Health inequality measurement using EuroQol-5D valuation weights

Objectives: Despite various government initiatives including expansion of health insurance coverage, health inequality has been a key health policy issue of South Korea in the past decade. This study describes and compares the extent of total health inequality and income-related health inequality among Korean adults. Methods This study uses the 1998, 2001, and 2005 Korean National Health and Nutrition Examination Survey (KNHANES) data. The KNHANES is a national household survey that provides comprehensive information on health status, health care utilization, and sociodemographics of nationally representative samples. The surveys include a general question on self-assessed health (SAH) measured on a five- point ordinal scale. The SAH ordinal responses were rescaled to cardinal values using ordered probit and interval regression. Interval regression can provide a more efficient alternative to ordered probit models when the values of thresholds are known. The boundary of each threshold for interval regression was obtained from the empirical distribution of EuroQol-5D (EQ-5D) valuation weights estimated from the 2005 KNHANES. We matched the cumulative frequency of each SAH category and EQ-5D valuation weights to find the threshold. The final model predicting individuals' health status includes age, gender, educational attainments, occupations, income, and regional prosperity index. Concentration index (CI) was used to measure and decompose health inequality. Results: Koreans appear to report their health status (SAH) in non-extreme terms. Only 6.8% and 2.2% of Koreans responded excellent and very poor SAH, respectively. This seems to be related to the results of the EQ-5D questionnaire, where most respondents checked the first category of each EQ-5D question. This discrepancy may originate from the differences in wording between the SAH question of the KNHANES and the EQ-5D question. The KNHANES data show unequal distribution of total health inequality in favor of higher income groups, getting worse over time (0.0309 in 1998, 0.0383 in 2001, and 0.0931 in 2005). The income-related health inequality (CI) in 2005 is 0.0324, indicating 34.2% of total health inequality attributed to income. About a half of income-related inequality in health is related to unavoidable variables such age and gender. The interval regression approach performs better than ordered probit models, and the rescaling of the SAH categories to cardinal utility with EQ-5D provides more consistent and reliable outputs. Conclusions: The findings indicate continued health inequalities across sociodemographic and income groups despite the recent government's efforts. Further research is warranted to investigate potential policy actions to decrease health inequality.