Realist-informed review of motivational interviewing for adolescent health behaviors

Motivational interviewing (MI) is a client-centered, collaborative approach that explores inconsistencies between clients’ beliefs and behaviors in order to elicit intrinsic motivation, reduce ambivalence, and evoke natural behavioral change [20]. Though a directed therapy, MI is respectful of individuals’ autonomy and ability to make decisions for themselves, and accordingly, maladaptive beliefs or cognitions are not confronted [20‐22]. Therapists rely on an empathetic, empowering, and encouraging approach [12, 21, 23], where reflective listening, positive regard, and other Rogerian skills are used to support clients’ sense of self-efficacy and reinforce the therapeutic relationship, which is considered an essential element for MI effectiveness [13, 20, 24]. Self-efficacy, or the confidence that one can achieve their goals, is considered a reliable outcome predictor and is a key factor in a client’s readiness to change [24, 25].

Readiness to change is defined as a process or state of moving between no intention of making a behavior change to committing to and maintaining behavior change [23, 26]. Based on the Transtheoretical Model (TTM), individuals are at various points along this continuum of readiness to change [27‐29]. Theoretically, individuals move through the stages sequentially; however, in practice, the stages are not mutually exclusive and individuals often move back and forth between stages or move through more than one stage simultaneously [22, 30]. The TTM has been criticized regarding the notion that change is predicated on the notion of stable stages or plans [31], the suggested discrete nature of the stages of change [32, 33], and the assumption of conscious decision-making and the exclusion of the role of rewards and punishment in the development of habits [34].

Research using MI to target problem health behaviors has focused on diet, nutrition, exercise, oral health, sedentary behavior, screen time, and monitoring of glucose levels [11, 15, 35]. Several reviews and meta-analyses have examined the effectiveness of MI for health behaviors and found that it is an effective intervention for diet, exercise, obesity, and diabetes among adults [12, 13‐16, 31]. More specifically, MI is associated with changes in body mass index (BMI), systolic blood pressure, and total body cholesterol [16].

Adolescence is characterized as a time of significant physical, personal, and social development [36], coupled with a shift from relying heavily on parents to make decisions to engaging in more independent decision-making [19]. Health behaviors acquired in adolescence are likely to be maintained into adulthood [9, 36‐38], making adolescents more responsible for their own health and lifestyle [19]. This newfound freedom may lead teenagers to eat foods low in nutrition, high in fat and/or “empty calories”; exercise well below the daily recommended amount [36, 39]; and poorly manage diabetes [40]. Thus, adolescence is a critical period for introducing interventions to halt problematic behaviors and reduce the risk of poor health outcomes or chronic disease in adulthood [41].

Research in the area of health behavior interventions for adolescents, specifically those related to obesity and diabetes, is relatively new [41]. The current literature is heavily dominated by “risky” health behaviors such as alcohol consumption, illicit drug use, cigarette smoking, and unsafe sex [12, 15]. Yet, some are beginning to add disordered eating and sedentary behavior into the fold as obesity and diabetes continue to rise in this population [39, 42‐44]. MI is one intervention gaining attention in this area, though much is still unknown regarding its efficacy and the mechanisms and contexts by which it works [12, 17, 45]. Initial evidence for the effectiveness of MI for childhood and adolescent obesity and diabetes is promising [12, 46].

Two recent meta-analyses have been conducted in the area of MI for adolescent health behavior change [17, 18]. Cushing and colleagues [17] found 15 studies from their systematic search, which yielded a small but significant effect size favoring MI interventions, g = 0.16 (95% CI [0.05, 0.27]). The authors note the importance of reporting the training received by clinicians, in order to assess whether fidelity to MI impacts adolescent health outcomes, as the majority of studies did not report training. Additionally, they recommend that future studies analyze the effects of interventionist training on outcomes.

