(S-119) Glance, L.G., Sunday 9:15

TITLE: EFFECT OF CASE MIX VARIATION ON THE MORTALITY ODDS RATIO BASED ON APACHE II USING A LOGISTIC REGRESSION APPROACH

AUTHORS: Laurent G. Glance, MD¹, Turner Osler, MD²
AFFILIATION: ¹University of Rochester Medical Center, Rochester, NY; ²University of Vermont Medical College, Burlington, VT

INTRODUCTION: APACHE II is used to adjust crude mortality rates for severity-of-disease in order to allow comparisons across ICUs. The use of APACHE II and other predictive models for severity adjustment is predicated on the assumption that these models can be transported to ICUs with different case mixes. The purpose of the study was to test the hypothesis that the mortality odds ratio is independent of case mix.
METHODS: APACHE II data was collected prospectively from the SICU at a single institution between 1990 and 1997 (cardiac, burn and pediatric patients were excluded). The patients in this data set were first ranked according to their APACHE II predicted mortality and then divided into ten groups (decile of risk) of equal size: 0-3.1%, 3.2-4.7%, …, 39- 100%. A computer simulation was used to create 300 different cases mixes with mortality rates between 6% and 15%. The number of patients in each decile (x₁ from the first decile, x₂ from the 2^nd decile, …, and x₁₀ from the 10^th decile) constituted a distinct case mix. A virtual ICU (VICU) with a given case mix was created by randomly resampling (with replacement) x₁ from the first decile, x₂ from the 2^nd decile, …, x₁₀ from the 10^th decile. Twenty different VICUs with identical case mixes were created for each of the 300 case mixes. Logistic regression analysis was used to derive the mortality odds ratio in order to measure the relative risk of death in each of the simulated data sets compared to the original database. Bootstrapping was used to estimate the mean value and 95% confidence intervals of the mortality odds ratio from the 20 simulated data sets for each case mix.
RESULTS: Linear regression analysis did not show a significant coorelation between the mortality odds ratio and the simulated mortality rate (R sq = 0.04).
CONCLUSION: Since all of the patients in the simulated data sets were cared for at a single institution, it was assumed that the quality of care remained uniform. As would be expected, changes in the case mix in this simulation did not lead to significant changes in the relative risk of death across ICUs after adjusting for severity of disease using APACHE II. Based on these findings, it may be possible to compare ICU performance using the mortality odds ratio.