Paul T. Fuglestad, Jody S. Nicholson, and Lauren James
Department of Psychology, University of North Florida, Jacksonville, FL, USA
Most modern diseases and causes of death have substantial contributing factors that are behavioral. The Center for Disease Control and Prevention estimates that premature death and unnecessary healthcare costs can be greatly ameliorated by enacting behavioral changes. For decades, researchers across disciplines in the social and health sciences have sought to create theories of health behavior that can be used to encourage healthy change and inform intervention design and implementation. Although these theories and approaches to intervention share underlying similarities, they vary widely in terms of their behavioral foci, focal constructs, and processes and stages of change. For example, Michie, West, Campbell, Brown, and Gainforth (2014), in attempting to catalog all behavior change theories relevant to interventions, identified over 80 theories with more than 1,000 constructs. Furthermore, health behaviors are numerous and varied, including diet, physical activity, alcohol use, tobacco use, vaccinations, medication adherence, sexual behavior, toxin exposure, and screenings. Fortunately, many theories are applied to numerous health behavior domains, and there is substantial overlap in mechanisms of change among the most widely used health behavior theories (Sheeran et al., 2016; Sheeran, Klein, & Rothman, 2017). This entry begins by reviewing numerous influential theories and approaches that have been used to understand and influence health behavior change and concludes with a discussion of challenges to health behavior change and promising new directions.
Although health behavior theories can be categorized in many different ways (e.g., Do they focus on intentional/explicit processes or implicit/habitual processes? Do they emphasize cognitions regarding a threat or cognitions regarding a specific behavior?), a useful distinction can be made between continuum and stage‐based models (Rothman, Baldwin, Hertel, & Fuglestad, 2011). Continuum models explain change as a steady process, whereas stage‐based theories assume change is nonlinear and consists of steps. Continuum models specify a set of target variables that are thought to always predict behavior regardless of individual differences or situations, and a person's behavior is thought to be a function of his/her standing on these specified variables. For example, a continuum model such as the health belief model might attribute changes in smoking cessation to incremental changes in the perception of susceptibility to a health threat (e.g., cancer) and barriers or benefits to quitting (e.g., quitting will reduce the likelihood of lung cancer). On the other hand, stage theories propose that people move sequentially through various stages in changing their health behavior and different variables are thought to be important for certain stages, but not others. For example, programming based on a stage‐based model like the transtheoretical model would focus on different change variables for someone who is not planning to quit (precontemplation stage) and someone who is actively quitting (action stage). Using this model, a program might focus on consciousness raising to increase knowledge for someone in the precontemplation stage while using stimulus control to break association cues that illicit smoking behavior for those in the action stage of quitting.
The first major health behavior theory was the health belief model developed by Irvin Rosenstock and colleagues in the 1950s. This model attempts to specify when people will take action against some health threat, such as getting a vaccination to prevent an illness, using a condom to prevent STDs, or wearing sunscreen to prevent skin cancer. It specifies that people will be more likely to take action if they perceive that they are vulnerable to a threat (i.e., perceived susceptibility), that the threat would have negative consequences (i.e., perceived severity), and that the benefits of action outweigh the costs or barriers.
Similar to the health belief model, the protection motivation theory, developed by Ronald Rogers, proposes that taking action to change a health behavior is dependent on constructs such as perceived susceptibility, severity, and whether the recommended behavior actually prevents the threat. It was largely developed in the context of using fear appeals to encourage people to engage in protective action against health threats. Building on stress and coping research, the protection motivation theory focuses on the cognitive processes of threat appraisal and coping appraisal that feed into one's protection motivation. One's emotional response to a potential threat (i.e., fear) was added to the threat appraisal process to go along with perceptions of susceptibility and severity. Similarly, one's confidence to enact a particular behavior (i.e., self‐efficacy) was added to the coping appraisal process to go along with perceptions of benefits and barriers. In a meta‐analysis of the protection motivation theory, Milne, Sheeran, and Orbell (2000) found that variables related to threat appraisal (e.g., susceptibility, severity, fear) and coping appraisal (e.g., self‐efficacy, response benefits and costs) were predictive of intentions and behavior. The impacts of threat and coping appraisals on health‐related intentions and behavior were small to medium in effect size and generally less predictive of subsequent rather than concurrent behaviors. Small to medium effect sizes were also evident in a meta‐analysis of the influence of experimentally manipulated perceptions of how likely one was to get an illness (i.e., perceived susceptibility) and how bad it would be to get it (i.e., perceived severity) on people's intentions and behaviors (Sheeran, Harris, & Epton, 2014).
