Changing health behaviors using financial incentives: a review from behavioral economics — TEI-tagged view

Ivo VlaevpersName, Dominic KingpersName, Ara DarzipersName, Paul DolanpersName (2019date). “Changing health behaviors using financial incentives: a review from behavioral economics.” BMC Public Healthtitle·j.

Object of study — how financial-incentive design features shape health behavior across populations and contexts

Research question — Which behavioral-economics principles can inform the configuration and implementation of financial incentivesterm for health behavior?

Contribution — a descriptive theoretical reviewterm·mth of behavioral-economics principles, incentiveterm examples, policy implications, uncertainties, and ethical constraints

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

Changing health behaviors using financial incentivesterm: a reviewterm·mth from behavioral economicsterm Abstract Incentivesterm are central to economicsterm and are used across the public and private sectors to influence behavior. Recent interest has been shown in using financial incentivesterm to promote desirable health behaviorsterm and discourage unhealthy onesterm.

If we are going to use incentiveterm schemes to influence health behaviors, then it is important that we give them the best chance of working. Behavioral economicsterm integrates insights from psychologyterm with the laws of economicsterm and provides a number of robust psychological phenomena that help to better explain human behavior. Individuals’ decisions in relation to incentivesterm may be shaped by more subtle features – such as loss aversionterm, overweighting of small probabilitiesterm, hyperbolic discountingterm, increasing payoffsterm, reference pointsterm – many of which have been identified through research in behavioral economicsterm. If incentivesterm are shown to be a useful strategy to influence health behavior, a wider discussion will need to be had about the ethical dimensions of incentivesterm before their wider implementation in different health programmes.

Policy makers across the world are increasingly taking note of lessons from behavioral economicsterm and this paper explores how key principles could help public health practitioners design effective interventions both in relation to incentive designsterm and more widely.

Background The effect of individual behaviorterm on health outcomesterm is considerable with estimates that up to 40% of premature deathsterm in the developed world are attributable to unhealthy behaviors, such as smokingterm, poor dietterm and sedentary lifestyleterm [ 1 ]. Reducing morbidityterm and mortality lossesterm in the future is likely to depend as much on motivating changes in behavior as on developing new treatments or technologies and many countries and health systems are now directing resources to this end [ 2 , 3 ].

A range of tools are at the disposal of policy makers seeking to influence behavior including legislationterm, price signalsterm (taxes and subsidiesterm) and information campaignsterm. Although the use of incentivesterm in wider public policy is nothing new, their role in encouraging health behaviors is a relatively recent phenomenon [ 4 ]. Incentivesterm can take a number of forms including cashterm or alternatively vouchersterm that can then be exchanged for desirable items. The apparent enthusiasm for using incentivesterm to influence health behaviors has come about as the full economicterm and social costs of unhealthy behaviors have become apparent and with the finding that health behaviors can be significantly affected by the structure of economic incentivesterm that individuals face [ 5 , 6 ]. Examples of incentiveterm schemes recently implemented in the United Kingdom include the ‘Give It Up For Baby’ programme in Tayside, Scotland to encourage pregnant smokers to quit the habit and the ‘Pounds for Pounds’ scheme in Kent, England to influence weight controlterm [ 7 ]. In diabetes care, incentiveterm programs directed at patients show promise as a means to influence patient behavior and intermediate outcomes such as weight lossterm [ 8 ].

But widespread concerns do exist and tend to center on the potentially coercive impact of using incentivesterm and the ‘unfairness’ of rewarding people for doing things that are already in their own interest [ 9 , 10 ]. We share some of those reservations, but rather than adding further to this normative debate, we will focus herein on positive ways in which we might give incentivesterm the best opportunity to work if and when they are considered appropriate.

