Interventions to support shared decision making for hypertension: A systematic review of controlled studies

Abstract Background Hypertension (high blood pressure) is a common long‐term health condition. Patient involvement in treating and monitoring hypertension is essential. Control of hypertension improves population cardiovascular outcomes. However, for an individual, potential benefits and harms of treatment are finely balanced. Shared decision making has the potential to align decisions with the preferences and values of patients. Objective Determine the effectiveness of interventions to support shared decision making in hypertension. Search strategy Searches in MEDLINE, EMBASE, CINAHL, Web of Science and PsycINFO up to 30 September 2017. Eligibility criteria Controlled studies evaluating the effects of shared decision‐making interventions for adults with hypertension compared with any comparator in any setting and reporting any outcome measures. Results Six studies (five randomized controlled trials) in European primary care were included. Main intervention components were as follows: training for health‐care professionals, decision aids, patient coaching and a patient leaflet. Four studies, none at low risk of bias, reported a measure of shared decision making; the intervention increased shared decision making in one study. Four studies reported blood pressure between 6 months and 3 years after the intervention; there was no difference in blood pressure between intervention and control groups in any study. Lack of comparability between studies prevented meta‐analysis. Conclusions Despite widespread calls for shared decision making to be embedded in health care, there is little evidence to inform shared decision making for hypertension, one of the most common conditions managed in primary care.


| BACKG ROU N D
Shared decision making is a process by which clinicians and patients work together to make health-care choices, based on clinical evidence and the patient's informed preferences. 1 Shared decision making is viewed as an ethical imperative by health-care professional regulatory bodies 2 and is embedded in health policy in several countries, including the UK and the United States. 3,4 It is increasingly advocated in the care of all conditions, including chronic health-care conditions such as hypertension (high blood pressure) 5 Implementing shared decision making in routine care has proven challenging, and many barriers have been identified from both patient and healthcare professional perspectives. 6,7 Interventions to support shared decision making include those which prepare health-care teams, individual clinicians or patients before consultations (e.g patient coaching interventions, decision aids, clinician or health-care team training interventions), and those which help practitioners and patients make decisions together during consultations, notably decision aids. There is evidence from conditions other than hypertension that shared decision making can lead to more appropriate care, 8 reduce overtreatment, 9 improve health outcomes 10 and may reduce health-care treatment costs. 11 A systematic review of interventions to support the adoption of shared decision making by health professionals 12 was unable to draw conclusions about the most effective interventions for supporting health professionals' adoption of shared decision making, due to the paucity of evidence. None of the studies in that review focused on people with hypertension. A recent systematic review of randomized controlled trials, including one study that did focus on hypertension management, found that people exposed to decision aids feel more knowledgeable, clearer about their values and may make choices more in line with their values. 8 Hypertension affected 31% of the world's adult population in 2010 13 ; it increases the risk of cardiovascular conditions such as strokes and heart attacks and is the leading preventable cause of premature death worldwide. 14 Observational studies show a progressive rise in cardiovascular risk as systolic blood pressure rises above 115 mmHg. 15 Hypertension is diagnosed when a person's blood pressure (BP) exceeds a threshold, typically 140/90 mmHg. 16 Management is characterized by monitoring of blood pressure alongside other cardiovascular risk factors and the use of lifestyle measures, usually combined with antihypertensive drug treatment to reduce blood pressure below treatment thresholds. Optimal treatment targets vary and are the subject of vigorous debate. 17 Treatment is typically lifelong with adjustment and, often, intensification of antihypertensive treatment over time. Hypertension control is frequently considered suboptimal, that is it fails to reach specified treatment targets. 18 Achieving blood pressure control has the potential for improved outcomes and cost savings at the population level. 19,20 However, from an individual patient's perspective, the potential benefits are less certain. Options to reduce blood pressure include a choice of medications and lifestyle changes. Potential benefit will vary with an individual's overall cardiovascular risk, and potential disbenefits include medication side-effects and the burden of having to take daily medication. Patients making decisions about antihypertensive drug treatment require discussions about treatment to be personalized in order for the decisions to make sense to them. 21 Shared decision making for hypertension has the potential to address this challenge, yet it is unclear how best to support shared decision making for hypertension, and the effect of shared decision making on outcomes is unknown. Given the high prevalence of hypertension and its impact on cardiovascular risk, shared decision making for hypertension may have profound impacts at both individual and public health levels.

