Preventive care: underused even when free. Is there something else at work?

Abstract

Explaining the rationale of preventive care underuse is a difficult task considering its great benefits for health. Underuse is even more difficult to explain in countries like Italy where preventive care can be obtained for free. In this article we investigate the determinants of prevention underuse with an empirical model based on human capital theory which also includes three factors to which little attention has been paid so far: role played by the General Practitioner (GP), nonmonetary barriers to access and health beliefs. We apply a recursive probit model explaining both recourse to prevention and to the GP which allows us to adequately measure the effect of the latter on the former and to quantitatively compare the determinants of curative and preventive care. We find that the GP plays a minor role in prevention use but that nonmonetary barriers to access and health beliefs are strong determinants of preventive care demand. Finally, we also find support for both Grossman's capital depreciation theory (at younger ages) and Cropper's shorter pay-off period theory (at older ages).

Keywords:

• preventive care
• General Practitioner
• nonmonetary access barriers
• health beliefs
• recursive probit model

JEL Classification:

• C30
• D12
• I11
• I18
• R53

Acknowledgements

We thank Donald Kenkel, Maarten Lindeboom, Willard Manning, Owen O'Donnell, Eddy van Doorslaer and all participants at the Annual Phd Seminar in health economics and policy organized by Swiss School of Public Health (Crans Montana 2009) for valuable comments. We also thank Elena Granaglia, Leandro Elia and two anonymous referees for helpful comments. The usual disclaimer applies.

Notes

¹Prevention is also an important health policy concern in developing countries but, given that their epidemiological transition is not yet complete, there are probably more urgent public health issues such as reducing the prevalence of infectious diseases.

²Even if contracts giving complete cover of monetary costs in case of illness were available, the utility losses associated with pain and suffering would hardly be insurable and, generally, the value attributed by the individual to such losses greatly exceeds the monetary cost and loss of productivity caused by illness (Tolley et al., 1994).

³Essential levels of assistance is the package of minimum level of care that the NHS provides to Italian citizens.

⁴ The two latent variables, PREV* and GP*, represent the propensity to use preventive care and GP services, respectively. The observed outcomes, PREV and GP, equal to one when there is a positive propensity to use the corresponding health service and zero otherwise.

⁵ The same model is assumed for all women and individual subscript i is omitted for readability.

⁶Note that we have compared the regional density of GP with the average number of patients per GP, and opted for the latter as it is more strongly correlated with the use of GP services.

⁷Our trust in the GP variable equals one if the respondent trusts the GP more than any other health provider and zero otherwise.

⁸ We also tried the alternative estimation of an IV LPM. We found that the instruments (i.e. the average number of patients per GP and trust in the GP) where reasonably strong with a F statistic of 12.3 and 22.6 for the mammography and pap test models, respectively. However, we also found that the instruments are not suitable for the pap test model as they fail the over-identifying restrictions test. A likely reason is that our trust in the GP variable is, by construction, correlated with the unobservable trust in the specialist who performs the pap test. In addition, the IV LPM for mammography produced very inefficient results. As discussed in Section IV, though, this does not invalidate the results obtained with the recursive bi-probit, but rather illustrate the greater practicability of this model in our setting.

⁹We have also estimated our empirical model for each geographical region separately. However, we did not find any important difference in the effects of the main covariates and in particular of supply variables. These considerable differences in preventive care utilization between geographical areas might be explained by contextual differences with respect to socio-economic conditions.