Another widely used set of health behavior theories are the theory of reasoned action and the theory of planned behavior developed by Martin Fishbein and Icek Ajzen. In the theory of reasoned action, behavioral intention (e.g., one's intention to eat more fruits and vegetables) is posited to be the most proximal predictor of health behavior and is predicted by an individual's attitudes and subjective norms. Attitudes about the behavior are based on one's expectation of behavioral outcomes and the value of those outcomes (e.g., eating more fruits and vegetables will have positive effects on health). Subjective norms regarding the behavior are based on one's perceptions of norms regarding a behavior and whether one is motivated to comply with those norms (e.g., important others think it is a good idea to eat more fruits and vegetables). In recognition that some behaviors are less volitional or controllable, the theory of planned behavior adds a construct very similar to self‐efficacy, perceived behavioral control, which is based on how easy/hard it would be to engage in the behavior (e.g. eating more fruits and vegetables is something that can easily be done).
Sheeran et al. (2016) summarized 18 meta‐analyses that examined the associations of attitudes, norms, and self‐efficacy with health‐related intentions and behaviors (e.g. condom use, physical activity, smoking cessation, alcohol use, and screening behaviors). Averaged across the meta‐analyses, these constructs were all moderately to strongly associated with behavioral intentions and actual behavior. Sheeran et al. (2016) also conducted a meta‐analysis to examine the experimental impact of these constructs. To be included in the meta‐analysis, an intervention had to experimentally manipulate attitudes, norms, or self‐efficacy related to a health behavior. Importantly, all of these constructs were related to intentions and behavior. Although the effect sizes were somewhat smaller than those reported in meta‐analyses of correlational studies discussed above, the effect sizes were generally in the medium range, suggesting that these common constructs are indeed important for promoting health behavior change.
Building on the theory of planned behavior, the prototype willingness model was developed by Meg Gerrard and Frederick Gibbons to more fully understand non‐intentional but volitional behaviors (e.g., risky sexual or substance use behaviors). It adds the variable of social prototypes that measures one's perception of the prototypical person who engages in a behavior (e.g., someone who engages in unsafe sexual practices), including one's similarity to the prototype and/or one's overall evaluation of this prototypical person. Social prototypes are thought to primarily influence another additional construct—behavioral willingness—which is one's willingness to engage in a risky/unhealthy behavior under certain circumstances (e.g., an attractive partner wants to have sex without a condom). Evidence suggests that these constructs are important for predicting behavior when one is engaging in less thoughtful, more heuristic‐based cognition (Gerrard, Gibbons, Houlihan, Stock, & Pomery, 2008).
Social cognitive theory, developed by Albert Bandura, is a continuum model that explains how humans develop, learn, and maintain certain behaviors and is often described in relationship to a model of self‐efficacy. Bandura defined self‐efficacy as one's belief in one's ability to succeed in specific situations and viewed people as more self‐regulatory than reactive to their environment. For example, how people perceive their situations can inform their decisions to behave in certain ways that can lead to an eventual change in their environments. Therefore, people's behavior can be predicted by knowing the beliefs they have about their own capabilities and environments. People's self‐efficacy beliefs are thought to be dependent on four main sources of influence: (a) coaching and feedback (i.e., social persuasion), (b) observational learning and modeling (i.e., vicarious experience), (c) efficaciousness when agitated or depressed (i.e., emotional/physiological states), and (d) prior success with a task (i.e., personal mastery experiences). Beyond self‐efficacy, behavior is also posited to depend on one's perceptions of environmental constraints that facilitate or prevent behavior (i.e., barriers to change), as well as one's attitude toward the behavior and beliefs in the likelihood that good or bad things will occur as a result of performing the behavior (i.e., outcome expectations and evaluations). As noted above, meta‐analysis suggests that self‐efficacy and attitudes toward the behavior are causally linked to health behavior change (Sheeran et al., 2016).