| Objective
The main objective of this study was to determine the effectiveness of interventions, including but not limited to decision aids, to support shared decision making in hypertension. A second objective was to describe the outcomes that have been used to evaluate interventions supporting shared decision making for hypertension.

| ME THODS
The protocol for this systematic review was registered on PROSPERO (CRD42015014143). 22

| Search strategy
We used search strategies incorporating subject heading and text word searches focused on shared decision making and hypertension (see Appendix 1 for MEDLINE searches). The search was developed in MEDLINE and adapted for subsequent databases. We searched MEDLINE, EMBASE, CINAHL, Web of Science, PsycINFO and the Cochrane library from their inception to September 2017. We identified further potentially relevant articles from forward (via Google Scholar) and backward (reference list of paper) citation tracking of included studies, applying the same inclusion criteria.

| Eligibility criteria
Following Cochrane Effective Practice and Organisation of Care (EPOC) guidance, 23 we included randomized controlled trials (RCTs), nonrandomized controlled trials, controlled before-after studies and interrupted time series studies. We included published studies reporting on interventions supporting shared decision making for adults (>18) with hypertension. Eligible comparator interventions were control or any other interventions. Interventions could be delivered in any health-care setting, either before or during consultations with any health-care professionals. We included studies describing interventions that supported shared decision making by supporting one of the two following processes of shared decision making: supporting a patient's consideration of their options in relation to a health-care choice; or supporting a patient to consider their values and preferences in relation to a health-care choice. We included studies in which only a proportion of participants were hypertensive, if study outcomes were reported separately for the hypertensive group. We excluded studies reporting interventions unrelated to health-care decisions, for example, purely educational interventions that aimed to increase hypertension knowledge without reference to health-care choices faced by the patient. We excluded interventions that aimed to increase the involvement of patients in their own care generally, but not in health-care decisions specifically. To develop an understanding of how interventions to support shared decision making were evaluated, we included studies regardless of the outcomes assessed. No date or language restrictions were applied.

| Reference management and study selection
EndNote X7.7 and Access 2013 were used to manage the references.
Duplicates were removed from the EndNote file. Titles and abstracts, and subsequently full texts, were screened independently by two reviewers (RJ, BP or AH); disagreements were resolved by discussion with reference to a third reviewer where necessary (KT, GF and HC).
If there was insufficient detail on potentially relevant studies within the report abstract, it was screened as full text. Reasons for exclusions of full-text reports were documented.
We scrutinized the text and reference lists of relevant systematic reviews for potentially eligible studies. Conference abstracts and relevant study protocols were followed up either by contact with the author where possible or by searching for subsequent publications in PubMed.

| Data extraction and risk of bias
Data were extracted into a custom-designed table which had been previously piloted by one reviewer (RJ). All data were extracted by one reviewer and checked by a second. Data were extracted on study type, setting, participants, interventions, controls, type of decision supported and outcome measures. Our prespecified primary outcome was any measure of shared decision making. Consistent with our objective of documenting what outcomes have been used to evaluate interventions to support shared decision making, all other reported outcomes were extracted as secondary outcomes. We extracted estimated effect sizes with 95% confidence intervals for each outcome assessed, using odds ratios for binary variables and mean differences for continuous variables. Risk of bias was assessed independently by two reviewers using the Cochrane EPOC risk of bias tool 23 ; disagreements were resolved by discussion. Risk of bias in some domains varied with the type of outcome measure; risk of bias grouped by type of outcome is presented in Figure 2.

| Data synthesis
For data pooling, where outcomes were assessed using different measures, we planned to calculate standardized mean differences (SMDs). Meta-analysis was planned if there were at least three studies with comparable interventions and outcomes at low risk of bias.
If meta-analysis was appropriate, we planned to assess heterogeneity amongst studies using the I 2 statistic. Analyses were carried out using Stata version 14. 1. 24 As meta-analysis did not prove possible, we present a narrative synthesis of the studies. 25 The included studies are summarized in the text, in a table of study characteristics and in a risk of bias summary table. The outcomes reported by included studies, grouped by type of intervention, are reported in Figure 3. Outcomes reported by at least three of the included studies are compared across the studies in forest plots and in the text.

