Chapter  101:  Economic Incentives for Preventive Care

Conceptual Framework

Disease prevention and health promotion cover a wide spectrum of behaviors for both consumers and providers, from simple, one-time vaccinations to complex behavioral changes such as weight control. Where a particular preventive concern falls on the spectrum is determined by: 1) the complexity of attitudinal or behavioral changes and commitment required by the consumer; 2) the complexity of attitudinal or behavioral changes and commitment required by the provider; and 3) the relative complexity of the prevention technology itself, e.g., nicotine patches or counseling. For the purposes of this study, the determining factor for whether a study was classified as a simple or complex preventive health concern was whether a sustained behavior change over time was required of the consumer.

From an economist's point of view, the proper question to ask is not whether economic incentives work, but rather how large they need to be, when they should be used, and to whom they should be given. It is likely that the impact of a given incentive on preventive behavior will depend on whether the prevention is simple or complex. If one views the creation of an incentive as overcoming barriers of some sort, the magnitude of those barriers will likely be complex functions of social, psychological, and philosophical values, beliefs, attitudes, conscious and sub-conscious cognitive processes. Contextual and economic issues help define and shape the understanding of the perceived problem the incentive is intended to address, and appropriate attention to these issues would improve the potential success of the incentive.

Figure 1. Conceptual framework

   Figure 1. Conceptual framework

Figure 1

Figure 1

Preventive services offered by a provider are likely influenced by many factors in addition to reimbursement concerns. These include factors such as convenience, held beliefs regarding the appropriateness and effectiveness of the preventive service, patient volume, external pressures on clinic office procedures arising from quality programs such as HEDIS, or uncoordinated practice guidelines from multiple managed care organizations. Barriers to changing levels of preventive services offered may include such concerns as resistance to changes in information technology systems, lack of organizational support, incongruence between economic and other philosophical views held by the provider, the opportunity cost of office visit time, and social cognition processes. Even if undertaken, the consequences of preventive care and health promotion counseling may create further unexpected barriers to change. For example, Bowman et al. found that patients reported declines in social and emotional functional status, including mental health, social health, self-esteem, and higher anxiety and depression scores for up to three months after physician recommendations for behavioral change during routine health maintenance visits and suggested such declines may inhibit a physician from making such recommendations.⁴

Preventive services demanded by the consumer are likewise influenced by many factors. Convenience, out-of-pocket outlays, and held beliefs about the effectiveness of preventive care and health promotion efforts are clearly important, perhaps particularly for the simple preventive concerns. Complex preventive concerns and health promotion, however, requiring sustained effort, over time, may be more strongly impacted by the readiness of the consumer to embrace a behavior change, self-discipline, feelings of self-efficacy and the belief that change is possible, and the willingness of the consumer to accept the provider's authority or collaboration in matters of personal lifestyle. Again, social psychological processes and potential incongruence between economic, patient autonomy and agency, and other philosophical views of the consumer would be expected to come into play.

Table 1. Conceptual framework: Level of action

Table 1. Conceptual framework: Level of action

	Participate	Act
Consumer	Attend educational sessions	Accept service
		Change behavior
Provider	Identify prevention needs	Give service

Figure 1

The financial incentive to provide a service may extend beyond direct reimbursement. The provision of preventive care may enhance the provider's quality reputation and that in turn may lead to increased payments from insurers or patients. Incentives offered to providers could include direct payments to the provider or their group, bonuses, or enhanced status by virtue of profiling, or meeting external regulatory pressures (e.g., HEDIS). Incentives for consumers could include monetary payments and gifts, lower costs for designated services, coverage of necessary components, or lower premiums for those who participate.

The overall perceived value of an incentive may reflect a number of expectations and assumptions held by the provider and consumer. A provider may perceive the incentive, aggregated over the caseload, as inadequate given the case mix as defined by the stages of change. Or, a consumer may give more salience to economic incentives for simple preventive services because the decision to utilize such services requires relatively less cognitive processing. Alternatively, economic incentives may be less salient for complex decisions that require significant cognitive processing which draws on multiple sources of input. Incentives offered to consumers may affect the provider's beliefs and behaviors as well as the consumer's.

Economic and contextual issues may also affect the impact of economic incentives, even well-designed ones, through unnoticed or unintended disincentives. Perhaps of largest concern is the effect of practice environments, including managed care, group practice, and individual practice arrangements, on economic incentives. There is also the potential impact of incentives on trust and patient autonomy within the provider/patient relationship.

Applicable Theory

The purpose of financial incentives is to induce behavior that otherwise would not occur. If appropriately designed, financial incentives move providers and consumers toward better (in a societal sense) behaviors.^a Financial incentives can take many forms. Providers can be paid per service provided (often called fee-for service, FFS), paid per service with a bonus or penalty paid based on assessed performance (often called fee-for-service with withhold) on the basis of their costs (often called cost-based or retrospective payment), a fixed payment (often called capitation or prospective payment), or a mixture of payment types.

The ability of several academic disciplines to provide meaningful insight into the question of effective economic incentives touched upon in this introduction, and the complexity of the processes and structures of decision making and processes of care, point out the importance of social ecological approaches to health care improvement. Addressing macro-level economic forces embedded within the health care system, and broad environmental and social supports for healthy and unhealthy behaviors, it is argued, requires multi-component interventions that address multiple factors.²

In a simple world, basic economic theory suggests that providers should be paid the value of their marginal product for the provision of preventive services. However, there are several real world complications that prevent this optimal payment structure from being implemented. First, there are important externalities in preventive care. Disease prevention on the part of one patient (e.g., stop smoking, vaccinations) reduces the health risk to the population as a whole.⁵ Thus, paying the provider the value of the marginal benefit to a patient would lead to an under-provision of preventive care. Second, the importance of third party payers implies that patients rarely pay the full cost of their care. Third, there are important informational asymmetries between providers and patients and providers and payers. Economic theory has long recognized that the presence of informational asymmetries can result in a wedge between the actual and optimal payment levels. Agency theory and mechanism design branches of economics have arisen to study the role of information and the optimal incentive structures.

Agency theory studies the behavior of two or more groups whose welfare is interdependent but whose incentives are not naturally aligned. An example is the patient-physician relationship. Physicians and patients both need each other, however, physicians may not have the incentive to prevent illness since, in general, the sicker the patient is the more they can earn. Agency theory steps into the problem and designs a payment mechanism that brings the interests of the physician into alignment with the patient. In this example, a solution to this agency problem might be to pay the physician a capitated payment regardless of whether the patient is sick or well. Under this arrangement it is in the physician's interest to have the patient avoid illness thereby reducing the effort and expenditure the physician must spend on treating an ill patient. However, as we show in the following discussion on the basic forms of provider economic incentives, capitation generates other incentives that may be less desirable.

The payment for services is the basic form of economic incentives for providers. Roughly speaking, provider incentives may be divided into two types, FFS or capitation for some portion of the medical services under their control. These two payment methods give physician organizations differing incentives to provide preventive services.

Under FFS it may be difficult for the physician to get reimbursed for implementing preventive services (e.g. smoking/alcohol counseling). Health plans have been disinclined to cover preventive services under FFS payment system because it is difficult to verify if they have been performed. If physicians can attract more patients by performing preventive care, they may be more inclined to do so. In addition, it is difficult ex ante for patients to judge the quality of preventive services provided by physicians and thus they may be unable to select physicians on that basis. Thus, physicians may have little incentive to provide these services if they do not gain more (or better paying) patients for this provision.

Under capitation, since the risk is shifted to the provider, the physician faces many of the same financial incentives to provide preventive care as a health plan. Providers functioning as insurers face a complex web of often offsetting incentives to cover and promote the use of preventive services. They must consider the net cost of its potential preventive care promotion versus the gain in revenue of a given program. Most preventive care, even if cost effective, is not cost saving. Thus, the incentives for a provider to cover a particular preventive service will depend on their ability to recoup the increased cost of preventive care coverage through increases in revenue or through decreased costs due to changing the risk profile of its enrollees. If enrollees appropriately value cost-effective preventive care (and there are several reasons why they may not), then the insurer may indeed increase revenues sufficiently to cover its increased costs through the provision of preventive services.

The financial incentives to provide preventive services under capitation are tied to the ability of the physician to recoup preventive care expenditures. If an enrollee stays with the provider for a long time, it is immediately in the provider's interest, independent of how the enrollees value the service, to cover and promote the use of cost-saving preventive services. However, the time lag between preventive care expenditure and benefits can be long (e.g. diet and exercise).⁶ Because the provider may not keep an enrollee long enough to benefit from the provision of cost-saving care, they may have little financial incentive to provide this care in spite of the potential societal gain from the provision of these services. Given the disincentive caused by enrollee churning, it likely makes more sense for large employers to engage in preventive interventions. However, while employees stay longer with employers than with health plans, there is still significant turnover of employees and the employer faces the same dilemma as the insurer for the provision of preventive services.

The use of bonuses as a means to direct the attention of the provider to preventive services is a relatively new development within the overall provider remuneration structures. While bonus qualification may be directly linked to a specific procedure, such as immunization, it may also be linked to quality of care measures, of which preventive concerns are only contributing factors.

