Goal setting and strategies to enhance goal pursuit for adults with acquired disability participating in rehabilitation
Abstract
Background
Goal setting is considered a key component of rehabilitation for adults with acquired disability, yet there is little consensus regarding the best strategies for undertaking goal setting and in which clinical contexts. It has also been unclear what effect, if any, goal setting has on health outcomes after rehabilitation.
Objectives
To assess the effects of goal setting and strategies to enhance the pursuit of goals (i.e. how goals and progress towards goals are communicated, used, or shared) on improving health outcomes in adults with acquired disability participating in rehabilitation.
Search methods
We searched CENTRAL, MEDLINE, EMBASE, four other databases and three trials registers to December 2013, together with reference checking, citation searching and contact with study authors to identify additional studies. We did not impose any language or date restrictions.
Selection criteria
Randomised controlled trials (RCTs), cluster‐RCTs and quasi‐RCTs evaluating the effects of goal setting or strategies to enhance goal pursuit in the context of adult rehabilitation for acquired disability.
Data collection and analysis
Two authors independently reviewed search results for inclusion. Grey literature searches were conducted and reviewed by a single author. Two authors independently extracted data and assessed risk of bias for included studies. We contacted study authors for additional information.
Main results
We included 39 studies (27 RCTs, 6 cluster‐RCTs, and 6 quasi‐RCTs) involving 2846 participants in total. Studies ranged widely regarding clinical context and participants' primary health conditions. The most common health conditions included musculoskeletal disorders, brain injury, chronic pain, mental health conditions, and cardiovascular disease.
Eighteen studies compared goal setting, with or without strategies to enhance goal pursuit, to no goal setting. These studies provide very low quality evidence that including any type of goal setting in the practice of adult rehabilitation is better than no goal setting for health‐related quality of life or self‐reported emotional status (8 studies; 446 participants; standardised mean difference (SMD) 0.53, 95% confidence interval (CI) 0.17 to 0.88, indicative of a moderate effect size) and self‐efficacy (3 studies; 108 participants; SMD 1.07, 95% CI 0.64 to 1.49, indicative of a moderate to large effect size). The evidence is inconclusive regarding whether goal setting results in improvements in social participation or activity levels, body structure or function, or levels of patient engagement in the rehabilitation process. Insufficient data are available to determine whether or not goal setting is associated with more or fewer adverse events compared to no goal setting.
Fourteen studies compared structured goal setting approaches, with or without strategies to enhance goal pursuit, to 'usual care' that may have involved some goal setting but where no structured approach was followed. These studies provide very low quality evidence that more structured goal setting results in higher patient self‐efficacy (2 studies; 134 participants; SMD 0.37, 95% CI 0.02 to 0.71, indicative of a small effect size) and low quality evidence for greater satisfaction with service delivery (5 studies; 309 participants; SMD 0.33, 95% CI 0.10 to 0.56, indicative of a small effect size). The evidence was inconclusive regarding whether more structured goal setting approaches result in higher health‐related quality of life or self‐reported emotional status, social participation, activity levels, or improvements in body structure or function. Three studies in this group reported on adverse events (death, re‐hospitalisation, or worsening symptoms), but insufficient data are available to determine whether structured goal setting is associated with more or fewer adverse events than usual care.
A moderate degree of heterogeneity was observed in outcomes across all studies, but an insufficient number of studies was available to permit subgroup analysis to explore the reasons for this heterogeneity. The review also considers studies which investigate the effects of different approaches to enhancing goal pursuit, and studies which investigate different structured goal setting approaches. It also reports on secondary outcomes including goal attainment and healthcare utilisation.
Authors' conclusions
There is some very low quality evidence that goal setting may improve some outcomes for adults receiving rehabilitation for acquired disability. The best of this evidence appears to favour positive effects for psychosocial outcomes (i.e. health‐related quality of life, emotional status, and self‐efficacy) rather than physical ones. Due to study limitations, there is considerable uncertainty regarding these effects however, and further research is highly likely to change reported estimates of effect.
PICOs
Plain language summary
Goal setting for adults receiving clinical rehabilitation for disability
Background
Goal setting is considered a key part of clinical rehabilitation for adults with disability, such as in rehabilitation following brain injuries, heart or lung diseases, mental health illnesses, or for injuries or illnesses involving bones and muscles. Health professionals use goals to provide targets for themselves and their clients to work towards. In this review we summarise studies that have investigated what effect, if any, goal setting activities have on achieving good health outcomes following rehabilitation.
Results
This review found 39 studies published before December 2013, involving a total of 2846 participants receiving rehabilitation in a variety of countries and clinical situations. The studies used a wide range of different approaches to goal setting and tested the effectiveness of these approaches in a number of different ways. Overall these studies provide very low quality evidence that goal setting helps patients achieve a higher quality of life or sense of well‐being and a higher belief in their own ability to achieve goals that they choose to pursue. There is currently no consistent evidence that goal setting improves people's functional abilities after rehabilitation or how hard they try with therapeutic interventions during rehabilitation.
Insufficient information exists to say whether goal setting increases or reduces the risk of adverse events (such as death or re‐hospitalisation) for people involved in rehabilitation. Because of the variety of approaches to studying goal setting in rehabilitation and because of limitations in the design of many studies completed to date, it is very possible that future studies could change the conclusions of this review. We also need more research to improve our understanding of how components of the goal setting process (such as how difficult goals are, how goals of therapy should be selected and prioritised, how goals are used in clinical practice, and how feedback on progress towards goals should be provided) contribute or do not contribute to better health outcomes.
Authors' conclusions
Summary of findings
| Goal setting with or without strategies to enhance goal pursuit compared to no goal setting for adults with acquired disabilityparticipating in rehabilitation | |||||
| Patient or population: adults with acquired disability participating in rehabilitation | |||||
| Outcomes | Illustrative comparative risks* (95% CI) | No of Participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | ||||
| No goal setting | Goal setting (with or without strategiesto enhance goal pursuit) | ||||
| Health‐related quality of life or self‐reported emotional status | The mean Physical Component Summary Scores on the Short Form‐36 for the control group was | The mean Physical Component Summary Scores on the Short Form‐36 for the intervention group was | 446 | ⊕⊝⊝⊝ | Higher scores indicate better outcomes. Scores estimated using a SMD of 0.54 (95% CI 0.17 to 0.88), indicative of an effect size that may range from small to large.Two additional studies with 142 participants however, reported no means or SD, but indicated that goal setting may lead to little to no difference in health‐related quality of life or self‐reported emotional status |
| Participation | See comment | See comment | 254 | ⊕⊝⊝⊝ | Outcomes unable to be pooled due to lack of reporting of data and lack of similarities in the types of measures used. We are uncertain whether goal setting improves participation‐level outcomes |
| Activity | The mean Barthel Index score for the control group was | The mean Barthel Index score for the intervention groups was | 223 | ⊕⊕⊝⊝ | Higher scores indicate better outcomes. Scores estimated using a SMD of 0.04 (95% CI ‐0.22 to 0.31). This evidence suggests that goal setting may not improve activity‐level outcomes |
| Body structure and body function | See comment | See comment | 235 | ⊕⊝⊝⊝ | Unable to pool outcomes due to lack of similarities in the types of measures used. We are uncertain whether goal setting improves outcomes at the level of body structure and body function |
| Engagement in rehabilitation (motivation, involvement and adherence) | The mean number of hours worked on a 26‐week support work placement programme for the control groups was | The intervention groups worked | 369 | ⊕⊝⊝⊝ | Higher scores indicate better engagement. Scores estimated using a SMD of 0.30 (95% CI ‐0.07 to 0.66). One additional study with 27 participants reported no means or SD but indicated that goal setting may lead to little to no difference in engagement in rehabilitation. One further study with 367 participants measured medication regime adherence as a dichotomous variable, and reported that the odds for the goal setting group adhering was 1.13 times higher (95% CI 1.08 to 1.19) than that of the no goal setting group. Overall, we are uncertain whether goal setting improves engagement in rehabilitation |
| Self‐efficacy | The mean Task Self‐efficacy score for the control groups was | The mean self‐efficacy in the intervention groups was | 108 | ⊕⊝⊝⊝ | Higher scores indicate better self‐efficacy. Scores estimated using a SMD of 1.07 (95% CI 0.64 to 1.49), indicative of a moderate to large effect size. One additional study with 88 participants reported no means or SD, but suggested that goal setting after rehabilitation may lead to little to no difference in self‐efficacy |
| *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). | |||||
| GRADE Working Group grades of evidence | |||||
| 1 The Physical Component Summary Score on the Short Form‐36 was used for this illustrative comparative risk as this was deemed to be the most common, most general measure of quality of life used in the studies included in the meta‐analysis for this outcome. The data on assumed risk for the Physical Component Summary Score on the Short Form‐36 was taken from control group data in the study that used this measure (Harwood 2012). | |||||
| Structured goal setting with or without strategies to enhance goal pursuit compared to 'usual care' that involved some goal setting but where no structured approach was followed for adults with acquired disabilityparticipating in rehabilitation | |||||
| Patient or population: adults with acquired disability participating in rehabilitation | |||||
| Outcomes | Illustrative comparative risks* (95% CI) | No of Participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | ||||
| 'Usual care' | Structured goal setting (with or without strategiesto enhance goal pursuit) | ||||
| Health‐related quality of life or self‐reported emotional status | The mean Mental Component Summary Scores on the Short Form‐36 for the control group was | The mean Mental Component Summary Scores on the Short Form‐36 for the intervention group was | 441 | ⊕⊕⊝⊝ | Higher scores indicate better outcomes. Scores estimated using a SMD of 0.18 (95% CI ‐0.19 to 0.56). One additional quasi‐RCT with 201 participants reported no means or SD, but indicated that usual care may lead to higher quality of life than structured goal setting. One further study with 122 participants reported that participants in the structured goal setting group were more likely to report being more satisfied or much more satisfied with their daily life compared to participants in the usual care group 3 months post intervention (RR 1.44, 95% CI 1.09 to 1.88), but not 2 years later (RR 1.27, 95% CI 0.94 to 1.70). Overall, this evidence suggests that structured goal setting in rehabilitation may result in little to no improvement in health‐related quality of life or self‐reported emotional status |
| Participation | See comment | See comment | 201 | ⊕⊝⊝⊝ | One quasi‐RCT reported no means or SDs for this outcome, but did not suggest that structured goal setting improves participation‐level outcomes. We are uncertain whether structured goal setting improves participation‐level outcomes |
| Activity | The mean Functional Indepdence Measure score in the control groups was | The mean Functional Independence Measure score in the intervention groups was | 277 | ⊕⊕⊝⊝ | Higher scores indicate better outcomes. Scores estimated using a SMD of 0.17 (95% CI ‐0.15 to 0.49). This evidence suggests that structured goal setting in rehabilitation may not improve activity‐level outcomes. One additional quasi‐RCT (201 participants) measured functional independence and reported no means or SD, and two further studies (118 participants) measured activity levels as ordinal data, but overall these studies also indicated that structured goal setting in rehabilitation may not improve activity‐level outcomes |
| Body structure and body function | See comment | See comment | 229 | ⊕⊝⊝⊝ | Unable to pool outcomes due to lack of similarities in the types of measures used. We are uncertain whether structured goal setting improves outcomes at the level of body structure and body function |
| Engagement in rehabilitation | See comment | See comment | 32 | ⊕⊝⊝⊝ | One study reported data on patient motivation in rehabilitation. A small difference in favour of structured goal setting in comparison to usual care was reported in terms of patient‐rated motivation (MD 1.40 on a 10‐point scale of self‐reported motivation, 95% CI 0.43 to 2.37) but not for therapist‐rated score of motivation (MD 0.48 on an 8‐point scale of therapist‐rated patient motivation, 95% CI ‐0.41 to 1.37) |
| Self‐efficacy | The mean self‐efficacy in the control groups was | The mean self‐efficacy in the intervention groups was | 134 | ⊕⊝⊝⊝ | Higher scores indicate better self‐efficacy. Scores estimated using a SMD of 0.37 (95% CI 0.02 to 0.71), indicative of an effect size that may range from small to large |
| *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). | |||||
| GRADE Working Group grades of evidence | |||||
| 1 The Mental Component Summary Score on the Short Form‐36 was used for this illustrative comparative risk as this was deemed to be the most common, most general measure of quality of life used in the studies included in the meta‐analysis for this outcome. The data on assumed risk for the Mental Component Summary Score on the Short Form‐36 was taken from control group data in the study that used this measure (Parsons 2012). | |||||
Background
Goal setting is considered an essential part of clinical rehabilitation. It has been described as a core practice within rehabilitation (Wade 2009), a requirement for effective interdisciplinary teamwork (Schut 1994), and an activity that specifically characterises both rehabilitation services and those who provide them (Barnes 2000; Scobbie 2009; Wade 1998). In clinical practice there has been growing emphasis on the need for interventions with patients to be goal oriented. Goal terminology is becoming integral to discussions of guidelines, policies and professional requirements at both regional and international levels (e.g. Duncan 2005; Evans 2001; Randall 2000; RCP 2003; RCP 2004; Rothstein 2003).
Some authors have suggested that evidence for the effectiveness of goal setting in improving patient outcomes has already been firmly established, and that this evidence can now direct how goal setting in rehabilitation should be implemented (Black 2010; Marsland 2010; Wilson 2008). However, a systematic review of randomised controlled trials (RCTs) concluded that the evidence regarding any generalisable effect of goal setting on patient outcomes following rehabilitation was inconsistent at best, and greatly limited by the quality of studies published at the time (Levack 2006a). Given that this review is now over nine years old, there is a need to update this work.
Description of the condition
This review focuses on the application of goal setting in the context of rehabilitation for adults with acquired disability. The term 'disability' is defined according to the World Health Organization's (WHO) International Classification of Functioning, Disability and Health (ICF) as an 'umbrella term for impairments, activity limitations or participation restrictions' (WHO 2001a, p.3) that result from interactions between a person (with a health condition) and that person's contextual factors (environmental factors and personal factors). For the purposes of this review, the term 'acquired disability' is used to refer more specifically to disability that arises during a person's adult life (i.e. after 16 years of age) following an accident, illness or development of a health condition. This term therefore excludes disability associated with health conditions arising prenatally or in childhood.
Description of the intervention
Reviews of literature on goal setting in rehabilitation are complicated by a number of factors, one of which is the difficulty that exists in describing what might (or might not) constitute 'goal setting' in a rehabilitation context. The terms 'goals', 'goal setting' and 'goal planning' have been used to refer to many different constructs with little current consensus around key terminology (Levack 2006b; Playford 2009). A range of different approaches to goal setting has been described in the literature, with various similarities and differences in the recommended process and content of each. These include (but are not limited to):
-
Goal Attainment Scaling (GAS) (Kiresuk 1968; Turner‐Stokes 2009);
-
goal setting based on the Canadian Occupational Performance Measure (COPM) (Pendleton 2005; Phipps 2007; Trombly 2002; Wressle 2002; Wressle 2003);
-
'SMART' goal planning (Barnes 2000; Bovend'Eerdt 2009; Mastos 2007; McLellan 1997; Monaghan 2005; Schut 1994);
-
'RUMBA' goal planning (Barnett 1999);
-
Self‐Identified Goal Assessment (Melville 2002);
-
Goal Management Training (Levine 2000);
-
approaches to goal planning from the Wolfson Neurorehabilitation Centre (McMillan 1999)
-
contractually‐organised goal setting (Powell 2002);
-
Collaborative Goal Technology (Clarke 2006);
-
goal setting as part of the Progressive Goal Attainment Programme (Sullivan 2006);
-
patient‐centred functional goal planning (Randall 2000); and
-
goal setting based on the Patient Goal Priority Questionnaire or Patient Goal Priority List (Asenlöf 2009).
Note: 'SMART' and 'RUMBA' are not abbreviations, but mnemonic acronyms for key components of goal setting, promoted by various authors. Interpretations of these acronyms differ (McPherson 2014; Wade 2009). One interpretation of the 'SMART' acronym is that it stands for Specific, Measurable, Achievable, Relevant, and Time‐limited goals (Barnes 2000). Similarly, it is suggested that 'RUMBA' refers to Relevant, Understandable, Measurable, Behavioural, and Achievable goals (Barnett 1999).
While these different approaches to goal setting frequently include common features, such as having measurable goals, or patient involvement in goal selection, few such features are universal to all recommended approaches. Indeed, all approaches to goal setting in rehabilitation differ from one another across a number of variables, including:
-
the group intended to use the approach (i.e. for use by a single, specific profession or for use by an interprofessional team);
-
the intended patient population for the approach;
-
the process by which goals are selected (e.g. who is involved; how goals are identified and prioritised);
-
the recommended characteristics of the actual goals set (i.e. how goals are written; whether they need to be phrased in a certain way);
-
the recommended content of goals set (i.e. what is considered an acceptable topic for a goal; whether goals need to be set at a particular level of the ICF);
-
the way the goals are subsequently used in clinical environments (e.g. the way goals are used in team meetings or meetings with patients; how feedback on progress towards goals is presented and used in clinical interactions); and
-
the intended purpose(s) of setting and having goals.
Even for individually‐named approaches to goal setting, opinions can differ in terms of how each approach should be implemented. For instance, multiple variations on the original GAS approach (Kiresuk 1968) exist, such as: involving greater patient participation in goal selection (Cytrynbaum 1979; LaFerriere 1978; Malec 1999; Turner‐Stokes 2009); having the treating therapist rather than an independent third party select and re‐evaluate the GAS goals (Cytrynbaum 1979; Turner‐Stokes 2009; Willer 1976); using a different number of 'levels' of goal achievement and a different scoring system than was originally proposed (LaFerriere 1978; Turner‐Stokes 2010; Willer 1976), or using standardised rather than individualised wording to indicate the extent of goal achievement (Turner‐Stokes 2009). Similarly, there is no one agreed 'SMART' approach to goal setting; the 'SMART' acronym has been interpreted to refer to a range of goal‐related concepts, and there is no consensus regarding the 'correct' interpretation of this approach (McPherson 2014; Wade 2009).
Goal setting is also often presented as a core component of a whole programme of intervention (e.g. Stuifbergen 2003). However, a systematic review of research into the effectiveness of goal setting needs to be able to separate out the independent effects of goal setting from those of other variables associated with these programmes of intervention (e.g. the amount of therapeutic activity, amount of additional education and information, or other behavioural interventions that are not related to the setting of rehabilitation goals). For more information on the history of goal setting and its application in rehabilitation please refer to Levack 2014a.
Definition of 'rehabilitation goal'
Within the field of psychology there is an enormous body of literature describing and analysing goal constructs from many perspectives. In this context, the term 'goals' has been defined as 'internal representations of desired states, where states are broadly construed as outcomes, events, or processes' (Austin 1996, p.338). This definition allows for goals that are consciously set as well as goals which are not; goals for individuals as well as goals for whole organisations or populations of people; biological goals (such as to change one's body temperature or reproduce); complex cognitive or aesthetic goals (such as to live a moral life or achieve a career objective); goals that relate to a moment in time and goals that relate to a lifespan. From this perspective, all human behaviour is goal directed.
In the context of rehabilitation however, the term 'goal' is generally used to mean something much more specific, and more explicitly linked to clinical work. For the purpose of this review we use the term 'rehabilitation goal' to refer to the concept of a 'goal' set for the purposes of clinical work in rehabilitation, in order to make a clear distinction between this type of goal and colloquial use of the term 'goal' or broader definitions of 'goals' from psychology.
One proposed definition of the term 'rehabilitation goal' has been 'a future state that is desired and/or expected. The state might refer to relative changes or to an absolute achievement. It might refer to matters affecting the patient, the patient's environment, the family or any other party. It is a generic term with no implications about time frame or level' (Wade 1998, p.273). Other authors, focusing on describing an approach to goal setting intended for a particular patient population or for use by one professional group, have been more specific in their definition of goals for rehabilitation. For example, Randall 2000 defined a 'functional goal' within the context of physical therapy as 'the individually meaningful activities that a person cannot perform as a result of an injury, illness, or congenital or acquired condition, but wants to be able to accomplish as a result of physical therapy' (p.1198).
In contrast, many 'goals' in the psychological sense of the word are implicit (i.e. goals which are implied without being directly stated or even necessarily consciously set). For example, the act of reaching for a cup is a motor activity with an implicit goal. Asking a patient to reach for a cup versus reaching into mid‐air is an example of using implicit goals to influence behaviour (Trombly 1999). However, using such activities as a clinical intervention (e.g. for exercise therapy after a stroke) is not an example of 'goal setting' in rehabilitation in its usual sense. While (as stated above) all human behaviour is arguably goal directed and rehabilitation cannot therefore occur without having 'goals' of some kind, it is not true that all goals are 'rehabilitation goals' in the sense usually intended by rehabilitation teams.
Furthermore, the concept of a 'rehabilitation goal' usually refers to a relationship between an individual patient and an individual or group of health professionals (and/or others). This excludes goals set at an organisational level (e.g. in the case of health service management) or community level (e.g. in the case of public health policy) from the definition of 'goal setting' in a rehabilitation context. In other words, while goals such as 'to reduce the incidence of falls in hospital' may be an important key performance indicator for a particular rehabilitation service, these types of organisational goals are not what is usually being discussed in the literature on goal setting in rehabilitation.
Therefore, for the purpose of this review, we define 'rehabilitation goal' as: a desired future state to be achieved by a person with a disability as a result of rehabilitation activities. Rehabilitation goals are actively selected, intentionally created, have purpose and are shared (wherever possible) by the people participating in the activities and interventions designed to address the consequences of acquired disability.
Definition of 'goal setting'
From a literal perspective, the term 'goal setting' refers solely to the selection of goals. For the purposes of this review, we define 'goal setting' more broadly as: the establishment or negotiation of rehabilitation goals. Consistent with other clinical researchers publishing in this area (Wade 1998), we will consider 'goal setting' to be synonymous with 'goal planning'.
Definition of 'goal pursuit'
In addition to the establishment or negotiation of rehabilitation goals, there are a number of activities related to how rehabilitation goals are communicated, used or shared that are intended to enhance how effective or successful people are in working towards those goals. For the purposes of this review we will use the term 'goal pursuit' to refer to these additional goal‐related activities. These activities include: development of a plan or strategy to achieve stated rehabilitation goals, provision of explicit feedback (oral or written) on a person's progress towards their rehabilitation goals, and use of strategies to maintain or enhance commitment to set goals (such as peer discussion of progress toward an individual's rehabilitation goals, or use of posters and electronic diaries reminding people about their rehabilitation goals). As the behavioural effects of having a goal are often moderated by a number of factors (e.g. people's ability to develop a plan to reach their goal, their awareness of how their current abilities or performance compares with that required to achieve their goal, and their level of commitment to specific goals) it is important not to exclude these factors from a systematic review of the therapeutic effects of goal setting in rehabilitation contexts.