| RE SULTS
Searches were run in December 2014 and updated in September 2017. A total of 6424 unique articles were screened, of which 91 full-text articles were assessed, and 11 reports of 6 studies were included in the review ( Figure 1). 26-34

| Included studies
Eleven papers were published from six studies, all based in primary care (Table 1). Five studies reported randomized controlled trials, 26,27,29,30 of which two were cluster randomized. 29,30 The remaining study was a nonrandomized controlled study. 28

| Profile of patients
The range of mean age of study participants was 58.5-64.5 years, and the range of female participants was 32.5%-66.0%. In five studies, all recruited patients had hypertension. [26][27][28][29]34 In the remaining study, 30 only a proportion of participants were hypertensive, although all had raised cardiovascular risk. Only results relating to the hypertensive patients within this study are included in this review. 30

| Profile of interventions
The interventions were heterogeneous in their content and often multicomponent ( Table 1). Intervention components included training interventions for clinicians, 28,29,34 coaching for patients, decision aids and written materials for patients. 26, 34 Tinsel and colleagues 29,32 evaluated a shared decision-making training programme for general practitioners, to understand whether it increased patients' perceived participation, optimized blood pressure values, enhanced patient knowledge of hypertension and improved adherence. Deinzer and colleagues 28,35 evaluated a shared decision-making training intervention for general practitioners, 28 testing the hypothesis that shared decision making would lead to more effective lowering of hypertension. In the study by Cooper and colleagues, 34,36 a communication skill training intervention for physicians and a coaching intervention for patients were evaluated, separately and in combination with each other, for their impact on patient-physician communication and care processes, patient adherence to medication and lifestyle recommendations, and blood pressure control. In two studies, the main intervention component was a decision aid. 27,30 In the first of these, Denig and colleagues 30,33 set out to support interactions between patients and health-care providers using a decision aid focusing on shared goal setting and decision making for patients with diabetes considering their treatment options, including for management of hypertension. In the second study, Montgomery and colleagues 27,31 set out to evaluate the effect of decision analysis as an aid to patient decision making for newly diagnosed hypertension on decision quality, treatment choices, clinical outcomes, and treatment and consulting behaviour. 31 In the final study, 26 the intervention was a leaflet distributed to patients with hypertension and hypothesized to lead to greater involvement of patients in their health-care choices, with the potential for improving on blood pressure control.
In four studies, 26,28,29,34 interventions supported the involvement of patients with established hypertension, without specifying which treatment choices were being supported. In one study, 27 the decision supported was whether to commence antihypertensives in newly diagnosed hypertensive patients. The intervention was an approximately hour-long session of decision analysis which took place outside of the clinical encounter. One intervention aimed to support shared decision making in consultations where multiple treatment options to lower cardiovascular risk were being considered, including decisions about commencing antihypertensive therapy. 30

| Risk of bias
Risk of bias assessment is reported in Figures 2 and 3. One nonrandomized controlled study was included in the review and was at high risk of bias for most domains. Two of the RCTs were at uncertain or high risk of bias for the majority of domains. 26,34 Three RCTs were at low risk for most domains. 27,29,30 However, the two RCTs reporting shared decision making were at uncertain risk of bias for this outcome because of the impossibility of blinding for, as well as the subjectivity of, this outcome.

| Outcomes
The included studies assessed a range of outcome measures.
Outcomes reported, by intervention type and risk of bias, are shown in Figure 3. Four studies reported a measure of shared decision making. 27,28,32,34 Clinical outcomes reported were as follows: blood pressure (five studies), 26  smoking status (one study) 31 and intention to start treatment (one study). 31 Anxiety was the only psychological outcome reported (one study). 27 Cognitive outcomes were hypertension knowledge (four studies) [26][27][28][29] and intention to start treatment (one study). 27 Only one study reported a measure of health-care use. 30 Other outcomes included health-related quality of life (one study) 28 and clinician communication (one study). 33 Here, we discuss our primary outcome (shared decision making), and the outcomes reported in at least three of the included studies (blood pressure, hypertension knowledge and medication adherence). The decision to limit our discussion to the most commonly reported one was a post hoc decision, as detailed reporting of all of the outcomes reported was not practical. All outcomes are reported in Table 2. None of the outcomes met our prespecified criteria for meta-analysis of at least three studies with comparable interventions and outcomes at low risk of bias; therefore, we did not pool data for any outcome.

| Primary outcome: shared decision makingrisk of bias (Figure 2) and results (Table 2 and Figure 4)
The four studies measuring shared decision making 27-29 used different patient self-report measures; measures are described in Table 2.

