Measuring the impact of Voluntary Health Insurance on out of pocket costs and socioeconomic-related inequality: methodological challenges and potential solutions with an application to Vietnam

Aims: This study has three aims: 1) to measure the impact of the Vietnamese Voluntary Health Insurance (VHI) programme on out-of-pocket (OOP) costs of health care after correcting for care-seeking and insurance-seeking self-selection biases; 2) to measure the effect of the VHI programme on socioeconomic-related inequality in out-of-pocket costs; and 3) to measure the role of VHI in preventing catastrophic health care costs. Data: This study is based on cross-sectional household survey data collected from three provinces of Vietnam: Hai Phong, Ninh Binh and Dong Thap. A total of 1,650 adults and 1,101 children were randomly selected and interviewed during the year 1999. Individual level data were available on the cost of health care in the last three months, the insurance status, personal and socioeconomic variables, health status and health care utilisation. In the sample, 1,192 individuals felt sick at least once in the last three months, and 985 of them sought care. Methods: The standard regression approach of measuring the average impact of VHI does not correct simultaneously for care-seeking and insurance-seeking biases. Also, the standard approach of measuring vertical equity in financing fails to account for the unmet need for care. This thesis proposes an improved approach, based on Heckman’s selection model, to estimate the impact of insurance on the cost of health care, after correcting for self-selection biases. To measure socioeconomic-related inequality in health care costs, a need standardised concentration index was proposed. This approach standardises for differences in the level of need between individuals, in turn controlling for the unmet need for care. Progressivity analysis was carried out using Kakwani’s index of progressivity. Finally, the incidence of catastrophic health care costs was modelled using probit equations that accounted for self-selection biases. Result: Analysis shows that insurance is negatively associated with expected cost of care, and this effect becomes more pronounced after correcting for selection biases. Need-standardised concentration indices demonstrate that insurance makes the distribution of health care costs more pro-poor. Kakwani indices suggest that insurance reduces the regressivity of financing. Finally, the study finds that VHI is associated with a lower probability of financial catastrophe. Conclusion: Membership in the Vietnamese VHI appears to have a protective effect on health care costs; this effect is augmented after controlling for selection biases due to unobserved characteristics. Insurance membership also appears to reduce the regressivity of health financing and the incidence and intensity of catastrophic health care costs.