Self-formulated conditional plans for changing health behaviour among healthcare consumers and health professionals

  • Protocol
  • Intervention

Authors


Abstract

This is the protocol for a review and there is no abstract. The objectives are as follows:

Our primary objective is to assess the effects of self-formulated conditional plans to change the health-related behaviours of health consumers and healthcare professionals. Our secondary objective is to explore if the effects of self-formulated conditional plans are modified by: (i) the form of the plan (e.g., implementation intention, action plan, coping plan); (ii) fidelity to plan development (did the individual participate in its development as planned?); and, (iii) complexity of the targeted behaviour (e.g., number of actions required; regularity of the behaviour).

Background

The problem

Health consumers’ behaviours are a major determinant of their health (Fisher 2011; Swann 2010) and are the largest source of variance in their health outcomes (Ryan 2008; Schroeder 2007), meaning an individual's behaviour largely determines his or her health. Similarly, healthcare professionals’ behaviours are major determinants of whether they deliver evidence-based care and advice to health consumers (Eccles 2005). Yet, behaviours that reflect the use of best evidence by health consumers and healthcare professionals are sub-optimal. For example, only one-third of consumers experiencing a cardiac event take up cardiac rehabilitation (British Heart Foundation 2009) and up to one quarter of individuals in commercial weight management programmes drop out (Tsai 2005). This can lead to serious health consequences for consumers. For example, cardiac rehabilitation participation is consistently associated with reduced all-cause and cardiovascular mortality in people who have experienced myocardial infarction, percutaneous coronary intervention, or coronary artery bypass graft surgery (Hammill 2009; Heran 2011; Suaya 2009; Witt 2004).

Comparable findings are seen with healthcare professionals, which in turn also impact on health outcomes for consumers. Results of studies in the USA and the Netherlands (Grol 2001; Schuster 1998) suggest that about 30% to 40% of consumers do not receive care according to present scientific evidence, and that about 20% to 25% of care provided is not needed or is potentially harmful. McGlynn and colleagues (McGlynn 2003) observed that patients in the USA received only 55% of recommended care, and that quality varied by medical condition, ranging from 79% of recommended care for senile cataract to 11% of recommended care for alcohol dependence. Similar findings are reported globally (Grol 2001). Our understanding of how to improve health consumers’ and healthcare professionals’ use of best evidence is also incomplete. This has resulted in considerable interest in knowledge translation (KT). KT is concerned with the assessment, review, and utilisation of scientific research. As a result, strategies to improve KT often require an increase in knowledge by health consumers and healthcare professionals, and that they change some aspect of their behaviour.

Description of the intervention

'Conditional plans' comprise statements that specify specific cues and behaviour(s) that one will undertake to realise a (health) intention or goal. The behaviour will occur in response to these cues; meaning it is the cues rather than the process of deliberation that results in the behaviour. With conditional plans, individuals form a mental image of some combination of when (time), where (cues), and how (response) they will execute the intended behaviour (Gollwitzer 1993). Conditional plans can be either provided to the person (e.g., in the form of a guideline, e.g., ‘if (when) X occurs, I will do Y (behaviour)’/ If you develop hypoglycaemia (the sugar in your blood drops to less than 4 mmol/L), take a fast-acting carbohydrate such as 15g of glucose or sucrose in the form of tablet or solution) or can be self-formulated (i.e., the individual participates in developing the plan). Self-formulated plans can be formed entirely by the individual completing the health behavior or in conjunction with another person. In the case of conditional plans focused on patients, the other person may be a healthcare professional (e.g., physician, nurse, pharmacist, allied health professional, etc) or a researcher. In the case of conditional plans focused on healthcare professionals, i.e., plans aimed at changing the professional's own behaviour rather than the behaviour of consumers, the other person will often be a researcher.

We will illustrate this with two real examples. With respect to self-formulated conditional plans focused on consumers, Jackson and colleagues (Jackson 2006) asked consumers to follow a set of instructions to form a conditional plan (in the form of an if...then…statement) for each daily dose of antibiotic they were prescribed, to determine whether self-formulated conditional plans increased adherence to short-term antibiotics. With respect to self-formulated conditional plans focused on healthcare professionals, Casper (Casper 2008) conducted a study with mental health practitioners, who were instructed (by a researcher) to create a conditional plan that identified the best time and place ('when' and 'where') to use a psychiatric care directive ('how').

In this review, we will focus on self-formulated conditional plans because there is evidence to suggest that involving consumers in their own health care leads to the development and provision of interventions that are both more responsive to their needs and offer better outcomes of care (Liddell 2008). There is also less risk of interventions being designed inappropriately if health consumers are involved in the development and planning phase (Elwell 2013), and interventions are more likely to be provided in a way that they want (Elwell 2013). Furthermore, organisations that support healthcare professionals and others in providing high quality care to health consumers (such as The National Institute for Health and Clinical Excellence (NICE) in the UK) recommend both consumer and healthcare professional input for best outcomes when planning behaviour change interventions (Elwell 2013; NICE 2007). Finally, consumers have a legal right to, and should participate in, the development of plans related to their health.