Conditional upon their health insurance status, consumers typically face a less complex array of direct financial incentives for obtaining preventative services. They may have to pay a co-pay to visit the provider and, if the service is not covered under their insurance or they do not have health insurance, they may have to pay a substantial amount to receive preventative care. Perhaps more relevant for consumers in determining their preventive behavior is the opportunity cost. If the care requires a trip to the physician, the transportation, opportunity cost in time, and the psychological cost of the physician visit may loom larger than the out-of-pocket expenditures. If the preventive activity requires a sustained behavior change(e.g. weight loss), the primary disincentive to engage in this behavior may simply be the cost of the time devoted to preventive care and the discomfort that the new behavior may induce. For example, the financial incentives designed to induce weight loss will have to be sufficient to overcome these time costs and the psychic cost of discomfort.

The first managed care organizations emphasized the role of preventive services in maintaining the overall health of their enrollee population. However, there is a sense that the emphasis on preventive care that was one of the hallmarks of the first HMOs is no longer common among managed care organizations. Thus, it is reasonable to ask: “If health care consumers value preventive services, what might deter health plans from offering these services?” Below we discuss some of the market imperfections that may impede individuals from getting the right amount of care.

First, we note that individuals appear to be price sensitive towards preventive services. Kenkel finds that women with insurance are five percentage points more likely to have breast exams and cervical screenings.⁷ Mullahy finds that those with insurance are three percentage points more likely to get a flu shot.⁸ Cherkin et al. finds that $5 co-pay resulted in fewer physical exams.⁹ In the Rand Health Insurance experiment women with free care were three to seven percentage points more likely to receive preventive care.¹⁰

Individuals seeking preventive care face monetary, time, and psychological costs. Poor segments of the population may simply not have the monetary means as they may not have health insurance or the access to the necessary infrastructure (e.g. transportation) to obtain preventive services. Many preventive behaviors by individuals require them to make significant time investments. Changing life style by exercising or quitting smoking is psychologically more costly than getting an immunization. It means giving up things that are pleasant. Sustained change is harder to achieve than one-time events.

Recent developments in economic theory have emphasized the importance of observed departures from the rational decision making model.^11–13 These models highlight that individuals may have time-inconsistent preferences and are therefore procrastinators. That is, individuals know that certain behaviors are bad for them, but they are unable to change their behavior even though they realize that changing their behavior would ultimately make them better off. Therefore, in these models, individuals who engaged in unhealthy behaviors may be better off if commitment mechanisms are put in place to help them change their behavior.¹¹

Since many individuals have access to health insurance (or certainly will at age 65) they have less incentive to engage in preventive care, as they will not have to pay the direct cost of treatment. However, the evidence on this indicates that insured individuals are less likely to engage in “dangerous” health behaviors.¹⁴ That is, those who select into insurance are likely to be less risk averse or have lower discount rates and therefore invest more in their health. Individuals may already have an incentive to engage in healthy behavior to avoid the disutility of illness.

Information may be an important barrier for individuals in seeking out preventive care. Many individuals rely on their physicians to inform them on what preventive services they should have performed and, as discussed above, physicians may not have the appropriate incentives to inform their patients. It is possible that individuals may not know about the health benefits/risks of certain behaviors/screening. However, there is evidence that people do respond to information regarding health consequences of behavior (e.g., Surgeon General's Report on Smoking, 1962) and diet.^{15, 16} Furthermore, individuals may not be able to assess their risk of various diseases. There is mixed evidence on whether individuals under or over estimate the risks of particular behaviors. Viscusi et al. find that smokers overestimate the risks associated with smoking.^17–19 However, Schoenbaum finds that heavy smokers over-estimate the likelihood they will live to 75.²⁰

Designing the appropriate incentive structures are non-trivial problems requiring an understanding of the information structures (e.g. who knows what and what behaviors are observable or measurable) and the responsiveness of providers, consumers, and payers to different incentive structures. The economics literature within the field of “mechanism design” has devoted significant amounts of effort to understanding the structure of the optimal incentive design. Viewed somewhat differently, instituting a new incentive structure is presumably a response to a perceived problem in how actors in the health care system are currently behaving. It is important to recognize that understanding the source of the problem is required in order to create an appropriate incentive design.

The question necessarily arises then - what would an effective economic incentive arm of a multi-component intervention for preventive care and health promotion look like. This is itself a complex question. While remaining mindful of the larger health care quality context, we can deepen our understanding of the role of economic incentives for preventive care by confining an examination of the literature to those studies which allow us to directly address the effectiveness of economic incentives for preventive care and health promotion.

This review focuses on evaluating the evidence for an intervention intended to modify behaviors, not health states. That is, we are looking for marginal changes in specific behavior embedded within a complex context. As such, study outcome measures should focus on measuring behaviors, such as utilization of preventive services, rather than, or in addition to, patient health outcomes.

This review does not directly assess the impact of HMOs and other managed care organization risk-sharing and payment mechanisms as compared to FFS. While there is considerable and understandable interest in the effects of these larger economic incentives, the focus is on explicit economic incentives for preventive care. In addition, the potentially numerous confounding factors derived from different patient populations, physician populations, and structures and processes of different systems might overwhelm the potential usefulness of a review of explicit incentives.²¹

Cost Effectiveness

A frequent criticism of the US health care system is that not enough preventive care is performed. In these critiques, several reasons for the dearth of preventive care are usually offered, and the list of causal forces that cap the level of care below what it “should” be is that providers and consumers are not given the appropriate economic incentives to perform or seek out preventive care. However, while the calls for more preventive services have been made, it is unclear whether more preventive services would make the health care system more cost-effective.²²

Cost-effectiveness can be thought about in two basic ways: 1) achieving a given preventive goal at a lower cost (sometimes called productive efficiency) and 2) achieving an improvement in health per unit of cost that exceeds a societally acceptable threshold (sometimes called allocative efficiency). Most of the work addressed here is restricted to the first definition, although ultimately the cost effectiveness of the incentive will depend on the cost effectiveness of the underlying preventive behavior it is designed to encourage. Screening for diseases and prevention of illness is very expensive. Russell estimates that if the recommendations of the US Preventive Services Task Force and the American Cancer Society were followed, screening for cervical cancer, prostate cancer, and cholesterol alone would cost as much as $80 billion per year.²² Tengs and Graham conclude that the most cost-effective interventions are not those most frequently used, suggesting that what is needed is not more preventive care, but rather a reallocation of preventive services.²³ The Tengs and Graham finding also suggests that the appropriate incentives (financial or otherwise) are not currently in place for providing the right mix of preventive services.

Definitions of Terms for the Study

Prevention. Clinical preventive medicine is that part of preventive medicine concerned with the maintenance and promotion of health and the reduction of risk factors that result in injury and disease.²⁴ The three main types of prevention are primary prevention aimed at not allowing a disease or adverse event to occur, secondary prevention aimed at the early detection of an asymptomatic disease, and tertiary prevention aimed at reducing the adverse consequences of a diagnosed disease. Immunizations are primary preventive acts. So is health promotion, such as smoking cessation, exercise, and attention to nutrition.

In the current environment of growing chronic illness burden and improving identification of risk factors for major diseases such as heart disease, the boundaries between primary, secondary, and tertiary prevention begin to blur. For example, high cholesterol and hypertension, though risk factors, are identified as treatable conditions. However, the purpose of treating high cholesterol is prevention of full-blown heart disease.

The fuzziness of the distinction between prevention and treatment compelled us to place specific boundaries on the literature search in order to narrow the scope of the project. For the purposes of this report, we defined preventive care and health promotion as involving those situations where consumers may consider themselves healthy or physically at risk but not yet labeled with a diagnosis. This would include individual-based health promotion and preventives services as defined in Healthy People 2000 and 2010, but excluding mental health, substance abuse, and health protection concerns such as injury prevention, occupational health and safety, environmental health, and oral health. Tertiary care, including self-care and management of diagnosed chronic illnesses such as diabetes and heart disease, was also excluded. The scope definition accords with the recent US Preventive Services Task Force's scope and selection of topics.¹²

Prevention and health promotion is the purview of both medical and public health professionals. We recognized this shared contribution by including non-clinical settings, such as worksite and community-based health promotion settings.

Economic Incentive. This review examines explicit incentives targeted at specific individuals, either providers or consumers. Consumer incentives are fairly straightforward and include cash, gifts, lotteries, and other free or reduced price goods and services for the benefit of the specific consumer.

Any provider fee or reimbursement system within health care is by definition an economic incentive. Financial incentives are paid at the organizational level, both at health plan to independent practice associations (IPA), and from IPAs to medical groups. Financial incentives are paid at the direct provider level as well. Financial incentives can be designed to direct attention to different measures of quality of care, utilization management, and other administrative and organizational concerns regarding shared risk and profits. As we are focused on target incentives for specific individuals for preventive care, this review examines incentives that are below the organizational level, directly payable to physicians or other health care staff members, and affect the providers' marginal income, not base income.

Incentives offered to providers could include direct payments or bonuses to the provider or his/her group. It was expected that economic incentives would vary considerably by the nature of the incentive, the components involved, size, frequency, duration, and the conditions that triggered payment of the incentive.

The definitions of preventive care and economic incentives given above are not easily applied cleanly to the literature. Research questions regarding the effectiveness of financial incentives are conceptualized such that they shade into questions regarding insurance status, access to care, and utilization of resources, rather than preventive care behaviors and outcomes. For example, Adams investigated whether fee generosity for Medicaid's Early and Periodic Screening, Diagnostic, and Treatment (EPSTD) program affects the number of physicians providing the services and improved access to care for an underserved population.²⁵ Also, as stated previously, prevention is often used as only one component of quality of care when measuring effectiveness of incentives. For example, Dudley conducted a literature review on the effect of financial incentives on quality of care that relied on IOM's definition of quality (the degree to which health services for individuals and populations increase the likelihood of desired health outcomes and are consistent with current professional knowledge) and included a general category of preventive services utilization as one of eight study outcome measures.^{26, 27}

Inclusion Criteria

• Published between 1966 and Oct 2002.

