hence, it was not possible for either participants or interviewers to be aware of the randomisation sequence."
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 5 (172) "interviews terminated, call times exhausted, one man with prostate cancer accidentally included, but data is excluded from results" pg. 7 baseline characteristics equally distributed; pg. 6 fig. 1 - flow diagram
Selective reporting (reporting bias)Unclear riskpg.13 (180)par. 5 "success of study protocol" "limitation to protocol: men not confronted with actual decision to undergo PSA screening; No indication that trial registered in central trials registry
Other biasLow riskpg. 13 (180) par. 5 "high follow-up rate and allocation concealment; study not subjected to selection bias" Appears to be free of other sources of bias
Blinding of participants and personnel (performance bias)
All outcomes
Low riskParticipants & interviewers were blinded
Blinding of outcome assessment (detection bias)
All outcomes
Low riskAt post-test, it was not possible to blind the interviewers but outcomes were objectively measured and not subjective to interpretation

Goel 2001

Methods

Randomised to detailed vs simple decision aid

Cluster randomised trial

Participants86 + 50 women considering surgery for breast cancer (cluster RCT with 57 surgeons randomised) in Canada
InterventionsDA: audiotape + booklet on options' outcomes, clinical problem, outcome probability, values clarification, other's opinions, coaching/guidance (Ottawa Decision Support Framework)
COMPARE: simple DA pamphlet with clinical problem, options outcomes
Outcomesknowledge, decisional conflict*, decisional regret, anxiety
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 2 "blocks of 8 based on a random number generator"
Allocation concealment (selection bias)Unclear riskpg. 2 - Prerandomisation was done to eliminate opportunities to select into the study intervention arm. The allocation was not revealed to the surgeon until after agreement to participate in the study was obtained.
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 3 - Baseline characteristics included. pg. 3 - results
Selective reporting (reporting bias)Unclear riskno information provided
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskpg. 2 - unclear on whether participants were blinded. The allocation was not revealed to the surgeon until after agreement to participate in the study was obtained
Blinding of outcome assessment (detection bias)
All outcomes
Low riskunclear blinding but outcomes were objectively measured and not subjective to to interpretation

Green 2001a

MethodsRandomised to decision aid + counselling vs counselling alone vs usual care
Participants29 + 14 women with a first degree relative with breast cancer interested in learning about genetic testing in the USA
Interventions

DA: CD-ROM plus counselling on options' outcomes, clinical problem, others' opinions, guidance/coaching

COMPARE: counselling
COMPARE: usual care

Outcomesknowledge, preferred options*
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 2 "block randomization schedule to one of 3 groups in a 2:2:1 ratio"
Allocation concealment (selection bias)Unclear riskno information provided
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskpg. 5 table "Values do not always add up to the number of participants due to missing data" Reasons not mentioned. pg. 4 "Participants' baseline knowledge was reflected in the control group's answers" Participants balanced in study groups.
Selective reporting (reporting bias)Unclear riskno information provided
Other biasLow riskappears to be free of other sources of bias
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskpg. 2 "genetic counsellor blinded to randomization until just prior to the session", unclear if participants were blinded
Blinding of outcome assessment (detection bias)
All outcomes
Low riskunclear blinding but outcomes were objectively measured and not subjective to to interpretation