There are a variety of forms that conditional plans can take. The most commonly reported form is implementation intentions that formulate conditional plans as specific 'if/then' statements; for example, if X (I have breakfast), then I will do Y (take my insulin). Intervention strategies using implementation intentions have been successful in changing a variety of health consumer behaviours (e.g., improved dietary practices (Adriaanse 2011), increased physical activity (Belanger-Gravel 2011), increased attendance at cervical cancer screening (Sheeran 2000), and performing breast self-examination (Sheeran 1999)). They have also been used successfully, albeit less often, with healthcare professionals (e.g., the use of a psychiatric care directive by mental health practitioners (Casper 2008), and improving hand hygiene compliance by professional nurses (Erasmus 2010)).

Other complementary conditional plan approaches to implementation intentions include action plans and coping plans.  Action plans link goal-directed responses to situational cues by specifying when, where, and how to act in accordance with one’s goal intention; they are not however presented as if/then statements like implementation intentions (Kwasnicka 2013; Sniehotta 2006). Coping plans are designed specifically to overcome the barriers that are perceived to prevent intentions and action plans from working (Kwasnicka 2013; Sniehotta 2006).

How the intervention might work

Most people intend to change an aspect of their health-related behaviour at some point in time, whether it is to quit smoking, or obtain screening for a disease by health consumers, or to follow established best practice guidelines by healthcare professionals. Systematic reviews illustrate that such intentions reliably predict behaviour. For example, in an overview of meta-analyses of the intention-behaviour relationship across 422 studies involving 82,107 individuals, Sheeran (Sheeran 2002) showed that intentions accounted for 28% of the variance in health consumers’ behaviours, on average. Similarly, in a recent review of healthcare professionals, Godin and colleagues (Godin 2008) demonstrated that across 12 studies involving 1754 professionals, intentions accounted for 22% of the variance in professional behaviour. Further support for the importance of intentions to enacting behavior can be seen in a recent review showing that an actual change in intention can lead to a change in health behaviour. In a meta-analysis of 47 studies of consumer health intention-behavior relations, a medium-to-large change in intention (d = 0.66) led to a small-to-medium change in behavior (d = 0.36) (Webb 2006). Thus, for some people health intentions can be and are translated into successful behaviour change. However, for many others, intentions never result in the targeted behaviour (Ogden 2007). A meta-analysis of 51 studies (involving 8166 health consumers) showed almost half (47%) of individuals with positive intentions to engage in health-related behaviours did not realise these intentions  (i.e., they did not change their behaviour to meet the intention) (Sheeran 2002). This is known as the ‘intention-behaviour gap’ (Orbell 1998; Sheeran 2005).

Why is it so difficult for people to enact their intentions?

Several methodological and measurement reasons (e.g., Sutton 1998) as well as substantive explanations (e.g., Sheeran 2005) for the intention-behaviour gap are suggested in the health psychology literature. Common measurement problems include, for example: violation of the Principle of Compatibility (i.e., not measuring intention and behaviour at the same level of specificity or generality); lack of scale correspondence (i.e., use of different magnitudes, frequencies, or response formats for the assessment of intention and behavior); and unequal number of response categories when measuring intention and behavior ( Sutton 1998). While these and other methodological issues are important, in this review, we focus on substantive reasons for the intention-behaviour gap.

In performing behaviours, individuals are theorised to pass through two phases: a motivational (pre-decisional) phase which includes obtaining the required knowledge to change, followed by a volitional (post-decisional) phase (Gollwitzer 1993; Heckhausen 1991). The motivational phase encompasses an individual’s orientation toward engaging in the behaviour and culminates in the formation of a behavioural or goal intention. Intentions are the instructions that people give themselves to perform particular behaviours or to achieve certain goals (Triandis 1980) and are characteristically measured by items of the form ‘I intend to do X’ (Sheeran 2005). Intentions represent the culmination of the decision-making process; that is, they signal the end of deliberation about behaviour and capture the standard of performance that a person has set for him or herself, their commitment to the performance, and the amount of time and effort that will be expected during action (Ajzen 1991; Gollwitzer 1990; Sheeran 2005; Webb 2005). The second phase of behaviour (the volitional phase) refers to the actual performance of the targeted behaviour (i.e., realisation of one’s intention).  

Two central processes are suggested to underlie why people sometimes do not progress past the motivational (intention) phase into the volitional (behaviour) phase: (1) intention elaboration and (2) intention activation (Sheeran 2005). Intention elaboration refers to specifying, in sufficient detail, the particular actions and contextual opportunities that will permit realisation of an intention (Sheeran 2005). This requires identifying both the means (actions) and the context (internal or external cues) that will permit intention realisation. In the absence of such intention elaboration, people are likely to miss the opportunity to act or even know how to act if the opportunity arises (Sheeran 2005). Intention activation refers to the extent to which contextual demands alter the salience, direction or intensity of a focal intention relative to other intentions (Sheeran 2005). For example, for most health behavior intentions, an individual is likely to have multiple (and often conflicting) goals pertaining to that exact time. This can result in prospective memory failure (where one forgets to do the behaviour) and/or goal reprioritisation (where the intention fails to attract sufficient activation to permit its realisation, and is postponed/abandoned) (Sheeran 2005). In addition, several other factors or barriers to change can come into play, including contextual factors such as structural or environmental factors and the availability of specific resources required to carry out the behaviour, e.g., lack of access to affordable opportunities for physical activity, domestic responsibilities, or lack of information or resources.