• Address primary preventive care defined as vaccination, screening, and health promotion behaviors such as smoking cessation and weight loss.

• Be a primary study.

• Take place in an industrialized country.

• Be written in English.

• Examine preventive care as at least one primary outcome.

Exclusion Criteria

• Patient adherence to drug therapy. For example, the treatment of tuberculosis, which has been the subject of many adherence studies and could be considered a preventive step because it prevents the spread of the disease, is excluded. For our purposes, primary prevention is defined as occurring pre-diagnosis.

• Financial rewards for participating in a research study. The economic incentive must be conceived as part of the intervention.

• Multiple component interventions in which the economic incentive is only one component and the study design precluded analyzing the effect of the incentive separate from the other components. Education and outreach efforts are examples of components in addition to economic incentives. This was most strongly noted with community and education-based prevention programs and worksite health promotion programs.

• Studies examining payment forms provided by more than one payment system, i.e., HMO vs. FFS were excluded because there are too many potential confounding factors. It was therefore too difficult to isolate the effect of specific economic incentives.

Selection bias is a major concern for outcomes research on economic incentives. The opportunity for self-selection of consumers, in particular, is high, given all of the unmeasured economic and contextual issues that may come into play for any specific economic incentive intervention. A strong research design including randomization greatly minimizes this concern. Therefore, we included only RCTs, time series, and prospective quasi-experimental designs for the structured literature reviews. However, we also provided information from relatively well-designed econometric cross-sectional studies for the provider incentives as another perspective for consideration.

Figure 2. Economic incentives search result flow chart (more...)

   Figure 2. Economic incentives search result flow chart

Figure 2

Econometric Research

Studies that rely on non-randomized trials (we will refer to these as observational studies) can teach us a great deal about health care provider and consumer behavior.^b In fact, data from observational studies in which subject are not randomized into “treatment” categories form the foundation of the empirical body of knowledge for both health economics and medical sociology. However, because the subjects are not randomized, the data analysis often requires special statistical attention to address potential biases induced by the nonrandomization. If the data meet certain criteria and appropriate statistical methods are applied given the nature of the data, observational studies can yield inferences that are equally valid (and in some circumstances the inferences are more easily generalized to real world practice) than those inferences made from a randomized trial. While observational studies can yield valid inferences if the appropriate statistical methods are applied, often those methods are not applied, or there are no methods that can correct for the biases given the nature of the data, and the resulting inferences are incorrect. Thus, in assessing the validity of the conclusions from a body of observational studies, careful consideration of the nature of the data and the appropriateness of the statistical methods is necessary. In order to provide a systematic method for assessing the strength of the evidence for literatures in which observational studies comprise an important component of the body of work, we developed an instrument to assess the validity of a given study's conclusions. This algorithm is provided in Appendix E.

The instrument attempts to jointly assess the quality of the data used in the analysis and, given the nature of the data, the appropriateness of the statistical methods applied to the data. Specifically, the instrument places studies into one of three categories: Very Informative, Informative, and Not Informative. The underlying logic of the instrument is that the greatest risk in making biased inferences from observational studies is from poor identification of the variables of interest from the lack of appropriate exogenous variation in the variables of interest. Thus, the instrument measures two categories of characteristics: ones that may bias the coefficient estimates, and dimensions of the study that affect the generalizability of the findings.

In order to place a given study in one of these three categories the algorithm first assesses if there is the appropriate variation in the data to measure the coefficients of interest. Then, the instrument assesses the nature of the variation in the data. That is, observational studies are at risk for “endogenous” variation in the variables of interest. An endogenous variable is one in which there is correlation between the error term and the right hand side variables. That is, there is some unobserved factor that impacts both the outcome of interest and the intervention thereby confounding inferences regarding the independent impact of the intervention on the outcome. The instrument ascertains whether the data generating process can be viewed as exogenous of the unobserved factors. It is often the case in observational data the variables of interest are endogenous and the study uses an appropriate statistical method to correct for any bias that would result from Ordinary Least Squares analysis. In a given study, if data is likely endogenous, then the instrument asks if appropriate statistical methods have been employed to correct for this potential bias. For a study to be categorized as ‘Very Informative’ it must have both appropriate variation in the data and either exogenous variation in the data or employ an appropriate statistical method to correct for the potential bias (e.g. instrumental variables with appropriate instruments). In addition, for the study to score in the ‘Very Informative’ category it must also score well in the other study criteria.

The other criteria attempt to measure the generalizability of the findings to populations outside of the sample population analyzed in the study and the appropriateness of the statistical methods.

Key Question 1. How have “preventive care” and “economic incentive” been defined in the literature?

Provider. In general, definitions for neither term are specifically addressed in the literature.

All studies approached preventive care from the perspective of the study's targeted preventive care service. Both Fairbrother, Hanson, Friedman et al. and Fairbrother, Siegal, Friedman et al. explicitly based the operational definition of the immunization outcome on expert panel guidelines.^{43, 44} Studies by Hillman, Ripley, Goldfarb et al. and Morrow, Gooding, and Clark referenced expert panel guidelines as justification for the chosen outcomes. ^45–47 No study provided a conceptual discussion of prevention or the study's outcome measures with regard to the simplicity or complexity of the preventive processes under examination.

While intending to examine the effects of incentive interventions on preventive behavior changes, most studies used operational definitions of prevention that relied on measuring change by recorded activity rather than actual observed behavior changes. Recorded activity allowed for preventive services obtained from other sources as well as from the provider subject to the economic incentive. Thus, apparent increases in preventive services may have reflected improvement in chart documentation procedures rather than actual increases in inoculations. Only Fairbrother, Hanson, Friedman et al. and Fairbrother, Siegal, Friedman et al., included a bonus targeted at increasing inoculations given by a provider, or the provider's staff, as one intervention arm under investigation, and Grady, Lemkau, Lee et al. included a behavioral outcome of mammography referrals.^{43, 44, 48}

No study based the outcome measure on evidence from the literature, nor discussed whether any relevant evidence literature was available to benchmark the preventive behavior of the provider participating in the study. Thus, it is not surprising no study reported examining providers' pre-test performance to verify the level of preventive care provided according to criteria for appropriate levels of preventive care as established by empirical evidence of expert panels. Without this examination, we are unable to assess the potential for ceiling effects.

Only Grady, Lemkau, Lee et al. included a conceptual discussion of economic incentives from the behavioral psychology theoretical perspective.⁴⁸ No report provided a definition or discussion of financial incentives as a map to locate the chosen intervention design for the study on such attributes as size, eligibility (threshold levels, individual performance, tournament style), or form (fixed or continuous payments).

Consumer. No study provided a conceptual discussion of prevention or health promotion for the study's outcome measures with regard to the simplicity or complexity of the preventive measures being examined. It appears consumer research is generally motivated by policy considerations such as responding to the federal Healthy People initiatives or achieving recommendations of national expert panels.

As was discussed in the Methods chapter, given the lack of consensus and implied, rather than specifically reported, concepts and assumptions underlying the research, we created categorization schemes based on the themes which emerged from the analysis. These categorization schemes were used to assist in answering Key Questions 2–4.

Key Question 2. Do incentives work?

To answer this question, we first discuss observed study design issues, that is, who was being studied, both the intervention target and the patients involved, what was the nature of the interventions applied, and how was effectiveness measured. We then turn to what can and cannot be said, based on the patterns of outcomes.

Nature of the intervention

Provider. In general, studies did not include justification for the specific design of the economic incentive.

Types of incentives: As shown in Evidence Table 6 in Appendix F, most explicit incentives were bonuses potentially payable to all qualifying physicians. The exceptions are the Hillman studies which paid bonuses in a tournament-style manner.^{45, 46} That is, only the top performing providers are rewarded with bonus incentives. Providers would need to estimate their ability to win such tournaments.

Only a few studies provided data on the range and mean size of bonus payments. Kouides, Bennett, Lewis et al. reported an average $242 bonus.⁴⁹ This seems to be a very small amount compared to typical physician incomes. In this study physicians created the patient list to target for immunizations. This collaborative involvement in the study may have increased the salience and personal motivation of the provider to reach the immunization goal. Hillman et al. reported bonus averages and ranges that were slightly more significant—up to $4,682 for one study.^{45, 46} The overall amounts are still small, and since bonuses were paid tournament style to only the top performing or improving sites, the marginal benefit of the bonus would need to be tempered by the physician's expectation of receiving the bonus. Grady, Lemkau, Lee et al. provided token amounts of $50 for achieving the target behavior goal.⁴⁸

No study provided information on the frequency and timing of payments. Reports also did not include the investment costs the physician or provider may have faced to establish the clinical and office procedures necessary to support production and behavioral changes, or whether they had the time, staff, or office system supports available for such changes to be feasible. We cannot assess from the information provided how these factors, plus the anticipated length of time for the bonus program, would have influenced the physician's decision on whether the incentive was considered financially beneficial.