How the intervention might work
Goal setting has been attributed with multiple purposes (or functions). Levack 2006b presents a brief typology of purposes from the clinical literature, and Levack 2006c provides an overview of purposes attributed to goal setting by health professionals working in rehabilitation environments for people with acquired brain injury. These papers highlight a number of reasons why rehabilitation professionals might believe goal setting is important in clinical practice.
-
Goal setting might improve patient outcomes, by:
-
improving the patient's motivation to engage in therapeutic activities;
-
improving clinical teamwork (providing teams with shared direction; focusing collaborative interprofessional practice);
-
enhancing the working relationship between patients, families and health professional (e.g. through development of a shared language and shared understanding of a health condition and the rehabilitation process);
-
improving the patient's ability to self‐regulate desirable behaviour (e.g. by retraining self‐awareness or addressing goal neglect in patients with problems in those areas);
-
assisting patients (and their family) to adapt psychologically to the consequences of disability; or
-
enhancing specificity of training (e.g. focusing therapy for an individual on performance of a specific activity in a specific environment relevant to that individual's daily life).
-
-
Goal setting might enhance patient self‐determination (i.e. autonomy) – considered by some to be an important reason to undertake goal setting regardless of other outcomes achieved or not achieved in terms of health and functioning.
-
The degree of goal attainment might be a useful measure of health outcome.
-
Goal setting is a contractual or legislative requirement of service delivery.
While clinicians, patients or family members may have different opinions about the main reason for undertaking goal setting in rehabilitation, for this review the improvement of patient outcomes is of greatest interest. One important point here is that goal setting as an intervention for improving health outcomes for patients should be considered separately from goal setting for the purpose of outcome evaluation (where 'outcomes' are evaluated in terms of 'goal achievement'). In other words, goal setting as an intervention (i.e. as a way of engaging with people with acquired disability) may be effective in terms of achieving higher levels of improvement in a person's functional abilities (to pick just one type of outcome) without the specific goals of rehabilitation for that person necessarily being reached.
In terms of how goal setting might influence patient motivation or self‐regulation in clinical environments in order to achieve improvements in patient outcomes, a number of additional theories from psychology have been suggested as relevant to rehabilitation, including:
-
Bandura's Social Cognitive Theory;
-
Locke and Latham's Goal Setting Theory;
-
Schwarzer's Health Action Process Approach;
-
Aspin and Taylor's Proactive Coping Theory;
-
Leventhal's Self‐Regulation Model of Illness Behaviour; and
-
Carver and Scheiers' Control‐Process Model of Self‐Regulation.
An overview of these theories and their application to rehabilitation has been given elsewhere (Scobbie 2009; Siegert 2004; Siegert 2014a). Broadly speaking, these theories describe: how people use and respond to goals in order to monitor, alter or adapt their behaviour; how emotional responses to goals or progress toward goals influence future goal‐oriented behaviour; how perceptions of illness and perceptions of the effect of interventions influence goal‐oriented behaviour; and how the effects of goals can be moderated by various factors such as personal goal commitment, beliefs in one's ability to achieve a goal (self‐efficacy), task complexity, and the way goals are presented or worded.
In clinical practice, several other variables may influence the success of goal setting interventions. These include how meaningful the goals are to the individual patient (how committed patients are to the goals; how well they relate to their higher‐order life priorities) and how involved patients are in the selection of goals, factors that might reasonably be considered to contribute to the 'person‐centredness' of the goal‐setting approach (Wilson 2008). Also, how 'reasonable' or 'realistic' a goal is to achieve (Wilson 2008) and how involved family members and significant others are in the selection of goals have been considered important (Levack 2009; McMillan 1999; Visser‐Meily 2006; Wade 1999a). Lastly, some researchers and clinicians have proposed that goal setting is likely to be more successful when goals are set at the level of 'activity' and 'participation' than when goals are established to address impairments at the level of body structure and body function (Marsland 2010; Randall 2000). For example, a goal to be able to transfer independently from a wheelchair to a toilet (an activity‐level goal) or to return to paid employment (a participation‐level goal) would be considered, in general, more effective for improving patient outcomes than a goal to improve muscle strength of the quadriceps by 150% (a goal set at the level of body structure and body function).
Why it is important to do this review
There is extensive research from education (Boekaerts 2000; Pintrich 2000), industrial‐organisational psychology (Latham 2007; Locke 2002), cognitive psychology (Austin 1996; Custers 2010; Moskowitz 2009) and sport psychology (Burton 2010; Hall 2001; Wilson 2006) which has demonstrated the effect that goals can have on human behaviour. It seems reasonable to assume that goal setting could have a similar type of effect in populations of people participating in clinical rehabilitation. However, what this broad body of research has also demonstrated is that the effectiveness of goal setting and the mechanism by which goals achieve these effects can be highly dependent on context. For instance, it has been found that theories of goal setting from industrial‐organisational psychology are not as effective when applied to goal setting in the context of sport psychology, leading to the development of new theories of goal setting specific to the sport environment (Hall 2001).
As highlighted above, multiple approaches to goal setting in rehabilitation exist and several different mechanisms are suggested by which goals might affect patient outcomes. Furthermore, there is debate about the evidence for the effectiveness of goal setting for improving patient outcomes, and the most recent systematic review of this literature is now several years old (Levack 2006a). There is a need for a Cochrane review regarding the effects of goal setting or strategies to enhance goal pursuit to influence patient outcomes in rehabilitation for adults with acquired disability. This review is beneficial for determining whether the evidence shows that goal setting or strategies to enhance goal pursuit are effective interventions, as well as providing possible directions for future research into the use of rehabilitation goals in clinical work.
Objectives
To assess the effects of goal setting, and strategies to enhance goal pursuit, on health outcomes in adults with acquired disability participating in rehabilitation. To test the following comparisons:
-
a structured approach to goal setting, with or without strategies to enhance goal pursuit versus no goal setting;
-
a structured approach to goal setting, with or without strategies to enhance goal pursuit versus 'usual care' that may involve some goal setting but where no structured approach was followed;
-
interventions to enhance goal pursuit versus no interventions to enhance goal pursuit; and
-
one structured approach to goal setting and/or strategies to enhance goal pursuit versus another structured approach to goal setting and/or strategies to enhance goal pursuit.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs), cluster‐RCTs, or quasi‐RCTs (where allocation to study groups was by a method that was not truly random, such as alternation, assignment based on date of birth, case record number or date of presentation, or due to use of stratification or minimisation).
Types of participants
People receiving rehabilitation for disability acquired in adulthood (e.g. after 16 years of age).
For the purposes of this review 'disability' was defined according to the ICF as an 'umbrella term for impairments, activity limitations or participation restrictions' (WHO 2001a, p.3) that result from interactions between a person (with a health condition) and that person's contextual factors (environmental factors and personal factors). Thus, we excluded studies investigating the application of goal setting to health interventions for non‐disabled people (e.g. in public health or obstetric contexts). More specifically, this review included people with disability arising from injuries, illnesses or disorders, as categorised by the WHO (WHO 1992), involving:
-
the musculoskeletal system or connective tissue;
-
the skin or subcutaneous tissue;
-
the cardiac system (including the cerebrovascular system);
-
the respiratory system;
-
the nervous system;
-
the sensory system (e.g. eye, ear etc);
-
the endocrine, nutritional or metabolic system;
-
the genitourinary system; and
-
mental or behavioural function.
For the purposes of this review 'rehabilitation' was defined as 'a process aimed at enabling persons with disabilities to reach and maintain their optimum physical, sensory, intellectual, psychiatric and/or social functional levels, thus providing them with the tools to change their lives towards a higher level of independence. The rehabilitation process does not, however, involve initial medical care' (WHO 2001b, p.290). Thus, we excluded studies investigating the effects of goal‐directed decision‐making by medical staff in emergency or intensive care settings, or in the management of acute medical conditions such as sepsis.
Types of interventions
We included studies that investigated the effects of establishing and negotiating rehabilitation goals, with or without strategies to enhance goal pursuit. For the purposes of this review, the term 'rehabilitation goals' refers to an actively selected and desired future state to be achieved by a person with a disability as a result of rehabilitation activities.
We included studies that investigated:
-
a structured approach to goal setting with or without strategies to enhance goal pursuit in comparison to no goal setting; or
-
a structured approach to goal setting with or without strategies to enhance goal pursuit in comparison to 'usual care' that may involve some goal setting but where no structured approach was followed; or
-
interventions to enhance goal pursuit in comparison to no interventions to enhance goal pursuit; or
-
one structured approach to goal setting and/or strategies to enhance goal pursuit in comparison to another structured approach to goal setting and/or strategies to enhance goal pursuit.
For the purposes of this review, approaches to goal setting were considered to differ if they involved different methods for:
-
identification, negotiation, or selection of rehabilitation goals; or
-
documentation of rehabilitation goals; or
-
involvement of health professionals, patients, family members or other significant people in the selection of rehabilitation goals.
Approaches to enhancing goal pursuit were considered to differ if they involve different methods for:
-
developing a plan on how to attain rehabilitation goals;
-
providing feedback to patients on their performance towards rehabilitation goals; or
-
enhancing patient commitment to attain rehabilitation goals.
We excluded studies investigating approaches to goal setting as an intervention compared to some other intervention intended to influence human cognition or behaviour (e.g. priming for pain attention in the case of Stenstrom 1994). We excluded any study that did not adequately control for additional treatment variables separate to the goal setting intervention. Hence we excluded studies in which goal setting formed only part of a whole programme of rehabilitation, where the outcomes of the intervention could not be specifically attributed to goal setting or to components of the goal setting process (e.g. Glasgow 2000).
Types of outcome measures
We excluded studies investigating only the immediate effects of goal setting. Studies were categorised as investigating the immediate effects of goal setting if they involved implementation of goal setting and collection of data on the effects of goal setting (e.g. in terms of immediate improvements in effort or performance on a set task) during only one session for each study participant, carried out over the course of less than one day (e.g. Gauggel 2001).
We prioritised the following outcomes.
Primary outcomes
Secondary outcomes
-
Outcomes at the level of body structure and function as defined by the ICF (WHO 2001a).
-
Patient self‐belief and engagement in rehabilitation, e.g. adherence, patient motivation, self‐efficacy.
-
Individual goal attainment.
-
Evaluation of care, e.g. satisfaction with care.
-
Service delivery level, e.g. cost of care, length of stay.
-
Adverse outcomes, e.g. complications, morbidity, mortality, readmission rate.
N.B. Individual goal attainment was not included as a primary outcome measure in this review as achievement of individualised goals is in part based on changes in health status achieved by rehabilitation patients, and in part based on the level of difficulty of the individually‐selected goals. It is possible therefore for two people to achieve the same degree of functional recovery (or gain in other outcomes) following rehabilitation, but score differently on measures based on achievement of individualised goals. As there is scope for debating what such differences in individualised goal attainment mean, we chose to select individual goal attainment as a secondary outcome.
Search methods for identification of studies
Electronic searches
We searched the following electronic databases in September 2012, with an updated search conducted in January 2014 for articles published to the end of December 2013.
-
The Cochrane Central Register of Controlled Trials (CENTRAL; 2013, Issue 12).
-
MEDLINE (OvidSP) (1950 to December 2013).
-
EMBASE (OvidSP) (1988 to December 2013).
-
PsycINFO (OvidSP) (1967 to December 2013).
-
CINAHL (EBSCOhost) (1981 to December 2013).
-
AMED (OvidSP) (1985 to December 2013).
-
Proquest Dissertations and Theses database (1673 to December 2013).
Detailed search strategies are presented in Appendices 1 to 7. We did not impose any language restrictions.
We searched the ProQuest Dissertations and Theses database for grey literature. We also searched databases in the WHO Clinical Trial Search Portal (www.who.int/trialsearch), Australian New Zealand Clinical Trials Registry (http://www.anzctr.org.au/), and Current Controlled Trials (www.controlled‐trials.com) to identify ongoing or recently completed studies (Appendix 8).
Searching other resources
We contacted experts in the field and authors of included studies for advice as to other relevant studies. We also searched reference lists of relevant studies and personal collections of articles. We sought full research reports of any potentially eligible studies that were published as abstracts or conference proceedings only. We included studies only published as abstracts or conference proceedings in the review where sufficient information about the study methods and data could be extracted from the abstract, published proceedings, or poster presentation, supplemented by author communication, with all sources of information noted in the section in Included studies. Where sufficient information could not be uncovered on potentially eligible studies, we excluded these studies, giving lack of information as the reason for this in the section in Excluded studies.
Data collection and analysis
The data collection and analysis methods were described in the review protocol (Levack 2012).
Selection of studies
Two review authors (WL and RS) independently screened all search results (titles and abstracts) for possible inclusion, and those selected by either or both authors were subject to full‐text assessment. The two review authors then independently assessed the selected articles for inclusion. The same two review authors resolved differences in the first instance by discussion, and then by input from a third review author from the review team. The whole review team debated particularly difficult decisions regarding inclusion. Any studies thus excluded but considered near the boundary for possible inclusion have been reported in Characteristics of excluded studies with the reason for exclusion given. We also present relevant ongoing studies in Characteristics of ongoing studies.
Data extraction and management
We used a standard data extraction form adapted from the Cochrane Consumers and Communication Review Group's Data Extraction Template for all included studies. Two review authors (WL and RS) independently assessed the risk of bias in included studies and independently extracted data for each study. The two review authors resolved differences in the first instance by discussion, and then by input from a third author from the review team. Review authors were not blinded to the names of study authors, journals or institutions.
For included studies, we extracted data on the intervention aims, study aims, study design, methods used, characteristics of participants (e.g. age, gender, ethnicity, socioeconomic status, principal health condition, inclusion of people with multimorbidities), characteristics of the study setting (e.g. geographic location, specific clinical context, co‐interventions being provided alongside goal setting), characteristics of the approach(es) to goal setting and strategies to enhance goal pursuit under investigation, outcome measures used, and reported findings.
Data extracted to categorise the approach to goal setting or goal pursuit under investigation included (if specified in the study's method):
-
the name of the approach to goal setting (e.g. GAS, SMART, COPM);
-
the health professional(s) involved in goal setting;
-
the training of health professionals for their involvement in goal setting;
-
the level of patient and/or family involvement in goal selection (e.g. whether goals were: prescribed with no input from study participants; selected through discussion and negotiation with the patient; selected through discussion and negotiation with the patient and their family; or selected by the study participants with no involvement of other parties);
-
the type of communication used for making selected goals explicit (e.g. written, oral);
-
whether or not the intervention involved an explicit process for developing a plan to achieve the stated goal(s), and if so what this was;
-
whether or not goals for study participants were made public to others (e.g. other patients);
-
whether or not study participants were reminded about their goals during the course of rehabilitation;
-
whether or not study participants were provided with feedback on their progress towards goals during the course of rehabilitation;
-
whether or not there was the development of written 'contracts' with participants to pursue specified goals;
-
assessment of the participants' level of commitment to attain their goals;
-
the level of goal difficulty (and how this was specified or quantified by the researchers); and
-
the targeted level of functioning for specific goals (e.g. if goals were set at the level of body structure and body function, activity, or participation, as defined by the ICF, WHO 2001a).
The first author (WL) then entered the data into Review Manager (RevMan 2014), with another author (RS) checking the accuracy of data entry.
Assessment of risk of bias in included studies
We assessed and reported on the risk of bias of included studies in accordance with the guidelines of the Cochrane Consumers and Communication Review Group (Ryan 2011), which recommends the explicit reporting of the following individual 'Risk of bias' elements for RCTs: random sequence generation; allocation concealment; blinding (participants, personnel), blinding (outcomes assessment); incomplete outcome data (attrition bias; adequacy of intention‐to‐treat analysis); selective outcome reporting; other sources of bias (e.g. recruitment bias, baseline imbalance, loss of clusters, incorrect analysis, and comparability with RCT based on per person randomisation in the case of cluster‐RCTs; and suitability of cross‐over design, management of carry‐over effect, incorrect analysis, comparability of results with parallel‐group trials, treatment period effects, and randomisation of order of treatments in the case of cross‐over design RCTs).
We conducted 'Risk of bias' assessments in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), with risk of bias being rated as high risk, unclear risk or low risk for each element and for each study overall. We used the same criteria for assessment of risk of bias for quasi‐RCTs; these studies were rated as being at high risk of bias both for random sequence generation and allocation concealment.
In all cases, two authors (WL and RS) independently assessed the risk of bias in included studies. These two review authors resolved differences in the first instance by discussion, and then by input from a third author from the review team. We attempted to contact study authors for additional information about the included studies or for clarification of the study methods as required. In the case of studies where one of the study authors was also an author of this review (i.e. McPherson 2009 and Taylor 2012), another review author took the lead on the 'Risk of bias' assessment.
Measures of treatment effect
Three categories of primary outcomes were the focus of this review:
-
health‐related quality of life or self‐reported emotional status;
-
participation outcomes; and
-
activity outcomes, as defined by the ICF (WHO 2001a).
We adopted the approach used by Brennan 2009 and Horvat 2014 for selection and extraction of primary outcomes from included studies. We included any primary outcome identified by study authors that fell within the scope of the primary outcomes categories listed above. If multiple primary outcomes were identified within any category, we ranked the reported effect estimates for each of these outcomes and selected the outcome with the median effect estimate. If no primary outcome within our categories was specified, we adopted the following strategy. First we used any outcome within our categories specified in sample size calculations; then, if necessary, we ranked relevant intervention effect estimates, as reported, and selected the median effect estimate. If the number of outcomes was even (n), we included the outcome whose effect estimate was ranked n/2. We have reported in our results whether we used the primary outcome or the outcome with the median effect estimate. Where possible, we also verified whether the specified primary outcomes in included studies were consistent with those identified in trial protocols, trial registry entries or both.
We extracted the intervention effect estimate reported for all included outcomes (both primary and secondary outcomes) with the associated P value and 95% confidence intervals (CIs), and the method of statistical analyses used to calculate them. For continuous data, where outcomes were measured in a standard way across studies, we reported the mean difference (MD) and 95% CI. Where outcomes were measured using different scales (e.g. for quality of life) we calculated a standardised mean difference (SMD) and 95% CI. When calculating a SMD within each outcome category, we multiplied all mean values for reversed scored outcomes (where lower scores indicate a better outcome) by ‐1 to ensure that the direction of all scales (from better to worse outcomes) were consistent.
For dichotomous data, where outcomes were measured in a standard way we reported the risk ratio (RR) and 95% CI. For categorical outcomes (such as employment outcomes) we related the numbers reporting an outcome to the numbers at risk in each group to derive a RR. We dichotomised ordinal data (such as Likert scales for symptom improvement) and managed them as a categorical outcome. We treated GAS scores as ordinal rather than interval data, as recommended by Steenbeek 2007 and Tennant 2007, and we treated count data as continuous data.
Unit of analysis issues
The primary analysis was planned on the basis of per person randomisation. For all studies we considered the possibility of unit of analysis issues arising from the inclusion of cluster‐randomised designs, repeated measurements and studies with more than two treatment groups. When applicable, we dealt with unit of analysis issues by analysing the data according to recommendations in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). In cluster‐RCTs, we first sought to use effect estimates and standard errors that were adjusted for clustering, combining the studies using the generic inverse‐variance method. When analysis in a cluster‐RCT did not take account of clustering, then we attempted to approximate the cluster‐adjusted effect size and standard error based on available data if the unadjusted effect estimate, the number or size of clusters, and the intraclass correlations were provided. If the intraclass correlation coefficient could not be obtained then we endeavoured to use an estimate from similar studies. If none of these options were possible, we included the studies unadjusted for clustered in the analyses, but then tested the effect doing this by examining the results of analyses with these studies removed. In studies with repeat observations (collecting data using the same measures on participants at a number of different time points) we selected the longest follow‐up data from each study. If studies had more than two groups we combined all relevant experimental intervention groups of the study into a single group, and combined all relevant control intervention groups into a single control group.
Dealing with missing data
If data were missing from the relevant comparisons we attempted to contact the study authors to obtain the information. Where studies did not state that results were reported using an intention‐to‐treat analysis for primary outcomes, we contacted the study authors to request data to enable us to conduct such an analysis. If no response from authors was provided, we analysed results as reported.
Assessment of heterogeneity
Given the potential for clinical and methodological diversity in studies that might have been eligible for inclusion, it was important to consider heterogeneity in the data analysis. Clinical heterogeneity was determined before analysis of data by extracting and considering information on each study's patient populations, clinical contexts, approaches to goal setting, and outcome measures used.
We identified statistical heterogeneity in studies thought to be clinically and methodologically similar by visual inspection of forest plots and by using a standard Chi2 test and a significance level of alpha = 0.1, in view of the low power of such tests. We also examined heterogeneity with I2, where I2 values of 50% or more were deemed to indicate a substantial level of heterogeneity (Higgins 2003). We used a random‐effects model to assess heterogeneity as, prior to conducting the review, we had anticipated finding substantive differences in the patient populations, rehabilitation settings and approaches to the selection and use of goals in the included studies.
Assessment of reporting biases
We assessed the extent of publication bias through visual inspection of asymmetry and running the regression‐based method for a funnel plot in Review Manager 5 (RevMan 2014). We considered other forms of reporting bias (e.g. multiple publication bias, location bias, language bias, outcome reporting bias) on review of the full papers for each included study. The possibility of reporting bias is presented in the results below.
Data synthesis
We began the data synthesis with a narrative overview of the findings in the form of a table. The review authors, as a team, considered the comparability of the participants, clinical contexts, approaches to goal setting, and types of outcome data in order to determine whether statistical pooling of results was appropriate. Where appropriate, we used meta‐analytical methods to pool outcome data from sufficiently homogeneous studies to calculate effects in the comparisons outlined in our Objectives. Following data extraction, but prior to data analysis, we made the post‐hoc decision to combine measures of self‐reported emotional status with self‐reported measures of health‐related quality of life. We did this because few studies reported measures of health‐related quality of life and because the two concepts were deemed to be sufficiently similar for the results of a meta‐analysis to be clinically meaningful: quality of life, for instance, often has an emotional health subscale. We conducted all analyses according to guidelines in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We assessed quality of evidence using GRADE, and have presented a summary of the results of the data synthesis and assessment of the quality of the evidence in a 'Summary of findings' table. In 'summary of findings Table for the main comparison' and in 'summary of findings Table 2' we included summary information on the following: health‐related quality of life or self‐reported emotional status, participation outcomes, activity outcomes, outcomes at the level of body structure and function, patient engagement in rehabilitation, and self‐efficacy.