We will look to see whether it is possible to provide further guidance on how best to configure and implement incentiveterm schemes using evidence from behavioral economicsterm. Behavioral economicsterm has come to prominence following the publication of Richard Thaler and Cass Sunstein’s book Nudgeterm [ 11 ], but the science underlying it is built on decades of applied research by the likes of Daniel Kahneman and Amos Tversky [ 12 , 13 ]. In contrast to economicterm models of rational choiceterm suggesting that we respond to information and price signalsterm, insights from across the behavioral sciences suggest that human behavior is actually influenced greatly by the context or environmentterm within which many of our decisions are taken. Nowadays ‘dual process’ theories can be found in social, personality, cognitive, clinical and health psychologyterm [ 14 ]. Two general paradigms for behavior change have emerged over the years. The first type aims to change high-order cognitions such as beliefs and attitudes as a route to influencing deliberate responses by using persuasion and education campaigns. The second approach aims to influence lower-order mental processesterm, thus triggering spontaneous responses, by changing the context or environmentterm within which the person acts without necessarily changing underlying higher-order cognitionsterm such as beliefs and attitudes. In other words, the distinction is between behaviors resulting from internally cued, reflectiveterm, and intentional changes versus behaviors resulting from externally cued, automatic and (often subconscious) reactiveterm changes [ 15 ].

Governments around the world are taking notice of the potential role behavioral economicsterm could play in designing more effective public policy. A prominent example is the coalition Government in the United Kingdom, who have so far made recommendations on how public behavior around charitable giving and preventive health could be influenced using the Mindspace frameworkterm for behavior change [ 16 – 19 ]. Intelligent design of incentiveterm schemes is one of the key tools in this new ‘behavioral’ approach.

Financial incentivesterm in healthcare It is well established that a higher priceterm reduces consumptionterm [ 20 ]. So, we see that smokingterm consumptionterm in Europe has been seen to decrease by about 5% for every 10% increase in the real price of cigarettes and that increasing the price of alcohol is among the most effective options for reducing consumptionterm [ 21 , 22 ]. Price signalsterm to influence behavior in healthcare can take forms besides traditional ‘sin taxes’ for example by subsidising healthy behaviors or rewarding adherence to a treatment programme. The increase in the last decade of schemes aimed at changing the health-related behavior of the public has been accompanied by evidence that even small incentivesterm can positively influence choices [ 23 – 25 ]. Having said that, a recent reviewterm·mth of the evidence for the effects of economicterm instruments (prices or income) on dietary and physical activity behaviors and corollary outcomes [ 26 ] revealed that evidence is limited in terms of the potential for causal inferenceterm and yields ambiguous or inconsistent findings (the evidence is also mostly about impacts on diet, with very limited evidence for impacts on physical activity). Those findings highlight the need to implement robustly designed interventionsterm and evaluations of the logic modelsterm and programme theoriesterm involved.

In clinical psychologyterm, the contextual-change route has taken a substantial share of research, because classical behavior therapy and cognitive-behavioral therapy focus on underlying learning processes and environmental contingencies of reinforcement [ 27 , 28 ]. Schedules of contingent reinforcementterm are also a key to the success of rewardsterm (materialised incentivesterm) and often used in behavioral psychologyterm circles. For example, providing contingent rewardsterm is used in interventions developing constructive habitsterm. This reinforcement principle has been successfully employed to treat drug addictionterm and substance misuseterm (including smokingterm and alcohol consumptionterm) and to improve medication complianceterm [ 29 ]. Such interventions usually include earning money or points contingent on the patients’ specific behaviors.

The effectiveness and long-term sustainabilityterm of behavioral change when incentivesterm are targeted at the more challenging and complex behaviorsterm such as smokingterm and obesity is less well known [ 30 – 33 ]. Also, given the opportunity costs of the changed behavior is the same across income groups, we may also expect to see that as £1 to a poor person is worth more than £1 to a rich person, small incentivesterm are also likely to be more effective in low-income groupsterm [ 34 ].

A reviewterm·mth from behavioral economicsterm We identified interesting examples from the literature through a search strategyterm of electronic databases including (PubMed, EMBASE, PsychInfo) using keywords. We reviewed Longitudinal, cross-sectional and retrospective studies of interventions using incentivesterm to change health behaviors, including patients and/or providers of healthcare. We also included systematic reviewsterm·mth of such interventions. Our team independently reviewed articles and compared and discussed interpretations. Results were synthesized under thematic classification and abstractionterm. Because this is a debate articleterm, we have not provided a reviewterm·mth of all those studies. We do no aim to synthesize the evidence. Rather, in our manuscript we offer a theoretical investigationterm of incentive designterm and behavior change.

