Categorical Apache IV prediction of ICU and 90 day mortality
Critical illness; prediction; mortality; ICU; Acute Physiology and Chronic Health Evaluation
Published online: Apr 21 2022
Abstract
The APACHE IV score is used to predict hospital mortality and for case-mix adjustments in benchmarking initiatives in critically ill patients. The timing of the evaluation of survival may be critical. Hence we evaluated categorical APACHE IV predictive scoring on ICU and 90 day mortality.
In a single center retrospective analysis all patients, admitted to the ICU of ZOL-Genk, Belgium, from 01- 01-2019 to 01-01-2020 were included into the analysis. Data were verified by a single, trained medical doctor, who was blinded to the patient outcome. Mortality data were retrieved from the national death register. The subgroups were defined by their proper APACHE diagnostic category. Quantitative analysis of ICU and 90 day mortality by logistic regression based on APACHE IV predictions per diagnostic category and the 10 most important features responsible for ICU and 90 day mortality.
Over the 1 year period 2816 ICU admitted patients were included in the analyses. The mean age was 62.66 ± 16.94 years. The patients could mainly be classified into cardiovascular (34.10%), neurologic (25.58%) and respiratory (11.83%).
Mean APACHE IV scores were 55.73 ± 23.21 and the overall APACHE mortality prediction was 13.86% ± 18.26. The mortality rate during ICU admission and after 90 day respectively was 5.11% and 10.90%. APACHE mortality prediction on the total population and Apache IV categories relative to the recorded ICU and 90 day mortality demonstrated a considerable overlap. The percentage of post-ICU mortality (<90 day post ICU admission) relative to ICU mortality for each Apache category was notable. Every cohort studied (total population, categorical data) revealed other weights to the 10 most important features responsible for ICU and 90 day mortality.
In conclusion the APACHE IV score underestimates the 90 day mortality for a mixed ICU population. This was the case for all diagnostic subgroups. This difference could possibly be explained by the relative weights of the ten most important variables in the patients APACHE IV score.
