Accuracy of Claims-Based Measures of Severity of Childhood Illnesses

Abstract

Background

The use of electronic health records to conduct comparative effectiveness studies requires accurate measure of severity of patients’ illness.

Objectives

This brief report provides data on relative accuracy of claims-based severity indices for childhood diseases.

Measures

We compared the accuracy of All Patient Refined Diagnosis-Related Groups (APR-DRG), All Payer Severity-adjusted Diagnosis-Related Groups (APS-DRG), Alemi and Walters Severity across Episodes of Illness, and count of diagnoses.

Methods

The accuracy of each measure was calculated using the percent of deviance explained in mortality and percent of variation explained in length of stay (a surrogate measure of resource utilization).

Subjects

Data were obtained from the 2006 Kid’s Inpatient Database of the Healthcare Cost and Utilization Project of the Agency for Healthcare Research and Quality. We examined data on 3.1 million patients across 38 states.

Results

Alemi and Walters’ formula-based severity score explained 34% of variation in length of stay and 32% of variation in mortality. This index was more accurate than other indices, especially in predicting mortality, where it was 5-fold more accurate than APS-DRG and 3-fold more accurate than APR-DRG. The difference in accuracy was not only statistically significant but also large enough that it could change conclusions of comparative effectiveness studies.