Recently, Flottorp and colleagues (Flottorp 2013) conducted a systematic review of frameworks of determinants of professional practice behaviour followed by a consensus process, resulting in the development of a checklist with 57 potential behavioural determinants grouped in 7 domains: guideline factors, individual health professional factors, patient factors, professional interactions, incentives and resources, capacity for organisational change, and social, political, and legal factors. These studies have led to an interest in KT interventions that can bridge the intention-behaviour gap in both consumer and health professional behaviours. Conditional plans are one strategy for this gap and thus increase the likelihood of individuals enacting targeted health behaviours (Gollwitzer 1993; Gollwitzer 1996; Gollwitzer 1998; Gollwitzer 2005).

Why it is important to do this review

Despite the widespread use of conditional plans to change health-related behaviours, there remains a paucity of synthesised evidence summarising their effects with health consumers and healthcare professionals broadly (i.e., across a range of behaviours). A few syntheses exist of the effects of a single form of conditional plans, including self-formulated plans, that target a specific condition or narrowly-defined set of behaviours; for example, action plans for chronic obstructive pulmonary disease (Walters 2010), implementation intentions for physical activity (Belanger-Gravel 2011), and implementation intentions for healthy eating (Adriaanse 2011)). In these reviews, the form of conditional plan was reported as effective overall, providing support for the intervention in a specific context.

Two additional reviews have examined a single form of conditional plan across multiple behaviours. Gollwitzer and Sheeran (Gollwitzer 2006) reviewed the evidence for implementation intentions across different behaviours, and reported a positive effect overall (effect size = 0.59, 95% CI 0.52 to 0.67). This review however has several limitations and requires updating. First, a narrow search strategy was used: (i) a single health database (MEDLINE) was searched; (ii) the search was restricted to the period 1993 and 2003, so is now 10 years old; and, (iii) search terms were limited to ‘implementation intentions’ and ‘plans’. Second, few of the included studies focused on health consumers or healthcare professionals (n = 23). Third, no assessment of methodological quality (risk of bias) was reported. Fourth, forms of conditional plans other than implementation intentions were excluded. Fifth, assessment of ‘self-formulated’ plans was not a focus.

Recently, Kwasnicka and colleagues (Kwasnicka 2013) conducted a review of the effects of "promoting individuals to form coping plans" to encourage health-related behaviour change. This review also has limitations: the search was limited to three databases (MEDLINE, EMBASE, PsycINFO), one form of conditional plans (coping plans) was included, and it only targeted health consumers. Eleven studies were included; the authors concluded that coping plans "appear" to be effective when individuals are "supported" in the process of their development (Kwasnicka 2013), offering additional support for using self-formulated conditional plans. Based on our preliminary scoping of the literature (Appendix 1), a large unsynthesised body of randomised controlled trial (RCTs) evidence currently exists that examines the effects of different forms of conditional plans across a variety of behaviours with both health consumers and healthcare professionals. We will conduct a synthesis of this literature.

Objectives

Our primary objective is to assess the effects of self-formulated conditional plans to change the health-related behaviours of health consumers and healthcare professionals. Our secondary objective is to explore if the effects of self-formulated conditional plans are modified by: (i) the form of the plan (e.g., implementation intention, action plan, coping plan); (ii) fidelity to plan development (did the individual participate in its development as planned?); and, (iii) complexity of the targeted behaviour (e.g., number of actions required; regularity of the behaviour).

Methods

Criteria for considering studies for this review

Types of studies

We will include randomised controlled trials (RCTs), with randomisation at individual or cluster level. We will only include real-world studies, and will exclude laboratory studies. An example of a laboratory study would be university students randomised to make a conditional plan (to reduce the intake of high calorie products or control) following an educational session on nutrition; the intervention is carried out in a university psychology laboratory.

Types of participants

We will include studies that target:

  1. non-vulnerable adult health consumers (i.e., persons aged 18 years or older who: (a) do not have psychiatric, cognitive, or developmental disorders; (b) are not mentally incompetent; or (c) do not have similar histories which raise concern)

  2. healthcare professionals (a person who by education, training, certification, or licensure is qualified to and is engaged in providing health care).

Types of interventions

All forms of conditional plans that focus on changing a health-related behaviour, where the participant was involved in the formulation of the plan to change their own behaviour, will be considered. A conditional plan refers to a plan that specifies the precise behaviour(s) that will be undertaken in response to specific cues (e.g., some combination of if, when, where, how they will execute the intended behaviour) to realise an intention.

We will include all forms of conditional plans, including:

  1. implementation intentions – if/then plans that contain when/where, and how to act in accordance with an intention i.e., ‘if X or when X, I will do Y’ (Gollwitzer 1993);

  2. action plans- a plan that links goal-directed responses to situational cues by specifying when, where, and how to act in accordance with one’s intention (Sniehotta 2006); and

  3. coping plans- a plan that specifies how to deal with anticipated barriers that prevent intentions and action plans from being realised (Sniehotta 2005; Sniehotta 2006).

Conditional plans in which the individual was not involved in its formation (e.g., use of instructions/ guidelines) will be excluded; previous syntheses address this topic (e.g., (Grimshaw 2004; Thomas 1999)).