Although success has been seen with a number of financial incentive schemesterm targeting preventive health behaviors, some financial incentivesterm have not worked at all well, or even at all in the case of reducing levels of obesity [ 33 ]. It may be that incentivesterm are unlikely to work at a cost-effectiveterm level in changing certain complex behaviorsterm. Another contributing factor could be that too little thought has been put into the design of incentiveterm schemes previously implemented. Roland Fryer has demonstrated how important design is when thinking about how financial incentivesterm can be used to improve educational achievement [ 35 ]. In a series of school based randomized trials, incentivesterm were only found to be effective when they were given for inputs to the educational production functionterm. Incentivesterm tied to educational outputsterm were not effective. Qualitative data suggested that because students do not understand the education production function, they were not able to ‘turn their excitement about rewardsterm’ into meaningful achievement. The same may be true for people offered incentivesterm to lose weight or stop smokingterm.

Those choosing the format of incentiveterm schemes will have many options available to them when thinking about design. Let us take a theoretical example of an incentiveterm scheme to encourage participation in a weight-loss programme. The rewardterm may be given for attending classes and it could be given at the beginning of the programme, at its completion or in increasing or decreasing increments as classes are attended. Alternatively, the incentiveterm could be given dependent on actual weight lossterm targets that result from following the programme. Different scheme designs are likely to lead to different outcomes.

It is important to think about the most effective design of incentiveterm schemes, as our responses can be shaped by a range of predictable biases and heuristicsterm [ 12 , 13 , 36 ]. Behavioral economicsterm provides us with a number of robust psychological phenomena that help explain the decisions we make in a range of settings, including savings, health and education [ 15 , 19 ]. Evidence suggests that human behavior is led by our very human and fallible brain and the context or environmentterm in which many of our choices are taken. The finding being that small changes in context (nudgesterm) can affect behavior as much as large price changes [ 11 , 37 ]. Such effects or ‘nudgesterm’ can be applied to the design of more effective incentiveterm schemes and include the following.

Lossesterm loom larger than gainsterm It has been demonstrated that we react more to lossesterm than to gainsterm of equivalent magnitude [ 38 ], which is embedded into the well-known (Nobel prize winning) Prospect Theoryterm of risky choice. Loss aversionterm implies that someone who loses £10 from his or her pocket will lose more satisfaction than another person would gainterm satisfaction from a £10 windfall. Most incentiveterm schemes tend to offer rewardsterm to participants but inducing some feeling of lossterm if they fail to do something may be more motivating than rewarding them by the same amount. So, instead of providing a £10 rewardterm for each of the ten sessions of a weight lossterm programme, it may be more effective to provide £100 at the end of the programme, with all missed sessions attracting a more salient and painful £10 lossterm. Using a behavioral economicsterm framework, one randomised study has shown the short-term effectiveness of such a scheme. Individuals in this programme contributed to a matched deposit contractterm that rewarded them if they met or exceeded their weight lossterm goal but led to the lossterm of the rewardterm if they failed [ 39 ]. Loss aversionterm is one of the most robust phenomenon from behavioral economicsterm and could be used more widely across incentiveterm schemes.

An interesting intervention tested the impact of financial incentivesterm framed as a gainterm or lossterm to promote Chlamydia screening in students, mimicking the standard outreach approach to student in halls of residence [ 40 ]. This was a in a cluster randomized trial ( N = 1060; age 18–24 years). The students were offered (depending on condition) £5 voucherterm vs. £200 lottery. In the controlterm group the screening rate was 1.5%, while the lottery increased screening to 2.8% and the voucherterm increased screening to 22.8%. Incentivesterm framed as gainsterm were more effective than loss-framed incentivesterm (10.5% vs. 7.1%, respectively). This work contributes to the literature by testing the predictive validity of Prospect Theoryterm to change health behavior in the field.