Subgroup analysis and investigation of heterogeneity
Where there were sufficient data (i.e. at least 10 studies) and where it was appropriate in the context of the study, we planned to conduct subgroup analysis on the basis of four factors:
-
level of patient and/or family involvement in goal selection;
-
level on the ICF at which rehabilitation goals were set;
-
level of goal difficulty; and
-
presence of cognitive or psychiatric impairments in study populations.
However, we did not identify sufficient studies to permit any of these subgroup analyses.
Sensitivity analysis
We undertook sensitivity analyses to examine the influence of risk of bias associated with including the studies in each meta‐analysis. We removed studies at the greatest risk of bias (i.e. those that failed to randomise adequately or failed to conceal random allocation) from the analysis in order to test the strength of evidence for the various effect estimates.
Consumer participation
We invited consumer referees to comment on the protocol and on the completed review through standard Cochrane Consumers and Communication Review Group editorial processes.
Results
Description of studies
Results of the search
We ran searches in September 2012, and again in January 2014, generating 9019 records, after removing duplicates. We screened the titles and abstracts of all citations and identified 151 articles that were potentially eligible for inclusion. We reviewed these in full text against the selection criteria, and identified 39 studies that met the inclusion criteria. Seven of these 39 studies were reported in multiple publications (see Table 1).
| Study | Other papers reporting study |
| Asenlof 2006; Asenlof 2009 | |
| Blair 1995 | |
| Duncan 2002 | |
| Jonsdottir 2012b | |
| Ostelo 2000; Ostelo 2004 | |
| Ranta 2000; Setter‐Kline 2007; Watson 2001 | |
| Sewell 2001 |
Included studies
Thirty‐nine studies met the selection criteria for this review (see Characteristics of included studies).
Comparison groups
Comparison 1
Of the 39 trials, 18 compared a structured approach to goal setting, with or without strategies to enhance goal pursuit, to no goal setting (Bassett 1999; Bell 2003; Blair 1991; Blair 1996; Coote 2012; Coppack 2012; Cross 1971; Duncan 2003; Evans 2002; Fredenburgh 1993; Harwood 2012; Howell 1986; Iacovino 1997; Mann 1987; O'Brien 2013; Scott 2004; Sewell 2005; Stanhope 2013).
Comparison 2
Fourteen trials compared a structured approach to goal setting, with or without strategies to enhance goal pursuit, to 'usual care' that may have involved goal setting but where no structured approach to goal setting was followed (Arnetz 2004; Asenlof 2005; Cheng 2012; Gagné 2003; Hart 1978; Holliday 2007; Jonsdottir 2012; LaFerriere 1978; McPherson 2009; Oestergaard 2012; Ostelo 2003; Parsons 2012; Taylor 2012; Woltmann 2011). Four of the studies in this comparison group were described by the authors as being pilot studies or feasibility studies (Gagné 2003; Jonsdottir 2012; McPherson 2009; Taylor 2012).
Comparison 3
Two trials investigated an intervention where the only difference to a control group was the use of a strategy to enhance goal pursuit (Culley 2010; Hart 2002).
Comparison 4
Nine trials compared one structured approach to goal setting and/or strategies to enhance goal pursuit to another structured approach to goal setting and/or strategies to enhance goal pursuit (Bassett 1999; Blair 1991; Blair 1996; Conrad 2000; James 1993; McPherson 2009; Miller 2012; Richardson 2007; Webb 1994). Four of these nine trials had more than two goal setting intervention groups, permitting their inclusion in more than one comparison in this review (Bassett 1999; Blair 1991; Blair 1996; McPherson 2009). For studies that had more than one intervention group, we only included the groups that met the inclusion criteria for analysis.
Communication with study authors
Of the 39 included studies, we found that 5 were reported in sufficient detail such that we required no further information about these studies for the purpose of this review (Hart 2002; Harwood 2012; James 1993; Ostelo 2003; Taylor 2012). We attempted to contact the authors of each of the other 34 studies to obtain additional information, and were successful with 19 (Arnetz 2004; Asenlof 2005; Bassett 1999; Bell 2003; Coote 2012; Conrad 2000; Culley 2010; Duncan 2003; Evans 2002; Holliday 2007; Jonsdottir 2012; LaFerriere 1978; McPherson 2009; Miller 2012; O'Brien 2013Parsons 2012; Scott 2004; Sewell 2005; Stanhope 2013), although full details were only available for 15 of these studies (Asenlof 2005; Bassett 1999; Coote 2012; Conrad 2000; Culley 2010; Duncan 2003; Evans 2002; Jonsdottir 2012; LaFerriere 1978; McPherson 2009; Miller 2012; O'Brien 2013; Parsons 2012; Sewell 2005; Stanhope 2013). We had at least one unanswered question about methods or outcome data for 24/39 (62%) of the included studies.
Types of studies
Unit of randomisation
Six studies used a cluster‐RCT design, randomising groups of participants clustered on the basis of the residential facility in which they lived (Blair 1991), the family physician with whom they were registered (Parsons 2012), the healthcare organisation or hospital providing their services (Cheng 2012; Stanhope 2013; Taylor 2012), or on the basis of the case manager who was managing their rehabilitation planning (Woltmann 2011). For three of these studies (Parsons 2012; Stanhope 2013; Taylor 2012) the effects of clustering on means and SDs in the outcome measures reported could be estimated from the information provided in the paper or from additional information provided by the study authors. For the other three studies (Blair 1991; Cheng 2012; Woltmann 2011) information on the effects of clusters on the data reported could not be accessed or estimated from other related publications. For these three studies we chose to include the data as reported in the published paper, but to report in our results where this occurred, and to test the influence of these decisions in our sensitivity analyses. For two other studies, randomisation occurred at the level of the participants' goals rather than at the level of the participant, with each participant having multiple goals being randomly allocated to either an intervention condition for enhancing goal recall or a control condition (Culley 2010; Hart 2002). For these two studies raw outcome data were available from the published paper or via author communication, so our analysis could be conducted at the same level as the unit of randomisation.
Sample sizes
Sample sizes ranged from 7 to 367, with a total of 2846 participants in the 39 studies addressing the 4 main comparisons in the review. Six of the included studies were reported as being pilot studies or feasibility studies (Gagné 2003; Jonsdottir 2012; McPherson 2009; Miller 2012; Richardson 2007; Taylor 2012) and therefore were not designed to have a sufficiently large sample size to detect statistically significant differences.
Types of settings
Seven of the included studies were conducted within inpatient hospital settings. This included three studies conducted within multidisciplinary units for people with neurological conditions (Holliday 2007; Jonsdottir 2012; Taylor 2012), one study conducted within a rheumatology unit (Arnetz 2004), two studies conducted within orthopaedic surgical units (Cross 1971; Oestergaard 2012), and one study conducted in a number of different hospital ward settings (Gagné 2003). An eighth study was conducted in both residential and outpatient‐based rehabilitation services for people with brain injury (Culley 2010). Eleven additional studies were conducted in outpatient or primary care settings, including four studies conducted within a cardiovascular rehabilitation service (Conrad 2000; Duncan 2003; Iacovino 1997; Mann 1987), one study conducted within a pulmonary rehabilitation service (Sewell 2005), one study conducted within an exercise laboratory (O'Brien 2013), four studies conducted within physiotherapy services (Asenlof 2005; Bassett 1999; Evans 2002; Ostelo 2003), and one study conducted within a cognitive behavioural therapy programme for chronic pain (James 1993). One additional study was undertaken in a short‐term residential rehabilitation unit for chronic pain (Coppack 2012).
The remaining 19 included studies were conducted in community and/or residential care settings. These included eight studies conducted within community‐based mental health services (Bell 2003; Coote 2012; Fredenburgh 1993; Hart 1978; Howell 1986; LaFerriere 1978; Stanhope 2013; Woltmann 2011), two studies conducted within residential care services for older adults (Blair 1991; Blair 1996), two studies conducted within community‐based diabetes services (Miller 2012; Richardson 2007), three studies conducted as part of home‐based nursing services (Cheng 2012; Parsons 2012; Scott 2004), three studies conducted within residential and home‐based service for people with brain injury (Hart 2002; McPherson 2009; Webb 1994), and one study conducted in people's homes after stroke (Harwood 2012).
Of the 39 included studies, 17 were conducted in the United States of America (USA) (Bell 2003; Blair 1991; Blair 1996; Cross 1971; Duncan 2003; Fredenburgh 1993; Gagné 2003; Hart 1978; Hart 2002; James 1993; LaFerriere 1978; Miller 2012; Richardson 2007; Scott 2004; Stanhope 2013; Webb 1994; Woltmann 2011), 7 in the United Kingdom (UK) (Coote 2012; Coppack 2012; Culley 2010; Evans 2002; Holliday 2007; Howell 1986; Sewell 2005), 6 in New Zealand (Bassett 1999; Harwood 2012; McPherson 2009; O'Brien 2013; Parsons 2012; Taylor 2012), 3 studies in Canada (Conrad 2000; Iacovino 1997; Mann 1987), 2 in Sweden (Arnetz 2004; Asenlof 2005), and 1 in each of Denmark (Oestergaard 2012), Switzerland (Jonsdottir 2012), Hong Kong (Cheng 2012), and the Netherlands (Ostelo 2003).
Types of participants
Participants (people receiving services) in the 39 included studies were adults receiving rehabilitation interventions for neurological conditions, including stroke in eight studies (Culley 2010; Hart 2002; Harwood 2012; Holliday 2007; Jonsdottir 2012; McPherson 2009; Taylor 2012; Webb 1994), musculoskeletal or chronic pain conditions in 10 studies (Arnetz 2004; Asenlof 2005; Bassett 1999; Coppack 2012; Cross 1971; Evans 2002; James 1993; O'Brien 2013; Oestergaard 2012; Ostelo 2003), mental health conditions in 8 studies (Bell 2003; Coote 2012; Fredenburgh 1993; Hart 1978; Howell 1986; LaFerriere 1978; Stanhope 2013; Woltmann 2011), cardiovascular conditions in 5 studies (Conrad 2000; Duncan 2003; Iacovino 1997; Mann 1987; Scott 2004), age‐related disability in 3 studies (Blair 1991; Blair 1996; Parsons 2012), diabetes in 2 studies (Miller 2012; Richardson 2007), and respiratory disorders in 1 study (Sewell 2005). The remaining 2 studies involved a mixed sample of patients with chronic disabling conditions (Cheng 2012; Gagné 2003).
Types of interventions
A range of different approaches to goal setting were employed in the included studies. Fifteen studies involved one or more named approaches to goal setting. Of these, seven employed GAS or a modified version of GAS as the method of goal setting under investigation (Arnetz 2004; Blair 1991; Blair 1996; Hart 1978; Howell 1986; LaFerriere 1978; Webb 1994). Three other studies investigated the effect of the COPM as a method of person‐centred goal setting (Oestergaard 2012; Sewell 2005; Taylor 2012). Five further studies investigated the effect of goal setting and strategies to enhance goal pursuit based on use of the 'ICF Rehab Cycle' (Jonsdottir 2012), the Patient Goal Priority Questionnaire (Asenlof 2005), the TARGET method of goal setting (Parsons 2012), Goal Management Training and Identity Oriented Goal Training (McPherson 2009), and the Goal Setting and Planning skills (GAP) programme (Coote 2012).
Twenty‐four other studies did not investigate the use of a specifically named approach to goal setting, but several of these did refer to theories of goal setting when describing the approach that was employed. Three of these studies used an approach to goal setting based on King's nursing theory of goal attainment (Cheng 2012; Fredenburgh 1993; Scott 2004). Eight other studies referred to Locke and Latham's goal theory and/or Bandura's Social Cognitive Theory when the background to goal setting in rehabilitation was described (Cross 1971; Culley 2010; Duncan 2003; Evans 2002; Holliday 2007; Iacovino 1997; Mann 1987; Miller 2012). However, of the studies that referred to Locke and Latham's goal theory, only two applied goal setting in a manner consistent with Locke and Latham's model, investigating the effect of difficult or challenging goals on human performance (Iacovino 1997; Miller 2012). One study investigated the effect of goal setting based on the theory of operant conditioning and its application via contingency management (Ostelo 2003), and another used Personal Construct Theory to guide goal setting in a rehabilitation context (Coppack 2012). Yet another study based strategies for enhancing goal pursuit on a theory of behavioural change called intentions implementation (O'Brien 2013). The remaining ten studies, did not employ a named approach to goal setting, or refer to any specific goal theory when the approach to goal setting under investigation was described (Bassett 1999; Bell 2003; Conrad 2000; Gagné 2003; Hart 2002; Harwood 2012; James 1993; Richardson 2007; Stanhope 2013; Woltmann 2011).
Involvement of participants in goal setting
Of the 39 included studies, 31 employed an approach to goal setting where the participants receiving services were actively involved in the selection of goals for therapy (Arnetz 2004; Asenlof 2005; Bassett 1999; Bell 2003; Blair 1991; Blair 1996; Cheng 2012; Coote 2012; Coppack 2012; Evans 2002; Fredenburgh 1993; Gagné 2003; Hart 1978; Harwood 2012; Holliday 2007; Howell 1986; Iacovino 1997; Jonsdottir 2012; LaFerriere 1978; Mann 1987; McPherson 2009; O'Brien 2013; Oestergaard 2012; Ostelo 2003; Parsons 2012; Scott 2004; Sewell 2005; Stanhope 2013; Taylor 2012; Webb 1994; Woltmann 2011). For most of these studies, selection of goals was described as a collaborative activity with therapists working as guides or coaches to assist the people receiving services set relevant goals for intervention. One of these studies compared this approach to goal setting to another approach where goals were prescribed by the treating therapist and to a third (control) group for whom therapy was provided without any goals being set (Bassett 1999). However, in three studies, while participants were involved in the selection of goals for therapy, their choice of goals was restricted to those from a prespecified list of goals (Mann 1987; Blair 1991; Blair 1996). Conversely, in two other studies, participants receiving the intervention had a much higher level of control over the selection of goals, with the researchers only providing them with training in goal setting, and ultimately leaving the application of this training up to the individuals themselves (i.e. self‐directed rehabilitation) (Coote 2012; Harwood 2012).
In six included studies, participants were not involved in goal selection at all (Conrad 2000; Cross 1971; Duncan 2003; James 1993; Miller 2012; Richardson 2007). Among these studies, four involved goals that were prescribed on the basis of group allocation, where the type of goal was manipulated as the independent variable in the study (Conrad 2000; James 1993; Miller 2012; Richardson 2007). The other two studies involved goals that were prescribed by the clinician providing the intervention (Duncan 2003), or where the same goal was allocated to all participants in the intervention arm of the study (Cross 1971). Two further studies did not report on who had been involved in goal selection or how the goals had been chosen (Culley 2010; Hart 2002).
Involvement of family members in goal setting
Only two studies reported family members as being actively involved in goal setting (Holliday 2007; Harwood 2012). In one of these studies, the involvement of family members was part of a wider aim to address cultural dimensions of the delivery of the intervention, with the researchers also ensuring that all interventions were delivered by a person of the same ethnicity as those receiving the interventions (Harwood 2012). In the other study, the involvement of family members in goal setting was encouraged but not required (Holliday 2007). In a third study, a 'collateral' person was involved in the goal setting process; however in this case, the relationship between this 'collateral' person and the person receiving care was not reported, and their role was solely to provide 'external validation of the patient's self‐report' (Hart 1978, p. 1244). A fourth study did not include family members in the goal setting process, but did involve them in the application of strategies to pursue goal achievement (Conrad 2000). The remaining 34 included studies did not report on the involvement of family members in goal setting or strategies to enhance goal pursuit.
Individual versus group‐based goal discussion
In one study, all goal selection and goal monitoring occurred within a small‐group context with peers who were also participants in the study (Bell 2003). For this study, all individual goals and goal achievements during the study were disclosed to these peers. In another study, a similar approach to group‐based goal discussion and goal monitoring was undertaken, but the group sizes ranged from one to four individuals, meaning that the trial involved some group‐based and some individual approaches to goal discussions (Iacovino 1997). In one other study, a family‐centred approach was taken to goal setting, meaning that again, all individual goals were shared with other people in a close relationship to the person receiving services (Harwood 2012). In a fourth study, all goal setting occurred on an individual basis, but all subsequence discussion of goals occurred within the context of group‐based rehabilitation classes (Sewell 2005). The other 34 included studies reported only on goal discussions occurring at an individual level with each study participant.
Goal difficulty
Six of the included studies emphasised the setting of realistic, achievable goals (Bassett 1999; Blair 1991; Cheng 2012; Duncan 2003; Holliday 2007; O'Brien 2013). One study emphasised the setting of challenging, difficult goals (Iacovino 1997). Two other studies investigated the effect of setting difficult, challenging goals versus less difficult goals (Conrad 2000; Miller 2012). One study involved therapists evaluating the level of difficulty of goals, but where goal difficulty was not apparently restricted by the goal setting process involved (Webb 1994). The remaining 30 studies did not report on whether any strategies were used to direct, monitor, or limit the level of difficulty of goals that were set.
Area or topic of focus for goals set
Two of the included studies focused on setting goals related to impairments of body structure and body function or on activity limitations (Arnetz 2004; Evans 2002). Nine studies focused on setting goals solely to do with addressing activity limitations (Bassett 1999; Blair 1991; Blair 1996; Duncan 2003; Gagné 2003; Oestergaard 2012; Ostelo 2003; Richardson 2007; Sewell 2005). Two other studies involved setting goals either to do with addressing activity limitations or participation restrictions (Holliday 2007; Taylor 2012). One study just referred to a focus on 'functional' goals (O'Brien 2013). Two studies focused entirely on goals related to work performance (Bell 2003; Iacovino 1997). Four studies involved goals about dietary behaviour (Conrad 2000; Cross 1971; Mann 1987; Miller 2012). One study only involved goals regarding the amount and frequency of practice of specific coping strategies for managing pain (James 1993). Four studies included a mix of topics as the focus of goal setting covering a wide area of future objectives such as functional abilities, emotional needs, information needs, financial needs, and so on (Cheng 2012; Coppack 2012; Harwood 2012; McPherson 2009). The topic of focus for goals was not specifically reported on in the 15 other included studies in this review (Asenlof 2005; Coote 2012; Culley 2010; Fredenburgh 1993; Hart 1978; Hart 2002; Howell 1986; Jonsdottir 2012; LaFerriere 1978; McPherson 2009; Parsons 2012; Scott 2004; Stanhope 2013; Webb 1994; Woltmann 2011).
Documentation of goals for participants
In 15 studies, written copies of goals were kept by the participants receiving the rehabilitation intervention (Asenlof 2005; Bassett 1999; Bell 2003; Conrad 2000; Coote 2012; Duncan 2003; Gagné 2003; Harwood 2012; Iacovino 1997; James 1993; McPherson 2009; O'Brien 2013; Ostelo 2003; Richardson 2007; Webb 1994). For four of these studies, the participants themselves were responsible for documenting their own goals (Bell 2003; Coote 2012; Harwood 2012; Iacovino 1997). In two studies all goals were recorded on a goal setting form designed for the study, but it was not clear whether this form was held by the patient or the clinician involved in delivery of the intervention (Cheng 2012; Evans 2002). For two other studies, the method of documentation of goals was the subject of the trial, so differed for the intervention and control goals (Culley 2010; Hart 2002). It was not reported in the remaining 20 studies whether participants were given a written copy of their goals (Arnetz 2004; Blair 1991; Blair 1996; Coppack 2012; Cross 1971; Fredenburgh 1993; Hart 1978; Holliday 2007; Howell 1986; Jonsdottir 2012; LaFerriere 1978; Mann 1987; Miller 2012; Oestergaard 2012; Parsons 2012; Scott 2004; Sewell 2005; Stanhope 2013; Taylor 2012; Woltmann 2011).
Evaluation of goal commitment
The level of commitment of participants to achieve goals that had been set was evaluated in only two studies (Iacovino 1997; Miller 2012). In one study the behavioural intentions of participants towards goal achievement were assessed using Likert‐type questions (Iacovino 1997). In the other study a questionnaire was used to evaluate the participants' level of determination to achieve goals at two time points: once following goal assignment and once at the end of the study (Miller 2012). The other 37 included studies did not report on the evaluation of goal commitment.
Development of a plan for goal pursuit
Twenty‐eight of the included studies described structured processes for the development of a plan to pursue goal achievement (Asenlof 2005; Bell 2003; Blair 1991; Blair 1996; Cheng 2012; Conrad 2000; Coote 2012; Cross 1971; Duncan 2003; Fredenburgh 1993; Hart 1978; Harwood 2012; Holliday 2007; Howell 1986; Iacovino 1997; James 1993; Jonsdottir 2012; Mann 1987; McPherson 2009; Miller 2012; O'Brien 2013; Oestergaard 2012; Ostelo 2003; Parsons 2012; Richardson 2007; Scott 2004; Sewell 2005; Stanhope 2013). Development of a plan for goal achievement was not relevant to two other studies because the aim of these studies was just to investigate strategies to enhance recall of goal over a period of time (Culley 2010; Hart 2002). The remaining nine studies did not report on strategies used to develop a plan for goal achievement (Arnetz 2004; Bassett 1999; Coppack 2012; Evans 2002; Gagné 2003; LaFerriere 1978; Taylor 2012; Webb 1994; Woltmann 2011).
Reminders for participants about goals
Twenty‐four of the included studies reported on the use of a strategy to assist participants (receiving services) to remember that they had rehabilitation goals and what these goals were (Asenlof 2005; Bassett 1999; Bell 2003; Cheng 2012; Conrad 2000; Coote 2012; Coppack 2012; Culley 2010; Duncan 2003; Evans 2002; Gagné 2003; Hart 1978; Hart 2002; Howell 1986; Iacovino 1997; James 1993; Mann 1987; McPherson 2009; Miller 2012; Ostelo 2003; Richardson 2007; Sewell 2005; Stanhope 2013; Webb 1994). For 15 of these studies, discussion of goals or reminders about goals featured as part of regular (usually weekly) therapy sessions (Asenlof 2005; Bassett 1999; Bell 2003; Cheng 2012; Coppack 2012; Evans 2002; Hart 1978; Howell 1986; Iacovino 1997; James 1993; Mann 1987; McPherson 2009; Ostelo 2003; Sewell 2005; Webb 1994). In 5 other studies, reminders about goals were provided less frequently: every 3 weeks over the study period (Duncan 2003), monthly for 11 months (Stanhope 2013), at 2 and 4 month meetings during a 7‐month study period (Conrad 2000), and just once midway through the study intervention (Coote 2012; Miller 2012). One study used a goal notebook to remind participants about their rehabilitation goals (Gagné 2003), and three further studies used electronic methods including text messaging (Culley 2010), portable voice organisers (Hart 2002), and the Internet (Richardson 2007). The remaining 16 studies included in this review did not report on the use of strategies to remind participants about their goals during the course of the intervention.