Green 2004

MethodsRandomised to detailed decision aid + genetic counselling vs routine genetic counselling
Participants106 + 105 women with first degree relative with breast cancer considering genetic testing in the USA
InterventionsDA: CD-ROM plus counselling on options' outcomes, clinical problem, others' opinions, guidance/coaching
COMPARE: genetic counselling
Outcomespreferred option, knowledge*, decisional conflict, satisfaction with decision, anxiety, counsellor/participant rating of effectiveness of counselling session, consultation length
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskGreen, 2004, JAMA (Primary Study):pg. 2 - used separate computer generated randomisation lists for low-risk and high-risk individuals at each study site; Green, 2005, Genet Med: no information provided
Allocation concealment (selection bias)Unclear riskGreen, 2004, JAMA (Primary Study): no information provided; Green, 2005, Genet Med: no information provided
Incomplete outcome data (attrition bias)
All outcomes
Low riskGreen, 2004, JAMA (Primary Study):pg. 4 figure; flow chart. Reasons for attrition/loss to follow-up not included. p.5 Baseline characteristics included; Green, 2005, Genet Med:pg. 4 - flow diagram; reasons for attrition mentioned and participants balanced in study groups. Baseline characteristics included.
Selective reporting (reporting bias)Unclear riskNo indication that the trial was registered in a central trials registry
Other biasLow riskGreen, 2004, JAMA (Primary Study): appears to be free of other potential biases; Green, 2005, Genet Med: appears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Low riskGreen, 2005, Genet Med: pg. 8 - this was not a blinded study "counselor's responses may have been biased" but primary outcome was objective, so it is unlikely to introduce bias.
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured and not subjective to interpretation

Hamann 2006

MethodsCluster randomised trial of decision aid vs usual care
Participants54 + 59 patients with schizophrenia considering treatment options (cluster RCT with 12 wards paired and randomised) in Germany
InterventionsDA: 16-page booklet on options' outcomes, outcome probabilities, explicit values clarification, coaching/guidance
COMPARE: usual care
Outcomesknowledge, participation in decision making (COMRADE - doctor gave me a chance to decided which treatment I thought was best for me), uptake of psycho education, rehospitalization, adherence, satisfaction with care, severity of illness (baseline only), attitudes about drug use, decision making preference
Notesprimary outcome was not specified
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskp266 "one member of each pair being randomly assigned to the control or to the interventional condition." Sequence generation method was not stated
Allocation concealment (selection bias)Unclear riskno mention of allocation concealment
Incomplete outcome data (attrition bias)
All outcomes
Low riskreasons for attrition mentioned
Selective reporting (reporting bias)Unclear riskno information provided
Other biasHigh riskclustering was not accounted for in the analysis
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskno information provided
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskno information provided

Hanson 2011

MethodsRandomised to decision aid vs usual care
Participants127 + 129 patients diagnosed with advanced dementia and eating problems considering long term feeding tube placement in the USA
Interventions

DA: booklet or audio recording on options' outcomes, clinical problem, outcome probabilities, explicit values clarification, others' opinion, guidance (steps in decision making, worksheet, summary)

COMPARE: usual care

Outcomessurrogate knowledge, risk perceptions, decisional conflict* (3 months post DA), frequency of communication with providers (3 months post DA), feeding treatment use (3, 6 and 9 months post DA), participation in decision making, satisfaction with the decision, decisional regret.
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg.2010 (randomization): computerized random number generation
Allocation concealment (selection bias)Unclear riskpg.2010 (randomization): no description of method used to conceal allocation
Incomplete outcome data (attrition bias)
All outcomes
Unclear risktable 3: intervention group missing data for 1 participant, reason for omission not reported
table 4: no explanation for number of participants in each group (127) given- numbers vary from those in 'recruitment and retention' figure
Selective reporting (reporting bias)Low riskregistered with clinicaltrials.gov, protocol on website
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskpg.2014 (discussion)
"Cluster randomization prevented double blinding and may have introduced bias due to site effects", study authors unsure of effect on study
Blinding of outcome assessment (detection bias)
All outcomes
Low riskpg.2010 (randomization) "because of cluster randomization, data collectors were not blinded to group assignment", authors believe has little impact on study

Heller 2008

MethodsRandomised to decision aid vs usual care
Participants66 + 67 breast cancer patients eligible for breast reconstruction in the USA
InterventionsDA: interactive software program on options' outcomes, others' opinions
COMPARE: standard patient education
Outcomesknowledge, anxiety, satisfaction with treatment choice, satisfaction with decision making ability
Notesprimary outcome was not specified
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 2 - "upon study entry, the participants were randomized (computer generated) to one of two groups"
Allocation concealment (selection bias)Unclear risknot enough information provided
Incomplete outcome data (attrition bias)
All outcomes
Low riskBaseline anxiety and knowledge included in graphs. pg. 3 Participant numbers between study groups balanced. Reasons for incomplete questionnaires and study withdrawals mentioned.
Selective reporting (reporting bias)Unclear riskno information provided re: protocol
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskno information provided
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskno information provided

