Total Healthcare Expenditures from the 2017 Iran’s National Households Income and Expenditure Survey: The Application of Two-Part Models

Introduction: Although the precise estimates of healthcare expenditures are critical for health policy-makers, the right-skewed distribution and a substantial number of zero values of the measures of healthcare expenditure make such estimates challenging. The present study used conventional two-part (CTP) and marginalized two-part (MTP) models to handle the skewness and zero-inflation in expenditure distribution as two serious challenges. Materials and Methods: Data was used from the 2017 Households Income and Expenditure Survey (HIES; 38,252 households), a national cross-sectional study in Iran. CTP and MTP models were utilized to estimate the medical supplies, outpatient, inpatient and total medical expenditures. The rural-urban difference in total medical expenditures and other health services were also examined. All data analyses were performed using SAS. For all tests, twosided p-values <0.05 were interpreted as statistically significant. Results: The mean (SD) out-of-pocket spending for total healthcare was $143 ($488) per capita, and $182 ($650) and $105 ($239) for urban and rural areas, respectively. The mean (SD) medical supplies cost per capita was $48 ($240), and the mean (SD) of outpatient cost per capita and inpatient cost per capita were $61 ($245) and $34 ($294), respectively. Both CTP and MTP models suggested that urban population spent more money on total expenditures than rural populations (p<0.05). Although both models gave the same set of parameter estimates, the AIC indicated that the MTP-GG model was a more appropriate fit. Conclusion: The marginalized models provided better estimates in documenting inequalities/healthcare expenditures. Unlike the CTP model, the estimation of covariate effects on the marginal mean of the whole population via using the MTP model is straightforward. However, the MTP model may not outperform the CTP model in all cases. The applications of such models need to be considered in the future research to provide better estimates/documentations of healthcare expenditure and healthcare inequalities. In addition, these findings suggest a substantial inequality in healthcare expenditures between urban and rural areas. Considering the differences in urbanity and rurality can be of interest to health economists and policymakers.