In another domain, such a framing manipulation was used to increase factory worker productivity in a field experimentterm [ 41 ]. They find conditional incentivesterm framed as both “lossesterm” and “gainsterm” increase productivity for both individuals and teams. In addition, teams more acutely respond to bonuses posed as lossesterm than as comparable bonuses posed as gainsterm. The total team productivity is enhanced by 1% purely due to the framing manipulation. Another interesting intervention tested the power of loss aversionterm to improve teacher performance [ 42 ]. During the 2010–11 Chicago school year teachers were randomly asked to participate in a pay-for-performanceterm program with “gainterm” and “lossterm” treatments. The “gainterm” group received traditional financial incentivesterm at the end of the year in the form of bonuses linked to student achievement. Those teachers in the “lossterm” group were paid a lump-sum in advance and asked to give back the money if their students did not meet performance targetsterm. Teachers in both conditions received the same monetary bonus if they reached the same performance targetsterm. This approach resulted in increases in math test scores for the lossterm condition by an equivalent of increasing teacher quality by more than one standard deviation. The gainterm treatment yields smaller and statistically insignificant results. The authors attribute the significant difference between the lossterm and gainterm condition to the loss aversionterm framing. Those intervention techniques could be used to improve the productivity of the healthcare workforce too.

Overweighting of small probabilitiesterm There is good evidence that people place more weight on small probabilities than standard economicterm theory would suggest [ 43 ], which is another essential element of Prospect Theoryterm. This helps explain the widespread popularity of lotteries and insurance. Although the tendency to overweight the probability of unlikely but salient outcomes can lead to problem gambling [ 44 ], it can also be used to positive effect using lottery based public policy interventions. Lottery based financial incentiveterm programs have been seen to be effective in a weight lossterm intervention and in improving warfarin adherence and anticoagulation controlterm [ 45 ].

Patel et al. [ 46 ] tested the effect of different types of lottery-based financial incentivesterm in increasing physical activity among University of Pennsylvania Employees with body mass index ≥27. All participants used smartphones to track their steps per day and received daily feedback on performance for 26 weeks (financial incentiveterm for 13 weeks and then were follow up for 13 weeks without incentivesterm). Daily lottery incentivesterm were designed as a “higher frequency, smaller rewardterm” (1 in 4 chance of winning $5), “jackpot” (1 in 400 chance of winning $500), or “combined lottery” (18% chance of $5 and 1% chance of $50). The outcome measure was the mean proportion of participants who achieved the daily goal of 7000 steps. During the intervention, only the combined lottery incentivesterm was significantly greater than controlterm (0.38 vs. 0.26 mean proportion of participant days that goal was achieved); and there were no significant differences during follow-up. This study shows that interventions need to experiment with designing different types of lottery schemes.

Living for today at the expense of tomorrow The third phenomenon is hyperbolic discountingterm also known as ‘present biasterm’ [ 47 , 48 ]. Economists assume that our preferences over today versus tomorrow are the same as those over this time next week and this time in eight days. Although standard discounting simply says that we use the same discount rate in each period, evidence tells us that today looms much larger so that we discount very heavily from the present and less heavily once we think about any time into the future. So given the option, some people would choose to take £18 today rather than £20 tomorrow but would be much less inclined to take the £18 in a week’s time than £20 a day later. It has been demonstrated that the immediacy of an incentiveterm can influence outcome of voucher-based incentive programmesterm for substance misuseterm [ 29 ], and an understanding of hyperbolic discountingterm should lead those designing schemes to think more carefully about when the actual incentiveterm is given.

The immediacy of financial incentivesterm also show potential for supporting smokingterm cessation in pregnancy. A reviewterm·mth of evidence found that providing vouchersterm contingent on testing for smokingterm were effective in reducing smokingterm rates in late pregnancy, compared to vouchersterm without testing [ 49 ]. Specifically, linking the incentiveterm to the desired outcome was clearly an important feature of the incentive designterm.

Similarly, an immediate financial incentiveterm has been shown to improve adherence to anti-psychotic medicationsterm [ 50 ]. This trial offered a £15 incentiveterm to one group of patients for each medication taken, whilst a second group received usual careterm. The majority of patients and clinicians felt positive about the use of incentivesterm, and the costs were relatively low. Patients receiving the incentiveterm were more likely to take the medication (85% vs 71%). When the incentivesterm stopped, adherence returned to the same level as those who had not received the incentivesterm.