Hess 2012

MethodsRandomised to decision aid vs usual care
Participants103 + 105 patients in the the emergency department with primary symptoms of nontraumatic chest pain and were being considered of admission to the emergency department observation unit for monitoring and cardiac stress testing within 24 hours.
Interventions

DA: one page printout on options' outcomes, clinical problem, and outcome probabilities

COMPARE: usual care

Outcomesknowledge*, risk perceptions, decisional conflict, actual choice, satisfaction with decision making process, patient-practitioner communication.
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg 253 - "Patients were randomized to either usual care or shared decision making through a Web-based, computer-generated allocation sequence in a 1:1 concealed fashion"
Allocation concealment (selection bias)Low riskpg 253 - "Patients were randomized to either usual care or shared decision making through a Web-based, computer-generated allocation sequence in a 1:1 concealed fashion"
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskSome of the numbers of patients reported in the results did not match the flow chart
Selective reporting (reporting bias)Low riskProtocol is available
Other biasLow riskappears to be free of other biases
Blinding of participants and personnel (performance bias)
All outcomes
Low riskpg 253. outcome measures.  Personnel were blinded, but unclear if patients were blinded.  However, the primary outcome is unlikely to be biased.
Blinding of outcome assessment (detection bias)
All outcomes
Low riskpg 253. outcome measures.  Investigators assessing outcomes were blinded. 

Hunter 2005

MethodsRandomised to decision aid with option to speak to genetic counsellor vs individual genetic counselling vs group counselling
Participants116 + 126 + 110 women of advanced maternal age considering prenatal diagnostic testing in Canada
InterventionsDA: audiotape workbook on options' outcomes, clinical problem, outcome probability, explicit values clarification, others' opinions, guidance/coaching (Ottawa Decision Support Framework)
COMPARE: individual counselling session on options' outcomes, outcome probability, values clarification
COMPARE: group counselling session on options' outcomes, outcome probability, others' opinions
Outcomesuptake of option, knowledge*, decisional conflict*, satisfaction with decision making process*, anxiety*
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 4 - randomised in blocks of 30, 10 to each intervention group
Allocation concealment (selection bias)Low riskpg. 4 - "the allocations were provided in opaque envelopes"
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskoutcomes reported on fewer participants but no rationale provided for missing data
Selective reporting (reporting bias)Unclear riskno information provided
Other biasLow riskappears to be free of other bias
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskunclear blinding
Blinding of outcome assessment (detection bias)
All outcomes
Low riskunclear blinding but outcomes were objectively measured and not subjective to to interpretation

Jibaja-Weiss 2011

MethodsRandomised to decision aid vs usual care
Participants51 + 49 women diagnosed with breast cancer considering surgical treatment in the USA
Interventions

DA: computer program on options' outcomes, clinical problem, outcome probabilities, explicit values clarification, others' opinion and guidance (step by step process for making the decision)

COMPARE: usual care + breast cancer treatment educational materials normally provided to patients

Outcomessurgical treatment preference (post DA), breast cancer knowledge (pre, post DA, post DA and consult), satisfaction with surgical decision (post DA), satisfaction with decision making process (post DA), decisional conflict (pre, post DA, post DA and consult), proportional undecided.
Notesprimary outcome was not specified
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg.42 (Methods section): "Patients at each hospital were randomized using permuted blocks"
Allocation concealment (selection bias)Unclear risknot addressed in the study
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskthere is no way to know if the plots are including all of the participants' data since they do not specify what was the number of patients used to obtain these mean scores
Selective reporting (reporting bias)Unclear riskno mention of protocol
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear risknot addressed in the study
Blinding of outcome assessment (detection bias)
All outcomes
Low riskunclear blinding but outcomes were objectively measured and not subjective to interpretation