Incentive conditions: In general, incentives were based either on patient outcomes or physician performance of a behavioral goal (Evidence Table 6 in Appendix F). Six studies used performance bonuses that rewarded the physician for achieving a target outcome rate based on patient utilization, in these instances immunizations^{43, 44, 46, 49, 51} and cancer screening.⁴⁵ Grady, Lemkau, Lee et al. applied a performance bonus that rewarded the physician for achieving a target behavior goal, a 50 percent mammography referral rate.⁴⁸ The Fairbrother studies also pitted the target outcome bonus against a per-input bonus, a more direct way of measuring physician behavior, which paid the physician for each additional shot administered and office visits that brought the child up to date in immunization coverage.^{43, 44} Morrow, Gooding, Clark et al. examined the effect of adjusted capitation rates partially based on a practice achieving preventive care outcome rates.⁴⁷ Fox and Phau examined the effect of raised delivery and prenatal visit fees, an incentive based on physician behavior.⁵⁰

Salience of the incentive: Only the Hillman studies attempted to directly assess awareness of the incentive program within a practice.^{45, 46} Only 56 percent and 67 percent of sites, respectively, responding to a survey indicated they were aware of the incentive program. As the incentive apparently was communicated to providers through regular HMO communication practices, this design aspect may be more indicative of “real world” responses to incentive programs than the remainder of the studies that involved more direct communication of the experiment to potential participants. Communications from HMOs may be buried under the busyness of a physician's normal daily practice.

The total incentive size paid to any one provider would arguably also impact the salience of an incentive program. While several studies did provide some payout information, as noted above, no study provided an assessment of the relative impact of the incentive on the provider's overall income.

Expectation of payment: Expectation to receive the bonus is also likely to impact a provider's decision to change preventive care behaviors. Physicians may have higher expectations to earn incentive payments if such payments are based on their own behaviors, not utilization behavior of the patients. Expectations would be affected not only by how the criteria of success are determined, but also whether the incentive is potentially payable to all physicians if they meet the target levels, or if the incentive is distributed in a tournament style.

As was mentioned earlier, because reports did not include information on pre-test assessments of preventive care performance against set criteria, it is uncertain whether physicians would have had low or high expectations of earning the incentive payments, regardless of tournament style or the potential for all to qualify.

In short, estimating the chance of receiving the bonus would likely be very difficult for the provider. The provider may not have been aware of his or her starting position when estimating the likelihood of success. Some of the bonus structures were very complex. The provider would have little to no control over the patient's choices regarding whether or not to accept the incented service.

Other factors: Timing of payment and the interplay between financial incentives and performance feedback may potentially impact the effectiveness of the incentive and should be taken into consideration. The studies did not generally address the average time lag between the provider activities which influence the potential for receiving a bonus and the actual receipt of the bonus. If everything else is constant, the longer the delay between the behavior and the reinforcement, the weaker the effect.

Performance incentives include an element of feedback. Feedback may be formal, through the use of reports, or informal, in that receipt of the incentive itself functions as feedback of performance levels. Feedback may also be private or public to all providers participating in the incentive program. As seen in Evidence Table 6 in Appendix F, some studies did include formal feedback reports. However, the studies often did not clarify whether such reports were private or public. Thus, it is difficult to assess the contribution of feedback to the effect of economic incentives. However, it should be noted Grady Lemkau, Lee et al. found a token bonus of $50, intended as a feedback mechanism, was not significant for changed behavior.⁴⁸

Consumer. Few studies provided justification for the specific design of the chosen economic incentive, although Smith, Weinman, Johnson et al.⁵⁵ and Kamb, Rhodes, Hoxworth et al.⁶⁴ noted using focus groups to establish the form and size of the incentive, while Laken and Ager⁶⁵ cited advice from the study population. Interestingly, while a few articles do mention Prochaska's meta-theory as a basis for the design of educational or motivational counseling sessions, no study discussed the impact of the theory on the design of the incentive.

Although several studies did provide some descriptive information on the uptake and desirability of the study incentive, only nine studies could be said to include direct tests of the uptake of an incentive (Evidence Table 2 in Appendix F). Another ten studies set different forms of economic incentives against each other as a direct test of the desirability of incentive form.

Table 2. Consumer incentive type

Table 2. Consumer incentive type

	Incentive Type			Positive Findings to Total Studies
	Lottery	Gift	Cash	Coupon	Free Medical	Punishment	Totals
Simple	2 of 5 (40%)	2 of 5 (40%)	5 of 5 (100%)	10 of 12 (83%)	3 of 4 (75%)	3 of 3 (100%)	25 of 34 (74%)
Complex	4 of 5 (80%)	2 of 2 (100%)	3 of 6 (50%)	2 of 3 (67%)	1 of 2 (50%) *	6 of 7 (86%)	18 of 25 (72%)
Totals	6 of 10 (60%)	4 of 7 (57%)	8 of 11 (73%)	12 of 15 (80%)	4 of 6 (67%)	9 of 10 (90%)	43 of 59 (73%)

*

One non-significant study mismatched test for effect of incentive. No direct measure of uptake.

Coupon = coupons, vouchers, gift certificates, free or reduced non-medical services

Free = free or reduced medical services

Punishment:

Punishment for simple category was clearly punishment.

Punishment for complex category was monetary return contracts that included a reward element. Those that measured followup periods showed rebound back to non-significant levels.

Non-significant test was for smoking cessation. The rest were weight loss.

Types of economic incentives: The 59 incentives offered in the studies were highly varied in type and size (Table 2). There were ten lotteries, seven gifts, 11 cash incentives, 15 coupons for free or reduced price goods or non-medical services, six free or reduced price medical services, and ten incentives involving negative reinforcement or the opportunity to avoid punishment.

All lotteries and raffles are included in the lottery category. Five of the ten lotteries were for cash prizes, ranging from $40 to $100. The remaining prizes included a microwave, dinner for two at a local restaurant, travel packages, and groceries. Few studies provided information regarding the expected value of the lottery per participant, leaving it difficult to compare the value against the incentive categories that did not involve uncertainty in receiving the incentive. In addition, lottery values depend on the cognitive processing of assessing expectations, which may or may not be ‘accurate.’ Lotteries were generally used as rewards for adhering, participating, and outcome categories.

Gifts were physical goods provided as incentives. Gifts ranged from a nutritional information package valued at $2 to inexpensive jewelry, baby blankets, and other infant products. Gifts within interventions for the two complex preventive concerns were accompanied by other intervention components such as lotteries, tournaments, and other educational components. Gifts were all given as rewards and evenly split between the adhering and participating categories.

Cash incentives ranged from $5 to return a tuberculosis skin test reading, to a potential $500 over ten months for smoking abstinence. Simple preventive incentives were $5, $10, or $15 per event. Complex preventive incentives ranged from $1 to $25 per payable event, payable from once per week to several months apart, with potential total payment per person, when the information was provided, of $50 to $500. Cash incentives were treated as rewards and were relatively evenly divided between the adhering, participating, and outcome categories.

Coupons included free or reduced-price coupons, vouchers, and gift certificates for non-medical goods and services. The coupon values were generally modest, from $2 to $15, when the values were provided. Coupons might be related to an encouraged healthy behavior, such as exercise passes, farmers markets, and condom purchase; to address barriers such as transportation; or were unrelated to the specific aim of the study, such as haircuts and infant formula. Given the flexible nature of this category, it is not surprising that it is well represented in virtually all categories, with coupons not used for only the outcome category, and more likely to be used for simple, rather than complex, preventive concerns.

Free or reduced cost medical services involved free influenza immunizations, reduced fees for clinic visits, and free or reduced price nicotine replacements. The studies providing this form of incentive viewed price as a barrier to care-seeking and healthy behavior. The values of incentives in this category were generally higher on a per event basis than the other incentive categories (assuming the expected value of an uncertain lottery is less than the face value), ranging from $10 to $25. All free or reduced medical service incentives were categorized as barrier removal and were evenly divided between the facilitating and adhering categories.

The last type of incentive included those interventions that intentionally incorporated a disincentive to exhibit behaviors counter to the desired behavior, including both punishments and negative reinforcement. Punishments were examined in three simple preventive concern studies, all related to vulnerable populations and the receipt of government subsidies. Birkhead LeBaron, Parsons et al. and Kerpelman, Connell, & Gunn required mothers to return to WIC offices more frequently to receive their benefits if the children were not up to date in immunization coverage.^{57, 58} Kerpelman, Connell, & Gunn tested whether the threat of loss of AFDC benefits for non-immunized children would effectively prompt parents to immunize.⁵⁸ All three incentives were represented in the adhering category.

The remainder of the studies in this category involved the use of monetary contracts; the participant deposited his/her own money in a contract whereby the money was returned in increments if agreed upon goals were met. The money was forfeited if the goals were not met. Such contracts were often made in the presence of other participants, perhaps working in teams, and the kitty of forfeited money was split among those participants who qualified to participate in the kitty, often by meeting their own health behavior goals. Incentives in this group were evenly split between adhering and outcome categories. No study provided information on incentives per individual when there was potential variability, or in total for the study.

Incentive condition: Seventy-eight percent of incentives required a target behavior of the participant as a condition for the distribution of the incentive. The remainder required the participant to attain a particular outcome. Marcus et al. is unusual in that the free bus-passes were essentially bribes, distributed without any required behavior or outcome from the participant.⁶⁰ This may have caused some cognitive problems for some participants in that the gift may have created a sense of obligation. However, only 33 percent of the participants reported actually using the bus pass (see Evidence Table 2 in Appendix F).

Multiple components: Several studies included other intervention components that potentially confounded the impact of the incentive. Of interest is the inclusion of other social pressures. Two studies provided incentives not only for the participant, but also for the significant other who was supporting the participant's efforts to change behaviors.^{66, 67} Another three studies introduced team competition into a worksite health promotion program, where cash prizes were distributed to the teams who best met the health promotion goals.^68–70 The reports of these studies did not address the potential effects of these social pressures, whether the results would be positively or negatively affected by them.

Consistency of outcomes

Table 3. Impact of provider incentives

Table 3. Impact of provider incentives

Study Characteristics	Frequency of Interventions with Positive Effects	Frequency of Interventions with No Effects
Total studies	4	5
Performance bonus - target outcome	3	3
Bonus structure - Tournament	0	2
Bonus structure - All providers have potential to qualify	4	3
Bonus structure - Reward/punishment	0	1
Performance bonus - target behavior	0	3
Adjusted capitation rates	1	0
Raised fees	0	1
Subgroup analysis - Group practice vs. solo	3	0
Source of bonus - Paid by HMO	1	3
Source of bonus - Not paid by HMO	3	2
Prevention target - Immunizations	4	2
Prevention target - Other	1	3

Provider. As shown in Table 3, outcome patterns were mixed. Target outcome performance bonuses showed positive results in three studies,^{43, 44, 49} but no results in three studies.^{45, 46, 51} The three studies showing positive results measured immunization compliance by chart documentation. Provision of the immunization services may have come from sources other than the study provider. Within the Fairbrother studies, outcome measures that directly addressed provider behavior, per-input bonuses, and percent of missed opportunities to provide immunizations, did not show positive results.^{43, 44} Thus it appears in these cases that the financial incentives were motivating providers to change chart documentation behaviors as opposed to increasing preventive services. In addition, the physicians participating in the Kouides study were required to create the patient list for targeted immunizations.⁴⁹ This up-front involvement may have increased the awareness of the bonus for the physicians.

Morrow, Gooding, & Clark did show significant increases in the percent of charts in compliance with MMR immunization and cholesterol screening indicators.⁴⁷ However, the capitation rate adjustments were based on both utilization and improvements in preventive care practices. The analysis did not include controls such as possible rate adjustments due to utilization. The study period also covered a time period when the awareness of prevention and quality of care, in which prevention is included, was increasing in the general physician population. Thus, it is difficult to determine whether the financial incentive itself was the causative factor.

Practice settings: Several studies included practice-level variables in the analyses. Group practices appeared to have better results than solo practices. The Hillman and Ritchie studies found evidence that increases in preventive care were greater for group practices.^{45, 46, 51} There is uncertainty regarding the effect of remuneration packages for physicians within group practices. Studies did not clearly report who was actually paid the incentives. No data was collected on physicians within group practices who may have been paid salary and whether they participated in the incentives.

Other factors: No other provider or patient characteristics included in the studies were found to be significant. Many factors of interest were not addressed in any study. These would include: how the readiness stage of a provider to change behavior may have affected the salience of the incentive program, the effects of non-economic barriers such as attitudes of the physicians regarding the perceived accuracy of the data, the effects of physician work-flow decisions when only a portion of the patient base may be affected by an MCO incentive program, the effect of the providers' level of knowledge and understanding of the preventive care guidelines, and what was necessary to reach incented target levels.

We also cannot speak to the effect practice area variations would have on effectiveness of incentives in other regions of the country or non-physician health care personnel.

Not all studies reported effect sizes or provided enough information to construct relative risk ratios. Attempts to contact lead authors to obtain such information were met with limited success. Based on what was provided in the articles, the effect size is moderate at best. Economic incentives were responsible for a seven percent increase in documentation of immunizations based on regression analysis in the Kouides, Bennett, Lewis et al. study.⁴⁹

Table 4. Potential explanations by category

Table 4. Potential explanations by category

Category	Number of Studies
Study design - Insufficient power	2
Study design - Inadequate size and duration of bonus	5
Study design - Historical trends	4
Salience - Physician beliefs and attitudes about the bonus	3
Salience - Lack of physician awareness	2

Given the mixed outcomes for the provider incentives, it can be informative to look at the explanations offered by the investigators for their findings (Table 4). Both Kouides, Bennett, Lewis et al. and Hillman, Ripley, Goldfarb et al. noted low power for detecting differences in absolute rates.^{46, 49} However, Hillman also noted it was unlikely that the small effect size a higher powered study could provide would be unlikely to motivate MCOs to commit significant financial resources to target economic incentives.

Five studies specifically pointed to inadequate size and duration of the bonus. Bonuses were too small in either absolute size or relative to the physician's overall income.^{43–46 48}

Four studies cited national trends increasing preventive care in general.^{45–47, 51} Ritchie, Bisset, Russell et al. noted jawboning of national preventive care goals had already begun to raise awareness and practice levels before the target payments were implemented.⁵¹ Likewise, Hillman noted increases in both intervention and control groups mirrored secular national trends.^{45, 46}

Several studies cited physician beliefs, attitudes, and inferences through social cognition processes. Morrow, Gooding, & Clark reported the potential factors of: peer pressure from physician committee or normalizing force of knowing how other offices perform; non-financial motivation of pride in job well-done; and fear of retribution for poor performance.⁴⁷ Likewise, Fairbrother's studies noted that physicians perceived that the incentive goals were not attainable.^{43, 44} In particular, the physicians saw their patients as subject to lack of continuity of care and themselves as “sick doctors,” which detracts from their ability to provide the preventive care. Since they were not the primary source for immunizations, they were less likely to take responsibility for patient immunizations. Lastly, Hillman's studies, noted a definite lack of awareness of the bonus on the part of the physician.^{45, 46}

Consumer.

Table 5. Overall positive findings to total consumer (more...)

Table 5. Overall positive findings to total consumer studies

	Facilitating	Participating	Adhering	Outcome	Structural Barrier Removal	Attitude Barrier Removal	Reward	Punishment	Purchasing Behavior	Income Behavior
Simple	6 of 6 (100%)	4 of 6 (67%)	9 of 12 (75%)	0 studies	6 of 6 (100%)	0 studies	10 of 14 (71%)	2 of 2 (100%)	6 of 7 (86%)	13 of 17 (76%)
Complex	1 of 1 (100%)	4 of 5 (80%)	5 of 8 (63%) *	10 of 12 (83%)	2 of 3 (67%) *	1 of 2 (50%)	11 of 15 (73%)	5 of 6 (83%)	7 of 8 (88%)*	10 of 15 (67%)

*

One non-significant study mismatched test for effect of incentive. No direct measure of uptake.

Outcome studies - 8 of the 10 positive studies either did not check for long-term results or showed rebound to original levels and non-significant.

Only 1 of the Attitude Barrier studies actually measured change in attitude - negative finding.

Four of the five positive findings for punishment showed rebound to original levels.

Complex reward changes to 12 of 18 (67%) and 9 of 13 (69%) if remove studies which included punishment

Complex income behavior drops to 8 of 19 (42%) and 6 of 15 (40%) with long term results.

Categories: (Table 5) The facilitating category, which included removal of structural barriers, showed significant positive findings. This group of studies demonstrated perhaps the tightest links between incentives, desired outcomes, and direct tests of the uptake of the incentives. However, Marcus' studies stand out in that only a minority of patients self-reported using the bus passes or were logged as turning in vouchers.^{60, 61} It is possible the population involved had much deeper cognitive and affective barriers, such as denial in a frightening situation where cancer might be detected, and that the structural barrier incentives were not strong enough to overcome the personal barriers.

The studies on disincentives found effective interventions. Caution should be used in interpreting the findings of these studies, however. For the simple preventive concerns, the Birkhead, LeBaron, Parons et al. study took place immediately following a measles epidemic when public attention and concern was running high.⁵⁷ The findings of Kerpelman, Connell, & Gunn may be overestimations as parental permission was needed to review charts and some families may have left the AFDC program due to the policy of linking immunizations to benefits.⁵⁸

As rewards, economic incentives for participating and adhering categories, whether for simple or complex preventive concerns, are, in general, effective in prompting people to change their behavior in order to attain the reward. Most studies matched a short-term incentive with a short-term behavioral change or outcome.

When looking at the outcome and penalty categories for complex preventive concerns, however, one begins to see a change. While many of the studies in the outcome and negative reinforcement categories showed positive effects in the short run, of the four studies that checked for long-term results, all the significantly improved measures had returned to original levels and non-significant findings.^{66, 71–73}

In modifying behavior, economic incentives may have unintended consequences which offset the ultimate aim of the incentives. For example, Jeffery and French found cash incentives to participate in group exercise sessions did increase participation.⁷⁴ However, overall exercise effort, when compared to the control group, was not different. Barring the possibility of over-reporting of exercise due to social desirability, it would appear the participants exchanged the convenience of walking at their choice of location on their own schedules for the group activity in order to maximize the cash incentives.

There may be evidence of a differential in the outcomes of economic definitions categories, although the evidence is too thin to speak with confidence. A significant finding when the majority of people do not actually collect or use the incentive suggests the possibility the impact of the incentive was informational rather than economic. Evidence Table 2 in Appendix F shows four studies that fit this case. Perhaps the incentive generated feelings of social support or cohesiveness, that they and their health are important, improving self-esteem or self-worth. Or perhaps the existence of the incentive sharpened the participant's attention to the health behavior in question.

Alternatively, a finding of not significant may have more to do with an incentive that is not large enough or meaningful enough to the target population than it does with the overall idea that economic incentives may or may not be effective.

Populations: As there is a definite correlation between vulnerable populations, low SES, and simple preventive care, it is difficult to infer whether the effects of incentives are consistent across populations. It would seem reasonable, as discussed above, to suggest that much more information is needed before any such inference could be made with confidence, as the nature of the intervention, type, duration, and frequency, would be dependent upon the target population.