| Secondary outcomes-risk of bias (Figure 2) and results (Table 2 and Figures 5 and 6)
Five studies evaluated the effect of the intervention on blood pressure 26,28,29,31 (Table 2); two studies were at low risk of bias, and three were at high risk of bias, for this outcome 26,28 ( Figure 2). Blood pressure was measured at different time points (range 6 months to 3 years). Four of the five studies (two at low risk of bias) report that there was no difference between blood pressure in the intervention and control groups; in the fifth study, intervention and control were not formally compared. The mean difference in change from baseline after 1 year in three studies with useable data is shown in Figure 5.
Hypertension knowledge was assessed in four studies, 26,28,29,31 at different time points (range 14 days to 18 months), using different scales in each study. Results were conflicting: two studies reported that the intervention increased hypertension knowledge, 26,27 and two studies 28,29 reported that there was no statistical evidence of a difference in hypertension knowledge between intervention and control. Two studies reported comparable scales at similar time points; SMDs for these studies are reported in Figure 6.
Adherence was assessed in three studies 27,29,34 at different time points (range 6 months to 3 years) and using different patient self-report measures; two studies 29,31,34 were at uncertain risk, and one study 43 29 and in the second study, decision analysis reduced decisional conflict at 14 days. 27 Of two further studies at high risk of bias, 34 only one provided useable data 28  Our definition of shared decision making builds on previous research in this area; our two core components of shared decision making were the elements that appear most frequently in conceptual definitions of shared decision making 44 and are central to the most frequently cited model of decision making. 45 To avoid missing eligible studies, we were inclusive at the title and abstract screening stage, where intervention descriptions were often sparse. No language restrictions were used, and screening was carried out in duplicate. Uncertainties about inclusion were discussed within a multidisciplinary team of GPs/health service researchers and social scientists to ensure validity of selection.

| D ISCUSS I ON
Using a narrative synthesis approach, we have been able to apply tools systematically resulting in a robust summary of the available studies, as well as highlighting where the evidence base is limited. To our knowledge, this is the first study to review interventions to support shared decision making for hypertension.
Limitations of this review include the small number of eligible studies, many of which were at uncertain or high risk of bias. Research in conditions other than hypertension has suggested that shared decision making has the potential to improve outcomes, 10 increase appropriateness of care, 8 reduce overtreatment 9 and reduce treatment costs. 11 Given the limitations of the studies within the review, the effects of shared decision making in hypertension remain uncertain, and none of these potential benefits can be confirmed. The interventions in several of the included studies [28][29][30]33 aimed to change the behaviour of clinicians in order to facilitate shared decision making. The challenges, for health professionals, in implementing shared decision making have been well described and include time constraints and the perceived lack of applicability of shared decision making to the particular clinical situation. 7 A recent review focussing on studies measuring shared decision making and patient outcomes found that shared decision making, when perceived to be happening by patients, tended to result in improved affective-cognitive outcomes, but that evidence was lacking for patient behavioural and health outcomes. 46 Consistent with this review, we found that all of our included studies that measured shared decision making used a patient-reported measure.
In the care of people with hypertension, there is a potential conflict between the aim of ensuring shared decision making occurs, and the aim of optimizing blood pressure control. Several of the included studies aimed to do both. The effect of shared decision making on clinical outcomes is important because, should it be implemented widely, it has the potential to impact on public health outcomes. 47 For example, should the consequence of shared decision making be that fewer people take antihypertensive medication, this will increase cardiovascular events. However, the rationale for shared decision making is not to improve compliance with clinical or public health priorities, and it is to achieve a decision which is congruent with the patient's personal priorities, values and beliefs. This potential conflict was not discussed in the study reports.

| CON CLUS ION
Hypertension is a long-term condition in which patients and their clinicians frequently face choices about starting or modifying hypertension treatment. Shared decision making is increasingly advocated for all health-care choices, including those taken in the care of long-term conditions. 5 Decision aids continue to proliferate, 48

ACK N OWLED G EM ENTS
We would like to thank Cath Borwick, University of Bristol Librarian, for her help with developing literature searches, and Andrea Waldeker and Anna King for their assistance with reference screening. We would also like to thank the following for their comments on the systematic review protocol and on earlier drafts of the manuscript: Dr Richard Lehman, Professor Nicky Britten, Professor Elizabeth Murray, Professor Glyn Elwyn.

CO N FLI C T O F I NTE R E S T
All authors declare that they have no conflict of interests.