In contrast to continuum models of health behavior change, the transtheoretical model, developed by James Prochaska and Carlo DiClemente, is a popular stage model that conceptualizes the process of intentional behavior change. The model proposes that people move through five stages of change: (a) no intention to change behavior in the near future (i.e., precontemplation), (b) thinking about changing but have not decided (i.e., contemplation), (c) deciding to change behavior but have not successfully done so (i.e., preparation), (d) successfully changed behavior (i.e., action), and (e) preventing relapse (i.e., maintenance). The model posits that different barriers—which are referred to as processes of change or behavioral strategies—affect movement through the various stages. Early on (precontemplation and contemplation stages), behavioral strategies such as increasing knowledge (i.e., consciousness raising) and increasing concern regarding unhealthy behaviors (i.e., dramatic relief) matter most. Later (action phase), processes such as providing rewards for positive behaviors (i.e., contingency management), helping relationships (i.e., social support), and avoiding or removing cues to unhealthy behavior and adding cues to health behavior (i.e., stimulus control) matter most. The transtheoretical model incorporates elements of self‐efficacy by taking into account the degree of confidence the individual has in their ability to maintain their desired behavioral change. This model can be helpful in determining the most appropriate interventions in health behavior change suited for the individual's current stage. As compared with the other models, it recognizes that there are individual differences that might make an intervention ineffective for some participants if they are not ready to engage in health behavior change. Meta‐analyses and systematic reviews of the transtheoretical model suggest that it is effective in modeling health behavior change (e.g., Norcross, Krebs, & Prochaska, 2011). However, like most stage theories, there is mixed evidence with respect to its effectiveness relative to continuum models, the number and sequence of stages, the process by which people move from one stage to the next, and applicability across various health behavior domains (e.g. Armitage, 2009; Riemsma et al., 2003). For example, in a review of the transtheoretical model literature, Armitage (2009) suggests that a two‐phase model incorporating motivational and volitional processes better conceptualizes the behavior change process than the five stages of the transtheoretical model.
The precaution adoption process model, developed by Neil Weinstein and Peter Sandman, provides a framework to describe the particular sequence of change in which an individual might engage. The model focuses on an individual's mental state that impacts why they decide to initiate action and how that decision translates into behavior. Individuals progress through seven stages that range from being unaware of an issue (stage 1) or unengaged by an issue (stage 2), to acting upon an issue (stage 6) and maintaining the intent to act (stage 7). At stage 3, undecided about acting, an individual can proceed to either deciding to act (stage 5) and proceed along the stages, or decide not to act (stage 4), in which case there is no continued advancement along the stages and the individual is considered in a state of moratorium. In general, the precaution adoption process model differs from the transtheoretical model in that it is more focused on the mental state of each stage and less focused on time until an intended action. In the precaution adoption process model, individuals may regress back in stages, and it is critical to not assume equal spacing between stages. Some individuals may be in one stage for such a brief period; it seems they skipped it altogether. For example, a doctor's advice may move an individual from unengaged (stage 2) to acting (stage 6). Moreover, the difficulty of the action of interest may impact the utility of this model, as behavior change that is easy (e.g., changing toothpastes, wearing seatbelts) may make the stages matter less, making it possible to combine stages with a more comprehensive treatment.
Although the precaution adoption process model has not been tested as widely as the transtheoretical model or health action process approach (HAPA) (see below), its utility has been demonstrated in work on radon testing, osteoporosis prevention, and dietary change (Weinstein, Sandman, & Blalock, 2008). For example, in encouraging undecided people to commit to radon testing (moving from stages 3 to 5), an intervention demonstrating high risk was more effective than an intervention demonstrating the ease of testing. Conversely, in encouraging decided people to act (moving from stages 5 to 6), the ease of testing intervention was more effective than the risk intervention. Future research, however, should focus on providing a more comprehensive understanding of why an individual would transition between stages.