Settings: As seen in Evidence Table 5 in Appendix F, the settings for the studies are quite diverse. It appears that the setting is less a concern than the general question of whether it is appropriate, to some consumers' minds, to offer economic incentives at all within the context of health care. Moran, Nelson, Wofford et al. provided anecdotal evidence from a few participants that questioned providing a lottery to encourage people to get a flu shot.⁶³

Key Question 3. Is there evidence of a dose/response curve?

Provider. Given the paucity of evidence for the effectiveness of provider incentives, we can not address this question for provider incentives at this time.

Table 6. Consumer incentive type by significance

Table 6. Consumer incentive type by significance

Non-Significant Findings Studies	Significant Findings Studies
Lotteries:
$50/session + $2,000 travel package. For middle-class people to attend group exercise program.91 (complex)	$100/month; a 1 in 10 chance to win over 3 years. For low SES and middle-class people to return monthly nutrition newsletter postcards.74 (complex)
Microwave. For people who showed for community program cholesterol retest.73 (simple)	Three travel packages: one week in Hawaii, weekend in San Juan, weekend at local hotel. For self-help smoking members of GHCPS.72 (complex)*
Dinner for two. For car dealership employees attending off-hour smoking cessation class.92 (simple)	Five $100 cash prizes; 5 in 29 chance if all participants make goal. For health fair participants with high cholesterol.93 (complex) **
$100 cash. For low SES women returning for post-partum check.65 (simple)	Two $40 cash for meeting three-month goal; 1 in 4 chance at $1,000 competitive lottery. For employees of ambulance services.69 (complex)*
	$25, $50, and $100 cash prizes. For immunizing parents of children/patients at public clinic.62 (simple)
	Three $50 grocery gift certificates. For low SES urban community health center.63 (simple)
Gifts:
Nutritional information package. For community women attending mammography.94 (simple)	Nutritional information package. For community women attending mammography.95 (simple)
Jewelry. For low SES women attending post-partum check.55 (simple)	Ceramic coffee mug (also lottery). For smoking members of GHCPS in self-help program.72 (complex)*
Baby blanket. For low SES women attending prenatal checks.77 (simple)	Turkey buffet and pooled kitty to winning worksites for competitive smoking cessation program. For employees of state government department.71 (complex)**
	Gerry Cuddler. For low SES women attending post-partum check.56 (simple)
Cash:
$50 payment. For middle-class worksite employees for cigarette abstinence.96 (complex)	$5 cash payment. For active drug users to return for TB skin test reading.53 (simple)
Potential of up to $266 over 18 months; $1 to $3 per exercise session. For middle-class people to attend group exercise program.97 (complex)	$10 cash payment. For active drug users to return for TB skin test reading.53 (simple)
$12.50 to $25 per week for 20 weeks. For middle-class people to lose weight.98 (complex)	$10 cash payment. For active drug users to return for TB skin test reading.54 (simple)
	Two $15 payments. For low SES people to attend STD prevention education.64 (simple)
	Three cash payments totaling $35. For active drug users to attend AIDS prevention education.52 (simple)
	$5 to $15 per participant for competitive team members. For employees of diverse firms for smoking cessation.70 (complex)*
	$1 to $10 per day for six weeks. For mostly hospital workers for decreased smoking.75 (complex)**
	$50 for participation, $15/month for abstinence, $1,980 competitive kitty. For employees of aerospace firms for smoking cessation.68 (complex)*
	$50/month for smoking abstinence, maximum ten months. For low SES pregnant women.67 (complex)***
Coupons:
Fitness center passes. For lower SES people for attending motivational counseling for exercise program.99 (complex)	Free postage. For lower SES people to return completed FOBT.59 (simple)
Bus passes. For lower SES women returning for abnormal pap.100 (simple)	Bus passes. For lower SES women returning for abnormal pap.60 (simple)
$5 department store gift certificate. For lower SES women returning for post-partum check.65 (simple)	75% off coupons for box of condoms. For young adults for STD prevention.101 (simple)
	Coupon for infant formula. For low SES women to attend post-partum check.55 (simple)
	Free taxicab fare. For low SES women to attend prenatal visit.77 (simple)
	$10 of free bus passes or fast food coupons. For active drug users to return for TB skin test reading.54 (simple)
	$10 grocery gift certificate. For active drug users to return for TB skin test reading.54 (simple)
	$20 in coupons for farmer's market fresh produce. For low SES women to improve nutrition.102 (complex)**
	Two $15 coupons for goods or services. For low SES people to attend STD prevention education.64 (simple)
	Free evening day care. For immunizing parents of children/patients at public clinic.62 (simple)
	Three grocery gift certificates totaling $35. For active drug users to attend AIDS prevention education.52 (simple)
	Multiple kinds, of high frequency and moderate to high value. For low SES women to breast feed.66 (complex)***
Free/Reduced Medical:
Free nicotine patches - did not test for uptake, only effect on cessation.103 (complex)	Free or reduced price of $6 or $20 (vs. $24) per pack of nicotine gum. For smoking rural family practice patients.76 (complex)
Voucher for reduced clinic visit. For low SES women for abnormal pap.100 (simple)	Voucher for reduced clinic visit. For low SES women for abnormal pap.61 (simple)
	Free flu shots. New Zealand104 (simple)
	Free flu shots. Denmark78 (simple)
Punishment
Return contracted amount, minimum $5 per paycheck. For worksite employees for smoking abstinence.105 (complex)	Return contracted amount, minimum $5 per paycheck. For worksite employees for meeting weight loss goals.105 (complex)**
	Return $20 per session for ten sessions. For overweight people to attend meet weight loss goals.106 (complex)**
	Return $20 per session for ten sessions. For overweight people to attend weight loss sessions.106 (complex)**
	Return $30 per every five pound reduction, up to $150. For overweight people to meet weight loss goals.107 (complex)**
	Return $1, $5, or $10 per session for 30 sessions. For overweight people to meet weight loss goals.108 (complex)**
	Return $5 per session, 14 sessions; forfeit money split. For mostly female employees of a hospital to attend weight loss sessions.109 (complex)
	Return to WIC offices monthly (not bimonthly) for vouchers. For low SES parents of non-immunized children.57 (simple)
	Return to WIC offices monthly (not quarterly) for vouchers. For low SES parents of non-immunized children.110 (simple)
	Lose AFDC benefits. For low SES parents of non-immunized children.58 (simple)

*

Studies showed rebound of outcomes back to pre-test levels

**

Studies did not follow up participants for long-term effects

***

Incentives also paid to participant's chosen Significant Other

Cash incentives have the expected rank ordering. Malotte, Rhodes, Mais showed a $10 cash incentive was more effective than a $5 incentive.⁵³ Stitzer and Bigelow also showed a rank ordering of $1, $5, and $10 per day incentives for reducing daily cigarette smoking.⁷⁵ Hughes, Wadland, Fenwick et al. showed price elasticity was higher for $6 vs. $20 than $0 vs. $6, suggesting larger incentive increases had a stronger impact on the participants to reduce daily cigarette intake.⁷⁶ Malotte suggested the street value of a coupon is half-price off face value, perhaps suggesting another interpretation of the finding that cash is preferred to coupons with an equivalent face value.⁵³

Although more open to interpretation, we might also suggest that coupons are preferred to gifts. Both studies that pitted a coupon incentive to a gift incentive found the coupon the more effective incentive.^{55, 77} In addition, while coupon incentives were in general effective, with 12 of 15 studies showing positive results, only four of seven gift incentives had positive results, and two of the positive results were potentially confounded by other intervention components comprised of a lottery or competition.

Key Question 4. What is the evidence for cost-effectiveness of economic incentive interventions?

Those limited number of studies that did address cost-effectiveness tended to do so from the perspective of productive efficiency, i.e., examining the cost of achieving the determined preventive goal.

Provider. Among the provider studies, only one study addressed cost-effectiveness and calculated a marginal cost of $3 per additional immunization.⁴⁹ The study did not provide any other figures against which to weigh this cost.

All of the calculations were relatively crude. No study included an attempt to generalize cost-effectiveness over time for the estimated potential population morbidity or mortality that would be affected by a policy decision to implement the economic incentive.

Other Perspectives

In this section we look at the limited pool of observational studies that have attempted to directly address whether preventive services are affected by economic incentives using econometric methods to analyze cross-sectional databases. We also look at what the Medicare/Medicaid demonstrations have to offer.

Provider

There is little evidence available to support the idea that explicit provider financial incentives, particularly of the perhaps modest and artificial nature such as were evaluated in the studies, are effective. The literature is scarce. Further, from the studies that have been performed, it appears bonuses don't work as simply and easily as some may have assumed they would. It would seem we have been functioning under core beliefs regarding the appropriateness and efficacy of financial incentives that have only recently begun to be subjected to examination through either experimentation or well-designed quasi-experimental or observational studies.

While there was some evidence that increases for preventive care were greater for group practices than solo practices, there is not enough evidence to sort out the causes. The improvements could signal increased staff and office system resources available to group practices. As the evidence isn't clear regarding whether the incentives were paid directly to the physician or to the group, the question remains open. Note Debrock and Arnould found that financial incentives were more effective if directed at individual physicians.¹¹¹

Even in its sparseness, some lessons can be gleaned from the literature. Perhaps most obviously, incentives need to be easily understood; simple, transparent, and have a clear connection between the incentive and the desired behavior. The complexity of the clinical practice and differing reimbursement systems should not be minimized. As MCOs are not homogenous, market penetration of MCOs in a local market is a significant factor in provider behavior.⁸³ Physicians may use a heuristic approach and manage their practices as an amalgam of the various guidelines and procedures required by plans and other reimbursement sources along with their own professional opinion.