Johnson 2006

MethodsRandomised to decision aid vs usual care
Participants32 + 35 patients considering endodontic treatment options in the USA
InterventionsDA: decision board on options' outcomes, clinical problem, outcome probability, guidance
COMPARE: usual care
Outcomesknowledge*, satisfaction with decision making process*, anxiety*
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskp.3 "four computerized random generation lists to assign to one of two groups"
Allocation concealment (selection bias)Unclear riskNO for residents: pg. 3 - 4 computer-generated randomisation lists (1 for each resident) were prepared by the PI; therefore residents would have had pre-generated lists; but UNCLEAR for patients p.3 "Allocation was concealed from patients" but does not explained how
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 6 fig. 3 - flow diagram; pg. 5 - all 40 patients agreed to participate in the study, but only 32 questionnaires were useable several residents did not understand need for entering data on the envelope and placing matched questionnaire in it
Selective reporting (reporting bias)Unclear riskNo indication that the trial was registered in a central trials registry
Other biasUnclear riskp.5 "baseline data obtained because possible that clinicians training in the EndoDB would alter usual care discussions" Mentions taking baseline characteristics, but not included in article
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskBlinding not mentioned. pg. 3 - allocation was concealed from patients only
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured and not subjective to to interpretation

Kasper 2008

MethodsRandomised to decision aid vs usual care
Participants150 + 147 multiple sclerosis patients considering immunotherapy in Germany
Interventions

DA: booklet and worksheet on options' outcomes, clinical problem, outcome probabilities, explicit values clarification (based on IPDAS)

COMPARE: information material on immunotherapy (80 pages)

Outcomesrole in decision making*, choice, feeling undecided, helpfulness with making a decision, attitudes toward immunotherapy, expectations of side effects realized at 6 months
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 2 - "allocation using computer generated random numbers"
Allocation concealment (selection bias)Unclear riskpg. 2 - Assignment - randomisation was carried out by concealed allocation, but method of concealment was not described
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 2 - Fig 1 flow of participants; baseline data/characteristics included
Selective reporting (reporting bias)Low riskpg. 2 "The protocol of this study has been published with the trial registration at http://controlled-trials.com/ ISRCTN25267500"
Other biasUnclear riskpg. 5 - difference in preferred interaction style between groups at baseline (P value 0.04)
Blinding of participants and personnel (performance bias)
All outcomes
Low riskpg. 3 - Masking - Participants were not told whether the information they received was standard information or the newly developed DA
Blinding of outcome assessment (detection bias)
All outcomes
Low riskpg. 3 - Masking - assessors were not told whether the information they received was standard information or the newly developed DA

Kennedy 2002

MethodsRandomised to decision aid + coaching vs decision aid only vs usual care
Participants215 + 206 + 204 women considering treatment for menorrhagia in the UK
InterventionsDA: video + booklet on options' outcomes, clinical problem, outcome probabilities, explicit values clarification, others' opinions, guidance/coaching
COACHING: ˜20 minute coaching with explicit values clarification by a registered nurse prior to see physician
COMPARE: usual care
Outcomesuptake of option, satisfaction, general quality of life*, menorrhagia severity, cost effectiveness
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 3 - allocation sequence was generated by computer and stratified by consultant and the age at which the woman left full-time education
Allocation concealment (selection bias)Low riskpg. 3 "Secure randomization ensured by using a central telephone randomization system"
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 4-5 - see Table 1 and Figure 1 flow diagram.
Selective reporting (reporting bias)Unclear riskno information provided
Other biasLow riskappears to be free from other risks of bias
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskpg. 6 possibility of contamination bias, clinicians could have applied the experience gained from consultations with the interventions groups in their consultations with the control group
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear if blinding used but most outcomes were objectively measured and not subjective to interpretation

Krist 2007

MethodsRandomised to decision aid booklet vs decision aid web-based vs usual care
Participants196 + 226 + 75 patients considering prostate cancer screening in the USA
Interventions