Increasing rather than decreasing payoffs Incentivesterm have generally been seen to be more effective for one off behaviorsterm like vaccinations [ 25 ]. An understanding of hyperbolic discountingterm is particularly useful for incentiveterm schemes where only a one-off rewardterm is offered. For complex behaviorsterm, multiple incentivesterm may need to be offered at intervals but how should they be given? It has been seen that people respond more to increasing payoffsterm, as opposed to decreasing or constant ones [ 51 ]. This principle has been used to develop successful interventions to treat drug addictionterm and substance misuseterm and to improve medication complianceterm [ 29 ]. In those interventions, the patients earn points contingent on submitting urine specimens that are drug-negativeterm or substance-negative. The reward pointsterm (incentiveterm) usually begin at a low value and increase with each successive negative test result. If the patient fails to provide a scheduled specimen or provides a drug-positive result, then the voucherterm’s value is reset back to the initial low value from which it could begin to increase again. Such incentiveterm contingency scheme can be used to improve outcomes across a wide range of different behaviors and populations.

Reference pointsterm matter A study from the developing world provides a further phenomenon which is that reference pointsterm matter when offering incentivesterm. Evidence from the field, suggests that people care more about what they gainterm or lose around what they already have rather than what they may end up with. It is known that many people tested for human immunodeficiency viruses (HIV) in the developing world do not pick up their results. This is a major challenge to prevention campaigns and has led to a variety of ‘know your status’ campaigns. A programme in Malawi has shown that offering incentivesterm can encourage people to pick up their HIV result [ 52 ]. The biggest increase in uptake, by around 50%, is observed when the incentiveterm changed from zero and one-tenth of a day’s wage. Offering more money still positively affects behavior but to a much lesser degree. This finding is consistent with a ‘concave’ utility function in economicsterm known as ‘diminishing marginal utility of incometerm’ (more income impacts us less), but the rate at which utility declines when income rises would have to be extremely steep which is not what traditional economicterm models would predict. The results suggest that the utility of moneyterm is judged relative to reference pointsterm that are very contextually and narrowly defined. This finding also suggests that such locally defined reference pointsterm could influence decisions around price and the cost-effectiveness of offered incentivesterm.

Ethical concerns Motivating behavior change in health is much more complex than can be accomplished with a single strategy and offering incentivesterm (both positive and negative) are just one route to achieving improvements in health outcomesterm. Financial incentivesterm are increasingly seen as an important vehicle to bring about changes in behavior that lead to healthier lifestyles and supporters and critics alike can be passionate about their use. Supporters of incentiveterm schemes generally believe that people should be encouraged where possible into behaviors that promote improved health outcomesterm and that appropriately targeted incentivesterm can reduce inequalities in health outcomesterm [ 4 , 25 , 53 ]. Incentiveterm programmes can be seen as an example of symmetric or libertarian paternalismterm that steer people towards better choices without limiting what those choices are [ 54 ]. But there are also legitimate ethical concerns. Monetary compensation has the potential to lead to intrinsic motivationterm being ‘crowded outterm’ or partially destroyed [ 55 ], so that when an activity is associated with an external rewardterm, a person may be less inclined to do the activity in the future without further rewardsterm. However, this concern is related to the efficacy of incentivesterm rather than the ethics. More concern comes from the perception that incentivesterm can be seen as a form of briberyterm and/or coercionterm and inconsistent with shred social values [ 56 ].

Marteau et al. [ 9 ] suggest a psychological perspective that can help us think about the appropriateness of using incentivesterm for encouraging participation screening tests. It is known that individuals do not always act according to their long-term goals and interests. We also know that in hindsight people often would prefer to have acted differently. For example, most of us would like to eat more healthy food, drink less, and stop smokingterm. Still, our behaviours do not match such intentions. Offering people a rewardterm/incentiveterm helps them to align their actions with such preferences. In this way, incentivesterm enhance their autonomy to act according to their true underlying preferences. This may help explain why governments and private organisations increasingly apply financial incentivesterm, or other extrinsic motivations, to improve health [ 57 ]. There is also evidence that the public supports incentiveterm schemes that are cost-effectiveterm [ 58 , 59 ].