Gayes and Steeles [18] conducted a meta-analysis for the efficacy of MI on adolescent health behaviors, delivered to the adolescent, a parent, or both. The authors synthesized 37 articles on this topic that targeted eight health conditions including obesity, HIV/AIDS, asthma, diabetes, infant health, dental health, accident prevention, sleep, and calcium intake. The analysis found a small overall effect size for MI when compared to other active treatments and no treatment, g = 0.282 (95% CI [0.242, 0.323]). Interestingly, this meta-analysis also analyzed moderator variables and found that interventions delivered by community health workers yielded the largest effect size, followed by health professionals, and lastly interventionists with master’s and doctorate degrees resulted in the lowest effect size. Additionally, studies that conducted MI with the adolescent and parent together yielded significantly larger effect sizes than MI delivered with either parent or child individually. These moderator variables suggest that specific contexts in which MI is delivered to adolescents give rise to different outcomes. Therefore, further investigation into the contexts by which MI works, and how it works within these contexts, is warranted.

Traditional systematic reviews focus on synthesizing literature and identifying whether or not interventions work. A realist approach goes beyond identifying the intervention’s efficacy to examine the underlying mechanisms and contexts in which the interventions work [45, 47]. Such an approach is predicated on the notion that an outcome occurs as a result of some events (or intervention), but there is a need to understand the generative mechanism that brings about the outcome in a particular context. These “mechanisms of change” are sometimes considered intervention or program theories that explain how and why outcomes occur. A realist approach states that contextual variables create the conditions by which generative mechanisms and outcomes can occur, which is explained through the construction of context-mechanism-outcome (CMO) configurations [48].

There are many differences between a realist review and a traditional systematic review, one of which includes its purpose. A realist review’s purpose is often exploratory and does not aim to determine whether an intervention works, as does a traditional systematic review. Rather a realist review aims to explore the process and contextual factors that play a role in the effectiveness or efficacy of an intervention [48]. Because of this, a realist review does not provide concrete and absolute answers and, as a result, the possibility for additional contributions to each theory may remain, as findings from a realist review are often provisional [48]. It should be noted, however, that provisional results and conclusions are common when examining mechanisms, since a sequence of studies must be completed in order to confirm the existence of particular mechanisms [49]. The present review was limited due to the small number of studies completed on the use of MI for adolescent health behaviors. Therefore, the present paper was adapted as a realist-informed review, due to the exploratory nature and, instead, is looking to develop a preliminary theory that includes some mechanism and contextual factors based on SDT. The provisional results of the realist review are often in the form of a conceptual model, which is the objective of the present realist-informed review.

Recent reviews have found that MI for adolescent health behavior change is an effective intervention; however, research has yet to uncover underlying mechanisms or examine how, why, and in what contexts MI works for addressing adolescent health behaviors [12, 21]. To address such process-related issues, this review synthesized the current literature and identified outcomes, contexts, and mechanisms related to the successful application of MI to target diet, exercise, and sedentary behaviors among adolescents. Our research questions included the following: What are the mechanisms or theories by which outcomes are achieved? How does the context of the MI intervention affect mechanisms and outcomes? In order to answer these questions, we tested self-determination theory (SDT) as the candidate theory.

Recent developments in research on MI have suggested that self-determination theory (SDT) provides a useful theoretical framework for understanding the efficacy of MI [50, 51]. SDT is a theory of personality development and behavior change originating from Deci and Ryan [52], which assumes that individuals innately strive towards personal growth. The theory stipulates that autonomous behaviors are more stable and have a more positive effect on the individual [53]. SDT posits that three conditions are necessary in order to achieve autonomous behavior change [54]. The first condition is the need for competence, defined as confidence in one’s abilities and that one’s abilities can affect outcomes. Competence has been compared to other constructs, such as the development of confidence, self-efficacy, and control [55]. The second condition is the need to feel autonomous over one’s behaviors and actions, rather than feeling controlled by external sources. Autonomy is achieved when an individual feels like they have choice or ownership over their actions [55]. Third is the need to feel relatedness, which involves experiencing a connection with others and having satisfying relationships. If individuals feel a sense of caring and commitment from important people in their life, they will be more willing to make a behavior change [55]. SDT suggests that all three of these conditions can be met if the individual is in an environment that allows for self-determination to foster [54]. If an individual is in an environment where they are being controlled, it is unlikely they will develop competence, autonomy, and relatedness.