DA: 4 page pamphlet with options' outcomes, clinical problem, outcome probability

COMPARE: web-site with same information as paper based DA

COMPARE: usual care

Outcomesrole in decision making*, knowledge, decisional conflict, time spent discussing screening, choice (PSA test ordered)
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 2 "coordinator referred to pre-generated randomisation tables to inform the participant to which arm he was randomised"
Allocation concealment (selection bias)Low riskpg. 2 - At the time of enrolment, the allocation was concealed from the coordinator
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 3 - results; pg. 4 - flow diagram
Selective reporting (reporting bias)Unclear riskno information provided
Other biasUnclear riskuneven groups but done intentionally, ration of 1:3:3 but appears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
High riskphysicians were not blinded - could affect decision making process and uptake of screening
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured and not subjective to interpretation

Kuppermann 2009

MethodsRandomised to decision aid vs simple decision aid
Participants244 + 252 pregnant women considering prenatal testing in the USA
Interventions

DA: computerized tool on options' outcomes, clinical problem, outcome probability, explicit values clarification

COMPARE: education booklet on the computer on options' outcomes, clinical problem

Outcomesknowledge*, decisional conflict*, accurate risk perception* (procedure related miscarriage, DS affected fetus), decision regret, satisfaction with the intervention*, satisfaction with involvement in decision making
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskno information provided
Allocation concealment (selection bias)Low riskpg. 3 - Materials and Methods - interviewer opened an opaque envelope containing the randomisation assignment
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 3 - Fig 1 flow diagram
Selective reporting (reporting bias)Low riskpg. 1 - Abstract CLINICAL TRIAL REGISTRATION: Clinicaltrials.gov, www.clinicaltrials.gov, NCT00686062
Other biasUnclear riskappears to be free of other potential biases p.10 "The effect of these potential selection biases not determined"
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskUnclear blinding
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured and not subjective to to interpretation

Labrecque 2010

MethodsRandomised to decision aid vs simple decision aid
Participants32 + 31 men considering vasectomy as an option for contraception in Quebec, Canada
Interventions

DA: booklet on options outcomes, clinical problem, outcome probability, explicit values clarification, guidance (step by step process for making the decision; one or more questions that asked patients to clarify their preferences; encourages patients to communicate with their practitioners)

COMPARE: booklet on options outcomes, clinical problem, outcome probability

Outcomesknowledge (pre and post DA), decisional conflict* (pre and post DA), preferred option (pre and post DA), proportion undecided.
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 557 (section 2.1): "computer-generated list"
Allocation concealment (selection bias)Low riskpg. 557 (section 2.1): "Randomization was stratified according to method of recruitment" "Each participant received either the full or the abridged PtDA inside an opaque, sealed, unmarked envelope"
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg.559 (Fig 1 flow diagram): p.560 Table 2 results
Selective reporting (reporting bias)Unclear riskno mention of examination of selective outcome reporting or study protocol
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Low riskpg. 557 (section 2.1): single-blinded- "researcher in charge of recruitment remained blind to the participants' group assignment". Participants were given an intervention but not aware of the alternative intervention.
Blinding of outcome assessment (detection bias)
All outcomes
Low riskpg.558 (section 2.4): "researcher blinded to the participant's group allocation when conducting the interviews", outcomes were objectively measured.

Lalonde 2006

MethodsRandomised to decision aid + pharmacist consultation vs simple decision aid + pharmacist consultation
Participants13 + 13 patients considering lifestyle changes and drug therapy to improve cardiovascular health in Canada
InterventionsDA: booklet and worksheet on options' outcomes, clinical problem, outcome probability, explicit values clarification, others' opinion, guidance/coaching (Ottawa Decision Support Framework)
COMPARE: personal risk profile with clinical problem, outcome probabilities
Outcomesknowledge, risk perception, decisional conflict, satisfaction with decision making process
Notesprimary outcome was not specified
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskp. 2 "research nurse randomly assigned participant" "stratified by community pharmacy" Does not indicate how randomisation occurred.
Allocation concealment (selection bias)Unclear riskno information provided
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskp. 4 does not indicate why participants did not complete post-intervention interview and post follow-up interview. Pre-intervention characteristics included on p. 6 in bar-graphs.
Selective reporting (reporting bias)Unclear riskprotocol not mentioned
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskunclear blinding
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskunclear blinding, interviews used open-ended question, do not know whether this contributes to risk of bias