To address our objectives, three comparisons are planned:

  1. self-formulated conditional plans alone or as the core/essential feature of a multifaceted intervention compared with no intervention;

  2. self-formulated conditional plans alone or as the core/essential feature of a multifaceted intervention compared with usual care;

  3. self-formulated conditional plans alone or as the core/essential feature of a multifaceted intervention compared with other interventions.

Types of outcome measures

Primary outcomes

Our primary outcome is health-related behaviour change, i.e., did the conditional plan result in the targeted (goal) behaviour of the consumer or the health professional? With health consumers, this could mean getting screened for a disease (e.g., cancer). With healthcare professionals, it refers to changing their own behaviour (i.e., not the behaviour of their patients). Examples of such behaviour of health professionals could be prescribing behaviours (e.g., reducing the number of prescriptions written for antibiotics), developing healthcare directives, and improving hand hygiene. The actual behaviours will vary across individual studies. For behaviours that require multiple sequential actions, the primary outcome will be the final (targeted) behaviour.

We do not anticipate multiple primary outcomes in a single study; however, if this occurs, our primary outcome will be determined by ranking the intervention effect estimates of these outcomes and selecting the outcome with the median estimate (Brennan 2009). We will indicate in our findings how the primary outcome for each study was selected. We will extract data on all reported measures of behaviour change, both objective (e.g., screening rates) and subjective (e.g., self-report), but will rely on objective measures where they exist in our analyses.

We will also include as primary outcomes any reported adverse events.

Secondary outcomes

Secondary outcomes will be health impact outcomes (positive and/or negative) where available in published reports. For example, if the change in behaviour (such as smoking cessation) subsequently led to improved health (such as improved lung function) or a change in medical outcomes (for example, blood pressure, weight), improved lung function and the specific medical outcome will be secondary outcomes.

Additional secondary outcomes include, for both consumers and clinicians: affective measures (e.g., satisfaction with completing the plan, anxiety associated with completing the plan, quality of life), use of health services, and a clearer identification of what are important patient outcomes (in clinician-focused studies).

Search methods for identification of studies

CF (co-author), an experienced health sciences information specialist, developed the following preliminary search strategy.

Electronic searches

Preliminary electronic search strategies will be produced by CF and tested through an iterative process with the research team. Extensive sensitive electronic searches will be conducted using both controlled vocabulary and text word terms appropriate to each database search. There will be no time or language restriction; all databases will be searched from their inception to the present. We will search:

  • The Cochrane Central Register of Controlled Trials (CENTRAL,The Cochrane Library) (latest issue),

  • MEDLINE (OvidSP)

  • MEDLINE In Process (OvidSP),

  • EMBASE (OvidSP)

  • CINAHL (EBSCO)

  • PsycINFO (EBSCO)

  • ASSIA (ProQuest)

  • Science Citation Index (ISI)

  • Social Science Citation Index (ISI) and

  • the Cochrane Effective Practice and Organization of Care (EPOC) Review Group specialised register.

The proposed MEDLINE search strategy is detailed in Appendix 1 and will be adapted for use in other databases.

We will also search the following trials registers: ClinicalTrials.gov and the WHO portal.

Searching other resources

We will search key journal websites for online pre-publications, including:

  • Implementation Science,

  • Psychology and Health,

  • Social Science and Medicine,

  • British Journal of Health Psychology,

  • Health Psychology,

  • Journal of Applied Social Psychology,

  • Journal of Personality and Social Psychology;

  • Personality and Social Psychology Bulletin,

  • Personality and Social Psychology Review,

  • American Psychologist,

  • Behavioural Medicine,

  • Annals of Behavioural Medicine, and

  • Health Psychology Review.

In addition, we will contact authors of relevant papers and scrutinise reference lists of these papers for details of additional published and unpublished reports. We will also search key websites for grey literature, including: National Implementation Research Network; National Institutes of Health Office of Behavioral & Social Science Research: Behavioral Research Program; National Institutes of Health Cancer Control & Population Sciences; and the UK National Institute for Health Research Health Services and Delivery Research Programme.

Data collection and analysis

Selection of studies

We will assess the results of the literature search using a two-step process. First, two review authors will screen records according to pre-specified screening questions (Level 1 screening). All potentially-relevant records and those records that do not contain enough information to determine eligibility (e.g., no available abstract) will be retained. Second, two review authors will independently perform full-text screening and discrepancies will be resolved by consensus or a third party (Level 2 screening). All potentially-relevant papers excluded from the review at full-text stage will be listed as excluded studies, with reasons provided in the 'Characteristics of Excluded Studies' table.

We will describe any ongoing studies that are identified, detailing the primary author, research question(s), methods and outcome measures along with an approximation of the expected reporting date. We will collate and report details of duplicate publications, so that each study (rather than each report) is the unit of interest in the review. We will report the screening and selection process using the ‘Preferred Reporting Items for Systematic reviews and Meta-Analyses’ (PRISMA) format (Moher 2009) and in an adapted PRISMA flow chart. To facilitate study selection, the literature search results (citation, abstracts, and full text PDFs) will be uploaded to Distiller Systematic Review Software© (DSR); an internet-based software program. We will develop screening questions based on the inclusion/exclusion criteria for both screening levels; preliminary questions are in Appendix 2. Before conducting the formal screening, a calibration exercise will be undertaken to pilot and refine our screening questions.

