Profiling Physician Practice Patterns Using Diagnostic Episode Clusters

Publisher Website
Google Scholar
Abstract
Health plans and providers need to profile current practice patterns to understand better the resources used in managing medical conditions. A profiling system is presented that groups International Classification of Diseases (ICD-9-CM) codes into 125 diagnostic clusters based on clinical homogeneity with respect to physician treatment response. For each diagnostic cluster, diagnostic episode clusters (DECs) are formulated. A DEC links all services incurred in treating a patient's medical condition within a specific period of time. Each DEC is marked with a severity-of-illness, comorbidity, and age indicator. To test the validity of the diagnostic cluster methodology, claims were analyzed from a preferred provider organization (PPO) and an independent practice association (IPA). PPO and IPA DEC charges and utilization were compared with t-tests. Physician practice patterns differed based on patient severity of illness, comorbidities, and age. Both PPO and IPA physicians delivered significantly more resources to patients in higher severity-of-illness categories. PPO physicians generally treated older patients with more resources than younger patients. Patient age did not have the same impact on IPA physicians' practice patterns. IPA physicians' average treatment pattern was about 22% less expensive than that of PPO physicians. IPA physicians decreased average expenses by reducing hospital days by about 73% (P < 0.01) and hospital outpatient visits by about 89% (P < 0.01) compared to the rates of PPO physicians. Ambulatory services among IPA physicians were not significantly higher than rates for PPO physicians. The DEC methodology is a valid approach for profiling patterns of treatment. The style of medicine in the IPA was less hospital intensive and, consequently, less expensive than that practiced by PPO physicians. PPO physicians also had greater practice pattern variations than IPA physicians.

This publication has 0 references indexed in Scilit: