Evaluation of alternative diagnosis based risk adjustment models and morbidity trajectories for application in Taiwan

Published: December 1, 2008
Category: Reports
Author: Chang HY
Country: Taiwan
Language: null
Type: Performance Analysis
Setting: Academic

Baltimore, MD, USA: Johns Hopkins University (doctoral dissertation).

Department of Health Policy and Management, Johns Hopkins University, Baltimore, MD, USA

OBJECTIVE: The overall goals of this study are to test the validity of the ACG (Adjusted Clinical Group) case-mix system in explaining healthcare costs, resource allocation, and predictive modeling using Taiwan’s National Health Insurance (NHI) data, and to empirically identify morbidity trajectories among the population and assess the effects of including morbidity trajectories into risk adjustment models.

DATA: A 1% random sample of NHI enrollees continuously enrolled in 2002 was used for concurrent analyses and evaluation of resource allocation (n= 173,234). A 2002 to 2003 cohort (n=164,562) and a 2002 to 2005 cohort (n=147,892) were used for prospective and longitudinal analyses.

METHODS: For concurrent analyses, health measures derived from 2002 diagnoses were used to explain 2002 resource utilization; for prospective analyses and predictive modeling, the outcome was 2003 resource utilization. For longitudinal analyses, health measures derived from 2002 to 2004 diagnoses were used to explain 2005 resource utilization. For evaluation of resource allocation, age/sex-adjusted self-reported health status, mortality rates, and the ACGs-based index in 2002 at district level were compared.

RESULTS: Comprehensive models performed better in explaining resource utilization. Adjusted R2 of total costs in concurrent/prospective analyses were 4% / 4% in the demographic model, 15% /10% in the ACGs or ADGs model, and 40% / 22% in the models containing EDCs. The ACGs morbidity index could accurately reflect geographic differences in morbidity burden given its high correlation with both self-reported health status and mortality rates at district level. In terms of predictive modeling, both diagnosis-based and prior cost models performed much better than the demographic model. Over a three-year period, most people’s morbidity levels tended to stay constant. There were significant differences in medical utilization across six morbidity trajectory groups. The effect of adding trajectory indicators differed substantially by risk adjustment models: the increase in adjusted R ranged from 0.3% in EDCs model to 5.7% in demographics model.

CONCLUSIONS: Given the wide-scale availability of claims data and the superior performance, Taiwan and other managers of health insurance plans should consider claims-based models for policy-relevant applications, such as cost prediction, resource allocation, predictive modeling, and quality improvement.

Resource Allocation,Predictive Risk Modeling,Cost Burden Evaluation,Morbidity Patterns,United States

Please log in/register to access.

Log in/Register

© The Johns Hopkins University, The Johns Hopkins Hospital, and Johns Hopkins Health System.
All rights reserved. Terms of Use Privacy Statement

Back to top