Active monitoring of progress towards goals
Sixteen of the included studies reported on the use of strategies to monitor progress towards achievement of goals and discussion of this information with study participants (Asenlof 2005; Bassett 1999; Bell 2003; Cheng 2012; Coppack 2012; Duncan 2003; Evans 2002; Hart 1978; Iacovino 1997; Mann 1987; McPherson 2009; Miller 2012; Ostelo 2003; Richardson 2007; Stanhope 2013; Webb 1994). Progress towards goal achievement was self‐rated by participants (receiving services) in the case of five studies (Iacovino 1997; Mann 1987; Miller 2012; Ostelo 2003; Webb 1994), with four of these studies employing self‐monitoring forms or charts to facilitate this process (Mann 1987; Miller 2012; Ostelo 2003; Webb 1994), and one study using an online Internet service with data on exercise performance uploaded directly from pedometers to facilitate participant self‐monitoring of progress towards goals (Richardson 2007). Progress towards goals was discussed with therapists at regular meetings (usually weekly) in the case of 11 studies (Asenlof 2005; Bassett 1999; Bell 2003; Cheng 2012; Coppack 2012; Evans 2002; Hart 1978; Iacovino 1997; Mann 1987; McPherson 2009; Ostelo 2003), with goals being revised as participants progressed in the case of two studies (Bassett 1999; Mann 1987). In one study, progress towards goals was discussed with therapists at just one point midway through the study (Miller 2012). Feedback on progress towards goals were provided graphically for participants (receiving services) in the case of three studies (Bell 2003; Duncan 2003; Richardson 2007). In one further study 'goal notebooks' were used to enhance discussions between patients and therapists about goals, but it was unclear whether this included measurement and monitoring of progress towards goals (Gagné 2003). The remaining 22 studies included in this review did not report on the use of strategies to monitor progress toward goals during the course of the intervention.
'Usual care' as a comparison intervention
Fourteen studies examined a structured approach to goal setting, with or without strategies to enhance goal pursuit, in comparison to 'usual care' where some goals may have been set but no structured goal setting approach was followed (Comparison 2 in this review; see Included studies: Comparison groups). Of these 14 studies, 5 provided at least some details regarding how goal setting in usual care was undertaken (Arnetz 2004; Asenlof 2005; Cheng 2012; Holliday 2007; Ostelo 2003), permitting analysis of how goal setting in the intervention arms of these studies actually differed from goal setting in the control arms. In two of these studies, treatment fidelity in the usual care group was formally assessed to ensure that the control intervention did indeed differ from the experimental intervention in the manner anticipated (Asenlof 2005; Ostelo 2003). For the other nine studies, goal setting in the usual care groups was characterised primarily by not involving the structured approaches to goal setting that the experimental groups received (Gagné 2003; Hart 1978; Jonsdottir 2012; LaFerriere 1978; McPherson 2009; Oestergaard 2012; Parsons 2012; Taylor 2012; Woltmann 2011). Despite this ambiguity, the usual care groups were broadly presented as having less patient involvement in goal selection, being less person‐centred, and being less focused on personally meaningful activities that patients wanted to pursue. In three studies, the usual care interventions were also presented as having differed from the experimental interventions because the experimental intervention included elements explicitly targeting behavioural change to help participants pursue goals when the usual care interventions did not (Asenlof 2005; McPherson 2009; Ostelo 2003). The specific approach to behavioural change differed from study to study, however. In Asenlof 2005, the experimental group received individual functional behavioural analysis in which behavioural barriers to goal achievement were identified and strategies were implemented to address these. In McPherson 2009, the additional behavioural intervention received by the experimental group focused on improving self‐regulation of goal‐oriented behaviour. In Ostelo 2003, the experimental groups were prescribed graded activity to achieve selected goals, with this intervention being based on contingency management and operant conditioning theory.
Excluded studies
Of the studies that we excluded from the review, 62 required additional discussion between the review authors before a consensus was reached regarding their exclusion. These are listed in the Characteristics of excluded studies. We gave more than one reason for exclusion for some of these studies but, broadly speaking, we excluded 23 studies because it was not possible to separate the effects of goal setting or strategies to enhance goal pursuit from other (frequently broader) interventions being provided to the experimental group but not the control group. Examples of broader goal‐directed interventions included: therapeutic coaching, counselling, work skills training, cognitive behavioural therapy, mindfulness meditation, self‐management skills training, and community rehabilitation.
We excluded an additional 18 studies because they did not involve, or it could not be confirmed that they involved, a population of people with disability acquired in adulthood. Among this group of excluded studies were trials involving goal setting interventions for people with subclinical health concerns (e.g. subclinical weight problems; subclinical high alcohol consumption), healthy older adults in residential care, and adults with congenital conditions or conditions most commonly acquired in childhood. We excluded 13 additional studies because they did not use an RCT or quasi‐RCT study design. We excluded 3 studies for other reasons. These other reasons included comparing a goal setting intervention with a different behavioural intervention (specifically instructions to attend to pain sensations and modify activity levels accordingly) (Stenstrom 1994), conducting the goal setting intervention and outcome measurement in a single clinical session (Wood 2012), and investigating a type of goal setting that did not align with the definition of 'rehabilitation goals' used for the purposes of this review (Adair 2013).
Risk of bias in included studies
We present 'Risk of bias' information in Characteristics of included studies and Figure 1. There were no noticeable trends in the 'Risk of bias' assessments when studies were grouped by comparison type, so the summary of information on risk of bias presented below is organised at the level of all included studies.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Allocation
We judged 21 trials (54% of included studies) as having low risk of bias for random sequence generation (Asenlof 2005; Bassett 1999; Bell 2003; Cheng 2012; Conrad 2000; Culley 2010; Duncan 2003; Fredenburgh 1993; Hart 2002; Harwood 2012; LaFerriere 1978; McPherson 2009; Miller 2012; O'Brien 2013; Oestergaard 2012; Ostelo 2003; Parsons 2012; Scott 2004; Sewell 2005; Stanhope 2013; Taylor 2012). Information was insufficient to permit a decision in relation to 12 trials (31% of included studies) (Blair 1991; Blair 1996; Coppack 2012; Cross 1971; Hart 1978; Howell 1986; Iacovino 1997; Jonsdottir 2012; Mann 1987; Richardson 2007; Webb 1994; Woltmann 2011). We judged six trials (15%) to be at high risk of bias due to problems with random sequence generation (Arnetz 2004; Coote 2012; Evans 2002; Gagné 2003; Holliday 2007; James 1993). These six studies employed a quasi‐RCT design, whereby a systematic method had been used for group allocation (e.g. alternate allocation or allocation based on room assignment or date of admission) or where group allocation had been manipulated in an attempt to balance the groups on the basis of preselected patient characteristics or to address patient preferences for group allocation.
With regards to allocation concealment, we judged 19 trials (49% of included studies) as having low risk of bias (Asenlof 2005; Bassett 1999; Bell 2003; Culley 2010; Fredenburgh 1993; Hart 2002; Harwood 2012; Jonsdottir 2012; LaFerriere 1978; McPherson 2009; Miller 2012; O'Brien 2013; Oestergaard 2012; Ostelo 2003; Parsons 2012; Scott 2004; Sewell 2005; Stanhope 2013; Taylor 2012) and seven trials (18% of included studies) to be at high risk of bias (Arnetz 2004; Cheng 2012; Coote 2012; Evans 2002; Gagné 2003; Holliday 2007; James 1993). The remaining 13 trials (33% of included studies) provided insufficient information to inform judgements (Blair 1991; Blair 1996; Conrad 2000; Coppack 2012; Cross 1971; Duncan 2003; Hart 1978; Howell 1986; Iacovino 1997; Mann 1987; Richardson 2007; Webb 1994; Woltmann 2011).
Blinding
Blinding of participants and personnel was not feasible in the majority (32/39) of studies because the goal setting intervention required active involvement of both parties in order to be correctly implemented. In the case of three studies, the delivery of the goal setting intervention was automated by the use of technology (phones, computers, and portable voice organisers), removing the need to blind personnel to group allocation (Culley 2010; Hart 2002; Richardson 2007). In the case of one of these studies, participants were also blinded to the aspect of goal setting under investigation (the type of goal set), so this study was effectively blinded for both participants and personnel (Richardson 2007). In two additional studies, while the participants in the experimental and control group were actively involved in the delivery of the intervention, they were not informed of the specific details regarding the hypothesis under investigation, partially blinding them to the experimental and control conditions (Coppack 2012; Evans 2002). Likewise, in yet another study, while therapists were aware of the differences in treatment protocol for the groups in the trial, they were not informed about the experimental hypothesis, partially blinding them to the anticipated patient response to the study conditions (James 1993). In two studies, different personnel were used for goal setting versus the other aspects of the rehabilitation intervention (e.g. exercise therapy), with those involved in goal setting not blinded to group allocation but those involved in delivery of therapy blinded to group allocation (Evans 2002; O'Brien 2013).
In terms of blinding for outcome assessment, five studies used adequate methods for blinding those people involved in outcome data collection (Coppack 2012; Gagné 2003; Jonsdottir 2012; O'Brien 2013; Richardson 2007). In ten studies the personnel involved in outcome data collection were not blinded to study group allocation (Arnetz 2004; Asenlof 2005; Bassett 1999; Blair 1991; Blair 1996; Holliday 2007; Howell 1986; James 1993; Stanhope 2013; Woltmann 2011). In eight other studies, outcome assessors were ostensibly blinded to group allocation but the overall risk of detection bias was deemed 'high' due to the outcome data collection being heavily dependent on self‐report by participants who were not (or could not be) blinded to group allocation, or due to other problems with maintaining the blinding of outcome assessors (Bell 2003; Conrad 2000; Coote 2012; Fredenburgh 1993; Hart 1978; LaFerriere 1978; Miller 2012; Oestergaard 2012). For the remaining 16 studies, the risk of detection bias was unclear. For four studies, this was because insufficient information was reported on the possible blinding of outcome assessors (Cross 1971; Iacovino 1997; Mann 1987; Webb 1994). For 12 studies, the risk of detection bias was unclear because, while the outcome assessor had been blinded to group allocation, the collection of data had been partially based on self‐report by participants who had not been blinded to group allocation (Cheng 2012; Culley 2010; Duncan 2003; Evans 2002; Hart 2002; Harwood 2012; McPherson 2009; Ostelo 2003; Parsons 2012; Scott 2004; Sewell 2005; Taylor 2012).
Overall, the majority of studies were at high risk of performance bias due to the nature of the intervention under investigation. However, 13% (5/39) of studies were at low risk of detection bias, 41% (16/39) had unclear risk of detection bias (16/39), and 46% (18/39) were at high risk of detection bias. Seventeen of the included studies (44%; 17/39) were deemed to have high risk for both performance and selection bias, while only one study (Richardson 2007) was deemed to be at low risk for both performance and selection bias.
Incomplete outcome data
Of the 39 included studies, 13 reported losses to follow‐up of less than 5%, one study reported loss to follow‐up of 7%, 11 reported losses to follow‐up of 10% to 20%, 12 reported losses to follow‐up of higher than 20%, and for the remaining two studies there was insufficient information available to judge the number of participants lost to follow up. There was no compelling evidence of an imbalance in losses to follow‐up across the intervention groups in any study, except for Woltmann 2011, where there was slightly higher attrition in the intervention group (17.5%; 7/40) versus the control group (10%; 4/40).
For the studies reporting moderate to high loss of participants to follow‐up (i.e. 10% to 20%), three reported following an intention‐to‐treat analysis approach (Asenlof 2005; Bell 2003; Ostelo 2003). Bell 2003 reported using an imputation method of analysis, bringing the last observed outcome forward for all missing data (raising the risk of attrition bias). Ostelo 2003 used a single value imputation method, substituting missing data with mean value, negative values and positive values if the reasons for participants dropping out were unexplained, an aggravation in symptoms, or a complete remission of symptoms, respectively. Asenlof 2005 compared multiple methods for management of missing data in their analysis: imputation methods from last observed outcome, substitution with the worst 10th percentile, and omission of missing data from the analysis. As these different methods were not found to alter the findings of the study, data from the latter approach were reported.
Overall, there was a spread of low to high risk of attrition bias, with 36% (14/39) of all studies being at low risk of attrition bias, 31% (12/39) having unclear risk of attrition bias, and 33% (13/39) being at high risk of attrition bias.
Selective reporting
Five studies had a protocol published in a peer‐reviewed journal or clinical trial registry prior to the trial being conducted (Harwood 2012; Ostelo 2003; Parsons 2012; Richardson 2007; Taylor 2012). For these studies we could confirm that there was no evidence of selective reporting. For the majority of trials (85%; 33/39) no published protocol could be identified, which meant that the risk of reporting bias for these studies was unclear. For four of these studies, data on multiple outcomes were collected and/or collected at multiple time points, with no named primary outcome, resulting in multiple tests for statistical significance and/or incomplete reporting of all results (Arnetz 2004; Duncan 2003; Iacovino 1997; Scott 2004). Cross 1971 reported some statistically significant findings, but appeared to have adjusted their analysis (e.g. combining groups; reporting change scores when end scores were not significant) in order to find these. Unadjusted SDs for the main outcomes in Cross 1971 were not reported. Finally, James 1993 was deemed to be at high risk of reporting bias as outcome data were explicitly dropped from the analysis when it was considered to be outlying.
Other potential sources of bias
For 26 of the included studies (67%) there was low risk of additional sources of bias, including low risk of cross‐group contamination. For another ten studies (26%), other sources of bias were noted that presented an unclear level of risk. Of these, seven were deemed to have had potential for cross‐group contamination, where the risk arising from this was unclear (Bassett 1999; Blair 1996; Fredenburgh 1993; Gagné 2003; Hart 1978; Howell 1986; McPherson 2009). For two studies, participants in the experimental groups appeared to receive much higher levels of contact time with the clinical providers than those in the control groups, resulting in an unclear level of additional risk of bias (Webb 1994; Woltmann 2011). Oestergaard 2012 used a novel, untested outcome measure to compare their two study groups where the construction of the measure itself appeared to favour the experimental intervention. Furthermore, Oestergaard 2012 appeared to introduce an outcome measure (the Dallas Pain Questionnaire) during the course of the study rather than decide on all outcomes and timing of data collection a priori, again introducing an unclear risk of bias.
Three studies (8%) were deemed to have involved a high risk of additional sources of bias (Arnetz 2004; Holliday 2007; Webb 1994). Of these, Arnetz 2004 used goal attainment as a main outcome measure to evaluate the effect of goal setting, but did not report when the therapists in each study group set the goals for therapy (i.e. before or after randomisation). Similarly, Webb 1994 used GAS as their only outcome measure. While the scoring of these GAS outcomes was reported as being undertaken by a blinded assessor, the construction of the individualised GAS scales in Webb 1994 was completed with involvement from an unblinded treating therapist after randomisation and allocation of participants to the study groups. Finally, Holliday 2007 reported the involvement of all staff in the study setting (an inpatient rehabilitation ward) in treating participants in both the experimental and control groups, sometimes concurrently, significantly increasing the risk of cross‐group contamination.
Effects of interventions
See: Summary of findings for the main comparison Goal setting with or without strategies to enhance goal pursuit compared to no goal setting for adults with acquired disability participating in rehabilitation; Summary of findings 2 Structured goal setting with or without strategies to enhance goal pursuit compared to 'usual care' that involved some goal setting but where no structured approach was followed for adults with acquired disability participating in rehabilitation
See also Data and analyses for pooled analyses; Table 2 for information about the selection and management of data for pooled analyses; Table 3 for data from single studies that were not pooled.
| Comparison 1: Structured goal setting with or without strategies to enhance goal pursuit versus no goal setting | |||||
| Health‐related quality of life and self‐reported emotional status | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Intrapsychic foundation and Interpersonal function subscales of the Quality of Life Scale | No, means and SDs not reported | n/a | n/a | n/a | |
| Rosenberg Self‐Esteem Scale | Yes | n/a ‐ only reported on one measure at one time point | Unable to adjust for effects of clustering; influence of study on findings tested in sensitivity analysis | Mutual goal setting groups with and without operant behaviour management were combined into a single experimental (goal setting) group | |
| Centre for Epidemiological Studies‐Depression Scale | Yes | Lowest of the two middle ranked effect estimates for four possible measures (other possible measures: Satisfaction with Life Scale, Positive Affect Scale, and Negative Affect Scale) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Minnesota Living with Heart Failure | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| 'Reorganisation' subscale of the Psychological Responses to Injury | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | The 'attention' control group (receiving social support but not goal setting from a sport psychologist) and the 'no additional input' control group were combined into a single control group | |
| Derogatis Stress Profile | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Physical Component Summary scores from the 36‐Item Short Form Health Survey | Yes | Selected as this was the primary outcome measure used in the sample size calculation | n/a ‐ not a cluster‐RCT | Four study groups. However, we judged the DVD inspirational video to be a separate intervention unrelated to goal setting, so only included data from the 'Take Charge' (goal setting) intervention as experimental group data and data from the 'usual care' group as control group data | |
| Satisfaction with Life Scale | No, means and SDs not reported | n/a | n/a | n/a | |
| Cardiac version of the Quality of Life Index | Yes | Lowest effect estimate from two possible measures (other possible measure: Mental Health Inventory‐5) | n/a ‐ not a cluster‐RCT | Three study groups. However, we judged the self‐management education to be a separate intervention, so excluded data from this group from our analysis | |
| Dyspnea component of the Chronic Respiratory Questionnaire | Yes | Lowest of the two middle ranked effect estimates for four possible measures (other possible measures: the Fatigue, Emotion, and Mastery components of the Chronic Respiratory Questionnaire) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Activity ‐ ability | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Six Minute Walk Test | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Barthel Index | Yes | Lowest effect estimate from two possible measures (other possible measure: Frenchay Activities Index) | n/a ‐ not a cluster‐RCT | Four study groups. However, we judged the DVD inspirational video to be a separate intervention unrelated to goal setting, so only included data from the 'Take Charge' (goal setting) intervention as experimental group data and data from the 'usual care' group as control group data | |
| Six Minute Walk Test | Yes | Median effect estimate out of five possible measures (other possible measures: Timed Up and Go; Activity of Daily Living subscale of the Lower Limb Task Questionnaire; Step Test; Ten meter Walk Test | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Continuous ambulatory activity monitor counts | Yes | Selected as this was the primary outcome measure used in the sample size calculation | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Engagement in rehabilitation | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Self‐reported percentage of recommended exercises session completed | Yes | Lowest effect estimate from two possible measures (other possible measure: self‐reported percentage of recommended repetitions of each exercise completed) | n/a ‐ not a cluster‐RCT | Three groups. We combined the two groups involving participant‐physiotherapist collaborative goal setting and physiotherapist‐mandated goal setting into a single experimental (goal setting) group | |
| Total number of hours worked during a work trial | Yes | Lowest effect estimate from two possible measures (other possible measure: Total number of weeks worked during a work trial) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Sports Injury Rehabilitation Adherence Scale | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | Three groups. We combined the groups involving therapist‐directed exercise without goal setting and non‐therapist‐directed exercise without goal setting into a single control group | |
| Percentage adherence to recommended food selection | Yes | n/a ‐ only reported on one measure | n/a ‐ not a cluster‐RCT | Three groups. However, we judged the group receiving no dietary education as irrelevant to the review question, so only included the groups receiving dietary education with and without goal setting | |
| Number of recommended exercise sessions completed | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Overall physiotherapist estimate of adherence | Yes | Lowest effect estimate from two possible measures (other possible measure: self‐reported percentage of recommended exercises completed) | n/a ‐ not a cluster‐RCT | Three groups. We combined the 'attention' control group (receiving social support but not goal setting from a sport psychologist) and the 'no additional input' control group into a single control group | |
| 'Motivation' subscale of the Griffiths Work Performance Scale | No, SDs not reported | n/a | n/a | n/a | |
| Percentage of eligible weeks worked during a work trial | Yes | No data were provided on the other measure referred to in the study: the Work Values Inventory | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| 24‐hour self‐reported sodium intake | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | Three groups. However, we judged the group receiving no dietary education as irrelevant to the review question, so only included the groups receiving dietary education with and without goal setting | |
| Sports Injury Rehabilitation Adherence Scale | Yes | Lowest of the two middle ranked effect estimates for four possible measures (other possible measures: number of classes attended; self‐reported adherence to stretching exercises; self‐reported adherence to walking exercises) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Self‐efficacy | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Sports Injury Rehabilitation Beliefs Survey | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | Three groups. We combined the groups involving therapist‐directed exercise without goal setting and non‐therapist‐directed exercise without goal setting into a single control group. | |
| Sports Injury Rehabilitation Beliefs Survey | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | Three groups. We combined the 'attention' control group (receiving social support but not goal setting from a sport psychologist) and the 'no additional input' control group into a single control group | |
| Task self‐efficacy | Yes | Median of three effect estimates (other possible measures: maintenance self‐efficacy, recovery self‐efficacy) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Self‐efficacy to Manage Disease in General | No, SDs not reported | n/a | n/a | n/a | |
| Comparison 2: Structured approach to goal setting with or without strategiesto enhance goal pursuit versus 'usual care' that may have involved some goal setting but where no structured approach was followed | |||||