Consequently, alongside further exploration of the appropriateness of using incentivesterm in health we also need to determine whether or not they work and are cost-effectiveterm. At present we do not know what impact different incentiveterm schemes have on various health behaviors or what impact they may have on different socioeconomic groups invited for appointments. If we learn that incentivesterm are unsuccessful and/or are not cost-effectiveterm then more broadly implementing these interventions would be considered unwise and would not be advised. Researchers have explored the use of such incentiveterm schemes in a range of settings, and with different populations. The results have varied, suggesting that incentivesterm are context-dependentterm, and need to be planned carefully according to the needs and preferences of different groups. For example, offering small financial incentivesterm – such as £10 in cashterm, or the opportunity to win £1000 in a lottery – did not improve attendance at eye screening for people with diabetes [ 60 ]. This study recruited people who had not attended their eye screening appointment in the last two years. Patients in the two incentiveterm groups were actually less likely to attend their appointments than those who received the standard invitation. The reasons for this unexpected result were not clear, but all the patients involved were from relatively deprived groups with a history of non-attendance. The researchers also suggested that the offer of an incentiveterm may cause a negative reaction, if the recipient believes that the screening must be unpleasant if they have to be paid to do it. This is supported by the finding that the lottery offer – to win a much larger sum – was associated with the lowest levels of attendance.

Given the promise of incentivesterm to positively motivate behavior, we should not necessarily use legitimate concerns about their wider use in public policy prevent us from studying them in earnest. This article seeks to do this while recognising the concerns about their use. If incentivesterm are shown to be a useful strategy to influence health behavior, a wider discussion will need to be had about the ethical dimensions of incentivesterm before their wider implementation in different health programmes.

Conclusions Financial incentivesterm are increasingly seen as an important vehicle by which to bring about changes in behavior that can lead to healthier lifestyles. Limited evidence also suggests that appropriately targeted incentivesterm could reduce inequalities in health outcomesterm [ 53 , 61 ]. There are lessons that can be learned from a range of disciplines and we have focused on those from behavioral economicsterm in this article.

Policy implications We cannot give specific advice to policy makers for all individual circumstances, but there are some clear examples of policy and program applications of the general ideas presented in this article.

Financial incentivesterm have been seen to be more effective in increasing performance of infrequent behaviorsterm (e.g. vaccinations, screening) rather than in more sustained behaviorsterm (e.g. smokingterm, dieting) [ 62 ]. Therefore, the incentives-related lessons from behavioral economicsterm should be more readily applied to infrequent behaviorsterm in the first instance. In relation to ‘overweighting small probabilities’, for example, policy makers and program designers could set up vaccinations incentiveterm programs as lotteries, that is, ‘if you vaccinate your child, we will enter your name into a lottery, and you just might win a lot of money’.

Financial incentivesterm for healthy behavior are already being used by large employers or health insurance providers [ 63 , 64 ]. For example, patients in some states in USA have some benefits reduced or eliminated if they do not participate in health care screenings, keep their medical appointments, take their medications and adhere to health improvement programs as directed by their health care providers [ 65 ].

The United States administration also issued a rule that employers can use financial rewardsterm and penalties for workers worth up to 50% of the health insurance premium as an incentiveterm to quit smokingterm, exercise, eat healthier food, lose weight, and lower cholesterol and blood pressure [ 66 ]. In particular, the rule allows, as part employee wellness programs, employers to rewardterm or penalize employees who meet specific standards related to their health (e.g., rewarding employees who do not use tobacco or who achieve a specific cholesterol level, weight or body mass index). In order to avoid wasting public resources, such programs will provide evidence how such incentivesterm schemes work in the longer-term and which method is most cost-effectiveterm. Employers can improve the effectiveness of such programs by incorporating the behavioral economicsterm principles discussed in this article.

Evidently, incentiveterm programs directed at both providers and patients have become increasingly widespread. Although we have focused on patients and the public in this article, incentivesterm are also frequently used to target provider behavior as a means to improve quality of care [ 67 ]. For example, in 2004 in the United Kingdom, a significant proportion of general practitioners’ pay became tied to a quality and outcomes framework [ 68 ]. The behavioral principles described here may also be applicable to incentive designterm on the provider side and have the potential to alleviate some of the problems identified with current programs [ 67 ]. Pay-for-performanceterm (P4Pterm) where providers receive financial incentivesterm to carry out specific care or improve clinical outcomes (i.e., performance targetsterm at which the incentiveterm is targeted) has been widely implemented. There is emerging important evidence on P4Pterm, which reveals a mixed pictureterm so far. Studies in both United States [ 69 ] and United Kingdom [ 70 ] have found that initially positive effects were short lived. In diabetes care for example, the existing literature indicates P4Pterm incentivesterm stimulate initial gainsterm that later tend to level off, while withdrawing the incentivesterm partially reverts the gainsterm [ 8 ]. We believe that behavioral re-design of some of those P4Pterm programs is very likely to improve their effectiveness.