Another lesson is that the effects of incentives need to be understood within the larger context of the process of preventive care and health promotion. How far down the stream of preventive care did the effects go? An intervention can change a facilitating preventive care behavior, yet not reach the outcome that is of most importance. Findings of a lack of office procedures to follow up on abnormal findings from preventive screening is an illustration of the importance of this issue.⁹⁰ Grumbach, Osmond, Vranizan et al. found that over one-third of physicians who reported facing incentives could not specify the amount of their overall income that was involved in bonus payments.¹¹² In addition, attention should be given to the role of other reinforcers such as profiling and feedback.

According to the National Health Care Purchasing Institute monographs,^{113, 114} the key factors for a successful implementation of explicit target economic incentives, as determined in an interview with a focus group comprised of physicians and plan administrators, included:

• The size of the financial incentive

• Peer and/or consumer knowledge of individual provider performance regarding the incentive target

• Recognition among the physicians of a need for change

• Support for the incentive program among medical leadership

• The practicing physician's knowledge and understanding of the performance incentives/sanctions

• Simplicity and directness of the incentive program

• Perceived and actual accuracy of the data on which the incentives are based. There is a lack of physician trust in data and data sources.

As a warning, it must be recognized that incentive models perceived as penalizing providers for patient behavior may inadvertently create incentives for providers to drop non-adherent patients.

Consumer

We may guardedly say that in the short run, for simple preventive care and distinct behavioral goals that are well defined, economic incentives are effective. There isn't sufficient evidence to say that economic incentives are effective for promoting the long-term lifestyle changes required for health promotion.

Funding bias may be contributing to our lack of information on the long-term effects of economic incentives. Research proposals to examine whether there is extinction of behaviors after the removal of economic incentives are not as exciting as testing a new intervention.

Perhaps most intriguing was the psychological impact of incentives. The Melnikow et al. study showed a positive effect for free taxi rides when only one of 34 vouchers was redeemed.⁷⁷ Marcus et al. showed similar findings with the incentive demonstrating a positive effect, yet only 33 percent of the patients used the bus passes provided, and 41.7 percent of those eligible redeemed the reduced clinic fee vouchers.^{60, 61} It has been noted that participants tend to underestimate the role of external incentives on their behavior.¹¹⁵ This may partly underlie reports that only a handful of participants were swayed by the lotteries or gifts.^{65, 92, 93}

There are also alternative views to the cognitive behavioral models in the studies which may help explain the effect of explicit economic incentives on consumer behavior. These views focus on expectations of future reinforcements or rewards as motivators and people's attributions about why they engaged in certain behaviors in the past. Deci and Ryan view the study of motivation as an “exploration of the energization and direction of behavior” and consider intrinsic motivation to be at least as important a factor in explaining human behavior as drives and external or environmental controls.¹¹⁶ They argue extrinsic motivators, such as economic incentives, actually inhibit intrinsic motivation, the spontaneous, internal experiences that accompany behavior. A person who is intrinsically motivated enjoys the rewards that are inherent in the activity and perform the activity for its own sake. This has the unintended consequence of reducing the likelihood of the desired behavior in the long run.

Curry, Wagner, & Grothaus attempted to address this question directly by testing a form of intrinsic motivation, represented by personalized feedback designed to improve self-efficacy and enhance a direct cognitive link between behavior and outcome, against an extrinsic motivation of gifts and lotteries for smoking cessation.⁷² While the extrinsic incentive improved participation in the smoking cessation program, abstinence rates post-program were not significantly different from the control group. The intrinsic motivation group, however, showed abstinence rates double those of the control group at followup.

Inhibition of intrinsic motivation, if this view is correct, is particularly important to the complex preventive health concerns found in the lifestyle changes for health promotion. Any economic incentive that is a reward must be viewed by the consumer as temporary support toward a personal goal. To do otherwise would not only impose an economic cost on an already strained system, and a cost that would be difficult to justify from a purely cost-effectiveness stance, it also creates a psychic environment which does not promote the personal responsibility and autonomy of the consumer for his own health.

On the other hand, if extrinsic motivations do not dampen intrinsic motivations, we are unable to infer from the given literature the rate of decline of preventive behavior back to baseline levels or the length of the time gap between the extinction of the incented preventive care behavior and the increase of the intrinsic motivation to whatever is the threshold level that would sustain the preventive or health promoting behaviors.

Special Populations Concerns

Pediatric population. The nature of incentives for children's preventive services differs from that for adults.¹¹⁷ Children's healthcare needs differ from adults. First, children are in a dependent relationship with adults who may not make choices based on the best interest of the child. Incentives need to “actively encourage such access.” Second, children's preventive services take place within a very dynamic context. Children are constantly developing and changing, and it is important to provide regular care and avoid pigeonholing into a specific disease category. Third, intertemporal relationships are even more relevant. A provider of children's preventive care will not reap benefits when a child reaches adulthood. Another provider will.

Vulnerable populations. It should be noted that the use of economic incentives, and the research regarding the incentives, is not without controversy. The ethical problem seen in public programs that threatened withholding public benefits as incentive to induce parents to immunize their children is the fact that the financial penalties were threatened only to people already at high risk for economic deprivation.¹¹⁸ Even simply requiring people to return more frequently to government offices to collect public benefits imposes additional costs. Further, the evaluation research of these programs is often conducted without the oversight of a Human Research Protection Review Board. The ethical imperative in research efforts is to decrease, not potentially increase, the overall risk borne by individuals already at higher risk for deprivation and poor health.¹¹⁸

It is also not surprising that studies that directly tested the desirability of different incentive were focused on high-risk and vulnerable populations, where public health policy and ethical considerations of economic incentives are at the fore. In fact, the opportunity for a natural experiment was created out of the policy decision to change from cash to non-cash incentives in AIDS/HIV prevention outreach programs, due to argument over the ethics of cash incentives to active drug users.⁵²

There is also concern that incentives may further fracture care for vulnerable populations. LeBaron noted in interview followups that low SES mothers did not wish to take their children to mobile clinics for free immunizations because they preferred to see their regular providers for shots.¹¹⁹ Ironically, the providers in the Fairbrother studies, who served a similar vulnerable population, didn't think of themselves as well-care providers, but rather sick-child providers.^{43, 44} This is a considerable disconnect between the stated needs of the parent and the perceptions of the providers.

Comparison of Provider and Consumer Incentive Research

Research shows a tendency to favor different rubrics for providers and consumers. Consumer research shows much more attention to psychological and behavioral theories. Provider research approaches research from an economics/business perspective. This is perhaps to be expected, given the roles consumers or providers play, and unfortunate in that both roles are played by humans who share in common many social, psychological, and philosophical qualities.

Both provider and consumer incentives evidenced “gaming the system,” that is, adjusting behavior to maximize income production without necessarily increasing the desired behavior. In physicians there was an adjustment in documentation and reimbursement behaviors.^{43, 44} In turn, consumers appeared to have adjusted where, rather than how much, they exercised.⁹⁷ Further, Breen, Feuer, Depuy et al. found little increase in reported mammography rates but an increase in reporting using public payment sources for the mammograms after Medicare extended its mammography benefit to reimburse for breast cancer screening mammograms.¹²⁰

It may seem reasonable to speculate that target incentives for preventive care might be generally welcomed by both providers and consumers as a structurally reinforced motivation for appropriate and quality care. There is evidence from the physician perspective that bonuses based on quality of care, including preventive measures, may promote job satisfaction among physicians.¹¹² Consumers may view bonuses for increasing a service, signaling the health of the consumer comes first, as inherently more trustworthy than financial incentives linked to physician resource use. Incentives of this kind lead to the Supreme Court case regarding the patient's right to sue HMOs for such cost containment efforts.^{29, 121}

In summary, the findings that “artificial” explicit economic incentives do work, but modestly and in the short term, fit well with the growing call for multi-component system changes are often needed to for prevention and health promotion.^{36, 122}

Overall

As this review highlights, the current evidence is extremely limited in its ability to inform future health care efforts. The limited success of modest and “artificial” incentives to induce long-term change supports the current push for multi-component interventions based on the full environmental or social ecological perspectives such as the McKinley model. There is a need for further studies.

The current work on economic incentives is scattered across topics and approaches. Other work suggests that simple preventive approaches are more cost-effective than complex ones, and economic incentives seem to work best in simple cases. On the other hand, such incentives have not been compared with other strategies, especially in terms of cost-effectiveness. For providers, these alternative approaches usually involve some sort of infrastructure change, such as creating registries or other information systems. Because the goal of such interventions is to create sustained, rather than simply transient, change, perhaps the most desirable research should employ economic incentives to encourage providers to adopt these new systems; but the research should be continued (if the initial results are promising) long enough to assess whether the innovations persist after the incentives are removed.

Future researchers need to be clear about the causal chain of prevention or health promotion being investigated and the purpose of the incentive intervention being considered. More careful definitions of the process of care for a given preventive concern, careful matching of the nature of the economic incentive in terms of type, size, duration, frequency, and the use of other components such as education, social support or competition, are all needed. The black box approach, which leaves the economic and behavioral assumptions underlying the research project unexamined, should be replaced with more well-developed conceptual models.

