BACKGROUND AND PURPOSE: Taiwan's National Health Insurance (NHI) program has considered the use of capitation payments to health care providers as a method for control of the rising costs of the system. The establishment of capitation payments usually requires the performance of risk adjustment. The purposes of this study were to develop a diagnosis-based risk adjustment model for the NHI and to evaluate its predictability. METHODS: Using a 2% random sample of 371,620 NHI enrollees, the authors developed a Taiwan version of the Principal Inpatient Diagnosis Cost Groups (TPIPDCGs) from 1996 claim records to predict an individual's expenditure in 1997. Weighted least squares regression models were built in an estimation sample (two-thirds of the study sample), and were cross-validated in a validation sample (the remaining one-third of the study sample). Predictive R2 and predictive ratios were used to evaluate the model's predictability. RESULTS: Only 7.88% of the study sample could be classified into 1 of the 16 TPIPDCGs. Combined with demographic variables, which alone could explain 3.7% of the variation in an individual's future expenditure, the risk adjustment model based on TPIPDCGs could explain 12.2% of expenditure variation. In addition, the finding that the predictive ratios of the TPIPDCG model approximated unity better than those of the demographic model in all subgroups indicates that the capitation payment as predicted by the TPIPDCG model for each subgroup would better correlate to the actual spending. CONCLUSION: Taiwan's risk-adjusted capitation model based on principal inpatient diagnoses has higher predictability on individual's future expenditure than its counterpart in the USA. This finding provides insight into not only the development of Taiwan's diagnosis-based risk adjustment models but also the necessity of modification when applying foreign-developed risk adjustment models to the NHI.
BACKGROUND AND PURPOSE: Taiwan's National Health Insurance (NHI) program has considered the use of capitation payments to health care providers as a method for control of the rising costs of the system. The establishment of capitation payments usually requires the performance of risk adjustment. The purposes of this study were to develop a diagnosis-based risk adjustment model for the NHI and to evaluate its predictability. METHODS: Using a 2% random sample of 371,620 NHI enrollees, the authors developed a Taiwan version of the Principal Inpatient Diagnosis Cost Groups (TPIPDCGs) from 1996 claim records to predict an individual's expenditure in 1997. Weighted least squares regression models were built in an estimation sample (two-thirds of the study sample), and were cross-validated in a validation sample (the remaining one-third of the study sample). Predictive R2 and predictive ratios were used to evaluate the model's predictability. RESULTS: Only 7.88% of the study sample could be classified into 1 of the 16 TPIPDCGs. Combined with demographic variables, which alone could explain 3.7% of the variation in an individual's future expenditure, the risk adjustment model based on TPIPDCGs could explain 12.2% of expenditure variation. In addition, the finding that the predictive ratios of the TPIPDCG model approximated unity better than those of the demographic model in all subgroups indicates that the capitation payment as predicted by the TPIPDCG model for each subgroup would better correlate to the actual spending. CONCLUSION: Taiwan's risk-adjusted capitation model based on principal inpatient diagnoses has higher predictability on individual's future expenditure than its counterpart in the USA. This finding provides insight into not only the development of Taiwan's diagnosis-based risk adjustment models but also the necessity of modification when applying foreign-developed risk adjustment models to the NHI.