Behavior change theory and behavior change technique use in cancer rehabilitation interventions: a secondary analysis

Abstract

BACKGROUND

There is limited evidence depicting ways that behavioral theory and techniques have been incorporated into cancer rehabilitation interventions. Examining their use within cancer rehabilitation interventions may provide insight into the active ingredients that can maximize patient engagement and intervention effectiveness.

AIM

This secondary analysis aimed to describe the use of behavior change theory and behavior change techniques (BCTs) in two previously conducted systematic reviews of cancer rehabilitation interventions.

DESIGN

Secondary analysis of randomized controlled trials (RCTs) drawn from two systematic reviews examining the effect of cancer rehabilitation interventions on function and disability.

SETTING

In-person and remotely delivered rehabilitation interventions.

POPULATION

Adult cancer survivors.

METHODS

Data extraction included: behavior change theory use, functional outcome data, and BCTs using the Behavior Change Technique Taxonomy (BCTTv1). Based on their effects on function, interventions were categorized as “very”, “quite” or “non-promising”. To assess the relative effectiveness of coded BCTs, a BCT promise ratio was calculated (the ratio of promising to non-promising interventions that included the BCT).

RESULTS

Of 180 eligible RCTs, 25 (14%) reported using a behavior change theory. Fifty-four (58%) of the 93 BCTs were used in least one intervention (range 0-29). Interventions reporting theory use utilized more BCTs (median=7) compared to those with no theory (median=3.5; U=2827.00, P=0.001). The number of BCTs did not differ between the very, quite, and non-promising intervention groups (H(2)=0.24, P=0.85). 20 BCTs were considered promising (promise ratio >2) with goal setting, graded tasks, and social support (unspecified) having the highest promise ratios.

CONCLUSIONS

While there was a wide range of BCTs utilized, they were rarely based on theoretically-proposed pathways and the number of BCTs reported was not related to intervention effectiveness.

CLINICAL REHABILITATION IMPACT

Clinicians should consider basing new interventions upon a relevant behavior change theory. Intentionally incorporating the BCTs of goal setting, graded tasks, and social support may improve intervention efficacy.

The growing number of cancer survivors around the world^1-3 have unique needs and face multiple physical, functional, and psychosocial challenges during and after cancer treatment.^{4, 5} Approximately 40% of cancer survivors report difficulties in activities of daily living (ADLs) (e.g., bathing, dressing), and 55-60% report limitations in instrumental ADLs (e.g., preparing meals, household chores).⁴ Adverse effects of cancer and its treatments often go undetected, undertreated, and can diminish survivors’ ability to participate fully in work and life roles^4-8 and reduce overall quality of life.^{9, 10}

Comprehensive cancer rehabilitation is considered an essential component of cancer care^11-14 and focuses on the prevention and treatment of immediate, persistent, or late effects of cancer and treatment.^15-17 Recent systematic reviews have concluded that cancer rehabilitation interventions can be efficacious in decreasing disability.^{15, 16} However, there is significant heterogeneity in both specific intervention strategies used and in the effects on functional outcomes.¹⁷ Variability in the approaches taken and implementation of these interventions can have an impact on their efficacy and the specific mechanisms of action regarding functional benefits are unclear.¹⁸ This limits our ability to draw from the existing cancer rehabilitation evidence to develop future interventions.

Cancer rehabilitation interventions are often complex and typically require some degree of active patient engagement for the therapeutic effect to be achieved and maintained.¹⁹ Active engagement in a therapeutic activity includes a patient’s ability to develop and adhere to new compensatory skills and adopt health-related behaviors.^{12, 20, 21} The incorporation of behavior change theory and related behavior change techniques (BCTs) has been recommended in the development and evaluation of these interventions^{19, 21-23} and can maximize the potential intervention efficacy.^{21, 24, 25} Behavior change theory provides a framework for understanding how people adopt new behaviors and aids in identifying constructs and/or mechanisms to be targeted.²² BCTs are the specific strategies and/or mechanisms of action that facilitate behavior change and are selected based on the theoretical constructs they are proposed to target.²⁶

Limited evidence shows how behavioral change theory and techniques have been incorporated into cancer rehabilitation interventions. Examining the use of behavior change theory and BCTs within cancer rehabilitation interventions may provide insight into the active ingredients that can maximize patient engagement and cancer rehabilitation intervention effectiveness. To build on this work, we conducted a secondary analysis of two recent systematic reviews examining efficacy of cancer rehabilitation interventions to: 1) identify the behavior change theories and BCTs used in the included studies, and; 2) examine the relationships between behavior change theory, BCTs, and intervention effectiveness.