| Health‐related quality of life and self‐reported emotional status | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Classification Committee of the World Organization of National Colleges, Academies and Academic Associations of General Practitioners/Family Physicians' chart of perceived health status | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Unable to adjust for effects of clustering; influence of study on findings tested in sensitivity analysis | n/a ‐ only two study groups | |
| General Health Questionnaire | No, means and SDs not reported | n/a | n/a | n/a | |
| Modified version of the Welsh Anxiety Scale | Yes | Lowest of the two middle ranked effect estimates for four possible measures (other possible measures: Anxiety Scale of the Today form of the Multiple Affect Adjective Checklist; Depression Scale of the Today form of the Multiple Affect Adjective Checklist; Rosenburg Self‐esteem Scale) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| General Health subscale of the 36‐Item Short Form Health Survey | Yes | Lowest effect estimate from two possible measures (other possible measure: Social Functioning subscale of the 36‐Item Short Form Health Survey). Data were collected from more than one time point. Data from the longest period of time following recruitment into the study were included in the meta‐analysis. | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Mental Component Summary score of 36‐Item Short Form Health Survey | Yes | Lowest effect estimate from two possible measures (other possible measure: Physical Component Summary score of the 36‐Item Short Form Health Survey) | Using the reported data on mean differences between the intervention and control groups with 95% CIs, we calculated adjusted means and SD that reproduced the 95% CI and P values in the paper to account for clustering | n/a ‐ only two study groups | |
| Schedule for Evaluation of Individual Quality of Life | Yes | Selected as this was named the primary outcome measure (other possible measure: 36‐Item Short Form Health Survey) Data were collected from more than one time point. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Using the reported data on mean differences between the intervention and control groups with 95% CIs, and the reported separate effects of clustering on variance plus the inter‐class correlations for clusters, we calculated adjusted means and SD that reproduced the 95% CI and P values in the paper to account for clustering | n/a ‐ only two study groups | |
| Activity ‐ ability | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Pain Disability Index | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Disability Index of the Health Assessment Questionnaire | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Unable to adjust for effects of clustering; influence of study on findings tested in sensitivity analysis | n/a ‐ only two study groups | |
| Functional Independence Measure | No, means and SDs not reported | n/a | n/a | n/a | |
| Roland Morris Disability Questionnaire | Yes | Lowest effect estimate from two possible measures (other possible measure: self‐reported severity of main activity limitation). Data were collected from more than one time point. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Functional Independence Measure | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Using the reported data on mean differences between the intervention and control groups with 95% CIs, and the reported separate effects of clustering on variance plus the inter‐class correlations for clusters, we calculated adjusted means and SD that reproduced the 95% CI and P values in the paper to account for clustering | n/a ‐ only two study groups | |
| Self‐efficacy | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Self‐efficacy Scale | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Chronic Disease Self‐efficacy Score | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Unable to adjust for effects of clustering | n/a ‐ only two study groups | |
| Satisfaction with service delivery | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Overall quality of physical therapy on a 0‐10 scale | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Satisfaction Scale for Community Nursing | Yes | n/a ‐ only reported on one measure at one time point | Unable to adjust for effects of clustering | n/a ‐ only two study groups | |
| Overall satisfaction on a 10 cm visual analogue scale | No, means and SDs not reported | n/a | n/a | n/a | |
| Satisfaction with counselling | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Patient Perception of Rehabilitation | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Using the reported data on mean differences between the intervention and control groups with 95% CIs, and the reported separate effects of clustering on variance plus the inter‐class correlations for clusters (the latter accessed by author communication), we calculated adjusted means and SD that reproduced the 95% CI and P values in the paper to account for clustering. | n/a ‐ only two study groups | |
| 5‐point client satisfaction questionnaire | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Adverse events (all‐cause) | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Number of participants withdrawn from the study due to death or hospitalisation | Yes | Withdrawal due to death or re‐hospitalisation combined. Numbers taken from last time point | Unable to adjust for effects of clustering | n/a ‐ only two study groups | |
| Number of participants withdrawn from the study due to worsening symptoms | Yes | Numbers taken from last time point | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Number of participants withdrawn from the study due to death | Yes | n/a ‐ only reported on one measure at one time point | Unable to adjust for effects of clustering | n/a ‐ only two study groups | |
CI = confidence interval; n/a = not applicable; RCT = randomised controlled trial; SD = standard deviation
| Comparison 1: Structured goal setting with or without strategies to enhance goal pursuit versus no goal setting | ||||||||
| Continuous data | ||||||||
| Trial | Outcome measure | Goal setting | No goal setting | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Total Work Behaviour Inventory | 136.5 | 22.2 | 30 | 120.5 | 25.5 | 33 | 16.0 (4.22 to 27.78) | |
| No. treatment sessions required for symptom relief | 12.56 | 6.86 | 45 | 13.29 | 11.01 | 21 | ‐0.73 (‐5.85 to 4.39) | |
| Baseline Dyspnea Index | 9.7 | 1.7 | 7 | 8.1 | 1.6 | 7 | 1.60 (‐0.13 to 3.33) | |
| Piper Fatigue Scale | ‐1.8 | 1.7 | 7 | ‐2.2 | 1.9 | 7 | 0.40 (‐1.49 to 2.29) | |
| Systolic blood pressure (mmHg) | 137.4 | 17.8 | 38 | 140.5 | 18.6 | 31 | ‐3.50 (‐12.15 to 5.15) | |
| Urinary sodium output (mmol/24 h) | 157.4 | 52.7 | 19 | 141.4 | 57.9 | 19 | 16.0 (‐19.20 to 51.20) | |
| Systolic blood pressure (mmHg) | 135 | 17.8 | 19 | 137.6 | 17.2 | 19 | ‐2.60 (‐13.73 to 8.53) | |
| Criterion referenced achievement test score | 18.8 | 2.7 | 19 | 17.1 | 3.3 | 19 | 1.70 (‐0.22 to 3.62) | |
| COPM satisfaction | 2.04 | 1.91 | 63 | 2.27 | 2.03 | 58 | ‐0.23 (‐0.93 to 0.47) | |
| Dichotomous data | ||||||||
| Trial | Outcome | Goal setting | No goal setting | Risk ratio (95% CI) | ||||
| Events | Total | Events | Total | |||||
| Dependency, based on modified Rankin Scores > 2 | 11 | 38 | 12 | 31 | 0.75 (0.38 to 1.46) | |||
| Death | 4 | 46 | 5 | 39 | 0.68 (0.20 to 2.35) | |||
| 'High' theoretical GAS achievement | 8 | 13 | 7 | 11 | 0.97 (0.52 to 1.80) | |||
| Return to work | 15 | 22 | 21 | 24 | 0.78 (0.56 to 1.08) | |||
| Comparison 2: Structured approach to goal setting with or without strategies to enhance goal pursuit versus 'usual care' that may have involved some goal setting but where no structured approach was followed | ||||||||
| Continuous data | ||||||||
| Trial | Outcome measure | Structured goal setting | 'Usual care' goal setting | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Pain ‐ Maximum | ‐3.6 | 3.17 | 28 | ‐4.9 | 3.1 | 37 | 1.30 (‐0.24 to 2.84) | |
| Tampa Scale of Kinesiophobia | ‐27.10 | 6.35 | 28 | ‐29.60 | 6.69 | 37 | 2.50 (‐0.69 to 5.69) | |
| Percentage of goals achieved by individual participants | 98.0 | 9.3 | 40 | 61.5 | 49.0 | 29 | 36.5 (18.43 to 54.57) | |
| Patient‐reported motivation | 8.93 | 1.1 | 15 | 7.53 | 1.66 | 17 | 1.40 (0.43 to 2.37) | |
| Therapist‐reported motivation of patients | 4.6 | 0.99 | 15 | 4.12 | 1.54 | 17 | 0.48 (‐0.41 to 1.37) | |
| Number of therapy sessions provided | 9.27 | 6.22 | 15 | 6.18 | 4.3 | 17 | 3.09 (‐0.66 to 6.84) | |
| Change in lumber spine range of movement (degrees) | 18.9 | 21.5 | 52 | 20.1 | 22.7 | 53 | ‐1.20 (‐9.66 to 7.26) | |
| Change in Tampa Scale of Kinesiophobia | 2.7 | 6.5 | 52 | 2.6 | 6.2 | 53 | 0.10 (‐2.33 to 2.53) | |
| Total healthcare costs in the 12 month follow‐up period (EUR) | 1978 | 1894 | 52 | 1339 | 1873 | 53 | 639 (‐81.61 to 1359.61) | |
| Percentage of goals recalled | 75 | 28 | 33 | 57 | 32 | 36 | 18.00 (3.84 to 32.16) | |
| Dichotomous data | ||||||||
| Trial | Outcome | Structured goal setting | 'Usual care' goal setting | Risk ratio (95% CI) | ||||
| Events | Total | Events | Total | |||||
| Met or exceeded a goal related to range of movement | 22 | 38 | 8 | 32 | 2.32 (1.20 to 4.47) | |||
| More satisfied or much more satisfied with daily living | 33 | 38 | 26 | 43 | 1.27 (0.94 to 1.70) | |||
| Healthcare use over two years (any visits to doctor, physiotherapists, or other care givers due to pain conditions) | 19 | 28 | 19 | 37 | 1.32 (0.88 to 1.98) | |||
| Achieved or exceeded at least one set goal | 7 | 13 | 7 | 9 | 0.69 (0.38 to 1.28) | |||
| Comparison 4: One structured approach to goal setting and/or strategies to enhance goal pursuit versus another structured approach to goal setting and/or strategies to enhance goal pursuit | ||||||||
| Continuous data | ||||||||
| Collaborative versus prescribed (therapist mandated) goal setting | ||||||||
| Trial | Outcome measure | Collaborative goal setting | Prescribed goal setting | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| No. of treatments required | 13.61 | 8.01 | 23 | 11.46 | 5.4 | 22 | 2.15 (‐1.83 to 6.13) | |
| No. of home exercise sessions complete | 75.43 | 20.88 | 23 | 65.82 | 23.65 | 22 | 9.61 (‐3.45 to 22.67) | |
| Goal setting with versus without operant conditioning | ||||||||
| Trial | Outcome measure | Goal setting plus operant conditioning | Goal setting without operant conditioning | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Rosenberg Self‐Esteem Scale | 1.185 | 1.111 | 27 | 1.577 | 0.703 | 26 | ‐0.39 (‐0.89 to 0.11) | |
| End‐only goals versus end goals with short‐term steps | ||||||||
| Trial | Outcome measure | End‐only goal | End goal with short‐term steps | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Self‐reported dietary fat consumption as a percentage of total energy consumption | 34 | 6 | 4 | 27 | 15.6 | 3 | 7.00 (‐11.61 to 25.61) | |
| Specific versus a non‐specific goal | ||||||||
| Trial | Outcome measure | Specific goals | Non‐specific goals | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Pain behaviour ‐ nonverbal complaint | 1.28 | 1.34 | 13 | 4.3 | 5 | 13 | ‐3.02 (‐5.83 to ‐0.21) | |
| Difficult versus easier goals | ||||||||
| Trial | Outcome measure | Difficult goal (8 servings) | Easy goal (6 servings) | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Total number of servings lower glycaemic index foods per day | 8.4 | 0.83 | 20 | 8.42 | 0.96 | 15 | ‐0.02 (‐0.63 to 0.59) | |
| Goal commitment | 4.09 | 0.12 | 20 | 4.5 | 0.14 | 15 | ‐0.41 (‐0.50 to ‐0.32) | |
| Satisfaction with goal achievement | 6.8 | 0.36 | 20 | 6.47 | 0.41 | 15 | 0.33 (0.07 to 0.59) | |
| Self‐efficacy ‐ total score | 9.3 | 0.28 | 20 | 9.33 | 0.32 | 15 | ‐0.03 (‐0.23 to 0.17) | |
| High intensity exercise goals versus non‐specific intensity exercise goals | ||||||||
| Trial | Outcome measure | High intensity exercise goal | Non‐specific exercise intensity goal | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Total steps during any walking | 6868 | 3751 | 13 | 6279 | 3306 | 17 | 589 (‐1985.37 to 3163.37) | |
| Steps counts during bouts of high intensity walking | 2616 | 2706 | 13 | 2070 | 2814 | 17 | 546 (‐1442.24 to 2534.24) | |
| Duration pedometers worn per day (hours) | 14.5 | 2.49 | 13 | 16.5 | 2.49 | 17 | ‐2.00 (‐3.80 to ‐0.20) | |
| Dichotomous data | ||||||||
| Goal Management Training versus Identity Oriented Goal Mapping | ||||||||
| Trial | Outcome | Goal Management Training | Identity Orientated Goal Mapping | Risk ratio (95% CI) | ||||
| Events | Total | Events | Total | |||||
| Achieved or exceeded at least one set goal | 3 | 8 | 4 | 5 | 0.47 (0.17 to 1.27) | |||
| High intensity exercise goals versus non‐specific intensity exercise goals | ||||||||
| Trial | Outcome | High intensity exercise goal | Non‐specific exercise intensity goal | Risk ratio (95% CI) | ||||
| Events | Total | Events | Total | |||||
| Would definitely recommend the programme to a friend | 8 | 13 | 17 | 17 | 0.62 (0.41 to 0.96) | |||
| Programme considered very useful | 4 | 13 | 12 | 17 | 0.44 (0.18 to 1.04) | |||
COPM = Canadian Occupational Performance Measure; GAS = Goal Attainment Scaling
Comparison 1: Structured goal setting with or without strategies to enhance goal pursuit versus no goal setting
Primary outcomes
Health‐related quality of life or self‐reported emotional status
Eight studies (n = 446) in this comparison group reported data on health‐related quality of life or self‐reported emotional status that could be pooled in a meta‐analysis using standard effect sizes (Blair 1991; Coote 2012; Duncan 2003; Evans 2002; Fredenburgh 1993; Harwood 2012; Scott 2004; Sewell 2005). The meta‐analysis showed an increase in health‐related quality of life or self‐reported emotional status when some form of goal setting (plus or minus strategies to enhance goal pursuit) was used in comparison to no goal setting (standard mean difference (SMD) 0.53, 95% confidence interval (CI) 0.17 to 0.88; Analysis 1.1; Figure 2). This mean effect estimate suggests a moderate clinical effect size in favour of goal setting (Higgins 2011). Substantial statistical heterogeneity in outcome was observed, but with seven out of the eight effect sizes favouring goal setting. Insufficient studies existed to permit meaningful subgroup analysis to further explore reasons for the heterogeneity. Removal of studies at the greatest risk of bias (Coote 2012; Evans 2002) resulted in a minor reduction in the pooled SMD and a widening of the 95% CI, but the result remained in favour of goal setting (SMD 0.45, 95% CI 0.02 to 0.87).
Forest plot of comparison: 1 Goal setting (with or without strategies to enhance goal pursuit) versus no goal setting, outcome: 1.1 Health related quality of life or self‐reported emotional status.
To test for the influence of the one unadjusted cluster‐RCT in this analysis we removed Blair 1991 from the pooled data. This widened the 95% CI for the SMD only a little (SMD 0.53, 95% CI 0.11 to 0.95), so we judged that not including an adjustment for clustering in this RCT had minimal effect on the result of the meta‐analysis.
Two additional studies in this comparison group collected data on health‐related quality of life, but could not be included in the meta‐analysis because no means or SDs were reported (Bell 2003, n = 74, 15% attrition; Iacovino 1997, n = 68, 32% attrition). Both of these studies reported no statistically significant results for these outcomes.
Participation outcomes as defined by the ICF
Four studies in this comparison group reported data on measures of participation as defined by the ICF (WHO 2001a), but the measures used were too dissimilar to permit meta‐analysis (Bell 2003; Howell 1986; Iacovino 1997; Sewell 2005). Bell 2003 (n = 11; 15% attrition) reported a difference in favour of their goal setting intervention in terms of improvements in work performances as measured by Total Work Behaviour Inventory scores (mean difference (MD) 16.0, 95% CI 4.22 to 27.78). Howell 1986 (n = 27; 11% attrition) did not report SDs, but stated that there were no statistically significant differences on the Griffiths Work Performance Scale between the treatment and control groups at the end of intervention. Iacovino 1997 (n = 68; 32% attrition) reported no difference between goal setting and non‐goal setting groups in terms of numbers returning to work three to six months after the intervention (RR 0.78, 95% CI 0.56 to 1.08). Sewell 2005 (n = 180; 33% attrition) also reported no difference between goal setting and non‐goal setting groups in self‐reported satisfaction with occupational performance after seven weeks of pulmonary rehabilitation (MD ‐0.23, 95% CI ‐0.93 to 0.47).
Activity outcomes as defined by the ICF
Four studies (n = 223) reported data from measures of activity as defined by the ICF (WHO 2001a), which we combined in a meta‐analysis using standard effect sizes (Duncan 2003; Harwood 2012; O'Brien 2013; Sewell 2005). The meta‐analysis showed no difference in activity levels when some form of goal setting (plus or minus strategies to enhance goal pursuit) was used in comparison to no goal setting (SMD 0.04, 95% CI ‐0.22 to 0.31; Analysis 1.2; Figure 3). There was no evidence of statistical heterogeneity in the observed outcomes.
Forest plot of comparison: 1 Goal setting (with or without strategies to enhance goal pursuit) versus no goal setting, outcome: 1.2 Activity ‐ ability.
Secondary outcomes
Outcomes at the level of body structure or body function as defined by the ICF
Data from measures at the level of body function as defined by the ICF (WHO 2001a) were reported in five studies (Bassett 1999; Mann 1987; Harwood 2012; Coppack 2012; Duncan 2003). The measures used were too dissimilar to permit meta‐analysis, and thus are presented descriptively instead.
Bassett 1999 (n = 74; 15% attrition) reported no difference in self‐reported symptom relief. These data were collected on a 3‐point Likert‐type scale, but analysed as a continuous variable (presented as means and SDs). Insufficient information about this data set was reported to treat it as categorical data as per our protocol. Mann 1987 (n = 66; 15% attrition) reported no difference between their groups receiving self‐management education with goal setting and self‐management education without goal setting in terms of urinary sodium output three months after the training (MD 16.0 mmol/24 h, 95% CI ‐19.2 to 51.2) or in terms of systolic blood pressure at three months following the training (MD ‐2.60 mmHg, 95% CI ‐13.73 to 8.53). Similarly, Harwood 2012 (n = 85 in the 'Take Charge' and control groups combined; 19% attrition) reported no difference between their 'Take Charge' (goal setting) group and control group on the basis of systolic blood pressure (MD ‐3.5 mmHg, 95% CI ‐12.15 mmHg to 5.15 mmHg).
Coppack 2012 (n = 48; no attrition) reported a difference between their goal setting and control groups both at baseline and at the end of treatment in terms of back muscle endurance (as measured by the Biering‐Sørensen test), with the goal setting group performing better on this test at both time points. However, Coppack 2012 reported no statistically significant difference between their three groups (one goal setting group and two control groups) in terms of change in back muscle endurance scores between baseline and treatment end on the basis of multivariate analysis of covariance. As no SDs were reported for change scores at a group level in Coppack 2012, a 95% CI could not be reported in this review for the MD in change in the Biering‐Sørensen test.
Similarly, Duncan 2003 (n = 16; 6% attrition) reported on data indicating a difference between their goal setting and control groups both at baseline and 24 weeks after study enrolment in terms of maximum ventilated oxygen, with the control group achieving higher maximum ventilated oxygen at both time points. However, a 95% CI could not be calculated for this review for the MD in change scores for maximum ventilated oxygen as no SDs were reported for change scores at a group level. Duncan 2003 also reported no differences between their two groups at 24 weeks after study enrolment on the basis of self‐reported dyspnoea as measured by the Baseline Dyspnea Index (MD 1.6, 95% CI ‐0.13 to 3.33) or fatigue measured by the Piper Fatigue Scale (MD ‐0.4, 95% CI ‐2.29 to 1.49).
Finally, one study also collected data from a criterion‐reference achievement test (Mann 1987; n = 66; 15% attrition). The test was designed for participants with hypertension and included questions testing knowledge of the condition and its management as well as affective responses (e.g. feelings of support; feelings of self‐competence). Three months after enrolment in the study, there was no difference between the dietary education plus goal setting group compared to the group receiving dietary education without goal setting on this test (MD 1.7 on a test with a maximum score of 23, 95% CI ‐0.22 to 3.62).
Patient self‐belief and engagement in rehabilitation
a) Adherence and engagement in rehabilitation
Nine studies (n = 369) in this comparison group reported data on measures of patient engagement that could be pooled in a meta‐analysis using standard effect sizes (Bassett 1999; Bell 2003; Coppack 2012; Cross 1971; Duncan 2003; Evans 2002; Iacovino 1997; Mann 1987; O'Brien 2013). The meta‐analysis showed no difference in patient engagement in rehabilitation when some form of goal setting (plus or minus strategies to enhance goal pursuit) were used in comparison to no goal setting (SMD 0.30, 95% CI ‐0.07 to 0.66; Analysis 1.3; Figure 4). There was evidence of substantial heterogeneity in the observed outcomes both in terms of the size and direction of effects. Insufficient studies existed to permit meaningful subgroup analysis to further explore reasons for the heterogeneity.
Forest plot of comparison: 1 Goal setting (with or without strategies to enhance goal pursuit) versus no goal setting, outcome: 1.3 Engagement in rehabilitation.
One additional study in this comparison group reported on engagement in rehabilitation, but could not be included in the meta‐analysis because no means or SDs were reported (Howell 1986; n = 27; 11% attrition). This study reported no statistically significant results for this outcomes.
One further study (n = 367; 30% attrition) in this comparison group reported on patient adherence as a dichotomous variable only, reporting the percentage of participants with a mental health condition who adhered to a prescribed medication regime (Stanhope 2013). These data were calculated as overall rates of medication adherence per month for 11 months. To accommodate unit of analysis issues in this‐cluster‐RCT, Stanhope 2013 reported using logistic regression models while including a random effect for site to calculate overall medication adherence. The reported odds ratio (OR) for medication adherence over time for the whole study population was 1.13 (95% CI 1.08 to 1.19) (author communication), indicating a very small effect size in favour of the goal setting intervention.
b) Self‐efficacy
Three studies (n = 108) in this comparison group reported data from measures of task‐specific self‐efficacy following rehabilitation that we could combine in a meta‐analysis using standard effect sizes (Coppack 2012; Evans 2002; O'Brien 2013). The meta‐analysis showed a difference in self‐efficacy at the end of rehabilitation in favour of goal setting (with and without strategies to enhance goal pursuit) in comparison to no goal setting (SMD 1.07, 95% CI 0.64 to 1.49; Analysis 1.4). This mean effect estimate suggests a large effect size (Higgins 2011). There was little evidence of heterogeneity in these data.
One additional study in this comparison group reported collecting data on general self‐efficacy, but reported no means or SDs for this outcome (Scott 2004; n = 88; 36% attrition). This study reported no statistically significant results for this outcomes.
Individual goal attainment
Three studies reported on individual goal attainment as a study outcome (Blair 1991; Blair 1996; Howell 1986). In these studies goals were set by researchers for the participants in both the intervention and control groups, but the control group participants (and, in the case of Howell 1986 their treating health professionals) were not made aware of these goals and the goals were not used during the delivery of rehabilitation interventions. In Howell 1986 these goals were thus called 'theoretical' goals. In all these studies goal attainment was reported as GAS scores and was treated as interval data only (reported as means and SDs) rather than ordinal data (as required by our protocol; see Measures of treatment effect), so no results could be extracted for the purposes of this review. However, Howell 1986 (n = 27; 11% attrition) also reported on goal attainment as a dichotomous variable, categorising participants as either having reached 'high' goal achievement or 'low' goal achievement on their theoretical GAS scale. While it was not clearly reported what 'high' goal achievement represented in this study, Howell 1986's data indicated no difference between the two groups at study end when categorised in this way (risk ratio (RR) 0.97, 95% CI 0.52 to 1.80).
Evaluation of care
No data were reported in studies in this comparison group on evaluation of care (such as satisfaction with service delivery).
Service delivery level
One study in this comparison group, set in the context of physiotherapy for musculoskeletal disorders, reported on the number of treatment sessions required to achieve symptom relief (Bassett 1999; n = 74; 11% attrition). No differences were founded when the pooled means and SDs for the two goal setting groups in this study (participant‐physiotherapist collaborative goal setting and the physiotherapist‐mandated goal setting) were compared to the control group who received no goal setting on the basis of the number of treatment sessions required to achieve symptom relief (MD ‐0.73 sessions, 95% CI ‐5.85 to 4.39).