The theory states that if the social environment provides structure, autonomy support, and involvement of others, then individuals will be able to achieve the three conditions of self-determination [56]. Structure facilitates competence; autonomy support facilitates the development of autonomy; and lastly, involvement, or the perception that others care about your well-being, facilitates feelings of relatedness [53]. Therefore, the present study adopts SDT as the candidate theory for MI for adolescent health behaviors. Specifically, the intervention would be hypothesized to be most effective in contexts that allow for structured setting, that permit autonomy support, and include involvement of others. These contexts will produce the hypothesized mechanisms, which include competence, relatedness, and autonomy. These mechanisms will result in the final outcomes of health behavior change.

A literature search was conducted for all eligible published studies from 1989 up to March 2017, with the earliest paper found from 1989. This systematic review included studies that examined or reviewed the effectiveness or efficacy of MI to change health behaviors among adolescent populations. Studies that combined MI with another intervention were included as well, as long as MI was a specific focus of the study. Health behaviors included in the systematic review were related to diet and exercise. Behaviors related to risk-taking (e.g., drug and alcohol abuse and sexual activity) were excluded because these behaviors were thought to involve different processes as related to MI. Participants were considered to be adolescent if they ranged in age from 12 to 18 years, which has been used in adolescent outcome studies by the World Health Organization [57]. Studies included in the systematic review were empirical papers, review articles, or case studies. Reviews and case studies were included because they were thought to offer insight and provide rich detail into the contexts and mechanisms through which MI works. Finally, studies had to be published in English. Relevant studies were identified through online searches of several relevant databases (i.e., PsychINFO, PubMed, Cochrane, and Healthstar). Search terms (with both American and British spelling) that were used to search the literature included (1) motivational interviewing or motivational enhancement; (2) health behavior, diet, exercise, obesity, diabetes, sedentary behavior, nutrition, OR physical activity; and (3) adolescent, teen, or youth. These terms were searched in the abstracts of studies. In addition, reference lists from review articles were checked for any other potentially eligible studies.

This search strategy generated a total of 185 abstracts that subsequently underwent abstract review. Two researchers independently reviewed the list of abstracts and initial decisions were made on whether to retrieve the full text article. In the case of disagreement, discrepancies were resolved through discussion with a third research team member. Abstracts were excluded if they failed to meet one or more of the previously cited inclusion criteria. Selected (n = 40) abstracts were then retrieved and two members of the research team independently reviewed every article to confirm inclusion, abstract information, and complete a quality appraisal. Articles were removed from analysis primarily for the following reasons: lack of focus on MI, the majority of participants were not within the age range, and lack of findings or conclusions related to the health behaviors of interest. If MI was used with teens with diabetes, the article was only included if changing health behaviors was an outcome of the intervention. As a result, the full text for 26 articles were retrieved and coded for relevant information. Two additional studies were added from the reference lists of the coded studies (see Fig. 1).

Quality appraisal of articles was evaluated using the recommendations for realist review from RAMESES [58]. The rigor and relevance of the article was considered, rated as low, medium, or high. An appraisal deemed “low” was used for articles that did not include any information or discussion of mechanism or contextual factors. A rating of “medium” was given to studies that provide information on either contextual or mechanism variables. A rating of “high” was given to studies that provide information on both mechanism and contextual variables. Rigor was assessed as low, medium, or high, based on the credibility and trustworthiness of the method that was used in each article [58]. Studies appraised as low in rigor or relevance were excluded from the realist review. Two authors appraised each article; discrepancies were discussed and resolved in order to facilitate an iterative process of quality appraisal.

Once the articles that met inclusion criteria were identified, full text articles were retrieved and relevant information for analysis was extracted. Extracted information was related to four topics: (1) a description of the population(s) that was included in the study and other contextual factors of the study (e.g., healthcare setting); (2) a description of the program or intervention of interest (MI or the combination of MI with another intervention), (3) with the exception of review papers, a description of the evaluation was required; and (4) a description of the mechanisms or the underlying causal factors that might be contributing to the success or failure of the program.

The research team initially met and discussed preliminary mechanisms or theories (self-determination theory; competence, autonomy, relatedness). Once the candidate theory was clarified, it was tested and refined by examining articles that the research team had extracted. Iterative theory development was done across the papers and was consistent with the data analysis standards from RAMESES [58]. The focus on the data extraction was on contexts, mechanisms, and outcomes. Then, through an iterative discussion process, the theory was further developed by the creation of demi-regularities that included the creation of context-mechanism, context-outcome, or mechanism-outcome links. Lastly, thematic overlap from the demi-regularities was further analyzed in order to create the final CMO configurations.