Materials and methods

This is a secondary analysis of articles drawn from two recent systematic reviews examining the effect of cancer rehabilitation interventions on cancer-related disability,^{15, 16} conducted by the Outcomes and Research Task Force of the Cancer Rehabilitation Networking Group of the American Congress of Rehabilitation Medicine. These two reviews were selected as an extension of the work conducted by the Outcomes and Research Task Force. While many reviews examine the role of behavior change theory and BCTs in improving some health behaviors (e.g., physical activity, diet) among cancer survivors, we were uncertain as to the extent of reporting and use of behavioral theory within the field of cancer rehabilitative medicine. Thus, we sought to build upon the previous work for the Task Force by exploring the use of behavior change theory and BCTs within cancer rehabilitative medicine, using the articles included within these previous systematic reviews^{15, 16} to establish the methodology that has been used in other behavioral fields within cancer rehabilitative medicine.

Inclusion criteria

Original systematic reviews

Detailed descriptions of the search strategies and findings of the original systematic reviews are provided in the original publications.^{15, 16} Briefly, the previous systematic reviews comprised articles that: 1) included cancer survivors ≥18 years old; 2) were controlled intervention trials comparing a cancer rehabilitation intervention to any type of control group (e.g., attention control, usual care); 3) included an intervention with >1 synchronous interactions with a rehabilitation professional; 4) measured disability as defined by the International Classification of Functioning, Disability, and Health Framework;²⁷ 4) were written in English. Articles were excluded from the original reviews if they: 1) focused on pharmacological interventions that did not report functional outcomes or were outside the scope of practice of the rehabilitation provider, or 2) only examined psychological outcomes (e.g., distress, depression).^{15, 16}

Secondary analysis

To ensure the articles for the current analysis were cohesive, additional inclusion criteria were applied. After deduplication, studies were included if they were: 1) a randomized controlled trial; 2) measured function as a primary outcome (e.g., functional mobility, ADLs or instrumental ADLs); 3) included an intervention that required patient participation (i.e., active patient engagement).

Data extraction and coding

Thirteen reviewers (MLV, RB, SW, YJ, GAT, SA, KLC, DEH, CK, AL, AS, AGS, JMJ) extracted study characteristics, the primary target behavior, behavior change theory, outcome data (within and between group differences), and coded BCTs for each study intervention in duplicate, following the guidelines from the Cochrane Handbook for Systematic Reviews of Interventions.²⁸ All reviewers completed the Behavior Change Taxonomy Training (www.bct-taxonomy.com) prior to extraction. Paired coders (MLV, RB, SW, YJ, GAT, SA, KLC, DEH, CK, AL, AS, AGS, JMJ) independently extracted the data into a spreadsheet developed for this review. An independent staff member compared the coding and data extraction completed by the paired coders to identify any discrepancies. Coders then met to identify and resolve disagreements after every 5-10 articles based on an a priori consensus process. If consensus was not reached, a third reviewer (JMJ) was consulted.

Intervention coding

Target behavior

Target behaviors were coded based on the following categories: 1) physical activity or exercise; 2) specific muscle strengthening (e.g., pelvic floor exercises/kegels, anal sphincter exercises, swallowing/voice exercises, breathing exercises); 3) engagement in daily activities (e.g., functional activity and ADLs); 4) Mood (e.g., relaxation, meditation, imagery); 5) symptom management (e.g., cognitive exercises, compensatory skills, self-lymphatic massage, self-management skills; 6) multiple behaviors.