Data extraction and management

Two review authors independently will extract and document data from each included study using a pilot-tested standardised data abstraction form in DSR to capture information on the descriptive and quantitative characteristics of each study. The data abstraction form will be based upon the Cochrane Consumers and Communication Group data extraction template; data will include, but not be limited to, the following:

  • Identification details of the study: authors, year of publication, country, setting, language, publication status, sources of funding, and study design (e.g. name of design, number of clusters or individuals, sample size, method of randomisation, blinding, and/or ‘control’ intervention).

  • Participant characteristics: type of participant (consumer or healthcare professional), profession/professional role, gender, age, literacy, education, health status, intention to engage in the health-related behaviour, etc.).

  • Intervention and control characteristics: form of conditional plan (e.g., implementation intention, action plan, coping plan; other 'forms' of plans will be extracted where they exist), purpose of the plan, content of the plan (planned and actually implemented), whether the plan that the individual forms is a new plan (to them) or a description of what they already do (in the case of behaviours that are not a new to the individual), fidelity to plan development (did the individual participate in its development as planned and what was their level of participation), delivery mechanism (planned and actually implemented, e.g., did the individual write or rehearse the plan where instructed to), enactment (did they make a plan), plan quality (did the plan have the recommended attributes), any support processes for the plan, time specification (when the plan was specified versus when the behaviour was to occur), unit of allocation, unit of analysis, study power, and presence and description of controls.

  • Outcomes. For our primary outcome (behaviour change) we will record: the targeted behaviour (i.e., the behaviour of focus in the study), results, methods for measuring the targeted behaviour, length of follow-up, and loss to follow-up data. We will extract data on complexity of the behaviour, for example: the number of behaviours required; the frequency of the behaviour (before and after the intervention was provided); frequency of opportunities for appropriately performing the behaviour; the extent to which complex judgements or skills were required to carry out the behaviour; whether other factors (e.g., barriers to organisational change) were required to be overcome for the behaviour change to occur; and, whether communication and/system change was also required in addition to individual change. We will also extract, where possible, whether the behaviour was new (i.e., the individual was not currently engaging in the behaviour) or was an existing behaviour. We will also extract data on adverse outcomes and health impact (the latter being secondary outcomes in the review) where available. For example, if the change in behaviour (e.g., smoking cessation) subsequently led to improved health (e.g., improved lung function), it will be recorded as a secondary outcome.

We will also extract data on equity in all included papers using the Equity Checklist for Systematic Reviews developed by the Campbell and Cochrane Equity Methods Group (Campbell and Cochrane Equity Methods Group 2011). Where necessary, additional data will be obtained through communications with the original trial authors. Disagreements in abstraction will be resolved through consensus and, if needed, consultation with a third review author. Prior to performing data abstraction, we will undertake a calibration exercise to pilot and refine our extraction form (Appendix 1). Published abstracts, where additional information cannot be obtained, will be listed under ‘Studies Awaiting Assessment’. One review author (JS) will enter the extracted data into the Cochrane Review Manager software, which will be checked by a second review author.

Assessment of risk of bias in included studies

Two independent review authors will assess the risk of bias (RoB) associated with each included study. A senior methodologist (co-author, JG) will oversee this task. We will assess and report on the RoB in accordance with the Cochrane Handbook (Higgins 2011a), using the Cochrane RoB tool. This tool evaluates seven domains: sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, missing outcome data, selective outcome reporting, and ‘other sources of bias’ (e.g., baseline imbalances on the targeted behaviour, contamination of intervention, etc) (Higgins 2011b). For cluster RCTs we will also assess the RoB associated with selective recruitment of participants. For each criterion listed above, we will describe the relevant information provided by the authors (in their report and through personal communication where necessary) and judge the criterion as being at a high, low or unclear RoB. Studies will be deemed to be at the highest risk of bias if they are scored as being at high or unclear risk of bias on either of the sequence generation or allocation concealment domains, with low risk of bias studies rated at low risk on both of these criteria.

For all included studies, two review authors will assess the RoB independently. Disagreements will be resolved through discussion and, if necessary, by consulting JG. We will report overall RoB at two levels: within studies (using the RoB tool criteria described above) and across studies (for our primary outcome of behaviour change). In determining overall RoB, we will consider magnitude and direction of the bias as well as whether the bias is likely to impact our findings. Further, consistent with recent reviews that contain only RCTs, any study rated at a high RoB on the random sequence generation item of the Cochrane RoB tool will be excluded from the review on the basis that it is not a truly-randomised trial.

Measures of treatment effect

We will present the results for all comparisons using a standard method of presentation where possible. For each study, data will be reported in natural units. Pre-intervention and post-intervention means or proportions will be reported for both the intervention and control groups, and the absolute change from baseline will be calculated with 95% confidence intervals (CIs). We will report the: median, inter-quartile ranges, and range of effect sizes across included studies.

We will express results of dichotomous outcomes (e.g., the proportion of individuals performing the behaviour) as adjusted risk differences (RD) and adjusted risk ratios (RR) with 95% CIs. RD refers to the difference in adherence to the behaviour after the intervention minus the difference before the intervention. A positive RD indicates that adherence to the behaviour improved more in the intervention group than in the control group. RR refers to the ratio of the relative probability of adherence to the behaviour after the intervention over the relative probability before the intervention. A RR greater than 1 indicates that adherence improved more in the intervention group than in the control group.