Langston 2010

MethodsRandomised to decision aid + coaching vs. usual care
Participants114 + 108 women pregnant women in their first trimester considering use of contraceptives in the USA
Interventions

DA: double-sided flip chart on clinical problem, outcome probabilities, guidance (administered by a research assistant), coaching (structured, standardized, non-directive contraceptive counselling) + usual care

COMPARE: usual care

Outcomesproportion of participants choosing very effective contraceptive method* (post DA and consult), actual choice on day of procedure (post DA and consult), adherence of very effective and/or effective methods at 3 months and at 6 months (post DA and consult)
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg.363 (Methods-study procedures): "Using a random-number table, we determined the sequence for 1:1 allocation constrained by blocks of 10"
Allocation concealment (selection bias)Low riskpg.363 (Methods-study procedures): "Randomization assignments were sealed inside numbered, opaque envelopes"
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskfor "method initiation on the day of the procedure" it is only said that the "Participants in the intervention group were not more likely to initiate the requested method immediately compared to those in the usual care group"; possible that the results contradicted the hypothesis and were excluded for this reason
Selective reporting (reporting bias)Unclear riskno mention of study protocol; not enough information to permit judgement
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Low riskpg.363 (Methods-study procedures): "No blinding of participants or coordinators was feasible due to the nature of the intervention. Physician-providers did not know the participant's allocation group, did not discuss the study with patients, and were asked not to change their counselling"
Blinding of outcome assessment (detection bias)
All outcomes
Low riskUnclear blinding but outcomes were objectively measured and not subjective to interpretation

Laupacis 2006

MethodsRandomised to decision aid vs usual care
Participants60 + 60 patients undergoing elective open heart surgery considering pre-operative autologous blood donation in Canada
InterventionsDA: audiotape booklet on options' outcomes, clinical problem, outcome probability, explicit values clarification, guidance (Ottawa Decision Support Framework)
COMPARE: usual care
Outcomesuptake of option, knowledge*, decisional conflict*, satisfaction with decision making process, satisfaction with decision, accurate risk perceptions
Notes*primary outcome
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskpg. 2 "Randomization envelopes were prepared centrally by a statistician"
Allocation concealment (selection bias)Low riskpg. 2 "The envelopes were labeled with identification numbers and contained a card specifying the patient’s group assignment. The envelopes were opened by the interviewer after completion of the baseline interview."
Incomplete outcome data (attrition bias)
All outcomes
Low riskpg. 4 results, fig 1 flow diagram
Selective reporting (reporting bias)Unclear riskno information provided
Other biasLow riskappears to be free of other potential biases
Blinding of participants and personnel (performance bias)
All outcomes
Unclear riskno information provided
Blinding of outcome assessment (detection bias)
All outcomes
Low riskunclear blinding but outcomes were objectively measured and not subjective to interpretation

Legare 2003

MethodsCluster randomised to decision aid vs simple decision aid
Participants97 + 87 post-menopausal women considering hormone replacement therapy (Cluster RCT with 40 family physicians randomised) in Canada
InterventionsDA: audiotape, booklet and worksheet on options' outcomes, clinical problem, outcome probabilities, explicit values clarification, others' opinions, guidance (Ottawa Decision Support Framework)
COMPARE: general information pamphlet on risks, benefits and side-effects of HRT
Outcomesdecisional conflict, satisfaction with decision making process, agreement between physicians' and patients' decisional conflict
Notesprimary outcome was not specified
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskno information provided
Allocation concealment (selection bias)Unclear riskno information provided
Incomplete outcome data (attrition bias)
All outcomes
Unclear risk</