Finally, behavioral economicsterm also provides some arguments against using incentivesterm in specific circumstances. Some economists claim that offering monetary incentivesterm may lead to destroying (‘crowding-out’) genuine (‘intrinsic’) motivation [ 55 ]. There is some evidence from a meta-analysis of experimental studies, which revealed that after a behavior is associated with an external rewardterm, people are less willing to enact the behavior without further external rewardsterm; in other words, extrinsic rewardsterm undermined motivation [ 71 ]. So, when designing behavior change interventions, we must consider that badly designed policies can exacerbate the challenges we currently face and lead to unintended consequences, such as when financial incentivesterm crowd-out/suppress intrinsic motivationterm for healthy behaviors [ 72 , 73 ].

Future directions This is a debate articleterm, so the presentation of the data is fundamentally descriptive and qualitative, because we do no aim to synthesize the evidence. Rather, we offer a theoretical investigationterm of incentive designterm and behavior change. Incentivesterm and behavior change perhaps present the best examples of where genuine attempts have been made to span the divide between behavioral science and health policy, because incentivesterm is the ‘tool’ most often used by the policy makers to influence behavior. However, despite the recent flurry of scientific theories and reports, there are not enough examples of successful transition from science to policy. This article aims to demonstrate that the evidence now exists where the ‘behavior incentive designterm’ can stimulate intervention projects informed by a scientific understanding of behavior.

This approach is useful in social sciences but is limited. For example, nudgesterm or economicterm interventions that show strong effect on the collective behavior of populationsterm are put at the same level with others that show only a tiny, and probably statistically insignificant, trend. At present, we do not have a complete enough picture of which configuration of incentivesterm works best. So, it is worth pinpointing areas which need more research and areas where plenty of robust studies support a given recommendation.

After reviewing the mounting evidence on the uncertain effect of financial incentivesterm to improve health behaviors, Thirumurthy, Asch and Volpp [ 74 ] conclude that the principle that individuals respond to incentivesterm has considerable empirical support, but the devil is in the detail, because the magnitude of effects differs substantially based on the nature of the behavior, the size of the incentiveterm, the population involved, the social contextterm, and the design (subtle design choices in how incentivesterm are situated, framed, or deployed can have substantial effects on their success). Therefore, investigating different ways of joining up some of the psychological mechanisms of actionterm – loss aversionterm, hyperbolic discountingterm, and increasing payoffsterm – may lead to real improvements in the efficiency and effectiveness of existing and novel incentive designsterm. Incentivesterm based on combinations of individual and group goals have also shown promising results. Furthermore, a potentially powerful approach is combining both patient and provider incentivesterm, but whether this is cost-effectiveterm has yet to be determined [ 8 ].

Further insights from the behavioral sciences could allow us to combine incentiveterm schemes with other policy tools to ensure long term effectiveness. One example are commitment devicesterm [ 15 ], which involves individuals making a public decision about the future that results in some additional (social and/or financial) cost if they fail to follow through with that decision (or additional benefit if they successfully do so). As many people revert to past behaviors once the incentiveterm is withdrawn, combining commitment devicesterm with incentivesterm may be useful in ensuring long-term behavior changeterm. There is limited evidence from a Cochrane reviewterm·mth that commitment contracts can potentially contribute to improving adherence to diagnostic procedures, therapeutic regimens or health promotion and illness prevention initiatives [ 75 ].

The only real way to test further interventions of this ‘joined-up’ kind is through the proper design and use of field experimentsterm [ 76 ]. We would encourage those designing incentiveterm schemes to consider how some of the lessons learnt from smaller scale and more experimental studies can be examined in a more naturalistic setting, thereby avoiding randomisation biasterm. Rigorous evaluationterm and dissemination of the outcomes from current and future schemes is also necessary to add to the limited evidence that tells us what does and does not work when using incentivesterm. The ultimate criterion by which to judge the merit of such interventions is whether they improve the health and well-being of the individual; and the more is the latter, the merrier is the incentiveterm scheme [ 77 , 78 ].