A Decision-Science Approach to Health-Behavior Change

Framing

Framing information in a new way can produce a different response even if two frames contain exactly the same information. For example, Li and Chapman (2009, Study 1, N = 470) presented descriptions of a vaccine that was “100% effective in preventing virus infections that cause 70% of known cases of a specific type of cancer” or that was “70% effective in preventing virus infections that cause all the known cases of a specific type of cancer” (p. 157). Both were accurate descriptions of the first-generation human papillomavirus (HPV) vaccine, but the former description elicited stronger judgments of intention to vaccinate. Decision makers are strongly influenced by a 100% figure, even if using that description does not change the net effectiveness of the vaccine, because of increased sensitivity around reference points, such as the 100% end of the probability scale. The effect is analogous to the certainty effect explained by prospect theory (Tversky & Kahneman, 1992) and fuzzy-trace theory (Corbin, Reyna, Weldon, & Brainerd, 2015).

Framing information in different ways can influence policy judgments as well as individual behavior. In a hypothetical scenario, participants allocated more transplant organs to patients with poor prognoses when the patients with both good and poor prognoses were presented in two different groups (grouped condition) than when all patients were presented in the same group (unified condition; Colby, DeWitt, & Chapman, 2015, Study 3, N = 1,000). In the grouped condition, decision makers spread the scarce resource across the groups, whereas in the unified condition, decision makers focused on using the scarce resource efficiently. In another example, when the benefits of a vaccination program were presented in terms of lives saved, people prioritized beneficiaries who had many life years left. In contrast, when the framing was in terms of lives lost, people prioritized younger beneficiaries, whether or not they had more years left to live (Li, Vietri, Galvani, & Chapman, 2010, Study 2, N = 503). Framing can determine the mental representation of the problem: either as a search for the most efficient use of resources (gain frame or unified presentation) or as a moral dilemma about fair distribution (loss frame or grouped presentation).

Information order

The order in which information is presented can modify its impact. Menu items at the top of the list tend to be chosen more often, and so menus in which selections are in order from healthiest to least healthy result in people making healthier choices (e.g., Policastro, Smith, & Chapman, 2017, N = 9,765). Order effects likely reflect a combination of the ease of information processing and social norms that are subtly communicated by changes in the environment.

Comparisons

Information presentation that affords a comparison is often particularly influential. Comparing two options side by side emphasizes attributes on which the options can be easily compared, whereas evaluating one option in isolation privileges attributes that can be easily evaluated without a reference point (Hsee & Zhang, 2010). In a study on calorie labeling (Roberto, Larsen, Agnew, Baik, & Brownell, 2010, N = 303), participants ordered a meal with fewer calories when the menu included a numeric listing of calories. When considering how much participants ate not only during the meal but also later in the day, however, the researchers found that calorie labeling decreased consumption only when it was accompanied by a statement of the recommended number of calories to eat per day. This condition provided a comparison, allowing participants to place the calories from a particular entrée in context. Traffic-light nutrition labels are an inherently comparative form of calorie labeling and have proven effective (Thorndike, Riis, Sonnenberg, & Levy, 2014, N = 2,285), perhaps because they relay the gist of the comparison (whether the behavior is better or worse than a relevant standard; Reyna, 2008).

Behavioral incentives

Although incentives often do promote health behavior, such as exercise (Charness & Gneezy, 2009), behavioral-science research suggests that particular types of incentives should be more effective than others. For example, Volpp et al. (2008, N = 57) incentivized participants to lose weight using a regret lottery in which they could win money if they met their weight-loss goals. If a participant’s number was drawn but he or she had not met the weight goal, that person would be told that the money would have been awarded if only the goal had been met. This system capitalizes on the desire to avoid regret. In another condition in that same study, participants were offered the opportunity to put their own money on the line. If they lost the weight, they would get the money back plus 1:1 matching funds from the researchers. However, they would forfeit the money if they did not meet their weight goal. Research indicates that losses are more impactful than equivalent gains, and so the prospect of losing your own money can be more motivating than only gaining additional money. Both incentive groups in this study lost more weight than the no-incentive control group, but that advantage dissipated after the incentives were removed. Monetary rewards sometimes undermine intrinsic motivation for the behavior (e.g., Deci, Ryan, & Koestner, 1999).

Social norms

Financial or other tangible outcomes are not the only incentives that motivate behavior. Social norms can serve as powerful and long-lasting influences. People tend to behave in line with what others are doing (descriptive norms) because the actions of others are perceived as an accurate source of information about appropriate behavior (Cialdini & Goldstein, 2004). Consequently, one is more likely to perform a behavior after finding out that the majority of others are performing that action. For example, telling hotel guests that the majority of other guests reuse their towels during their stay increases this behavior (Goldstein, Cialdini, & Griskevicius, 2008, Study 1, N = 1,058), and telling physicians how their antibiotic-prescribing behavior compares with that of other physicians decreases inappropriate prescribing (Meeker et al., 2016, N = 248). Likewise, learning that a behavior is endorsed by other people is influential (injunctive norms) because compliance is perceived as a route to approval from others (Cialdini & Goldstein, 2004). Even a subtle communication about the expected health behavior can shift decisions. For example, children served themselves and ate more fruits and vegetables when the cafeteria trays included pictures indicating where fruits and vegetables could be placed, suggesting that taking fruits and vegetables is normative (Melnick & Li, 2018, N = 270).