We will express the results of continuous outcomes (e.g., the number of times a consumer took a medication) as standardised mean differences (SMDs) with 95% CIs. SMDs will be calculated by dividing the difference in mean scores between the intervention and comparison group in each study by an estimate of the pooled standard deviation. This results in a ‘scale free’ effect estimate for each study, which can be pooled across studies regardless of the original scale of measurement used in the studies (Laird 1990). If a study uses two methods to measure the same outcome (e.g., a dichotomous and a continuous measure), we will include both measures in our analysis.

Unit of analysis issues

We will correct any unit of analysis errors encountered where possible by obtaining relevant information from the trial authors or other means. For example, if effects of clustering have not been taken into account, we will adjust the standard deviations using the intra-class coefficient (ICC) where it is reported (Higgins 2011a). If the ICC is not known, we will extrapolate an estimate of it from similar cluster RCTs (Higgins 2011a) or by contacting the trial author. Where sufficient data are present to adjust for the error, we will recalculate results using the appropriate unit of analysis (Giguere 2012). Where sufficient data are not present to adjust for the error, we will report effect estimates and annotate them as a ‘unit of analysis error’.

Dealing with missing data

Where outcome data are unclear, incompletely reported, or not reported, we will attempt to contact the trial authors to obtain the data. We will conduct an intention-to-treat (ITT) analysis, meaning we will include in our analyses all participants randomised to each group and analyse data according to this group allocation irrespective of whether or not they received, or complied with, the intervention (Higgins 2011a). We will report on the levels of loss to follow-up and assess this as a source of potential bias as part of our RoB assessments.

Assessment of heterogeneity

Where studies are considered similar enough (based on consideration of populations, interventions, and the behaviour targeted) to allow pooling of data using meta-analysis, we will assess the degree of heterogeneity by visual examination of the scatter of effect estimates on forest plots and by using the Chi2 statistical test to assess whether observed differences across the studies might be due to chance. If the intervention effects are more different from each other than one would expect due to chance, we will assume statistical heterogeneity. We will quantify statistical heterogeneity using the I2 statistic, which describes the proportion of variability in the summary estimate that is due to heterogeneity rather than chance (Higgins 2003). We will also report and assess the 95% CIs for I2 in line with recent recommendations by Ioannidis and colleagues (Ioannidis 2007). I2 values of 50% or more indicate a substantial level of heterogeneity (Higgins 2011a).

Where we detect substantial clinical, methodological or statistical heterogeneity across included studies we will not report pooled results from meta-analysis but will instead use a narrative approach to data synthesis. In this event we will attempt to explore possible clinical or methodological reasons for this variation by grouping studies that are similar in terms of populations, intervention features, methodological features, or the targeted behaviour to explore differences in intervention effects.

Assessment of reporting biases

The tendency for negative or inconclusive results to remain unpublished may impact the findings of a systematic review. Therefore, we will investigate publication bias where feasible (i.e. when pooled effects are generated) graphically using funnel plots (Sterne 2011) and statistically with tests for funnel plot asymmetry where more than 10 studies reporting the same outcome are available (Higgins 2011a). Where asymmetry is found, we will also explore other plausible reasons, including heterogeneity or RoB. If a sufficient number of studies is not available to allow construction of funnel plots to examine publication bias, we will assess for the bias qualitatively based on: (i) the characteristics of the included studies (e.g., if only small studies that indicate positive findings are identified for inclusion), and (ii) information that we obtain from contacting experts and authors or studies that suggests there are relevant unpublished studies.

Data synthesis

Before conducting a synthesis of the evidence we will tabulate the study findings (Petticrew 2006); separate tables will be developed for health consumers and healthcare professionals. Study characteristics will be summarised qualitatively in summary tables and will include information pertaining to: sample size and participant demographics, setting, targeted behaviours, conditional plan intervention characteristics, comparator characteristics, study RoB, and funding source. This will assist us to identify possible heterogeneity (clinical, statistical, and methodological) across the included studies.

We will then conduct separate syntheses for health consumers and healthcare professionals. Where appropriate and possible (depending on the quantity of studies, RoB, and statistical and clinical homogeneity of the data), we will compute pooled estimates of intervention effects using standard random-effects meta-analytic methods (Arends 2008; DerSimonian 1986). We will pool RDs and RRs (for dichotomous outcomes) and SMDs (for continuous outcomes) for each comparison. The decision to meta-analyse or not will be made by consensus of all review authors based on whether the participants, intervention, comparison, and outcomes are sufficiently similar to ensure a clinically meaningful result. We will use a random-effects meta-analysis for combining data, as we anticipate that there may be natural heterogeneity between trials attributable to different forms of conditional plans, targeted behaviours, and groups of participants. For continuous variables we will use the inverse variance method for meta-analysis while for dichotomous variables we will use the Mantel-Haenszel method (Higgins 2011a). If cluster RCTs are included in the final data sets we will use the generic inverse variance method for meta-analysis (Higgins 2011a). Analyses will be conducted in Review Manager.