Adverse events
One study (Harwood 2012; n = 172) reported on the number of participants in each group who died prior to final data collection. No difference in risk of death was observed between the goal setting ('Take charge') group and control group in this study (RR 0.68, 95% CI 0.20, 2.35). However, as the 95% CIs for this comparison are wide, it is not possible to rule out differences in mortality between the two groups without considerably more data.
Comparison 2: Structured approach to goal setting with or without strategies to enhance goal pursuit versus 'usual care' that may have involved some goal setting but where no structured approach was followed
Primary outcomes
Health‐related quality of life and self‐reported emotional status
Five studies (n = 441) in this comparison group reported data on health‐related quality of life or self‐reported emotional status that could be pooled in a meta‐analysis using standard effect sizes (Cheng 2012; LaFerriere 1978; Ostelo 2003; Parsons 2012; Taylor 2012). The meta‐analysis showed no difference in health‐related quality of life or self‐reported emotional status when a structured approach to goal setting (plus or minus strategies to enhance goal pursuit) was compared to usual care without a structured approach to goal setting (SMD 0.18, 95% CI ‐0.19 to 0.55; Analysis 2.1; Figure 5). Substantial heterogeneity was observed. Insufficient studies existed to permit meaningful subgroup analysis to explore reasons for the heterogeneity. Removal of studies at the greatest risk of bias (Cheng 2012) shifted the SMD a little more in favour of structured goal setting, and widened the CIs for this estimate a little, but did not change the conclusion of this analysis (SMD 0.27, 95% CI ‐0.16 to 0.70). Removal of Cheng 2012 also suggested that lack of adjustment for the effects of clustering in this cluster‐RCT had minimal influence on the outcome of this meta‐analysis.
Forest plot of comparison: 2 Structured goal setting (with or without strategies to enhance goal pursuit) versus no structured goal setting, outcome: 2.1 Health related quality of life or self‐reported emotional status.
One additional study in this comparison group, Holliday 2007 (a quasi‐RCT; n = 201; no attrition), reported on health‐related quality of life as measured by the General Health Questionnaire, but could not be included in the meta‐analysis because no means or SDs were reported. Nonetheless this study reported a statistically significant difference in favour of their usual care group for health‐related quality of life (P = 0.037; no effect size reported).
In addition to studies that reported on continuous outcome data, one study in this comparison group reported on overall 'satisfaction with daily living' using ordinal data (Asenlof 2005). When these data were dichotomised, the participants in the structured goal setting intervention were more likely to report being more satisfied or much more satisfied with their life (versus no difference or less satisfied with their life) compared to participants in the usual care group three months after the intervention (RR 1.44, 95% CI 1.09 to 1.88). However, this difference did not remain two years after intervention (RR 1.27, 95% CI 0.94 to 1.70).
Participation outcomes as defined by the ICF
One study, Holliday 2007 (a quasi‐RCT; n = 201; no attrition), reported on outcomes from a single measure of participation as defined by the ICF (WHO 2001a) ‐ the London Handicap Scale. However, while the authors of this study reported that there was no difference between the two study groups on this measure, no means or SDs were published.
Activity outcomes as defined by the ICF
Four studies (n = 277) in this comparison group reported data on activity outcomes as defined by the ICF (WHO 2001a) that could be pooled in a meta‐analysis using standard effect sizes (Asenlof 2005; Cheng 2012; Ostelo 2003; Taylor 2012). The meta‐analysis showed no difference in activity levels when a structured approach to goal setting (plus or minus strategies to enhance goal pursuit) was compared to usual care without a structured approach to goal setting (SMD 0.17, 95% CI ‐0.15 to 0.49; Analysis 2.2). Moderate heterogeneity was observed. Insufficient studies existed to permit meaningful subgroup analysis to further explore reasons for the heterogeneity. Removal of studies at the greatest risk of bias (Cheng 2012) reduced the SMD (i.e. less in favour of goal setting) and widened the 95% CI for this estimate a little, but did not change the conclusion of this analysis (SMD 0.07, 95% CI ‐0.30 to 0.44). Removal of Cheng 2012 also suggested that lack of adjustment for the effects of clustering in this cluster‐RCT had minimal influence on the outcome of this meta‐analysis.
One additional study in this comparison group, Holliday 2007 (a quasi‐RCT; n = 201; no attrition), reported on activity levels as measured by the Functional Independence Measure, but could not be included in the meta‐analysis because no means or SDs were reported. This study reported no statistically significant difference between their structured goal setting and usual care groups.
Two further studies reported on activity data as median and percentile scores only, treating the data as ordinal measures (Gagné 2003; Oestergaard 2012). Gagné 2003 (n = 31; no attrition) reported a difference in favour of structured goal setting in comparison to usual care in Functional Independence Measure subscores for upper body dressing after two weeks of occupational therapy (Mann‐Whitney U test, P < 0.019), but not for five other subscores on activities of daily living: eating, grooming, lower‐body dressing, toileting, and bathing. Oestergaard 2012 (n = 87; 31% attrition) reported no differences between their structured goal setting and usual care groups in terms of self‐rated performance and satisfaction with performance on a list of 18 activities of daily living when using Wilcoxon rank sum to test for differences.
Secondary outcomes
Outcomes at the level of body structure or body function as defined by the ICF
Three studies reported outcomes at the level of body function as defined by the ICF (WHO 2001a) that could be reported descriptively but not pooled in a meta‐analysis (Asenlof 2005; Oestergaard 2012; Ostelo 2003). Asenlof 2005 (n = 122; 47% attrition) reported on maximum pain and pain control on 10‐point Likert‐type scales two years after physical therapy for a persistent pain problem. When the baseline values for these measures were used as covariates in the analyses, Asenlof 2005 reported that there was a difference in favour of the structured goal setting intervention for maximum pain but not pain control. When the reported data were analysed without adjustment for baseline values however, there was no difference for either measure at two years (maximum pain MD 1.30, 95% CI ‐0.24 to 2.84). Oestergaard 2012 (n = 87; 31% attrition at three years) reported on pain, as measured by the Dallas Pain Questionnaire, experienced by participants three years after multidisciplinary rehabilitation for lumbar spinal fusion for degenerative disc disease. This was reported as ordinal data, with a Wilcoxon rank sum test identifying no difference between the structured goal setting and usual care groups for total Dallas Pain Questionnaire scores at three years (P = 0.38). Ostelo 2003 (n = 105; 11% attrition) reported on three measures of body function (severity of back pain, severity of sciatica, and range of lumbar spine movement) in people following physiotherapy for persistent back pain following lumbar disc surgery. No differences were observed between the structured goal setting and usual care group for any of these measures of body function, with change in range of lumbar spine movement (in degrees) being the measure with the median effect size (MD ‐1.20°, 95% CI ‐9.66° to 7.26°)
Patient self‐belief and engagement in rehabilitation
a) Adherence and engagement in rehabilitation
Only one study (LaFerriere 1978; n = 65; 51% attrition) reported on measures of patient engagement in rehabilitation, specifically therapist‐rated scores and patient‐rated scores of patient motivation for people with mental health conditions participating in a series of individual behaviour therapy or psychotherapy sessions. LaFerriere 1978 reported a small difference in favour of structured goal setting for patient‐rated motivation (MD 1.40, 95% CI 0.43 to 2.37), but not for therapist‐rated scores of motivation (MD 0.48, 95% CI ‐0.41 to 1.37) after completion of the therapy.
b) Self‐efficacy
Two studies (n = 134) reported on general self‐efficacy as an outcome following rehabilitation (Asenlof 2005; Cheng 2012). When combined in a meta‐analysis using standard effect sizes and the last recorded, unadjusted self‐efficacy data from both Asenlof 2005 and Cheng 2012, a difference in favour of the structured goal setting interventions was found (SMD 0.37, 95% CI 0.02 to 0.71; Analysis 2.3), indicative of a small effect size. The accuracy of this estimate is questionable however, given the lack of ability to adjust for clustering in Cheng 2012.
c) Kinesiophobia
Kinesiophobia was reported in two studies involving people receiving physiotherapy for persistent pain problems (Asenlof 2005; Ostelo 2003). Neither study reported any difference between their structured goal setting group and usual care group in terms of Tampa Scale of Kinesiophobia scores, which were reported as end values for Asenlof 2005 (MD 2.5, 95% CI ‐0.69 to 5.69) and change from baseline values in the case of Ostelo 2003 (MD 0.1, 95% CI ‐2.33 to 2.53).
Individual goal attainment
Levels of goal attainment were reported in five studies (Arnetz 2004; Cheng 2012; Hart 1978; Jonsdottir 2012; McPherson 2009). However, the diversity of methods used for collection and reporting of goal attainment data meant that undertaking a meta‐analysis on this outcome was not possible. Arnetz 2004 (n = 77; 3% attrition) reported on the percentages of patients achieving their goals within subgroups of patients with similar types of goals (e.g. people with goals relating to pain, range of movement, strength, balance), but as information was missing on the number of patients in the structured goal setting group versus the usual care group within each subgroup, RRs could not be calculated for most of these data. For Arnetz 2004's data on attainment of goals related to range of movement however, we were able to extrapolate sample sizes from other data within the text and from a Chi2 test reported in the paper. From these we were able to calculate that, for those with range of movement goals, people in the structured goal setting group were more likely to achieve their goals when compared to people within the usual care group (RR 2.32, 95% CI 1.20 to 4.47). For Cheng 2012 (n = 96; 28% attrition), goal attainment was measured as the percentage of goals achieved by each patient within each group (structured goal setting versus usual care). On this basis, participants in the structured goal setting group were found to achieve a higher percentage of their goals than people in the usual care group at 24 weeks after enrolment (MD 36.5%, 95% CI 18.43% to 54.57%; results unadjusted for clustering effects as no intraclass correlations for this measure were available). For Hart 1978, goal attainment was recorded as GAS scores. However these scores were treated as interval data (reported as means and SDs) rather than ordinal data (as required by our protocol; see Measures of treatment effect), so no results could be extracted. Jonsdottir 2012 reported on goal attainment at the level of goals per group in the study, with some participants having more than one goal. As goal attainment was not reported at a participant level, however, no data could be extracted. McPherson 2009 (n = 34; 35% attrition) also reported on goal attainment on the basis of GAS scores. In this case however, raw data for each participant was available from the researchers (author communication), and were dichotomised. On this basis, seven out of 13 participants in the two structured goal setting groups (Goal Management Training and Identity Oriented Goal Mapping) achieved or exceeded at least one of their set goals (with each participant having between one and three goals) in comparison to seven out of nine in the usual care group. No significant difference between the groups in terms of goal attainment was observed (RR 0.69, 95% CI 0.38 to 1.28).
Goal recall was also reported in one study (Woltmann 2011; n = 80; 14% attrition) involving people participating in community‐based interventions for mental health conditions. Recall of goal was evaluated in terms of the percentage of goals that could be reported back to researchers by participants two to four days after a goal planning meeting. In this study, the participants in the structured goal setting group recalled a higher percentage of their goals when compared to participants in the usual care group (MD 18%, 95% CI 3.84% to 32.16%; results unadjusted for clustering effects as no intraclass correlations for this measure were available).
Evaluation of care
Five studies (n = 309) reported data on satisfaction with care that could be pooled in a meta‐analysis using standard effect sizes (Arnetz 2004; Cheng 2012; LaFerriere 1978; Taylor 2012; Woltmann 2011). The meta‐analysis showed a statistically significant difference in patient satisfaction in favour of structured goal setting (with and without strategies to enhance goal pursuit) in comparison to usual care (SMD 0.33, 95% CI 0.10 to 0.56; Analysis 2.4; Figure 6, indicative of a small effect size) (Higgins 2011). There was no evidence of heterogeneity in these data. When we excluded studies at high risk of bias due to lack of adequate randomisation (Arnetz 2004) or lack of concealment of random allocation (Arnetz 2004; Cheng 2012) from the meta‐analysis, the pooled SMD was no longer statistically significant. If we just excluded studies for which we could not account for the effects of clustering (Cheng 2012; Woltmann 2011) the pooled SMD remained statistically significant, albeit with a wider 95% CI (SMD 0.42, 95% CI 0.03 to 0.82).
Forest plot of comparison: 2 Structured goal setting (with or without strategies to enhance goal pursuit) versus no structured goal setting, outcome: 2.4 Satisfaction with service delivery.
One additional study, Holliday 2007 (a quasi‐RCT; n = 201; no attrition), reported on participant satisfaction scores, but could not be included in the meta‐analysis because no means or SDs were reported. Nonetheless, this study reported a statistically significant difference in favour of structured goal setting when compared to usual care (P < 0.001).
Service delivery level
Five studies evaluated the extent of use of healthcare services (Asenlof 2005; Cheng 2012; LaFerriere 1978; Oestergaard 2012; Ostelo 2003). However, the diversity of methods used for collection and reporting on healthcare utilisation meant that we were unable to undertake a meta‐analysis on this outcome. Asenlof 2005 reported on healthcare utilisation in the two years following intervention as a dichotomous variable (counting any visit to a doctor, physiotherapist or other health professional as evidence of ongoing access to healthcare). No difference was observed between the structured goal setting and usual care groups for this measure at two years (RR 1.32, 95% CI 0.88 to 1.98). Cheng 2012 also reported no statistically significant difference between structured goal setting and usual care group in the number of emergency rooms visits, number of unplanned hospital readmissions, and length of hospital stays within 24 weeks. However, no means and SD were reported for these measures by Cheng 2012 for this measure. LaFerriere 1978 reported no difference between their structured goal setting and usual care group for the average number of therapy sessions each group received (MD 3.09 sessions, 95% CI ‐0.66 to 6.84). Similarly, Oestergaard 2012 collected data on hours of occupational therapy received. Oestergaard 2012 did not publish the means and SDs, but did report that there was no statistically significant difference between their structured goal setting and usual care groups in terms of occupational therapy hours provided. Ostelo 2003 completed a full economic analysis of the direct and indirect costs associated with healthcare and disability in the year following completion of their intervention. They too reported no difference between their structured goal setting and usual care groups in terms of total healthcare costs one year following intervention (MD EUR 639, 95% CI ‐EUR 81.61 to EUR 1359.61).
Adverse events
Three studies (n = 406) reported on the number of participants who withdrew from the studies due to adverse events: death (Cheng 2012; Parsons 2012); hospitalisation (Cheng 2012), or worsening symptoms (Ostelo 2003). When combined in a meta‐analysis, there was no difference between structured goal setting and usual care groups in terms of the number of participants who withdrew from the studies due to adverse events of any cause (Peto Odd's Ratio 0.64, 95% CI 0.27 to 1.47; Analysis 2.5). However, given that this meta‐analysis was based on raw data from two cluster‐RCTs for which the effects of clustering could not be accounted (Parsons 2012; Cheng 2012), and given that the 95% CI was wide even without adjusting for the effects of clustering, it was not possible to rule out differences in mortality between the two groups without considerably more data.
Comparison 3: Interventions to enhance goal pursuit versus no interventions to enhance goal pursuit
We identified two studies where the intervention and control groups differed solely in terms of the strategies used to enhance goal pursuit; the methods used for goal selection were identical for all participants (Culley 2010; Hart 2002). Both studies were at unclear risk of bias, both involved participants with traumatic brain injury, and both investigated the effectiveness of an intervention involving information technology (a Portable Voice Organizer in the case of Hart 2002; text messaging in the case of Culley 2010) to improve patient recall of rehabilitation goals. In both cases randomisation occurred at the level of the goals rather than at the level of participants. Each participant was required to have six goals for rehabilitation, either in order to enrol in the study (Culley 2010) or set as part of the study (Hart 2002). These goals were then randomised (three to each condition) to the intervention condition (to be communicated three times daily to participants via the information technology under investigation) or to the control group (no additional prompting about these goals was provided). Outcomes were measured in terms of the participants' recall of goals.
Two forms of recall were tested: open recall (where patients reported on their goals without any prompting) followed by cued recall (where patients were given pre‐negotiated key words to help them remember their goals). Recall was scored on a 4‐point scale from 0 (no recall) to 3 (complete recall) for each goal, resulting in a total score from 0 to 9 for each condition, for each participant. A full set of raw data for this outcome was published in Hart 2002 (n = 10) and accessed via communication with the authors in the case of Culley 2010 (n = 11).
We combined raw data on open recall from these two studies (n = 21) using an individual subject meta‐analysis. In both cases, the data used in the meta‐analysis were from the longest period of time following participant enrolment (one week for Hart 2002; two weeks for Culley 2010). We used a mixed linear model to estimate the differences between the intervention and control conditions for these two studies, and found a difference in favour of the use of information technology to facilitate the participants' open recall of rehabilitation goals (MD 3.1, 95% CI 2.04 to 4.15; P < 0.001).
No data were reported in these studies on health‐related quality of life or self‐reported emotional status, participation level outcomes, activity level outcomes, outcomes related to body structure or function, patient self‐belief or engagement in rehabilitation, individual goal attainment, evaluation of care, service delivery level or adverse outcomes.
Comparison 4: One structured approach to goal setting and/or strategies to enhance goal pursuit versus another structured approach to goal setting and/or strategies to enhance goal pursuit
Nine trials compared different structured approaches to goal setting and/or strategies to enhance goal pursuit (see Included studies: Comparison groups).
Collaboratively set versus therapist‐mandated goal setting
Bassett 1999 (n = 74; 11% attrition) compared outcomes when musculoskeletal physiotherapy goals were set collaboratively between the patient and treating physiotherapist versus when goals were prescribed by the physiotherapist without patient input. No differences between these two groups were reported for the number of treatment sessions required to achieve symptom relief (MD 2.15 sessions, 95% CI ‐1.83 to 6.13) or in the patients' self‐reported home exercise adherence in terms of the percentage of prescribed sessions completed (MD 9.61% of sessions, 95% CI ‐3.45% to 22.67%). Also reported was no difference between the two groups for self‐reported symptom relief as measured on a 3‐point ordinal scale. However insufficient information was available to report relative risk arising from these data.
Goal setting with operant conditioning versus goal setting without operant conditioning
Both Blair 1991 (n = 89; 11% attrition) and Blair 1996 (n = 15; no attrition) were conducted within nursing care homes. Both included two intervention groups where the approach to goal setting was identical (based on GAS methods), but where one of these two groups received additional operant conditioning interventions to enhance goal pursuit and the other did not. All goals focused on increasing the residents' independence in morning activities of daily living, with targeted activities selected as goals following collaborative discussion between the nursing staff and residents. The operant conditioning interventions included prompting (i.e. reminders to do target activities), shaping (i.e. helping with set up of activities), and reinforcement. Reinforcement included smiles, praise, affectionate touches, and conversation when residents completed or partly completed the target activities, and a withdrawing of these responses when residents refused to do a task for no clear reason. After six weeks of nursing intervention and at a 22‐week follow‐up, GAS scores were collected for participants in both Blair 1991 and Blair 1996. However, in both these studies GAS measures were treated as interval data (reported as means and SDs) rather than ordinal data (as required by our protocol; see Measures of treatment effect), so no results could be extracted for the purposes of this review. In addition to GAS scores, Blair 1991 reported no difference between the two goal setting groups on the basis of Rosenburg Self‐Esteem scores at the 22‐week follow‐up (MD ‐0.39, 95% CI ‐0.89 to 0.11; results unadjusted for clustering effects as no intraclass correlations for this measure were available).
Setting an end goal only versus setting incremental short‐term steps towards an end goal
Conrad 2000 reported on a small RCT (n = 7; no attrition) where patients with coronary artery disease were randomised to receive either an immediate goal of decreasing their dietary fat to 10% of total energy consumption or a series of short‐terms goals gradually reducing their dietary fat to 10% of total energy consumption over a four‐month period. After seven months, no differences were observed between the two groups on the basis of self‐reported dietary fat consumption, cholesterol consumption, protein consumption, body weight, or serum cholesterol. For self‐reported dietary fat consumption as a percentage of total energy consumption (the primary outcome in the hypothesis for testing) the difference in means after four months was 7.0% (95% CI ‐11.6% to 25.6%).
Setting a specific goal versus setting a non‐specific goal
James 1993 (n = 37 in the two cognitive behaviour therapy groups; 30% attrition) recruited patients participating in cognitive behavioural therapy for chronic headache, and randomised patients to receive either a specific, time‐limited goal to practice pain‐coping strategies for a prescribed duration and frequency on a daily basis over a six‐week period or the non‐specific goal of practicing pain‐coping strategies 'as much as possible for as long as possible' (p. 310) over the six‐week period. Multiple measures were used, resulting in 21 different outcome variables that James 1993 tested for significance (one for pain severity, six for medication use, one for self‐efficacy, one for activity levels, four for pain behaviour, one for pain‐related disability, four for psychological states and traits, and three for coping). No differences between the specific and non‐specific goal groups were found after therapy except for three of the four pain behaviour measures, which all favoured the specific goal group. The pain behaviour measure with the median effect size was for the nonverbal complaints subscore on the Pain Behavior Questionnaire (MD ‐3.02, 95% CI ‐5.83 to ‐0.21).
Goal Management Training versus Identity Oriented Goal Training
McPherson 2009 described a pilot study (n = 34; 35% attrition) in which participants with traumatic brain injury were randomised to one of three groups: 1) Goal Management Training, 2) Identity Oriented Goal Training, or 3) usual care. The two goal setting approaches were based on self‐regulation theory and designed to address impairments in self‐regulation secondary to brain injury. Goal Management Training involved identification and documentation of a rehabilitation goal and the development and rehearsal of planned steps to achieve that goal, with an emphasis on errorless learning (i.e. avoidance of goal failure). Identity Oriented Goal Training involved development of an 'identity map' to use as a tool for identification and articulation of goals that would help participants connect with and progress towards a meaningful, higher‐order life goal. Outcomes from GAS scores were evaluated after eight weeks of therapy and at a three‐month follow‐up assessment. Based on raw data from each participant (accessed via author communication) three out of eight participants in Goal Management Training and four out of five participants in the Identity Oriented Goal Training groups achieved or exceeded at least one of their set goals (with each participant having between one and three goals), with no difference observed between the two groups in terms of goal attainment (RR 0.47, 95% CI 0.17 to 1.27).
