Predictive value of critical illness scores for outcomes of patients with community-acquired pneumonia: analysis of critical care data from the MIMIC-IV database

This study evaluated the predictive abilities of critical illness scoring instruments in intensive care unit (ICU) patients with community-acquired pneumonia. In this population-based retrospective study, patients with community-acquired pneumonia admitted to the ICU were extracted from the fourth edition of the Medical Information Mart for Critical Care (MIMIC IV) database from 2008 to 2019. Critical illness scores used were the Simplified Acute Physiologic Score II (SAPS II), Acute Physiology and Chronic Health Evaluation III (APACHE III), Sequential organ failure assessment (SOFA), Oxford Acute Severity of Illness Score (OASIS), and logistic organ dysfunction score (LODS), CURB-65, and CHA 2 DS 2 -VASc scoring systems. Patients’ mean age was 65.1 years, 56.7% were male, 66.1% were white, and 53.8% were admitted to the ICU from the emergency department. All seven critical illness scores were significantly higher in non-survivor patients who had cardiovascular events and those who developed atrial fibrillation than in those who did not. APACHE III (AUROC 0.761) and LODS (0.742) showed relatively higher AUROCs for predicting 90-day mortality than the other evaluated scores, although their discrimination remained fair. For predicting incident atrial fibrillation, CURB-65 (0.672), CHA₂DS₂-VASc (0.655), and APACHE III (0.625) had AUROCs above 0.6, indicating modest discrimination. APACHE III and LODS showed relatively higher, but still fair, discrimination for 90-day mortality compared with the other evaluated scores. All scores demonstrated poor discrimination for incident cardiovascular events. CURB-65 and CHA₂DS₂-VASc showed relatively higher but modest discrimination for incident atrial fibrillation, suggesting limited utility as stand-alone predictive tools.