The HAPA, developed by Ralf Schwarzer, is a stage theory of health behavior change that proposes that individuals pass through varying mindsets on their way to behavior change and therefore interventions that tailor to these changing mindsets will be most effective. In this manner, the adoption, initiation, and maintenance of health behaviors should be conceived of as a structured process, including a motivation phase and a volition phase, with changing perceptions of specific types of self‐efficacy throughout.
Similar to continuum models discussed above, constructs such as risk perceptions, outcome expectations (i.e., attitudes), and task self‐efficacy are thought to predict the formation of goals and intentions, an important aspect of the motivational phase. In contrast, the volitional phase is impacted by self‐monitoring behaviors and awareness of standards (i.e., planning, action control) and confidence in maintaining behavior change (i.e., maintenance self‐efficacy) and responding to behavioral lapses (i.e., recovery/coping self‐efficacy). The volitional phase is further differentiated between people that have translated their intent to action and those that have not (i.e., the inactive stage). Individuals in the inactive stage are motivated to change but do not act because they lack the skills and internal or external resources to translate intention to action. Therefore, planning becomes an important part of the process of change, and post‐intentional planning becomes a mediator between intent and action. Planning can further be divided into action planning and coping planning. Action planning involves the initiation of health behaviors, whereas coping planning is required for initiation and maintenance. According to the model, perceived self‐efficacy is required throughout the process of behavior change. However, the challenges that individuals face change during the process, and therefore the type of self‐efficacy required also changes. For example, changing diet and physical activity indicates the need for action efficacy early on (the confidence to make difficult changes) but maintenance or recovery efficacy later in the process (the confidence to resist temptation and respond to lapses).
In general, research is supportive of the HAPA model (e.g., Schwarzer, 2008), although certain tenets are not always supported (e.g., Barg et al., 2012; Sutton, 2008). For example, in a study of the determinants of physical activity intentions and behaviors, action self‐efficacy predicted behavioral intentions, whereas maintenance self‐efficacy predicted planning and behavior; however, contrary to the model, planning did not predict behavior (Barg et al., 2012). Some have argued that the HAPA model is best thought of as a continuum model that builds on prior models like social cognitive theory and the theory of planned behavior by elucidating the intention to behavior link (Sutton, 2008).
Although there is evidence for stage‐based models as discussed above, meta‐analyses of smoking cessation interventions and diet and physical activity interventions suggest that continuum and stage‐based models do not differ in terms of effectiveness (e.g. Prestwich, Whittington, Dombrowski, Rogers, & Michie, 2014; Riemsma et al., 2003). Interestingly, Riemsma et al. (2003) also found that studies explicitly referencing theory were not reliably more effective than studies that did not explicitly reference theory. Of course, the interventions that did not explicitly use a particular theory may have targeted theoretically important constructs (e.g., self‐efficacy, attitudes toward the behavior, etc.). Furthermore, many of the theory‐based interventions did not explicitly link intervention techniques to theory constructs or did not measure the purported theory‐based mediating constructs. As discussed by Sheeran et al. (2017), in order to promote health and advance health behavior theory, researchers will need to fully test mechanisms of change to know why an intervention succeeds or fails. Because stage models are inherently more complex and more difficult to use than continuum models, it is necessary to show that their effectiveness goes above and beyond continuum models.
As pointed out by many health behavior theorists (e.g., Milne et al., 2000; Sheeran et al., 2016), there is extensive overlap in the constructs of health behavior theories even though they are problematically assessed with different measures. For example, the construct of attitude in the theory of reasoned action/planned behavior is conceptually very similar to costs and benefits in the health belief model or outcome expectations in social cognitive theory. Similarly, concepts related to self‐efficacy are in the majority of health behavior theories. For example, Sheeran et al. (2016, 2017) identified constructs similar to attitude (i.e., expectations and evaluations of the behavior/outcomes of the behavior) and self‐efficacy in most major health behavior theories. An important goal in health behavior change will be to develop a unified way to classify and measure these constructs of interest (see Michie et al., 2014).