For comparisons where meta-analysis is not possible due to significant heterogeneity, we will perform a qualitative narrative synthesis and present the summarised results in a table. If a narrative synthesis is necessary we will group the data based on the category that best explores the heterogeneity of studies and makes most sense to the reader (i.e. by interventions (implementation intention, coping plan, action plan), populations or outcomes (targeted behaviour)). Within each category we will present the data in tables and narratively summarise the results.

Subgroup analysis and investigation of heterogeneity

We will explore heterogeneity by conducting subgroup analyses. If data are available we plan to carry out the subgroup analyses separately for both health consumers and healthcare professionals, by:

  1. form of the plan (e.g., implementation intention, action plan, coping plan);

  2. fidelity to plan formulation (e.g., did the individual participate in its development as planned);

  3. level of participation in developing the plan;

  4. enactment (did they make a plan);

  5. plan quality (did the plan have the recommended attributes);

  6. time specification (when the plan was specified versus when the behaviour was to occur);

  7. identified barriers to behaviour change; and

  8. complexity of the targeted behaviour.

For the eighth subgroup analysis (complexity of the targeted behaviour) we will develop a category schema based on our review of the papers. Currently, we anticipate the following characteristics to come into play: the number of behaviours required; the frequency of the behaviour (before and after the intervention is provided); frequency of opportunities for appropriately performing the behaviour; the extent to which complex judgements or skills are required to carry out the behaviour; whether other factors (e.g., organisational change barriers) are required to be overcome for the behaviour change to occur; and, whether communication and/system change is also required in addition to individual change. We will also conduct a subgroup analysis comparing high risk versus lower/average risk patients, if there are sufficient data.

We will limit all subgroup analyses to when three or more trials contribute data (Saeterdal 2012). We will examine differences between subgroups by visual inspection of the subgroups’ confidence intervals (CIs); non-overlapping CIs will suggest a statistically-significant difference in treatment effect between groups (Borenstein 2008).

Sensitivity analysis

We will consider how sensitive our results are to the way we did our analysis. We will adopt recommendations from Higgins 2011a including:

  1. we will repeat our analysis excluding unpublished studies (if there are any);

  2. if there are one or more very large studies, we will repeat our analysis excluding them to see if they dominated the results; and

  3. we will investigate how our pooled intervention effect (if we perform meta-analysis) is affected by the inclusion of RCTs at high and unclear risks of bias by performing two meta-analyses: one for RCTs at low RoB and another for all RCTs.

'Summary of findings' table

We will present the results for the major comparisons in the review in standard Cochrane ‘Summary of findings’ tables, for each of the major primary outcomes, including harms, as outlined in Types of outcome measures. The ‘Summary of findings’ tables will include the following six elements:

  1. a list of all primary outcomes,

  2. a measure of the typical burden of these outcomes (e.g. illustrative risk, or illustrative mean),

  3. absolute and relative magnitude of effect,

  4. numbers of participants and studies addressing these outcomes,

  5. a grade of the overall quality of the body of evidence for each primary outcome (which may vary by outcome), and

  6. additional comments.

We will provide a source and rationale for each assumed risk cited in the table(s). The ‘Summary of Findings’ tables will be prepared using GRADE profiler software (GRADEpro 2008).

Strength or quality of evidence determines the level of confidence authors place in the estimate of effect (i.e., their judgment that the evidence reflects the true effect). It also reflects the likelihood that future research might affect the level of confidence in the estimate or change that estimate (Owens 2010). We will grade the quality of evidence in this systematic review using GRADE profiler software (GRADEpro 2008). For each outcome, we will assess and report on the quality of the evidence using the GRADE approach (Balshem 2011), which involves consideration of five categories: RoB, imprecision, inconsistency, indirectness, and publication bias. We will rate the evidence in one of four categories – high, moderate, low, or very low quality (Balshem 2011). If we are not able to pool results statistically (or not able to pool results statistically for all comparisons or outcomes), we present narrative 'Summary of findings' tables.

Consumer participation

Involvement of governmental and non-governmental organisations that represent a range of potential user groups will be an important part of this review. To date, we have actively engaged four consumers in the development of this protocol: Alison Paprica, Michael Hillmer, Marcello Tonelli, and Catherine Yu. Michael Hillmer and Alison Paprica hold leadership roles with governmental organisations.

Michael Hillmer is the Director of the System Policy and Strategy Branch of the Ontario Ministry of Health and Long-Term Care (MOHLTC) in Canada that sets the strategic directions for Ontario's health system, and supports them with legislation and policy and selects and manages portfolios of initiatives within the ministry to further health system goals. Alison Paprica is the Director of the Planning, Research and Analysis Branch of the Ontario MOHLTC (Canada) that is the focal point for policy-relevant research evidence at the Ministry, is committed to health services and population health research, and has extensive knowledge transfer and exchange activities to ensure that research findings are understood and used in health system policy development and planning.

Marcello Tonelli and Catherine Yu are clinicians (physicians) and also hold leadership roles with non-governmental organisations. Marcello Tonelli is the Chair of the Canadian Task Force on Preventive Health Care, which comprises a panel of experts that make recommendations on clinical preventive services for Canada’s 36,000 family physicians. He is also a member of the Canadian Society of Nephrology Scientific Committee (Kidney Foundation of Canada), and the Past President of the Canadian Society of Nephrology. Catherine Yu is the Chair of the Canadian Diabetes Association’s Clinical Practice Guidelines Dissemination and Implementation Committee, which develops recommendations for managing patients with diabetes.