In this qualitative process, an immersion/crystallization qualitative approach was used and researchers independently immersed themselves in the articles in order to crystallize, or decide on reportable themes or mechanisms [59]. Crystallization of the final contexts and mechanisms, as they related to MI for adolescent populations, was done through legitimizing and corroborating [59]. The current review was not registered in PROSPERO.

As noted in Fig. 1, there were a total of 28 articles included in this review and the most common study design included in this review was randomized control trials (n = 13). The next most common type of article was pre/post, correlational, or repeated measures design (i.e., seven in total). As well, two protocols, four reviews (including two book chapters), and two papers reporting secondary analyses (one qualitative) were included in this review. Most of the studies were of medium or high relevance (i.e., 23 in total). Health behaviors of target by all of the studies included were diet and exercise (see Table 1).

Nearly every study focused on achieving outcomes related to increasing physical activity and changing eating habits, with varying results between the interventions. Olson and colleagues [60] reported increased physical activity at 6-month follow-up after an MI intervention and increases in milk servings at a 6-month follow-up; however, no changes were reported for fruit intake, vegetable intake, and sweetened beverage consumption. Black and colleagues [61] reported that although there were no changes in daily activity counts for the group of adolescents, those who were overweight or obese at the start reported increases in daily activity counts, compared to those who were in the normal range for weight. They also found decreases in caloric intake in general, as well as reduced consumption of snacks/desserts, marginal significant decreases in dietary fat and fried food, and marginally significant increases in fruit intake (in addition to the aforementioned activity counts examination) [61]. Resnicow and colleagues [35] reported that when asked about the telephone-based MI intervention specifically, 47% of teens reported it had a lot of influence on their physical activity and 42% reported it had a little influence. Additionally, 47% of teens reported that the MI intervention phone calls had a “lot of influence” on their diet and 45% of teens reported that it had a “little influence.” Forty-five percent of teens reported that the intervention helped them think differently about their eating habits, but no significant differences between groups were found for diet- or physical activity-related outcomes. Gourlan and colleagues [62] implemented two individual sessions with a healthcare provider discussing health behavior. In addition, the experimental group also received the same intervention plus six additional MI sessions, which was found to lead to a significant increase in physical activity at 6 months. Lee and Kim [63] reported that male students from a junior high school in Seoul who received ME twice a week for 16 weeks had a significant increase in physical activity.

Neumark-Sztainer and colleagues [64] found improvement regarding portion control among teenage girls at a 9-month follow-up. The authors also found significant improvements for sedentary activity, eating patterns, unhealthy weight control behaviors, and body image [64]. Pakpour and colleagues [65] implemented MI as adolescent only, or with parental involvement. They found that the group receiving MI with parental involvement had significantly different outcomes including BMI changes, diet, physical exercise, and self-efficacy for diet. The MI with parental involvement was not superior to MI alone for servings of fruits, vegetables, and milk products; waist circumference; or psychosocial functioning.

Two documents examined physical activity and diet outcomes in the context of diabetes management [66, 67]. These studies approached the treatment of diabetes from a self-management perspective and focused on supporting youth to achieve physical activity and nutrition outcomes as a means to self-manage their diabetes. Nansel and colleagues [67] found that at 18 months, scores on the Healthy Eating Index 2005 were 7.2 greater and scores on the Whole Plant Food Density scale were 0.5 greater in the intervention group versus control. Notably, there was no difference between groups in HbA1c levels across the study duration.

A number of studies also examined secondary outcomes related to the MI intervention. A study by Bean and colleagues [68] found that MI participants had greater 3-month adherence to dietitian and behavioral support visits, which was consistent at 6 months. Carcone and colleagues [69] implemented a secondary analysis of coded MI sessions and found that 62% of the time, open-ended questions elicited change talk by the adolescent. Additionally, statements emphasizing autonomy were more likely to elicit change talk. These authors also found that affirming statements were not effective in eliciting change talk [69]. Carcone and colleagues [70] carried out a secondary analysis of MI sessions with youth and their caregivers. The results indicated that caregivers reported more ambivalence than youth overall, but caregivers had less ambivalence than youth for nutrition-related changes [70]. Conversely, youth had less ambivalence for activity-related changes compared to their caregivers. The authors reported the greatest divergence in ambivalence was between youth and caregivers for physical activity-related changes [70].