Behavior change theory

Due to expected heterogeneity in intervention reporting, the use of behavior change theory was coded dichotomously and the specific theory was coded in Dedoose software (SocioCultural Research Consultants, LLC, Los Angeles, CA, USA) if applicable. If a specific behavioral approach based on a theory (e.g., cognitive behavioral therapy or acceptance and commitment therapy) was mentioned, it was assumed that the intervention was based on the associated theory.²⁹

Behavior change techniques

Reviewers then extracted and coded the BCTs mentioned in the interventions using the BCT Taxonomy v1 (BCTTv1).²⁶ The BCT Taxonomy is a reliable and widely used coding frame that includes 93 BCTs (https://www.bct-taxonomy.com/pdf/BCTTv1_PDF_version.pdf). Only techniques addressing the target behavior were coded and each BCT was only recorded once in Dedoose software. When clarity was lacking in the description of the intervention and the coders remained uncertain if the BCT was present, it was not coded.

For each study, the raw data for the primary outcome (means, standard deviations, sample size, statistical significance) were extracted for intervention and control groups. When no primary time-point was defined and multiple time points were reported, we extracted the data closest to the completion of the intervention period.

Study quality

Study quality was extracted using the ratings provided by the two original systematic reviews. As the risk of bias tools varied, study ratings using the American Academy of Neurology classifications³⁰ from Sleight et al. were re-categorized to match the Cochrane RoB2 ratings²⁸ provided in the Brick et al. article. The AAN classifications were recoded as Class I=Low risk of bias, Class II & Class III=Some concerns, Class IV=High risk of bias.³⁰

Data summary and analysis

Final extracted data for each article were compiled into a Microsoft Excel spreadsheet to create a database of results for analysis. Data were then summarized in tables and analyzed using a statistical software package (IBM, SPSS, version 26). Based on the widely applied^31-33 two-step method developed by Gardner et al., each study was assigned to a mutually exclusive promise rating based on their potential (or promise) to improve function (Table I).²⁸

Table I. —Promise category definitions.

Promise category	Definition
Very promising	Statistically significant between group difference reported for the primary functional outcome
Quite promising	Statistically significant within group difference reported for the primary functional outcome in the intervention group
Non-promising	No statistically significant improvement on the primary functional outcome within the intervention group or relative to the comparator group

To explore the potential impact of each BCT to intervention promise, a ‘promise ratio’, was calculated as the number of promising (very or quite) interventions utilizing the BCT divided by the number of non-promising interventions using the same BCT.²⁸ A BCT was considered promising if it was used in at least twice as many promising interventions as non-promising interventions (ratio >2).^{28, 34} If a BCT only appeared in promising interventions, a ratio was not calculated and the number of interventions in which it was coded was reported. We did not calculate a promise ratio if a BCT was only used in one study³² or in non-promising interventions.²⁸

Descriptive analyses were conducted to describe study quality and the frequency of theory use, BCT use, and BCT promise ratios. BCT number was not normally distributed so non-parametric tests were used with this data. Mann-Whitney U and Kruskal-Wallis tests were conducted to examine the distribution of BCT use in interventions by theory use (yes/no) and across the intervention promise categories, respectively. A Chi-squared Test was conducted to examine promise category and theory use. An exploratory post-hoc analysis examined the associations between study quality and BCT use, theory use, and promise ratio using Kruskall-Wallis and Chi-square Tests, respectively. BCT use and promise across target behaviors was also explored, however, inferential statistics were not conducted due to small cell sizes.

Results

Based on our inclusion criteria, 206/430 articles were considered from the original systematic reviews. The majority (N.=218) were excluded because they did not have function as a primary outcome. Following full text review, another 26 articles were removed due to: non-randomized design (N.=10), no functional primary outcome (N.=11), or the intervention did not require patient engagement (N.=5). The final sample comprised 180 eligible trials (Figure 1). The studies were published between 2007 and 2022 and primarily took place in North America or Europe (N.=120, 66%). Sample sizes ranged from 19 to 711 participants and the most frequently included cancer diagnoses were breast only (N.=73, 40%), prostate only (N.=27, 15%), or multiple tumor groups (N.=27, 15%).

Figure 1.

—PRISMA flow chart.