Another future direction for health behavior change will be to elucidate behavior change processes that are less intentional or deliberative. In a meta‐analysis of experimental studies by Sheeran et al. (2016), the associations of attitudes, norms, and self‐efficacy with behavior remained predictive when controlling for intentions, perhaps indicating that these predictors of behavior change may not be entirely deliberative or controlled. Although few moderators were identified by Sheeran et al. (2016), one finding of interest was that attitudes and self‐efficacy were less impactful for frequent versus infrequent behaviors. The authors suggest that these frequent behaviors may be more habitual or automatic. Newer health behavior theories do emphasize habit (e.g., habit theory; Wood & Neal, 2007) and implicit processes (e.g., health goal priming; Papies, 2016). Furthermore, many health behavior theories emphasize the processes that might enhance the associations between intentions and behavior. For example, the HAPA emphasizes action planning and coping planning (Schwarzer, 2008).
An important goal for health behavior theories going forward will be to specify what constructs are most important for what types of behaviors, what kinds of people, and what stage of change. For example, behavioral willingness (versus intention) is a better predictor of risky behaviors, whereas intention (versus willingness) is a better predictor of health protective behaviors (Gerrard et al., 2008). Sequential multiple assignment randomized trial (SMART) designs are becoming increasingly used to build individualized treatments to influence health behaviors. For example, Sherwood et al. (2016) are conducting a SMART intervention that is aimed at testing secondary intervention procedures with initial non‐responders in a weight loss study. In this way, more intensive procedures can be utilized with only those that need them, and the most optimal procedures can be identified with respect to timing and content (e.g., introducing portion controlled meals versus introducing self‐regulation training). Taking another tack, researchers have had success tailoring intervention messages and strategies to match people's underlying psychological dispositions (Rothman et al., 2016). In this way, interventions can capitalize on people's self‐regulatory strengths by encouraging them to pursue goals in ways that fit their natural inclinations.
As discussed above, models of health behavior change are generally effective in promoting initial changes in behavior and generally focus on the processes related to deciding and starting to make a change. However, even people who succeed in making changes in their behavior often fail to maintain those changes (Sarwer, von Sydow Green, Vetter, & Wadden, 2009). Strategies to promote maintenance have begun to emerge, such as continually reinforcing the determinants of initial change (a continued care model), instilling the strategies and skills necessary for maintenance at the outset of intervention, and targeting maintenance‐specific predictors once people have had initial success (Rothman, Baldwin, Burns, & Fuglestad, 2016). For example, Rothman et al. (2011) have specified which predictors are most important for initiating change (outcome expectations, self‐efficacy) and maintaining change (satisfaction with outcomes, satisfaction with the behaviors). Interventions on smoking cessation and weight loss provide support for these propositions (Rothman et al., 2011). A direct comparison of a continued care treatment with a satisfaction‐based maintenance treatment found that both were similarly effective in encouraging behavioral maintenance (Rothman et al., 2016).
Clearly, theory and research have made great progress in affecting positive changes in health behavior. As demonstrated in various meta‐analyses, constructs such as self‐efficacy, attitudes, threat susceptibility, self‐monitoring, and stages of change have been shown to effectively predict and influence healthful changes in behavior (e.g., Norcross et al., 2011; Sheeran et al., 2014, 2016). However, many challenges remain. For example, it is unclear whether continuum or stage‐based models are more effective and/or useful in influencing behavior change (Prestwich et al., 2014). It is also unclear as to how to best classify and operationalize health behavior theories and constructs (Michie et al., 2014). Going forward it will be important to conduct direct comparisons of competing theories and to experimentally examine how interventions affect theoretical mechanisms of change and how those mechanisms lead/do not lead to behavior change (Sheeran et al., 2017). Finally, although theories of health behavior change are generally effective in getting people to make initial changes in their behaviors, long‐term maintenance of those behavioral changes remains challenging.
Paul T. Fuglestad is an assistant professor at the University of North Florida. His work applies theory from personality and social psychology to understand and affect health behavior change.
Jody S. Nicholson is a developmental psychologist and an assistant professor at University of North Florida. Her research focuses on parents' role in children's health behavior decisions.
Lauren James is a clinical research coordinator in the Center for Healthcare Delivery Science at Nemours Children's Specialty Care and works to improve health outcomes for children, adolescents, and their families. Through several grant‐funded studies, her research involves working on the design, development, and testing of web‐delivered resources for youth with type 1 diabetes and their parents.