Health Insurance Is Associated With Preventive Care but Not Personal Health Behaviors

Discussion

As hypothesized, we found that individual change in health insurance status was associated with changes in preventive care but not with changes in health behaviors. Parameter estimates of the associations were similar in unadjusted and adjusted analyses, the latter including some possibly endogenous variables, underscoring the robustness of our findings.

Gaining (losing) health insurance from the baseline to follow-up year was associated with increased (decreased) adherence to influenza vaccination, CRC screening, Papanicolaou testing, mammography, and PSA testing. These associations, consistent with the HBM prediction that reducing barriers (in this case financial impediments to accessing care) should increase care utilization, likely simply reflect price–demand relationships, albeit modified by factors not addressed by health insurance (eg, lack of awareness of services, opportunity costs, copayments). Economists assert that such associations represent ex post moral hazard, given that they entail increased financial risk to insurers.^4,40 However, the findings also are consistent with an intended purpose of health insurance: to facilitate receipt of preventive care.⁴¹ Indeed, this purpose of health insurance sets it apart from other types of insurance (eg, fire, automobile), which are intended to lessen the financial burden of a catastrophic event after it happens. In this context, the connotation of the term moral hazard (implying something undesirable) is unfortunate. Further ambiguity in the moral hazard concept arises when one considers that ex ante moral hazard theory suggests the potential for decreased preventive care following insurance gain due to reduced personal costs of nonadherence to preventive care, possibly offsetting any ex post moral hazard effects.

One might have anticipated a priori larger changes in preventive care with more frequent receipt intervals than for preventive care with longer intervals. However, our findings suggest a more complex picture. For example, insurance gain was associated with a smaller increase in influenza vaccination, which is recommended annually, than in cancer screening tests, including CRC screening, which is generally performed in the United States using colonoscopy at 5- to 10-year intervals. Although our study was not designed to examine the reasons for these findings, we hypothesize that use of the relatively low-cost, ubiquitous influenza vaccine may be less affected by health insurance change than use of relatively more expensive and more difficult to access cancer screening tests. These findings may be useful in forecasting changes in various aspects of preventive care due to evolving health insurance expansion.

In contrast to our findings related to preventive care, we found that individual health insurance change was not associated with significant changes in health behaviors (becoming nonobese, quitting smoking, and adopting consistent seatbelt use). Stated in economic terms, we found no evidence of ex ante moral hazard resulting from insurance gain. Previous studies examining such behaviors, again involving restricted samples—mostly those eligible for Medicare—and considering only insurance gain, had inconsistent findings.^{5⇓⇓⇓⇓⇓–11} In this context, our findings, stemming from large and nationally representative samples, question the notion of ex ante moral hazard as an outcome of gaining health insurance, at least as it pertains to the health behaviors we examined. Our study cannot determine why we found no evidence of ex ante moral hazard, but, in the context of health behaviors, such hazard presumably has little influence.

Other explanations, also grounded in the HBM, may further help to account for the contrasting findings with regard to preventive care versus health behaviors. As noted previously, insurance gain (loss) increases (decreases) access to clinicians, whose actions are likely to be influenced by perceptions of accountability for various aspects of care. It seems plausible that clinicians may tend to feel more accountable for preventive care than for patient health behaviors. Indeed, empirical evidence suggests clinicians counsel more about preventive care than about health behaviors.⁴² Furthermore, while the health behaviors we studied can be favorably influenced by access to health care, in general they seem to be less influenced by health care access than preventive care and relatively more influenced by other barriers, such as the requirement of sustained patient motivation and effort. Research indicates that most people who quit smoking do so without help from a health care provider⁴³; that provider counseling regarding use of seatbelts has uncertain influence on actual use⁴⁴; and that routinely employed approaches to encouraging weight loss in the clinical setting are largely ineffective.^45,46 In the context of these observations, it would be of interest to examine the proportions of people who report adopting the health behaviors we studied with and without provider involvement. We hypothesize that, at a population level, improvements in such behaviors are largely independent of provider involvement.

Our study had some limitations. The analyses were observational, so caution is needed in deriving a causal interpretation of the findings. We examined a range of preventive care and health behaviors, which were chosen primarily based on their availability in the MEPS. It is uncertain whether similar findings would be observed for other care and behaviors. Although we detected significant individual changes in preventive care associated with changes in health insurance status, follow-up intervals longer than the 1-year period possible in the MEPS would likely be required to more fully capture such effects.

The summary estimates of the effects of changing insurance yielded by our conditional logistic regression models do not capture possible differences in the effects of gaining insurance (eg, pent-up demand) versus losing insurance (eg, catching up on care immediately before coverage lapses) but average out these unmeasured differences. Nonetheless, we believe it is reasonable to assume comparable insurance gain and loss effects, given prior research using the MEPS that indicates that the health care expenditures of people gaining insurance closely mirror the expenditures of those losing insurance.⁴ We also did not separately examine the effects of changing insurance stratified by insurance type (private vs. public), in part because of concerns about relatively smaller sample sizes, reducing the power to examine effects on preventive care and health behaviors. These issues may be worthy of examination in future studies.

While our analyses adjusted for numerous factors that may influence preventive care and behaviors, we could not adjust for other behavioral influences (eg, state seatbelt laws, local smoking ordinances). However, in a multiyear study such as ours, the average effects of gaining (or losing) insurance will be captured before, during, and after changes in such factors. All the study measures were subject to self-report errors, increasing the possibility of misreporting insurance status, preventive care, and health behaviors. Future studies should examine the degree to which health insurance gain or loss determined via other methods (eg, review of administrative records) is associated with self-reported and, when feasible, objectively measure preventive care and health behavior change (eg, change in BMI based on measured height and weight). Finally, nonresponse to the MEPS may produce bias, so generalizability to nonresponders is uncertain. Still, MEPS data are likely the most representative data available to examine the study research questions.