In terms of intervention target behavior, 9 studies (5%)^35-43 targeted more than one behavior. Exercise or physical activity was the intervention target in 59% of studies (N.=117, 64.6%).^31-113 Specific muscle strengthening exercises were the second most common (N.=29, 16%).^114-155 The remaining articles targeted symptom management skills (N.=23; 12.7%),^114-136 mood (N.=8; 4.4%),^{114, 137-144} diet (N.=8; 4.4%),^{36-42, 156-185} and engagement in daily activities (N.=3, 1.7%).^186-188

Use of behavior change theory

The inclusion of behavior change theory was reported in 14% (N.=25)^{48-50, 60, 61, 114, 117-119, 124, 129, 132-134, 137, 139-141, 144-150} of the articles. Twenty studies (80%) reported using one theory,^{35, 43, 63, 73, 74, 144, 158-160, 165, 174, 175, 179-181, 184, 188-192} three (12%) reported using two theories,^{42, 61, 173} and two (8%) reported three theories.^{62, 170}A total of 17 different theories were reported (see Supplementary Digital Material 1: Supplementary Table I). Cognitive Behavioral Theory was the most frequently reported (N.=9, 36%), followed by Social Cognitive Theory (N.=4, 16%) and the Transtheoretical Model (N.=4, 16%).

Behavior change techniques

Of a possible 93 BCTs, 54 (58%) were reported in at least one intervention.²⁷ The number of BCTs utilized per intervention ranged from 0-29. Studies reported a median of 3.5 BCTs. Fourteen studies (7.7%) did not report any BCTs.^{48, 65, 67, 94, 108, 109, 112, 114, 120, 121, 129, 193, 194} There were no apparent trends (primary outcomes, intervention types, etc.) in the studies that did not include BCTs (all p’s > 0.05). Most studies reported using two BCTs (N.=34; 18.8%), followed by 3 BCTs (N.=26; 14.4%). Studies that included behavior change theory reported a higher number of BCTs (median=7) compared to studies that did not report theory use (median=3; U=2827.00, P=0.001).^195-211

The most used BCT groupings included Feedback and Monitoring; Goals and Planning; Shaping Knowledge; and Social Support (see Figure 2).

Figure 2.

—Frequency of BCT Category Usage.

Overall, item 4.1 “Instruction on how to perform the behavior” was used the most frequently (N.=118; 65%), followed by 2.3 “Self-monitoring of behavior” (N.=65; 36%), 1.4 “Action planning” (n=48; 26%), and 3.1 “Social support (unspecified)” (N.=45, 24%) (Supplementary Digital Material 2: Supplementary Table II). The frequency of BCT usage is displayed in Figure 3. The distribution of BCT number did not differ across the intervention promise groups (H(2)=0.24, P=0.89).

Figure 3.

—Number of BCTs used overall and by theory use.

Promise ratings

Of the 180 studies, 54% (N.=98) were considered very promising,^{60-71, 147, 151-156}12% (N.=21) quite promising,^{48-59, 130, 131, 137, 144, 157-161} and 34% (N.=61) non-promising.^{31-40, 43, 45-47, 115, 116, 123, 129, 132-136, 138, 145, 146, 148-150, 162-191} A theory-based approach was used in 15% (N.=16) of the 104 studies that were very or quite promising and 17% (N.=9) of the 52 non-promising studies, however, this was not significant (χ² (2, 180)=2.35, P=0.3).

A BCT promise ratio was calculated for 36 BCTs, 16 of which were promising (Figure 4).

Figure 4.

—Promise ratio for behavior change techniques in descending order.

An additional three BCTs were found in only promising interventions (“Monitoring of emotional consequences” (5.4), 10.4 “Social reward” (10.4), and 1.9 “Commitment” (1.9)). The BCTs with the highest promise ratios were “Framing/reframing” (13.2) (N.=5 interventions; promise ratio=4.0), “Goal setting” (1.3) (N.=10; promise ratio = 4.0), “Graded tasks” (8.7) (N.=18; promise ratio=3.5), and 3.1 “Social support (unspecified)” (N.=43; promise ratio=3.4). Five out of 9 (56%) BCTs in the grouping “Goals and Planning” (1.1, 1.3, 1.6, 1.7, 1.9); two of three (67%) BCTs in “Social Support” (3.1, 3.3); and two out of three (67%) BCTs in “Comparison of Behavior” (6.1, 6.2) were considered promising.

