Download PDF

Participation in preventive care programs: individual determinants, social interactions and program design.

Publication date: 2014-10


Bouckaert, Nicolas


This doctoral research focuses on existing medical preventive care programs. Because of externalities (e.g. in the prevention of communicable diseases) or the program cost-benefit ratio, preventive care programs require high participation rates. In the United States, the Centers for Disease Control and Prevention have set clear participation objectives – next to quality targets – which are measured and evaluated over time (National Center for Health Statistics, 2012). For example, the 2010 participation target for cervical cancer screening in the past 3 years for women aged 18 or more is set at 90%; the target for breast cancer screening in the past 2 years for women aged 40 or more is set at 70% and the target for influenza vaccination is set at 90% of the over-65. Similarly, European countries define participation targets for their large scale prevention programs. First of all, we observe that participation rates rarely meet – let alone exceed – the desired targets, even for prevention measures that are generally considered to be cost-effective such as influenza vaccination for the elderly (Nichol, 2003). Secondly, participation in prevention is not equal among all targeted individuals and might give rise to health disparities by socioeconomic status, by education level, by geographic location, by sex or by other dimensions. The individual's decision to participate (or not) in preventive care programs is the main thread throughout this dissertation.The individual participation decision has been studied in the literature, both theoretically and empirically. In economic models, private incentives, i.e. broadly defined costs and benefits, are generally considered the main drivers of the preventive care participation decision. Participation benefits include reduced mortality and morbidity. Participation costs include financial, non-financial, social and psychological costs. Empirical applications relate the participation decision to individual characteristics and beliefs. Private incentives are discussed mainly in the first chapter. However, a broader understanding of non-participation behavior should equally include social interaction effects and policy design. The former relates to direct non-market interactions between individuals. The argument is that the individual's participation decision relates to the participation decision of others in the individual's peer group. The fact that other people participate or do not participate can have a social spillover effect. Kremer & Miguel (QJE, 2007) state that social effects can result from imitation behavior, social learning about e.g. how to use certain medication efficiently, social learning about the benefits and costs of preventive programs and epidemiological externalities. Depending on the dominating driving force, peer effects can be positive or negative. The policy design of a preventive care program is a second element that might affect the participation decision. Differences between programs may arise from eligibility rules, information campaigns, invitation procedures, the technology that is used, the level of organization, delays in enrollment or delays in learning about the screening results… By deciding upon the design of the prevention program, policy makers endogenously influence take-up rates, nudging people's behavior into a certain direction. Most policy reforms operate on an individual's private incentives by altering prices and (non-) monetary costs and benefits. In addition, this might indirectly generate social interactions effects. Understanding how these social interactions influence individual behavior is important for policymaking since they could reinforce or offset the policy effects on the individual's private incentives to take-up preventive care. Social interaction effects might therefore lead to higher or lower participation rates than otherwise expected and a social program might reach non-targeted individuals and households through social spillovers. On the other hand, policy interventions, such as mass-media campaigns, can also aim directly at changing social norms and social interaction transmission mechanisms. It is clear that private incentives, social interaction effects and policy design are interrelated and difficult to disentangle empirically. A carefully designed empirical strategy is necessary. This identification challenge is taken up in chapters 2 and 3. Chapter 1. Differing types of medical prevention appeal to different individualsIn chapter 1, participation in medical prevention is analyzed with an expected utility model that is sufficiently rich to capture diverging features of different prevention procedures and disorders. We distinguish primary and secondary prevention for both fatal or non-fatal diseases. Moreover, we introduce a flexible relationship between the specific disease for which the prevention procedure is set up and the general background health of the individual.We derive four main hypotheses from the theoretical model. First, current health is positively related to participation in prevention for fatal diseases (e.g. cancer) and negatively for diseases in which good current health mitigates the effects of the disease (e.g. the flu). Second, mortality risk, future costs and benefits only matter for fatal diseases. Third, decreases in program complexity and prevention costs positively correlate with participation for all disease and prevention types. Fourth, increases in current income positively affect participation for fatal diseases, but the income effect can be either positive or negative for non-fatal diseases. These hypotheses are analyzed empirically using European wide data of the Survey of Health, Ageing and Retirement in Europe (SHARE). We look at six types of preventive care (mammography, dental caries screening, influenza vaccination, blood pressure screening, blood sugar screening, cholesterol screening). The observed correlations provide support for the theoretical predictions.Chapter 2. Neighborhood peer effects in the use of preventive careIndividual participation in preventive care may depend on preventive health behavior in an individual's peer group. Chapter 2 analyzes the effects of policy changes and social interactions in preventive care participation in the context of new social policies in Mexico that aim at encouraging preventive care. The program is targeted at the extreme poor in rural areas and is designed as a conditional cash transfer program, meaning that families receive cash transfers conditional on the household engaging in a set of behaviors. Program requirements include, amongst others, participation in various types of preventive care, such as child growth monitoring, child immunization, blood pressure tests, usage of deworming drugs and cervical cancer screening. We follow the promising approach of analyzing social interactions in real world peer groups. Identification of social interactions is based on a partial-population design.Results indicate that the social program succeeded in increasing preventive care usage among program eligible households. In addition, endogenous social interactions increase preventive care usage for various types of prevention. The magnitude of the effects differs across prevention types. Effects are especially pronounced for annual child growth and weight monitoring. The overall treatment effect on prevention can be decomposed in a direct effect related to financial incentives and an indirect effect related to social interactions. The indirect effect accounts for 10% up to 60% of the total treatment effect. Chapter 3. Unintended spillover effects of influenza vaccination: a regression discontinuity approach.In chapter 3, we investigate direct and spillover effects of an extension of the target group for the Dutch influenza vaccination program to all Dutchmen aged 65 years and over in 1996. Members of the target group qualify for free influenza vaccination and receive a personal invitation letter from their GP. Using a rich dataset that combines survey data on health with administrative records from Statistics Netherlands, we exploit the quasi-random variation that was introduced at age 65 by the reform to analyze vaccination behavior and its impacts on the arguably even more important outcomes of morbidity, medical care use, sickness absence and mortality. While the effects on the targeted population are useful to evaluate direct policy effect, our primary aim is to estimate policy induced spillovers onto non-targeted individuals, in our setting the adult children of targeted individuals.Our results indicate a positive direct policy effect on vaccination coverage of the parents (an increase in vaccination rates from about 30% to 50%), accompanied by a negative spillover effect from parents to children (a decrease in vaccination rates from about 9% to 5%). In addition, we estimate that the influenza vaccination program saves 0.8 individuals out of 100,000 at the age threshold, and reduces the number of individuals consulting a GP and using prescribed medicines with 10 percentage points during the typical influenza months. Mortality and GP visits of the adult children are not affected, but the occurrence of influenza-like symptoms increases from 45% to 55% and sickness absence among this group increases from 14% to 22%. We explore several possible channels that might generate the negative spillover effects and find suggestive evidence that a social stigma costs is revealed to children – who are not targeted by the vaccination program – when their oldest parent crosses the age threshold. A potential trigger for the social stigma cost is the explicit framing of the target group in the invitation letter sent out to eligible parents. Our results also underline the importance of public health campaigns to pay attention to the effects of information dissemination on public perceptions and attitudes on (voluntary) preventive care participation.