Setting a difficult goal versus setting an easier goal
In Miller 2012 (n = 46; 24% attrition) people with type 2 diabetes participated in a 5‐week dietary intervention aimed at increasing their consumption of lower glycaemic index foods. Participants were randomised to receive a specific, difficult goal (to change their dietary behaviour to include eight servings per day of lower glycaemic index foods) or a specific, easier goal (to change their dietary behaviour to include six servings per day of lower glycaemic index foods). After eight weeks both groups had increased their consumption of lower glycaemic index foods from baseline (based on four‐day self‐report) and there was no significant difference between the two groups in terms of total servings per day of lower glycaemic index foods (MD ‐0.02 servings, 95% CI ‐0.63 to 0.59). Participants in the group receiving the specific difficult goal were less committed to their treatment goal, as measured on a 5‐point scale (MD 0.41, 95% CI 0.32 to 0.50), but there was no reported difference between the two groups for the participants' level of satisfaction or dissatisfaction with meeting their goal, as measured on a 9‐point scale (MD 0.33, 95% CI 0.07 to 0.59). Furthermore, there was no difference between the two groups in terms of task‐specific self‐efficacy (MD 0.03, 95% CI ‐0.17 to 0.23).
Setting a high intensity exercise goal versus setting a goal of non‐specific exercise intensity
Richardson 2007 (n = 35; 14% attrition) investigated the effect of different types of exercise goals for people with type 2 diabetes participating in a six‐week home‐based exercise programme delivered via the Internet with the use of a pedometer to monitor exercise duration and intensity. The participants were randomised to either a 'lifestyle goal' or a 'structured goal'. The lifestyle goal‐centred on increasing total daily step counts regardless of exercise intensity (a goal of non‐specific exercise intensity); the structured goal involved increasing total daily step counts for high‐intensity exercise only (a high‐intensity exercise goal). High‐intensity exercise was defined as exercise that occurred for a minimum of ten minutes at a time, with at least 60 steps per minute. Both lifestyle goals and structured goals were automatically assigned and altered weekly by the Internet interface based on the previous week's step count data (uploaded via the pedometer; starting with one week of baseline recording). Lifestyle goals were automatically set at 1200 steps more per day than the previous weeks' average daily step count, up to a maximum of 10,000 steps per day. Structured goals were automatically set at 800 steps more per day of high‐intensity exercise only, compared to the previous weeks' average daily step count for high‐intensity exercise, up to a maximum of 10,000 steps per day. At the end of the study, there was no difference between the two groups in terms of average total daily step counts (MD 589 steps, 95% CI ‐1985 to 3163 steps) or average total daily step counts involving high‐intensity exercise only (MD 546 steps, 95% CI ‐1442 to 2534). However, the participants in the structured goal group were found to be less satisfied with the exercise programme. Only 62% (8/13) of the structured goal group were inclined to definitely recommend the programme to a friend in comparison to 100% (17/17) of the lifestyle goal group (RR 0.62, 95% CI 0.41 to 0.96), and only 31% (4/13) of the structured goal group reported finding the programme 'very useful' compared to 71% (12/17) of the lifestyle goal group (RR 0.44, 95% CI 0.18 to 1.04). The lifestyle goal group also wore their pedometers for more hours each day than did the structured goal group (MD 2 hours, 95% CI 0.2 to 3.8 hours).
High level versus low level patient involvement in goal setting
In Webb 1994 (n = 16; no attrition) people participating in a day hospital or residential rehabilitation programme for traumatic brain injury were randomised to either a high level or a low level of involvement in goal setting. Both groups of participants were formally oriented to goal setting in rehabilitation at the beginning of the study and both were involved in prioritisation of their rehabilitation goals, with therapists converting the participants' chosen goals into a GAS scale to be achieved over the following eight weeks. The two groups differed in that the high involvement group also participated in a discussion of the importance of goal setting and were encouraged to ask questions about goal setting at the beginning of rehabilitation, whereas the low involvement group were not. Furthermore, the high involvement group had options for rehabilitation goals presented to them as possible goals written on wooden blocks that the participants could use to help order and communicate their preferences. In comparison, the low involvement group selected goals from a list written on paper. Finally, the high involvement group were encouraged to discuss their goals on a weekly basis for the duration of the study, with progress towards goals regularly documented by and for the participants on worksheets and in a goal diary. Outcomes on the basis of GAS scores were evaluated after eight weeks of rehabilitation. However, as GAS measures were treated as interval data (reported as means and SDs) rather than ordinal data (as required by our protocol; see Measures of treatment effect), no results could be extracted for the purposes of this review.
Discussion
Summary of main results
The results from data on the effects of goal setting interventions were divided into four sections. The first of these comprised studies where, within the context of a rehabilitation intervention, any type of goal setting (with or without strategies to enhance goal pursuit) was compared to the same rehabilitation intervention provided without goal setting.
Based on the GRADE assessment (summary of findings Table for the main comparison), there is very low quality evidence that goal setting results in a moderate increase in health‐related quality of life or self‐reported emotional status. There is also very low quality evidence that goal setting results in a large increase in patient self‐efficacy. However, we found no evidence of an effect of goal setting (with or without strategies to enhance goal pursuit) on activity, improvements in body structure or body function, or levels of engagement in rehabilitation, although these conclusions were based on a small number of studies with an overall moderate risk of bias. There is insufficient data to draw conclusions regarding the possibility of adverse events arising from goal setting interventions.
The second section of our results comprised studies where, within the context of a rehabilitation intervention, a structured approach to goal setting (with or without strategies to enhance goal pursuit) was compared to 'usual practice' in which some goals were set, but no structured or required approach to goal setting was followed. Conclusions from these studies are more difficult to reach as most studies were not explicit regarding what goal setting in usual care typically involved. This makes it difficult to be certain how the structured goal setting interventions that the experimental groups received differed from the process of goal setting provided to participants in the control groups. However, broadly speaking, the structured goal setting interventions appeared to involve more patient participation in goal selection, were more person‐centred, and focused more on personally meaningful outcomes that patients wanted to achieve, with some studies also including additional strategies to enhance patient behaviour directly related to goal pursuit.
Based on the GRADE assessment (summary of findings Table 2), there is very low quality evidence that the more structured approaches to goal setting result in a small to moderate increase in patient self‐efficacy. There is also low quality evidence that patients are more satisfied with service delivery if receiving a more structured approach to goal setting in comparison to usual care. However, no evidence was found for the effect of structured goal setting in comparison to usual care with regard to health‐related quality of life, patient‐reported emotional status, or activity levels, although these conclusions are based on a small number of studies with an overall moderate risk of bias. Insufficient information exists to draw conclusions regarding the effects of structured goals on outcomes at the level of social participation or patient engagement in rehabilitation, or regarding the risk of adverse events arising from structured goal setting.
The third section of our results included studies where, within the context of a rehabilitation intervention, one approach to enhancing goal pursuit was compared to another. We identified only two studies in this group of trials, both of which investigated the same strategy to enhance goal pursuit (i.e. use of information technology to help patients recall their goals). These studies were at unclear risk of bias but involved a total of only 21 participants. When individual patient data from these studies were pooled there was low quality evidence that using information technology (text messaging or portable voice organisers) improved the ability of people with brain injury to recall their goals for rehabilitation.
The final section of our results contained studies where one structured approach to goal setting (with or without strategies to enhance goal pursuit) was compared to another structured approach. Each of the studies in this section involved testing a different intervention regarding goal setting or goal pursuit (e.g. goal specificity, goal difficulty, interventions to address impairment of self‐regulation related to goal pursuit, and so forth), with outcomes being evaluated through a wide range of measures. All results for this group of studies are based on single trials, with small sample sizes, and are at an overall moderate risk of bias. As a consequence, there is insufficient information overall to draw any firm conclusions regarding the effects of various structured approaches to goal setting and goal pursuit on health outcomes.
Overall completeness and applicability of evidence
This review is based on a comprehensive search strategy without restrictions on language or publication status. The definitions we used for 'acquired disability' and 'rehabilitation' when making decisions about whether or not to include studies in this review were broad, so erred on the side of inclusiveness in terms of clinical context. Included studies involved participants with a wide range of disabling health conditions, receiving rehabilitation in inpatient through to community settings, from countries in North America, Europe, and Australasia. Only one study involved participants from Asia (specifically, Hong Kong). The scope of the review could have potentially been further broadened to also include healthcare interventions for people with disability that had been acquired in childhood (such as intellectual disability or congenital disorders), however this would likely have just resulted in new problems regarding what to set as boundaries for inclusion and exclusion of studies.
Convincing, high quality evidence on the effectiveness of goal setting interventions was lacking, so answers to the review questions remain incomplete or uncertain. Furthermore, the individual meta‐analyses we conducted for the primary outcomes of interest were each based on few studies, with the largest meta‐analysis involving nine studies. This meant that while we were able to reach some conclusions regarding our broad research question (e.g. Is any type of goal setting better than no goal setting?), we were unable to undertake subgroup analysis to examine the contribution of specific aspects of goal setting processes to the overall effects of goal setting on clinical outcome or to explore reasons for heterogeneity among studies reporting on similar types of outcomes. The following is a list of the types of questions that we were unable to address in this review, having insufficient studies on which to conduct subgroup analyses.
-
Do higher levels of patient involvement in goal setting or goal planning result in better clinical outcomes?
-
Do higher levels of family/carer involvement in goal setting or goal planning result in better clinical outcomes?
-
Do goals that focus on activity or participation domains, as defined by the ICF (WHO 2001a) result in better clinical outcomes in comparison to goals that focus on objectives at the level of body structure and function?
-
Do difficult, ambitious, or challenging goals result in better clinical outcomes in comparison to easily achievable goals?
-
Does the use of written or oral feedback to patients regarding progress towards goals result in better clinical outcomes?
Furthermore, there are very few studies that consider the cost of goal setting. We identified only one study which included a complete economic evaluation of the cost and outcomes of a structured approach to goal setting in comparison to usual care (Ostelo 2003). Goal setting often requires the input of additional time from health professionals, and the cost of this can multiply when teams of health professionals are involved in goal setting meetings, so studies that investigate whether the added benefits of goal setting (if these do indeed exist) merit the additional cost accrued are well worth undertaking.
Quality of the evidence
'Risk of bias' ratings (see Figure 1) show the variability across all studies. The criterion for which studies scored worst was the blinding of participants and personnel. This was not surprising given the nature of the interventions under investigation, which most often required active involvement of patients and their treating health professionals in order to be implemented. A small number of studies addressed (or partly addressed) the blinding of participants and personnel by automating the delivery of goal setting via information technology or by having a third party set goals with the patients and not informing the treating health professionals of these goals. However these approaches limit the generalisability of findings to only certain types of clinical contexts. In future trials, lack of blinding of participants and personnel will necessarily remain a limitation of studies in this area of clinical practice.
The next two criteria for which studies scored worst were the blinding of outcome assessment and incomplete collection of data, with 27 of the 39 included studies being at high risk of bias on one or both of these criteria. Unlike blinding of participants and personnel there is no reasonable methodological explanation for not blinding outcome assessors to group allocation. This is an issue that should be addressed in all future RCTs on rehabilitation goal setting. Incomplete data collection resulting from attrition of participants presents more of a challenge for rehabilitation studies, particularly when data are to be collected over a long period of time, and particularly for certain clinical populations with higher rates of itinerant lifestyles or mortality. However, researchers should consider and include strategies to maximise retention of participants in RCTs once recruited, even if only for outcome assessment, and should include intention‐to‐treat analyses within their study design.
For the main findings reported in this review, we downgraded the quality of evidence due to either overall unclear or high risk of bias, unexplained heterogeneity, wide confidence intervals, or small sample sizes. This means that overall we are uncertain about the outcome estimates reported, and further research is very likely to have an important impact on our confidence in the estimates of effect, including being likely to change these estimates.
Potential biases in the review process
The methods and searches for this review were rigorous and should be relatively low in bias. We made only minor amendments to our protocol (Levack 2012; see Differences between protocol and review) which should have had little impact on potential bias. The most significant of these amendments was the decision, after completing data extraction but prior to data analysis, to combine measures of self‐reported emotional status with measures of health‐related quality of life. Our justification for this was that insufficient studies reported data using measures of health‐related quality of life and that the two concepts were deemed to be sufficiently similar for the results of a meta‐analysis to be clinically meaningful. However, as this was a post‐hoc decision, the risk of bias arising from this aspect of the review process should be considered slightly higher than if these decisions had been made during the protocol stage.
One further consideration is that authors of this review were also authors of two of the included studies (McPherson 2009; Taylor 2012). Adherence to our published protocol however minimised risk of bias arising from this, and as both were pilot studies they have had little influence on the overall conclusions of this review.
Agreements and disagreements with other studies or reviews
To our knowledge there are three prior systematic reviews of goal setting in rehabilitation: two specifically investigating goal setting in stroke rehabilitation (Rosewilliam 2011; Sugavanam 2013) and one investigating goal setting in all rehabilitation literature (Levack 2006a). Based on appraisal of observational studies and qualitative research, the two reviews of goal setting within the context of stroke rehabilitation concluded that active patient participation in goal setting appeared to be something that patients value and that structured methods of goal setting seem to increase patients' perceptions of their level of involvement in clinical decision‐making (i.e. enhancing a sense of self‐determination) (Rosewilliam 2011; Sugavanam 2013). The third review concluded that some experimental studies had provided limited evidence that goal setting might increase patient adherence to treatment regimens (Levack 2006a). Overall however, these three systematic reviews concluded that there was insufficient experimental research of adequate quality to allow any firm conclusions to be drawn regarding what effect, if any, specific goal setting practices have on health outcomes following rehabilitation.
Our current review differs from these past reviews in four main ways. It has involved: 1) a published protocol with an a priori plan for analysis (Levack 2012); 2) a far more comprehensive search of the literature including the screening of over 9000 titles and the inclusion of non‐English and grey literature, research theses, and conference presentations; 3) categorisation of studies on the basis of comparison types, for which a particularly important distinction has been made between studies with no goal setting as a control group and studies with usual care goal setting as a control group; and 4) pooling of the main outcomes where possible with meta‐analyses.
Our review concurs with past reviews regarding the limited quality and quantity of experimental studies on goal setting in rehabilitation, although our latest review has resulted in the identification of substantially more RCTs. For instance, Levack 2006a identified 13 RCTs while in this latest review we have identified 33 RCTs and six additional quasi‐RCTs. Some of this increase in the number of trials is due to increased interest since 2006 in goal setting in rehabilitation as a research topic, but it is also the result of a more comprehensive search strategy, which also identified more relevant studies published prior to 2006.
This review, while still limited by the quality of evidence and diversity of studies underpinning it, provides a more robust and transparent evaluation of the evidence than past reviews. We have found some (very low quality) evidence of the value of any type of goal setting, in comparison to no goal setting, in terms of a potentially positive impact on health‐related quality of life or patient‐reported emotional status and in terms of higher patient self‐efficacy. We are more equivocal regarding the impact of goal setting versus no goal setting, or structured approaches to goal setting versus usual care, on patient engagement in rehabilitation than in our previous review (Levack 2006a). Evidence regarding the individual contribution of specific components of the goal setting process (e.g. levels of patient involvement, levels of goal difficulty) remains inconclusive.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Forest plot of comparison: 1 Goal setting (with or without strategies to enhance goal pursuit) versus no goal setting, outcome: 1.1 Health related quality of life or self‐reported emotional status.
Forest plot of comparison: 1 Goal setting (with or without strategies to enhance goal pursuit) versus no goal setting, outcome: 1.2 Activity ‐ ability.
Forest plot of comparison: 1 Goal setting (with or without strategies to enhance goal pursuit) versus no goal setting, outcome: 1.3 Engagement in rehabilitation.
Forest plot of comparison: 2 Structured goal setting (with or without strategies to enhance goal pursuit) versus no structured goal setting, outcome: 2.1 Health related quality of life or self‐reported emotional status.
Forest plot of comparison: 2 Structured goal setting (with or without strategies to enhance goal pursuit) versus no structured goal setting, outcome: 2.4 Satisfaction with service delivery.
Comparison 1 Goal setting (with or without strategies to enhance goal pursuit) versus no goal setting, Outcome 1 Health related quality of life or self‐reported emotional status.
Comparison 1 Goal setting (with or without strategies to enhance goal pursuit) versus no goal setting, Outcome 2 Activity ‐ ability.
Comparison 1 Goal setting (with or without strategies to enhance goal pursuit) versus no goal setting, Outcome 3 Engagement in rehabilitation.
Comparison 1 Goal setting (with or without strategies to enhance goal pursuit) versus no goal setting, Outcome 4 Self‐efficacy.
Comparison 2 Structured goal setting (with or without strategies to enhance goal pursuit) versus no structured goal setting, Outcome 1 Health related quality of life or self‐reported emotional status.
Comparison 2 Structured goal setting (with or without strategies to enhance goal pursuit) versus no structured goal setting, Outcome 2 Activity ‐ ability.
Comparison 2 Structured goal setting (with or without strategies to enhance goal pursuit) versus no structured goal setting, Outcome 3 Self‐efficacy.
Comparison 2 Structured goal setting (with or without strategies to enhance goal pursuit) versus no structured goal setting, Outcome 4 Satisfaction with service delivery.
Comparison 2 Structured goal setting (with or without strategies to enhance goal pursuit) versus no structured goal setting, Outcome 5 Adverse events (withdrawal due to death, re‐hospitalisation or worsening symptoms).