The large literatures in the social and behavioral sciences on incentives should be brought to bear. Most of the studies reviewed here have not evidenced use of this literature. The research reviewed here generally takes an atheoretical approach to studying incentives. However without a theoretical underpinning it is difficult to understand exactly why the incentive worked or didn't work. Perhaps, more importantly, without a theoretical underpinning the right questions do not get asked.

The first question a study should ask is “What problem is the incentive trying to solve and how is the incentive going to solve this problem?” It is not inherently interesting to ask: “Can I change behavior using incentives?” The answer is certainly “yes” if the incentive is great enough. Rather, a better question is, “Given that a preventive service is underprovided, what is the best incentive mechanism to bring the provision of that service up to its optimal level?” A second important question is then “Given this new incentive is in place, are there other, unintended consequences of this incentive?” Of course, anticipating unintended consequences requires a theoretical framework.

On the surface, it would seem wise to approach economic incentive intervention as mimicking a dose-response curve. That is, start with a dose that should be “therapeutic” and test for the possibility of lower doses. However, as the review highlights with the differential effects of modest incentives for providers and consumers, different levels of economic incentives may trigger different modes of decision-making processes. What are the trigger points which determine whether psychological models or economic models of decision-making are used? When does a person view an economic incentive as input into a cost/benefit equation, a reward as goal object, a removal of a barrier? While not mutually exclusive, each model suggests different sets of hypotheses.

Another important question is how economic incentives compare to and complement other strategies to improve preventive care, particularly with regard to long-term effects. There may be joint effects of intervention components, including incentives, within multi-component research. Do explicit incentives improve, impede, or are they unnecessary when a larger ecological effort is made, especially for consumers?

Attention should be given to measuring the direct impact of the incentive itself. Health outcomes may be too high of a measurement standard for incentives that are designed to address only one behavior pattern within a continuum of preventive care. Many other factors, both controllable and uncontrollable, may come to bear on the achievement of a health outcome even when the consumer or provider adjusted behavior in the manner exactly desired for the incentive.

Natural settings for social science research are important. Randomized controlled trials lose the primacy of gold standard with behavioral change research because many factors remain uncontrollable in the normal busyness of every day life. Adaptation to the environment by the provider or consumer may be either good or bad, but once a person is adapted, it is difficult to get him/her to move. Randomized controlled trials should come only after we can answer why the intervention is needed from this perspective. The difference between controlled trials and “real world” efficacy was demonstrated in the series of studies on the effectiveness of incentives and intensive followup for women with abnormal pap smears. Even though both controlled trials showed positive effects of the economic incentives,^{60, 61} a planned evaluation of the same interventions using a quasi-experimental design¹⁰⁰ was unable to replicate the positive effect. The potential cost-effectiveness of incentives would be erased if positive results of an incentive are fragile enough they only survive in controlled settings.

Mixed-method research projects would also seem an appropriate approach for understanding the effects of incentives within a health or health care setting. Qualitative components, particularly conducted retrospectively with a purposive sampling of participants that captures the range of outcome responses, would do much to advance our understanding and improve future research design. We know very little of the meaning and value of the incentives to the populations they are intended for or the attitudes and beliefs they hold. For example, Moran provided anecdotal evidence that some patients found a lottery inappropriate within health care.⁶³ The bias this belief may create on the findings could be mixed; a feeling of indignation could as easily prompt a person to not receive a flu shot from the clinic in protest as receive the shot while not participating in the lottery.

Personality research and other ways to understand who the people are may provide a step toward understanding what the “problem” is and how to address it. A phenomenological orientation, that it is the world of the perceiver that determines what he will do and not the physical environment, except as the physical environment comes to be represented in the mind of the behaving individual, also requires an understanding of how people differ from one another in how the events come to be represented in the mind. Certainly Prochaska & DiClementi's stages of change meta-theory provides one ready-made conceptual framework with potential usefulness that has not been fully explored.¹⁴⁰

Another question that remains unresolved is “What is the right metric for determining if preventive care is truly under-provided?” Possible perspectives are cost-effectiveness, effectiveness, consumer welfare, HMO welfare, or the opportunity cost of other types of care.

Providers. Physicians work in organizations and organizations matter. The dynamics of the organization affect the rules under which physicians work and they affect their financial incentives. Economic incentives do not live in a vacuum. They are often coupled with other incentives. Physicians can be paid a salary, which is viewed as a low-powered incentive, but face the prospect of being fired if they under perform. Physicians face many competing demands on their time and are under the influence of many different incentives (economic and otherwise). It may be difficult for outside actors (government, health plans) to affect the direct incentives of the physician.

Measurement also matters. You can base incentives only on things that you can measure. That means if incentives are paid on the basis of measurable performance there is the potential for slackening of effort in other unmeasurable but potentially important domains of care. The unintended consequences of an economic incentive may wipe out what appeared to be a cost-effective practice.

Paying incentives on health outcome measures for providers becomes a default choice when we cannot measure the process. We are often not able to determine with any sense of confidence what a unit of preventive care services is, such as providing health promotion counseling during both routine checks and provider visits for acute care. Furthermore, success in prevention is generally a non-event. It is much easier to count something that does occur than estimate the number of events that might have occurred but did not.

We need to tease out the differential effects of two major components of economic incentives for physicians: motivation and information content. Physicians often claim they are in a permanent state of stress. Given this stress, they can become resistant to change and incentives can shake up the patterns. When something changes, an incentive can redirect the physician's attention to the new “agenda concern,” often switching one behavior for another in the process. Using an incentive as an “attentional” device is most helpful when we understand what the initial barrier to change is: no desire to change; not enough knowledge; not enough time; overcoming heuristics that are a hindrance.

Future researchers should keep in mind scaleability matters. For an incentive system to be put to use widely, it has to be scaleable. There has to be a clean relationship between the size of the incentives and the behavior that is being encouraged. It must be large enough to make the task seem worthwhile. We know little about how large such incentives need to be.

Consumers. Questions such as who the people are and what motivates their actions are of great importance when designing an economic incentive. Personality research, with its focus on individual differences, offers a potentially useful approach to these questions. For example, some participants in a team competition for cash prizes reported disliking the inherent peer pressure.⁶⁸ Lotteries engendered feelings of unfairness and created tension among participants in another study.⁹¹ Competition and the openness to experiencing gambling uncertainty have differential effects for different personality types.¹¹⁶

Cultural difference is another potential subgroup categorization scheme. Smith et al. found African Americans and Hispanics responded differently to the different economic incentives of infant formula versus a gift of jewelry. They suggested there may be cultural differences in the meaning of gifts.⁵⁵

A major concern with economic incentives is the potential for the coercive effect of incentives on patient autonomy. No study has actually investigated this concern. As we saw, several studies cited interview or anecdotal evidence that the incentives themselves were not a factor in their decision to seek care. This may point to less concern regarding coercive effects for the likely value level of consumer incentives. On the other hand, the finding of positive effects for an incentive intervention, even when the incentive was not redeemed in several studies, could be interpreted as a psychological vulnerability to manipulation of social obligation through the social desirability response. In addition, Malotte et al. suggested paying people for preventive care behaviors may create an expectation for future payments that may result in even lower participation rates should the expectation be thwarted.⁵⁴ Smith suggested many of the WIC population may be becoming desensitized to “gifts” from public programs competing for the attention of the beneficiary.⁵⁵

The benefits from competitions, or tournaments, as an aspect of worksite economic incentive programs may be open to debate as well. Deci and Ryan state there are gender differences in responses to competition, and that competition against another person is less effective for women.¹¹⁶ Of the studies that included tournament-style competitions, 53 percent to 85 percent of the participants were male.

A potentially fruitful area is the possibility for cross-fertilization between health promotion activities and the extensive research conducted around disease management programs. Patient self-care for disease management requires sustained behaviors, also often without apparent immediate “gain” or improvement, and the similarities and differences in patient motivation and patient-provider relationships may be very informative.¹⁴¹

Doctor/Patient Interactions. The patient-provider relationship itself is also important. There is extensive literature in this area that would inform future research on the effects on incentives and how they might impact preventive services and the consumer's acceptance of the provider as a collaborator in health promotion activities. In turn, there may be the potential for differences in the provider's behavior when the provider is aware of lower costs to the patients. Future research should also investigate possible joint effects of coordinated incentives simultaneously applied to both providers and consumers for a particular preventive concern.

Conclusion

If we accept the value of preventive efforts, we must recognize the inadequacies of existing systems to encourage such practices. The literature reviewed here suggests that financial incentives have been used, in an uncoordinated fashion, at three levels in an attempt to increase prevention behaviors: 1) as motivators in the larger economic context at the health plan level, where savings associated with prevention is believed to be efficient, or where market interventions have instituted preventive care performance measures as quality indicators; 2) as provider incentives to induce discrete behavioral changes; and 3) as consumer incentives to remove barriers, improve health education, and reward healthy behavior. System-level economic incentives can help to change the larger health care environment, in turn prompting the individual providers and consumers to adapt to a new environment. Financial incentives, if they are big enough, can influence discrete behavior at the individual level in the short run. The benefits of such incentives may be magnified if they are coordinated with each other and with system level incentives, although this potential synergy remains untested. Whereas provider incentives do work, they may not provide a sustained behavior change. There is always a danger that they will be displaced by a new set targeted at a new topic. So questions remain regarding whether investing in office system changes, including information technology, which support long-term changes in practice, is a better choice than relying on incentives. More importantly, since various observers have noted that the business case for quality improvement is still weak, we must ask who is prepared to bear the cost of either strategy.