Hospital Financial Risk Stratification for Inpatient Service in the impending TW-DRGs Reimbursement System: A Classification and Regression Trees Modeling Approach

Presenter: Sheng-Hsiu Wu, Chang Jung Christian University

Abstract

Confronted with the escalating healthcare expenditures in the NHI system, the Bureau of NHI, Taiwan has been proposing to launch the Taiwan-edition diagnosis-related groups (DRGs) reimbursement system. Hospitals, on the other hand, have been concerned about the possibly adverse effect of the new payment system on the currently deteriorating financial status. So far no evidence showed whether they would truly become victims of the new payment system as they expected or, instead, they might gain some benefits from it. Therefore, the main purpose of this study is to simulate and predict the likely incurred financial risk for different categories of hospitals with specific characteristics under the proposed inpatient TW-DRGs payment system.

The claims data from the “Inpatient Medical Expense” files released by the BNHI for the years 2005 and 2006 were collected. A total of 297,235 records from more than 400 hospitals (472 in 2005 and 442 in 2006) nationwide were extracted. After excluding records of mental health care services, which were not the targeted payment services under the proposed TW-DRGs system, and those which could not be correctly coded, 218,990 records were retained for analysis. With the aid of the “DRG Coding Service” software package available from the BNHI, a specific DRG code was generated for each inpatient service record. The hospital-level characteristics, ALOS deviation (from the Taiwan norms) by DRG, case mix index, and Herfindahl-Hirschman Index (HHI) were calculated for each hospital and used as the independent variables in the classification and regression trees (CART) model to classify hospitals into different financial risk groups.

Results showed that all the hospitals could be classified into 12 financial risk groups. Among the rest, the medical centers (irrespective of their ALOS deviations) have the highest financial risk (average financial risk= -0.233), followed by the local hospitals with ALOS deviation 5.4 days more (average financial risk= -0.147) and regional hospitals with ALOS deviation 0.6 days more (average financial risk= -0.078). To a certain extent, the variable ALOS deviation plays a key role in the financial risk profiling for both the regional- and local-level hospitals—hospitals with an ALOS deviation 0.3 days more compared to the Taiwan norm or below will benefit from the proposed TW-DRGs payment system as indicated by the positive values of average financial risk. In general, the higher the ALOS deviation, the worse the financial risk. Meanwhile, the financial risk of general hospitals is relatively higher than that of specialty hospitals or chronic disease hospitals.

In summary, the prediction accuracy of the CART model reaches a satisfactory level of 96.7%. The medical centers might suffer from financial risk as they have been concerned partly due to their worse ALOS performance. To the contrary, hospitals will benefit from the new reimbursement mechanism as long as their ALOS deviations may be limited to 0.65 days less compared to the Taiwan norm or above. The implications derived from this study will serve as managerial decisions reference for the targeted hospitals.