Predictors of Medical Expenditures in Hypertension Patient in Taiwan－ a Quantile Regression Approach

Presenter: Hsiu-Ying Fang, BNHI, Taiwan

Abstract

Objectives: Treating hypertension and hypertension-related diseases require high medical cost in Taiwan. The proposed capitation payment must be able to respond to the expecting medical expenditure of the insured in order to uphold justice and efficiency. The study is to explore the important factors in the medical expenditures of hypertensive patients as well as to understand the medical situations (including those with extreme expenditures) by using a quantile regression approach. Methods：The study design was retrospective secondary data analysis. The analysis identified how gender, age, premium, number of complications, catastrophe diseases, previous medical expenditure, category and ownership affects medical expenditures. The study used quantile regression to reveal how independent variables influence medical expenditures in different percentiles, and the results are compared with the mean derived from the conventional regression method.Results：Age and gender could explain little variation in the medical expenditures of hypertensive patients, only 0.2% to 0.5%. When considering medical expenditures of the previous year, there was a greater variation in outpatient expenditure as well as in the expenditure of outpatient, inpatient and complications, with the ratio of 57.4% and 50.9%. When considering catastrophe diseases and category, the variability is 58.8% and 50.9%, whereas little variability showed in the expenditure of inpatient and complications, only 7.5%. Comparing quantile regression analysis with conventional regression analysis, there is a big gap between their coefficients. For example, when the quantile increases, the effect of the medical expenditure of the previous year on the expenditure of outpatient, inpatient and complications are stronger( NT＄0.67 increased based on NT＄1 raised at the 0.20 quantile and NT＄5.06 increased at the 0.99 quantile, while the mean is about NT＄1.05). Other variables that cause big differences include catastrophe diseases and category . Conclusions：The strongest determinant of medical expenditures of hypertensive patients is the medical expenditures of the previous year, followed by category, the number of complications, and catastrophe diseases. There is a large difference in the numerical analysis between the quantitle regression models and the conventional regression models. The average value derived from the conventional regression analysis is not able to predict the medical expenditures of all hypertensive patients. The capitation payment by diseases should be applied in the future, and it should start with the outpatient expenditure which is more likely calculable; quantile regression should be adopted to compare the differences between coefficients; extreme quantile groups and patients with catastrophe diseases should be excluded; and the payment scheme of different category should be considered.