All four consumers have been central to driving the formation of the review objectives as well as the development of this protocol. Using these established working relationships, we will continue to engage and utilise the expertise of these and other consumers throughout the research process to ensure the completion of a quality product with high potential for impact. In the peer review process we will also liaise with the Cochrane Consumers and Communication Review Group to seek external referees reflecting these interests.

Acknowledgements

We thank the editors of the Cochrane Consumers and Communication Review Group for their assistance and feedback in developing this protocol. We also thank Alison Paprica (Director of the Planning, Research and Analysis Branch of the Ontario MOHLTC (Canada)) and Michael Hillmer (Director of the System Policy and Strategy Branch of the Ontario MOHLTC (Canada)) for their input into the development of this protocol.

Appendices

Appendix 1. MEDLINE Search Strategy

1. intention/

2. cues/

3. (intention? or cue?).tw.

4. or/1-3

5. 4 and (plan or plans or plann*).tw.

6. 1 and implement*.tw.

7. (implement* adj2 (intention* or intend*)).tw.

8. ((implementation or action or coping or conditional or if then) adj (plan or plans or plann*)).tw.

9. (behavio?ral adj2 (plan or plans or plann*)).tw.

10. or/5-9

11. randomized controlled trial.pt.

12. controlled clinical trial.pt.

13. randomized.ab.

14. placebo.ab.

15. drug therapy.fs.

16. randomly.ab.

17. trial.ab.

18. groups.ab.

19. or/11-18

20. exp animals/ not humans.sh.

21. 19 not 20

22. 10 and 21

Appendix 2. Screening and Extraction Forms

Abstract screening form (Level 1)
1. Does this study target either health consumers or healthcare professionals?
a. Yes (include)
b. No (exclude)
c. Can’t tell (include)

2. Does this study examine use of a conditional plan to change a health-related behaviour?
a. Yes (include)
b. No (exclude)
c. Can’t tell (include)

Full text screening form (Level 2)
1. Is this study a randomised controlled trial in a real-world setting (i.e. not a laboratory setting)?
a. Yes
B. No

2. Does the study include the intervention of a self-formulated conditional plan*, focused on changing a health-related behaviour?
a. Yes
b. No

3. Is there a comparator group that either receives no intervention, or other intervention (not including a condition plan)
a. Yes
b. No

4. Does this study report the outcome of interest (i.e. a change in health-related behaviour)?
a. Yes
b. No

5. How are the outcomes reported?
a. Objective (e.g. screening rates)
b. Self-reported (e.g. survey)

* A conditional plan is defined as any plan in which an individual specifies the precise behaviour(s) they will undertake in response to specific cues (e.g., any combination of if, when, where and how they will execute the intended behaviour).
Self-formulated means the individual who is targeted by the plan was involved in some capacity in the development of the plan

Data extraction form

Study characteristics:

  1. Authors

  2. Year of publication

  3. Country

  4. Setting

  5. Language

  6. Publication status

  7. Sources of funding

  8. Study design (e.g. type of design, number of clusters or individuals, sample size, method of anatomizations, blinding, and/or “control” interventions)

Participant characteristics:

  1. Type of participant (health consumer of healthcare professional)

  2. Profession/professional role

  3. Gender

  4. Age

  5. Literacy

  6. Education

  7. Health status

  8. Intention to engage in the health-related behaviour

Intervention and control characteristics:

  1. Form of conditional plan (e.g. implementation intention, action plan, coping plan, etc. - other 'forms' of plans will be extracted where they exist)

  2. Purpose of plan

  3. Content of plan (planned and actually implemented)

  4. Whether the plan that the individual forms is a new plan (to them) or a description of what they already do (in the case of behaviours that are not a new to the individual

  5. Fidelity to plan development (did the participant participate in development of the plan as proposed and what was their level of participation)

  6. Delivery mechanism (planned and actually implemented)

  7. Enactment (did they make a plan)

  8. Plan quality (did the plan have the recommended attributes)

  9. Any support processes for the plan

  10. Time specification (when the plan was specified versus when the behaviour was to occur)

  11. Unit of allocation

  12. Unit of analysis

  13. Study power

  14. Presence and description of control

Outcome: behaviour change

  1. Targeted behaviour

  2. Results

  3. Methods for assessing/measuring the targeted behaviour

  4. Length of follow up

  5. Loss to follow up data

  6. Complexity of the targeting behaviour, for example:

  • number of behaviours required

  • frequency of behaviour (before and after the intervention is provided)

  • frequency of opportunities for appropriately performing the behaviour

  • the extent to which complex judgements or skills are required to carry out the behaviour

  • whether other factors (e.g., organisational change) are required for the behaviour change to occur

  • whether communication and system change is also required in addition to individual change

  • whether the behaviour is new (to the individual) or an existing behaviour (conducted prior by the individual)

Contributions of authors

All review authors have contributed to the production of the protocol. Janet Squires (JS) drafted the protocol, with input and amendments provided by all members of the review team.

Declarations of interest

No conflicts of interest are reported.

Ancillary