| Goal setting with or without strategies to enhance goal pursuit compared to no goal setting for adults with acquired disabilityparticipating in rehabilitation | |||||
| Patient or population: adults with acquired disability participating in rehabilitation | |||||
| Outcomes | Illustrative comparative risks* (95% CI) | No of Participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | ||||
| No goal setting | Goal setting (with or without strategiesto enhance goal pursuit) | ||||
| Health‐related quality of life or self‐reported emotional status | The mean Physical Component Summary Scores on the Short Form‐36 for the control group was | The mean Physical Component Summary Scores on the Short Form‐36 for the intervention group was | 446 | ⊕⊝⊝⊝ | Higher scores indicate better outcomes. Scores estimated using a SMD of 0.54 (95% CI 0.17 to 0.88), indicative of an effect size that may range from small to large.Two additional studies with 142 participants however, reported no means or SD, but indicated that goal setting may lead to little to no difference in health‐related quality of life or self‐reported emotional status |
| Participation | See comment | See comment | 254 | ⊕⊝⊝⊝ | Outcomes unable to be pooled due to lack of reporting of data and lack of similarities in the types of measures used. We are uncertain whether goal setting improves participation‐level outcomes |
| Activity | The mean Barthel Index score for the control group was | The mean Barthel Index score for the intervention groups was | 223 | ⊕⊕⊝⊝ | Higher scores indicate better outcomes. Scores estimated using a SMD of 0.04 (95% CI ‐0.22 to 0.31). This evidence suggests that goal setting may not improve activity‐level outcomes |
| Body structure and body function | See comment | See comment | 235 | ⊕⊝⊝⊝ | Unable to pool outcomes due to lack of similarities in the types of measures used. We are uncertain whether goal setting improves outcomes at the level of body structure and body function |
| Engagement in rehabilitation (motivation, involvement and adherence) | The mean number of hours worked on a 26‐week support work placement programme for the control groups was | The intervention groups worked | 369 | ⊕⊝⊝⊝ | Higher scores indicate better engagement. Scores estimated using a SMD of 0.30 (95% CI ‐0.07 to 0.66). One additional study with 27 participants reported no means or SD but indicated that goal setting may lead to little to no difference in engagement in rehabilitation. One further study with 367 participants measured medication regime adherence as a dichotomous variable, and reported that the odds for the goal setting group adhering was 1.13 times higher (95% CI 1.08 to 1.19) than that of the no goal setting group. Overall, we are uncertain whether goal setting improves engagement in rehabilitation |
| Self‐efficacy | The mean Task Self‐efficacy score for the control groups was | The mean self‐efficacy in the intervention groups was | 108 | ⊕⊝⊝⊝ | Higher scores indicate better self‐efficacy. Scores estimated using a SMD of 1.07 (95% CI 0.64 to 1.49), indicative of a moderate to large effect size. One additional study with 88 participants reported no means or SD, but suggested that goal setting after rehabilitation may lead to little to no difference in self‐efficacy |
| *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). | |||||
| GRADE Working Group grades of evidence | |||||
| 1 The Physical Component Summary Score on the Short Form‐36 was used for this illustrative comparative risk as this was deemed to be the most common, most general measure of quality of life used in the studies included in the meta‐analysis for this outcome. The data on assumed risk for the Physical Component Summary Score on the Short Form‐36 was taken from control group data in the study that used this measure (Harwood 2012). | |||||
| Structured goal setting with or without strategies to enhance goal pursuit compared to 'usual care' that involved some goal setting but where no structured approach was followed for adults with acquired disabilityparticipating in rehabilitation | |||||
| Patient or population: adults with acquired disability participating in rehabilitation | |||||
| Outcomes | Illustrative comparative risks* (95% CI) | No of Participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | ||||
| 'Usual care' | Structured goal setting (with or without strategiesto enhance goal pursuit) | ||||
| Health‐related quality of life or self‐reported emotional status | The mean Mental Component Summary Scores on the Short Form‐36 for the control group was | The mean Mental Component Summary Scores on the Short Form‐36 for the intervention group was | 441 | ⊕⊕⊝⊝ | Higher scores indicate better outcomes. Scores estimated using a SMD of 0.18 (95% CI ‐0.19 to 0.56). One additional quasi‐RCT with 201 participants reported no means or SD, but indicated that usual care may lead to higher quality of life than structured goal setting. One further study with 122 participants reported that participants in the structured goal setting group were more likely to report being more satisfied or much more satisfied with their daily life compared to participants in the usual care group 3 months post intervention (RR 1.44, 95% CI 1.09 to 1.88), but not 2 years later (RR 1.27, 95% CI 0.94 to 1.70). Overall, this evidence suggests that structured goal setting in rehabilitation may result in little to no improvement in health‐related quality of life or self‐reported emotional status |
| Participation | See comment | See comment | 201 | ⊕⊝⊝⊝ | One quasi‐RCT reported no means or SDs for this outcome, but did not suggest that structured goal setting improves participation‐level outcomes. We are uncertain whether structured goal setting improves participation‐level outcomes |
| Activity | The mean Functional Indepdence Measure score in the control groups was | The mean Functional Independence Measure score in the intervention groups was | 277 | ⊕⊕⊝⊝ | Higher scores indicate better outcomes. Scores estimated using a SMD of 0.17 (95% CI ‐0.15 to 0.49). This evidence suggests that structured goal setting in rehabilitation may not improve activity‐level outcomes. One additional quasi‐RCT (201 participants) measured functional independence and reported no means or SD, and two further studies (118 participants) measured activity levels as ordinal data, but overall these studies also indicated that structured goal setting in rehabilitation may not improve activity‐level outcomes |
| Body structure and body function | See comment | See comment | 229 | ⊕⊝⊝⊝ | Unable to pool outcomes due to lack of similarities in the types of measures used. We are uncertain whether structured goal setting improves outcomes at the level of body structure and body function |
| Engagement in rehabilitation | See comment | See comment | 32 | ⊕⊝⊝⊝ | One study reported data on patient motivation in rehabilitation. A small difference in favour of structured goal setting in comparison to usual care was reported in terms of patient‐rated motivation (MD 1.40 on a 10‐point scale of self‐reported motivation, 95% CI 0.43 to 2.37) but not for therapist‐rated score of motivation (MD 0.48 on an 8‐point scale of therapist‐rated patient motivation, 95% CI ‐0.41 to 1.37) |
| Self‐efficacy | The mean self‐efficacy in the control groups was | The mean self‐efficacy in the intervention groups was | 134 | ⊕⊝⊝⊝ | Higher scores indicate better self‐efficacy. Scores estimated using a SMD of 0.37 (95% CI 0.02 to 0.71), indicative of an effect size that may range from small to large |
| *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). | |||||
| GRADE Working Group grades of evidence | |||||
| 1 The Mental Component Summary Score on the Short Form‐36 was used for this illustrative comparative risk as this was deemed to be the most common, most general measure of quality of life used in the studies included in the meta‐analysis for this outcome. The data on assumed risk for the Mental Component Summary Score on the Short Form‐36 was taken from control group data in the study that used this measure (Parsons 2012). | |||||
| Study | Other papers reporting study |
| Asenlof 2006; Asenlof 2009 | |
| Blair 1995 | |
| Duncan 2002 | |
| Jonsdottir 2012b | |
| Ostelo 2000; Ostelo 2004 | |
| Ranta 2000; Setter‐Kline 2007; Watson 2001 | |
| Sewell 2001 |
| Comparison 1: Structured goal setting with or without strategies to enhance goal pursuit versus no goal setting | |||||
| Health‐related quality of life and self‐reported emotional status | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Intrapsychic foundation and Interpersonal function subscales of the Quality of Life Scale | No, means and SDs not reported | n/a | n/a | n/a | |
| Rosenberg Self‐Esteem Scale | Yes | n/a ‐ only reported on one measure at one time point | Unable to adjust for effects of clustering; influence of study on findings tested in sensitivity analysis | Mutual goal setting groups with and without operant behaviour management were combined into a single experimental (goal setting) group | |
| Centre for Epidemiological Studies‐Depression Scale | Yes | Lowest of the two middle ranked effect estimates for four possible measures (other possible measures: Satisfaction with Life Scale, Positive Affect Scale, and Negative Affect Scale) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Minnesota Living with Heart Failure | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| 'Reorganisation' subscale of the Psychological Responses to Injury | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | The 'attention' control group (receiving social support but not goal setting from a sport psychologist) and the 'no additional input' control group were combined into a single control group | |
| Derogatis Stress Profile | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Physical Component Summary scores from the 36‐Item Short Form Health Survey | Yes | Selected as this was the primary outcome measure used in the sample size calculation | n/a ‐ not a cluster‐RCT | Four study groups. However, we judged the DVD inspirational video to be a separate intervention unrelated to goal setting, so only included data from the 'Take Charge' (goal setting) intervention as experimental group data and data from the 'usual care' group as control group data | |
| Satisfaction with Life Scale | No, means and SDs not reported | n/a | n/a | n/a | |
| Cardiac version of the Quality of Life Index | Yes | Lowest effect estimate from two possible measures (other possible measure: Mental Health Inventory‐5) | n/a ‐ not a cluster‐RCT | Three study groups. However, we judged the self‐management education to be a separate intervention, so excluded data from this group from our analysis | |
| Dyspnea component of the Chronic Respiratory Questionnaire | Yes | Lowest of the two middle ranked effect estimates for four possible measures (other possible measures: the Fatigue, Emotion, and Mastery components of the Chronic Respiratory Questionnaire) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Activity ‐ ability | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Six Minute Walk Test | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Barthel Index | Yes | Lowest effect estimate from two possible measures (other possible measure: Frenchay Activities Index) | n/a ‐ not a cluster‐RCT | Four study groups. However, we judged the DVD inspirational video to be a separate intervention unrelated to goal setting, so only included data from the 'Take Charge' (goal setting) intervention as experimental group data and data from the 'usual care' group as control group data | |
| Six Minute Walk Test | Yes | Median effect estimate out of five possible measures (other possible measures: Timed Up and Go; Activity of Daily Living subscale of the Lower Limb Task Questionnaire; Step Test; Ten meter Walk Test | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Continuous ambulatory activity monitor counts | Yes | Selected as this was the primary outcome measure used in the sample size calculation | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Engagement in rehabilitation | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Self‐reported percentage of recommended exercises session completed | Yes | Lowest effect estimate from two possible measures (other possible measure: self‐reported percentage of recommended repetitions of each exercise completed) | n/a ‐ not a cluster‐RCT | Three groups. We combined the two groups involving participant‐physiotherapist collaborative goal setting and physiotherapist‐mandated goal setting into a single experimental (goal setting) group | |
| Total number of hours worked during a work trial | Yes | Lowest effect estimate from two possible measures (other possible measure: Total number of weeks worked during a work trial) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Sports Injury Rehabilitation Adherence Scale | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | Three groups. We combined the groups involving therapist‐directed exercise without goal setting and non‐therapist‐directed exercise without goal setting into a single control group | |
| Percentage adherence to recommended food selection | Yes | n/a ‐ only reported on one measure | n/a ‐ not a cluster‐RCT | Three groups. However, we judged the group receiving no dietary education as irrelevant to the review question, so only included the groups receiving dietary education with and without goal setting | |
| Number of recommended exercise sessions completed | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Overall physiotherapist estimate of adherence | Yes | Lowest effect estimate from two possible measures (other possible measure: self‐reported percentage of recommended exercises completed) | n/a ‐ not a cluster‐RCT | Three groups. We combined the 'attention' control group (receiving social support but not goal setting from a sport psychologist) and the 'no additional input' control group into a single control group | |
| 'Motivation' subscale of the Griffiths Work Performance Scale | No, SDs not reported | n/a | n/a | n/a | |
| Percentage of eligible weeks worked during a work trial | Yes | No data were provided on the other measure referred to in the study: the Work Values Inventory | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| 24‐hour self‐reported sodium intake | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | Three groups. However, we judged the group receiving no dietary education as irrelevant to the review question, so only included the groups receiving dietary education with and without goal setting | |
| Sports Injury Rehabilitation Adherence Scale | Yes | Lowest of the two middle ranked effect estimates for four possible measures (other possible measures: number of classes attended; self‐reported adherence to stretching exercises; self‐reported adherence to walking exercises) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Self‐efficacy | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Sports Injury Rehabilitation Beliefs Survey | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | Three groups. We combined the groups involving therapist‐directed exercise without goal setting and non‐therapist‐directed exercise without goal setting into a single control group. | |
| Sports Injury Rehabilitation Beliefs Survey | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study was included in the meta‐analysis | n/a ‐ not a cluster‐RCT | Three groups. We combined the 'attention' control group (receiving social support but not goal setting from a sport psychologist) and the 'no additional input' control group into a single control group | |
| Task self‐efficacy | Yes | Median of three effect estimates (other possible measures: maintenance self‐efficacy, recovery self‐efficacy) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Self‐efficacy to Manage Disease in General | No, SDs not reported | n/a | n/a | n/a | |
| Comparison 2: Structured approach to goal setting with or without strategiesto enhance goal pursuit versus 'usual care' that may have involved some goal setting but where no structured approach was followed | |||||
| Health‐related quality of life and self‐reported emotional status | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Classification Committee of the World Organization of National Colleges, Academies and Academic Associations of General Practitioners/Family Physicians' chart of perceived health status | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Unable to adjust for effects of clustering; influence of study on findings tested in sensitivity analysis | n/a ‐ only two study groups | |
| General Health Questionnaire | No, means and SDs not reported | n/a | n/a | n/a | |
| Modified version of the Welsh Anxiety Scale | Yes | Lowest of the two middle ranked effect estimates for four possible measures (other possible measures: Anxiety Scale of the Today form of the Multiple Affect Adjective Checklist; Depression Scale of the Today form of the Multiple Affect Adjective Checklist; Rosenburg Self‐esteem Scale) | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| General Health subscale of the 36‐Item Short Form Health Survey | Yes | Lowest effect estimate from two possible measures (other possible measure: Social Functioning subscale of the 36‐Item Short Form Health Survey). Data were collected from more than one time point. Data from the longest period of time following recruitment into the study were included in the meta‐analysis. | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Mental Component Summary score of 36‐Item Short Form Health Survey | Yes | Lowest effect estimate from two possible measures (other possible measure: Physical Component Summary score of the 36‐Item Short Form Health Survey) | Using the reported data on mean differences between the intervention and control groups with 95% CIs, we calculated adjusted means and SD that reproduced the 95% CI and P values in the paper to account for clustering | n/a ‐ only two study groups | |
| Schedule for Evaluation of Individual Quality of Life | Yes | Selected as this was named the primary outcome measure (other possible measure: 36‐Item Short Form Health Survey) Data were collected from more than one time point. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Using the reported data on mean differences between the intervention and control groups with 95% CIs, and the reported separate effects of clustering on variance plus the inter‐class correlations for clusters, we calculated adjusted means and SD that reproduced the 95% CI and P values in the paper to account for clustering | n/a ‐ only two study groups | |
| Activity ‐ ability | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Pain Disability Index | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Disability Index of the Health Assessment Questionnaire | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Unable to adjust for effects of clustering; influence of study on findings tested in sensitivity analysis | n/a ‐ only two study groups | |
| Functional Independence Measure | No, means and SDs not reported | n/a | n/a | n/a | |
| Roland Morris Disability Questionnaire | Yes | Lowest effect estimate from two possible measures (other possible measure: self‐reported severity of main activity limitation). Data were collected from more than one time point. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Functional Independence Measure | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Using the reported data on mean differences between the intervention and control groups with 95% CIs, and the reported separate effects of clustering on variance plus the inter‐class correlations for clusters, we calculated adjusted means and SD that reproduced the 95% CI and P values in the paper to account for clustering | n/a ‐ only two study groups | |
| Self‐efficacy | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Self‐efficacy Scale | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Chronic Disease Self‐efficacy Score | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Unable to adjust for effects of clustering | n/a ‐ only two study groups | |
| Satisfaction with service delivery | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Overall quality of physical therapy on a 0‐10 scale | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Satisfaction Scale for Community Nursing | Yes | n/a ‐ only reported on one measure at one time point | Unable to adjust for effects of clustering | n/a ‐ only two study groups | |
| Overall satisfaction on a 10 cm visual analogue scale | No, means and SDs not reported | n/a | n/a | n/a | |
| Satisfaction with counselling | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Patient Perception of Rehabilitation | Yes | Only one measure used but outcomes from more than one time point were reported. Data from the longest period of time following recruitment into the study were included in the meta‐analysis | Using the reported data on mean differences between the intervention and control groups with 95% CIs, and the reported separate effects of clustering on variance plus the inter‐class correlations for clusters (the latter accessed by author communication), we calculated adjusted means and SD that reproduced the 95% CI and P values in the paper to account for clustering. | n/a ‐ only two study groups | |
| 5‐point client satisfaction questionnaire | Yes | n/a ‐ only reported on one measure at one time point | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Adverse events (all‐cause) | |||||
| Study | Outcome measure | Pooled in meta‐analysis? | Selection from multiple measures and/or time points | Management of clustering | Management of multiple groups |
| Number of participants withdrawn from the study due to death or hospitalisation | Yes | Withdrawal due to death or re‐hospitalisation combined. Numbers taken from last time point | Unable to adjust for effects of clustering | n/a ‐ only two study groups | |
| Number of participants withdrawn from the study due to worsening symptoms | Yes | Numbers taken from last time point | n/a ‐ not a cluster‐RCT | n/a ‐ only two study groups | |
| Number of participants withdrawn from the study due to death | Yes | n/a ‐ only reported on one measure at one time point | Unable to adjust for effects of clustering | n/a ‐ only two study groups | |
| CI = confidence interval; n/a = not applicable; RCT = randomised controlled trial; SD = standard deviation | |||||
| Comparison 1: Structured goal setting with or without strategies to enhance goal pursuit versus no goal setting | ||||||||
| Continuous data | ||||||||
| Trial | Outcome measure | Goal setting | No goal setting | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Total Work Behaviour Inventory | 136.5 | 22.2 | 30 | 120.5 | 25.5 | 33 | 16.0 (4.22 to 27.78) | |
| No. treatment sessions required for symptom relief | 12.56 | 6.86 | 45 | 13.29 | 11.01 | 21 | ‐0.73 (‐5.85 to 4.39) | |
| Baseline Dyspnea Index | 9.7 | 1.7 | 7 | 8.1 | 1.6 | 7 | 1.60 (‐0.13 to 3.33) | |
| Piper Fatigue Scale | ‐1.8 | 1.7 | 7 | ‐2.2 | 1.9 | 7 | 0.40 (‐1.49 to 2.29) | |
| Systolic blood pressure (mmHg) | 137.4 | 17.8 | 38 | 140.5 | 18.6 | 31 | ‐3.50 (‐12.15 to 5.15) | |
| Urinary sodium output (mmol/24 h) | 157.4 | 52.7 | 19 | 141.4 | 57.9 | 19 | 16.0 (‐19.20 to 51.20) | |
| Systolic blood pressure (mmHg) | 135 | 17.8 | 19 | 137.6 | 17.2 | 19 | ‐2.60 (‐13.73 to 8.53) | |
| Criterion referenced achievement test score | 18.8 | 2.7 | 19 | 17.1 | 3.3 | 19 | 1.70 (‐0.22 to 3.62) | |
| COPM satisfaction | 2.04 | 1.91 | 63 | 2.27 | 2.03 | 58 | ‐0.23 (‐0.93 to 0.47) | |
| Dichotomous data | ||||||||
| Trial | Outcome | Goal setting | No goal setting | Risk ratio (95% CI) | ||||
| Events | Total | Events | Total | |||||
| Dependency, based on modified Rankin Scores > 2 | 11 | 38 | 12 | 31 | 0.75 (0.38 to 1.46) | |||
| Death | 4 | 46 | 5 | 39 | 0.68 (0.20 to 2.35) | |||
| 'High' theoretical GAS achievement | 8 | 13 | 7 | 11 | 0.97 (0.52 to 1.80) | |||
| Return to work | 15 | 22 | 21 | 24 | 0.78 (0.56 to 1.08) | |||
| Comparison 2: Structured approach to goal setting with or without strategies to enhance goal pursuit versus 'usual care' that may have involved some goal setting but where no structured approach was followed | ||||||||
| Continuous data | ||||||||
| Trial | Outcome measure | Structured goal setting | 'Usual care' goal setting | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Pain ‐ Maximum | ‐3.6 | 3.17 | 28 | ‐4.9 | 3.1 | 37 | 1.30 (‐0.24 to 2.84) | |
| Tampa Scale of Kinesiophobia | ‐27.10 | 6.35 | 28 | ‐29.60 | 6.69 | 37 | 2.50 (‐0.69 to 5.69) | |
| Percentage of goals achieved by individual participants | 98.0 | 9.3 | 40 | 61.5 | 49.0 | 29 | 36.5 (18.43 to 54.57) | |
| Patient‐reported motivation | 8.93 | 1.1 | 15 | 7.53 | 1.66 | 17 | 1.40 (0.43 to 2.37) | |
| Therapist‐reported motivation of patients | 4.6 | 0.99 | 15 | 4.12 | 1.54 | 17 | 0.48 (‐0.41 to 1.37) | |
| Number of therapy sessions provided | 9.27 | 6.22 | 15 | 6.18 | 4.3 | 17 | 3.09 (‐0.66 to 6.84) | |
| Change in lumber spine range of movement (degrees) | 18.9 | 21.5 | 52 | 20.1 | 22.7 | 53 | ‐1.20 (‐9.66 to 7.26) | |
| Change in Tampa Scale of Kinesiophobia | 2.7 | 6.5 | 52 | 2.6 | 6.2 | 53 | 0.10 (‐2.33 to 2.53) | |
| Total healthcare costs in the 12 month follow‐up period (EUR) | 1978 | 1894 | 52 | 1339 | 1873 | 53 | 639 (‐81.61 to 1359.61) | |
| Percentage of goals recalled | 75 | 28 | 33 | 57 | 32 | 36 | 18.00 (3.84 to 32.16) | |
| Dichotomous data | ||||||||
| Trial | Outcome | Structured goal setting | 'Usual care' goal setting | Risk ratio (95% CI) | ||||
| Events | Total | Events | Total | |||||
| Met or exceeded a goal related to range of movement | 22 | 38 | 8 | 32 | 2.32 (1.20 to 4.47) | |||
| More satisfied or much more satisfied with daily living | 33 | 38 | 26 | 43 | 1.27 (0.94 to 1.70) | |||
| Healthcare use over two years (any visits to doctor, physiotherapists, or other care givers due to pain conditions) | 19 | 28 | 19 | 37 | 1.32 (0.88 to 1.98) | |||
| Achieved or exceeded at least one set goal | 7 | 13 | 7 | 9 | 0.69 (0.38 to 1.28) | |||
| Comparison 4: One structured approach to goal setting and/or strategies to enhance goal pursuit versus another structured approach to goal setting and/or strategies to enhance goal pursuit | ||||||||
| Continuous data | ||||||||
| Collaborative versus prescribed (therapist mandated) goal setting | ||||||||
| Trial | Outcome measure | Collaborative goal setting | Prescribed goal setting | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| No. of treatments required | 13.61 | 8.01 | 23 | 11.46 | 5.4 | 22 | 2.15 (‐1.83 to 6.13) | |
| No. of home exercise sessions complete | 75.43 | 20.88 | 23 | 65.82 | 23.65 | 22 | 9.61 (‐3.45 to 22.67) | |
| Goal setting with versus without operant conditioning | ||||||||
| Trial | Outcome measure | Goal setting plus operant conditioning | Goal setting without operant conditioning | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Rosenberg Self‐Esteem Scale | 1.185 | 1.111 | 27 | 1.577 | 0.703 | 26 | ‐0.39 (‐0.89 to 0.11) | |
| End‐only goals versus end goals with short‐term steps | ||||||||
| Trial | Outcome measure | End‐only goal | End goal with short‐term steps | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Self‐reported dietary fat consumption as a percentage of total energy consumption | 34 | 6 | 4 | 27 | 15.6 | 3 | 7.00 (‐11.61 to 25.61) | |
| Specific versus a non‐specific goal | ||||||||
| Trial | Outcome measure | Specific goals | Non‐specific goals | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Pain behaviour ‐ nonverbal complaint | 1.28 | 1.34 | 13 | 4.3 | 5 | 13 | ‐3.02 (‐5.83 to ‐0.21) | |
| Difficult versus easier goals | ||||||||
| Trial | Outcome measure | Difficult goal (8 servings) | Easy goal (6 servings) | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Total number of servings lower glycaemic index foods per day | 8.4 | 0.83 | 20 | 8.42 | 0.96 | 15 | ‐0.02 (‐0.63 to 0.59) | |
| Goal commitment | 4.09 | 0.12 | 20 | 4.5 | 0.14 | 15 | ‐0.41 (‐0.50 to ‐0.32) | |
| Satisfaction with goal achievement | 6.8 | 0.36 | 20 | 6.47 | 0.41 | 15 | 0.33 (0.07 to 0.59) | |
| Self‐efficacy ‐ total score | 9.3 | 0.28 | 20 | 9.33 | 0.32 | 15 | ‐0.03 (‐0.23 to 0.17) | |
| High intensity exercise goals versus non‐specific intensity exercise goals | ||||||||
| Trial | Outcome measure | High intensity exercise goal | Non‐specific exercise intensity goal | Mean difference (95% CI) | ||||
| Mean | SD | n | Mean | SD | n | |||
| Total steps during any walking | 6868 | 3751 | 13 | 6279 | 3306 | 17 | 589 (‐1985.37 to 3163.37) | |
| Steps counts during bouts of high intensity walking | 2616 | 2706 | 13 | 2070 | 2814 | 17 | 546 (‐1442.24 to 2534.24) | |
| Duration pedometers worn per day (hours) | 14.5 | 2.49 | 13 | 16.5 | 2.49 | 17 | ‐2.00 (‐3.80 to ‐0.20) | |
| Dichotomous data | ||||||||
| Goal Management Training versus Identity Oriented Goal Mapping | ||||||||
| Trial | Outcome | Goal Management Training | Identity Orientated Goal Mapping | Risk ratio (95% CI) | ||||
| Events | Total | Events | Total | |||||
| Achieved or exceeded at least one set goal | 3 | 8 | 4 | 5 | 0.47 (0.17 to 1.27) | |||
| High intensity exercise goals versus non‐specific intensity exercise goals | ||||||||
| Trial | Outcome | High intensity exercise goal | Non‐specific exercise intensity goal | Risk ratio (95% CI) | ||||
| Events | Total | Events | Total | |||||
| Would definitely recommend the programme to a friend | 8 | 13 | 17 | 17 | 0.62 (0.41 to 0.96) | |||
| Programme considered very useful | 4 | 13 | 12 | 17 | 0.44 (0.18 to 1.04) | |||
| COPM = Canadian Occupational Performance Measure; GAS = Goal Attainment Scaling | ||||||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Health related quality of life or self‐reported emotional status Show forest plot | 8 | 446 | Std. Mean Difference (IV, Random, 95% CI) | 0.53 [0.17, 0.88] |
| 2 Activity ‐ ability Show forest plot | 4 | 223 | Std. Mean Difference (IV, Random, 95% CI) | 0.04 [‐0.22, 0.31] |
| 3 Engagement in rehabilitation Show forest plot | 9 | 369 | Std. Mean Difference (IV, Random, 95% CI) | 0.30 [‐0.07, 0.66] |
| 4 Self‐efficacy Show forest plot | 3 | 108 | Std. Mean Difference (IV, Fixed, 95% CI) | 1.07 [0.64, 1.49] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Health related quality of life or self‐reported emotional status Show forest plot | 5 | 441 | Std. Mean Difference (IV, Random, 95% CI) | 0.18 [‐0.19, 0.56] |
| 2 Activity ‐ ability Show forest plot | 4 | 277 | Std. Mean Difference (IV, Random, 95% CI) | 0.17 [‐0.15, 0.49] |
| 3 Self‐efficacy Show forest plot | 2 | 134 | Std. Mean Difference (IV, Fixed, 95% CI) | 0.37 [0.02, 0.71] |
| 4 Satisfaction with service delivery Show forest plot | 5 | 309 | Std. Mean Difference (IV, Random, 95% CI) | 0.33 [0.10, 0.56] |
| 5 Adverse events (withdrawal due to death, re‐hospitalisation or worsening symptoms) Show forest plot | 3 | 406 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 0.64 [0.27, 1.47] |
