The effect of casemix adjustment on mortality as predicted by APACHE II

Abstract

To describe the effect of casemix on mortality as predicted by APACHE II scoring. Retrospective analysis of an ICU database. All patients admitted to 19 ICU between 1 January 1992 and 31 May 1994 on whom data had been entered into a database. Excluded from the analysis were those readmitted to ICU, those aged under 16 years, those admitted after cardiac surgery or with burns, those for whom physiological data was incomplete and those for whom hospital outcome was unknown. Data on the remaining 6258 patients are reported. APACHE II scores were calculated from the worst values within 24 h of ICU admission. Hospital mortality was predicted with the published equation and coefficients. Mortality ratios (observed hospital deaths divided by predicted hospital deaths) were calculated for various groups. Mortality ratios varied widely by subgroup, and observed hospital deaths were greater than predicted, particularly for the following patient groups: those with predicted mortality of less than 70%, those with APACHE II scores in the range of 5–19, those older than 55 years, those with a Glasgow Coma Score of 15 or in the range 9–14, those not having emergency surgery, those with either 0 or 4 chronic health points and those in non-operative respiratory or neurological categories. The mortality ratio was markedly less than 1.0 only among non-operative cardiovascular patients. APACHE II did not accurately adjust for casemix in our data. Unless account is taken of differences in casemix, using mortality ratios to compare ICU is likely to be inaccurate and misleading.