Behavior change techniques by target behavior

Among the exercise interventions, BCTs from the categories of Shaping Knowledge, Feedback and monitoring, and Goals and Planning were used the most frequently while Identity and reward and threat were used the least (Figure 2). Similar trends were noted among the interventions targeting specific muscle strengthening exercises, symptom management, diet, and mood. All the included studies targeting activities of daily life included at least one BCT targeting Goals and planning.

Of the specific BCTs used, most of the BCTs included in exercise interventions were considered promising (Supplementary Table II), while “information on antecedents” (promise ratio=6), “prompts/cues” (promise ratio=6), “reduce negative emotions” (promise ratio=5) appeared to be the most promising. Among specific muscle strengthening interventions, “goal setting (behavior)” (promise ratio=2.78) and “biofeedback” (promise ratio=2.3) were the most promising. “Monitoring of behavior by others without feedback” (promise ratio=4) and “Self-monitoring of behavior” (promise ratio=3) were the most promising amongst symptom management interventions. Among the mood and diet-focused interventions, the majority of BCTs were only used in promising interventions, thus a promise ratio could not be calculated, and a score is reported instead. Based on these scores, “Self-monitoring of behavior” (score=6) and “Information about health consequences” (score=5) appeared to be promising for mood interventions. “Prompts/cues” (score=7) and “social support (unspecified)” (score=5) appeared to be the most promising among diet-focused interventions. None of the BCTs used in studies targeting engagement in ADLs (N.=3) were considered promising, however, this is may also be due to the small number of studies targeting this behavior.

Study quality

Overall, 137 (88.9%) studies were rated as “some concerns” (Supplementary Table I), with the remaining articles rated as a high risk of bias (n=13, 7.2%) or low risk of bias (N.=5, 3.9%). There was no significant difference between study quality ratings and theory use (χ² (2, 180)=3.39, P=0.18) or the number of BCTs reported (H(2)=1.624, p=0.44). Study promise ratings also did not vary across study quality ratings (χ² (2, 180)=1.65, P=0.80).

Discussion

This secondary analysis examined the use of behavior change theory and BCT in cancer rehabilitation interventions aimed at increasing function. We found that the reporting of behavior change theory was rare, with limited use of behavioral theory mentioned in cancer rehabilitation intervention development and implementation. While there was a wide range of BCTs utilized, they were rarely based on theoretically proposed pathways. Findings suggest that cancer rehabilitation intervention research would benefit from more explicit descriptions of the theoretical underpinnings and greater transparency between proposed theory, intervention ingredients (BCTs) and outcomes.

The lack of reporting on the behavior change theory likely has multiple causes. The simplest may be that many investigators do not utilize a theoretical conception of interventions. Alternatively, investigators may eliminate behavior theory details due to word/page limitations and publication guidelines that do not emphasize delineation of the theoretical mechanisms. Whatever the causes, there are consequences for translation of the findings. For researchers, explicit use of theory and understanding how specific theories support effective behavior change can help elucidate mechanisms of change in behavior and help with replication.²¹²

Of the small percentage of studies that did reference behavior change theory (14%), a wide range were applied and often lacked explicit justification of why a theory was selected and tested. The breadth of theory use makes it difficult to assess their association with behavior change. It also makes it challenging to provide recommendations for researchers on best theoretical approaches to use.¹⁹²

Previous studies have reported that the use of theory can increase effectiveness,^193-196 while others have not found this relationship.¹⁹⁷ The way in which theory is explicitly applied may provide some insight into the mixed literature and our findings. To begin, it may be that the researchers select an incorrect or misaligned theory. Choosing the most suitable theory can be challenging given the large number of available and overlapping theories.^{24, 198} In addition, there is a lack of guidance and supporting evidence, especially in the rehabilitation literature, on how to select specific theories for a particular purpose. Another explanation is that a theory, while appropriate, may not be properly applied nor linked to the specific intervention components or outcomes, which can impact intervention effectiveness. Michie and colleagues have termed these as ‘theory-inspired interventions’ rather than theory–based interventions. A review by Bluethmann et al. examined the application of theory and intervention effectiveness in behavioral interventions, it was found that studies that included explicit linkages between the selected theory and the intervention constructs, measurement, and interpretation of the results were more effective.¹⁹⁹ In the future, it is recommended that investigators delineate how the interventions components are tied to the theory, including models of proposed mechanisms of action that link theory and outcomes. Further, when summarizing the results, it is important to specifically interpret the findings while also considering their relevance to the theoretical framework This ensures that translation focused work emphasizes theory and mechanisms, not just the action and outcomes.