Very few studies examined readiness to change as outcome measures. Channon and colleagues [71, 72] found that the majority of patients (84%) reported changes towards the action stage at 12-month follow-up. Berg-Smith and colleagues [73] included participants who varied in their readiness to change at pre-intervention and found that readiness significantly increased by one point on a scale from one to 12; action plans were made by 94% and successfully implemented by 84% of participants. Neumark-Sztainer and colleagues [64] found that 13% more participants in the MI intervention group progressed from pre-contemplation to contemplation in comparison to the control group.

Although BMI and weight were typically included as an outcome measure in studies, there was little evidence to support the impact of MI or MI in combination with other techniques at reducing BMI [62, 74, 75]. One exception is research by Black and colleagues [61] who found the percentage of obese/overweight adolescents in their control group increased in weight by 11%, yet the percentage decreased by 5% in the intervention group. Irby and colleagues [76] noted an improvement in BMI; however, the study’s quality and rigor are weak because it is a case study. Additionally, Lee and Kim [63] found a significant reduction in BMI when MI was used in addition to an obesity intervention. The results of this study should also be interpreted with caution as no control group was used for comparison [63]. A reduction in BMI may not be considered an achievable outcome given that adolescence is a period of relative weight gain and growth and efforts to counteract this would be futile. Furthermore, BMI may not be an ideal outcome because adolescents undergo a number of developmental physical changes that can impact BMI scores [77].

A small number of studies mentioned contexts and mechanisms that were not found to fit within the SDT framework. For example, Olsen and colleagues [60] found that an interest in changing behavior at baseline was a significant predictor of outcomes at follow-up. Additionally, a study by Carcone and colleagues [70] of African American adolescents and their caregivers compared the frequency of ambivalence statements during an MI session with the dyad. The results found that overall ambivalence was higher for caregivers than for youth. The ambivalence between the dyad was similar for nutrition-related changes, but diverged for physical activity-related changes. Parents were more likely to be ambivalent towards their adolescent making physical activity-related changes compared to youth. The role of ambivalence within the parent-adolescent dyad may be an important variable to further explore in order to better understand how the ethnic minority context and the mechanism of relatedness is activated in this population.

Flattum and colleagues [78] monitored adolescent girls’ goal-setting behaviors and found that when receiving MI, girls were able to set goals 100% of the time and achieved goals 75% of the time. Neumark-Sztainer and colleagues [64] paralleled this result and found that the MI intervention increased physical activity goal-setting behaviors of adolescent females. The role of goal setting as a mechanism of change for health behaviors is still unknown. Lastly, Naar-King et al. [81] found that individuals with higher executive functioning had greater weight loss in the short term, but not the long term.

A number of limitations exist within the present review. Importantly, this realist review did not have differentiated pathways for specific outcomes. The majority of studies looked at a combination of diet, exercise, and weight loss outcomes making it impossible to separate CMO configurations out by outcome. Rather, this review focused on contexts and mechanisms that lead to health behavior change generally. More research is needed in order to identify the pathways that lead to each specific health behavior change.

Second, a number of contextual factors were mentioned in articles, with no observable patterns of outcomes. For example, a number of interventions took place within a health clinic, in a community setting, or via telephone. It is possible future research can examine these contextual factors in order to assess their efficacy and create different CMO configurations. For the purpose of this review, it was concluded that not enough evidence currently exists to understand the impact of those contextual factors.

Due to the nature of realist reviews, some material needs to be omitted in order to focus the theory building in a rich and cohesive manner. Therefore, other aspects of the data can become neglected and unexplored, which are described as insufficiently explored mechanisms and contexts. It should be noted that future syntheses might choose to focus on different contexts and mechanisms within MI or to use a comparison theory in order to build additional CMO configurations.