Social comparison

Humans are inclined to compare their situation with that of others. Consequently, social comparison can be employed to promote health behavior. For example, participants in a physical-activity program walked more steps when they received feedback about how their walking compared with that of others than if they only saw feedback about their own walking (Chapman, Colby, Convery, & Coups, 2016, Study 2, N = 64). This intervention makes use of loss aversion because knowing or fearing that one is doing worse than others can motivate one to improve. Indeed, social comparison with competition is particularly motivating. Competitive incentives, in which one is paid for doing better than others, fueled physical activity better than an individual incentive or a collaborative incentive in which teams were paid for achieving a joint goal (Zhang et al., 2016, N = 790).

Prosocial motives

Some health behaviors, such as blood or organ donation, reflect a prosocial motive to help other people. For example, vaccination protects not only the vaccinated individual but also nonvaccinated contacts. Evidence from some laboratory studies suggests that vaccination rates can be increased by emphasizing their benefits to people other than the person vaccinated (Betsch, Böhm, & Korn, 2013; Betsch, Böhm, Korn, & Holtmann, 2017; Chapman et al., 2012; Vietri, Li, Galvani, & Chapman, 2012). Responding to prosocial consequences may boost a decision maker’s self-image.

Cues and prompts

Behavior does not always match intentions, but people who already have favorable intentions to perform a health behavior can be cued to do so with a reminder or prime. A meta-analysis indicated that simply asking people whether they plan to exercise or attend a cancer screening can prompt action (Rodrigues, O’Brien, French, Glidewell, & Sniehotta, 2015). Providing a prompt to make a plan about where and when one will carry out the desired behavior increases follow-through on intentions (Gollwitzer, 1999). For example, prompting employees to write down the day and time they planned to get a flu shot from an on-site clinic increased vaccination uptake (Milkman, Beshears, Choi, Laibson, & Madrian, 2011, N = 3,272). Such planning prompts delegate control of the behavior to situational cues rather than relying on decision makers to remember their positive intentions and find an opportunity to act on them (Gollwitzer, 1999).

Defaults

Making the healthy option the easy choice increases the likelihood that it will be selected. Decision makers have a tendency to stick with the default, or the option they will receive if they do not make an explicit choice. For example, citizens are more likely to be registered organ donors when they must actively opt out if they do not wish to be a donor (whereas inaction leads to donor status) than when they must actively opt in to be a donor (whereas inaction leads to nondonor status; Johnson & Goldstein, 2003, N = 161). Both opt-in and opt-out presentations preserve the ability to choose either option, and yet the two presentations produce often radically different decisions. Physicians prescribe more generic medications when generics are the default in the computerized ordering system (Patel et al., 2014, N = 255). Patients who are prescheduled for a flu-shot appointment that they can cancel if they wish get vaccinated at a higher rate than patients who are informed that they can schedule an appointment if they wish (Chapman, Li, Leventhal, & Leventhal, 2016, N = 886). Default interventions may have their effect through several channels: implying a recommendation for the default course of action, endowing patients with the default state (e.g., an appointment) that patients may be reluctant to forgo, and setting the default action as the course of least resistance.

Recommendations

A recommendation from a health-care worker for a health behavior is strongly associated with adherence. Particularly potent are presumptive recommendations (Brewer, Hall, et al., 2017, N = 30 practices), in which the health-care worker announces that the needed medical intervention will be provided rather than asking whether the patient would like the intervention. This form of recommendation parallels the default effect described above because it sets adherence as the default so that only active hesitancy from the patient would result in deviation from the recommended action. Presumptive recommendations communicate a strong endorsement for the standard of care treatment.

Self-control

Even when people make plans to exercise, eat vegetables, or quit smoking, they may lack the self-control to carry through on those plans. A commitment device allows decision makers to bind themselves to their planned action, such as with the commitment contract mentioned above (Volpp et al., 2008). For example, shoppers who received a discount when buying vegetables and other healthy groceries were offered the option to lay their discount on the line; specifically, they would lose the discount if they did not increase their purchase of healthy foods (Schwartz et al., 2014, N = 6,570). One third of consumers accepted this challenge, and those who were offered the option increased their purchase of healthy food relative to those who were not offered the precommitment. The effect persisted long after the intervention concluded (Mochon, Schwartz, Maroba, Patel, & Ariely, 2017, N = 4,073). Similarly, when using temptation bundling (Milkman, Minson, & Volpp, 2014, N = 226), one can enjoy the tempting indulgence (e.g., listening to an audiobook) only when engaging in the desired behavior (e.g., working out at the gym). In this study, individuals assigned to the temptation-bundling condition exercised more than participants for whom the indulgence was not tied to the health behavior. These commitment devices make use of the tendency to place a high value on immediate outcomes while making little distinction between different levels of delay.