While a variety of BCTs were incorporated into the included intervention studies, we noticed overlap between commonly reported behavior change theories and the most frequently used BCTs. Techniques including goal setting, graded tasks, instruction, self-monitoring, and action planning, as well as social support and environmental modification are common interventions with the tenets of the most frequently used theories. These BCTs have been found to be effective in other reviews with cancer survivors and physical activity^{25, 200-206} and focus on enabling individuals who are motivated to make a conscious effort to self-regulate their thoughts, feelings and behaviors in order to achieve the target behavior.²⁰⁷ Given that rehabilitation typically involves the prescription of therapeutic exercise and activities to increase function, it is consistent with the more frequent use of these types of theories. Despite the number of promising BCTs identified, the number of BCTs used was not related to intervention promise, a finding that has been reported by others,^{208, 209} and suggests that more techniques in one intervention does not equate to better outcomes. As suggested by Keogh and colleagues,²¹⁰ it is likely that BCTs were simply used due to the nature of program design and delivery mode, rather than being strategically selected based on theory. Studies that have not selected an appropriate theory or have failed to explicitly apply theory in the development of the intervention would likely result in a mismatch of BCTs and would be less likely to modify the processes that result in improved outcomes.²¹¹

Given the number of theories and techniques that have been identified to change behavior, choosing the optimal approach can be challenging and time consuming and requires more specific knowledge and training in behavioral science.²¹² There is a need for further training in behavioral techniques which could be applied in both clinical and research applications.^{213, 214} Moving forward, in addition to formal training, the development of collaborations with behavioral and social scientists can foster more understanding and reporting of mechanisms. Proposed and existing frameworks and models should be considered and encouraged. In addition, the Rehabilitation Treatment Specification System (RTSS), which places emphasis on active ingredients, or mechanisms of change of interventions, provides a framework that can strengthen the reporting of treatment protocols and results.²¹⁵

Limitations of the study

The results from our study must be interpreted considering limitations. Based on the results from the systematic reviews, the includes studies are known to have moderate to high levels of bias and lower quality,^{15, 16} which may have impacted the overall effectiveness or promise ratios for the BCTs. Further, theory use was crudely coded, rather than utilizing a more detailed checklist (e.g., The Theoretical Domain Framework^{198, 216}), due to the heterogeneity between studies and poor intervention reporting. It is possible that theory was applied and/or BCTs implemented, however, they were not adequately described in the published main or supplementary materials, thus underestimating frequency of theory and BCT use. Our work provides preliminary evidence regarding BCT promise in cancer rehabilitation interventions addressing disability in terms of activity limitations and participation restrictions. However, our findings may not generalize to interventions addressing changes in body structures/functions, personal factors, and/or environmental factors. Future work should consider how BCTs are implemented in this alternate body of intervention research which encompasses another area of cancer rehabilitation research. While our initial exploration allowed us to examine the potential contribution of individual BCTs to intervention effectiveness, we were unable to examine potential multiplicative effects of interventions that contained multiple BCTs. As a result, the promise ratio can be considered a relatively crude measure. Future work should adopt methods that examine the complex and interacting nature of BCTs in these interventions. Finally, study quality was not associated with behavior change theory use and BCT promise. However, almost all the studies were categorized as “some concerns”, and the small cell sizes mean these results should be interpreted with caution. Further, the two original systematic reviews utilized different study quality checklists and ratings were recategorized for consistency; results may differ if the primary articles were coded using the same checklist.

Conclusions

This was the first study to describe the use of behavior change theory and BCTs in cancer rehabilitation interventions. The results indicate that the methodology described by Gardiner et al. can be applied to rehabilitation interventions, suggesting a systematic review can be conducted to better understand the application of behavior change theory, which may highlight a gap in the professional knowledge base and existing literature. Moving forward, there needs to be purposeful and detailed reporting of the intervention development, design, and results to guide future intervention development and refinement that enhance functional outcomes in cancer survivors.

Supplementary Digital Material 1

Supplementary Table I

Different theories reported.

Supplementary Digital Material 2

Supplementary Table II

All target behaviors