Comparative Analysis of Four Risk Stratification Models to Identify Patients with Acute Pulmonary Embolism at Risk of Short-term Mortality

Abstract

Acute pulmonary embolism (PE) is potentially life-threatening, with up to 15% risk of death. We compared four risk stratification models to identify outpatients at risk of mortality up to 90 days post acute PE. A retrospective cohort study included outpatients aged ≥18 years with confirmed PE from June 1, 2014 to May 31, 2019, identified via diagnostic imaging reports. Simplified Pulmonary Embolism Severity Index (sPESI) and Hestia scores were calculated as per original derivation methods. Patients were stratified by four models: sPESI alone, Hestia alone, sPESI plus right ventricular dysfunction (RVD), and Hestia plus RVD. Model accuracy and discriminatory power for 30- and 90-day mortality were assessed by area under the receiver operating curve (AUC). The study comprised 785 outpatients (mean age 65.0 years; 42.2% male). Overall mortality rates were 4.1% at 30 days and 7.8% at 90 days. sPESI identified 31.5% as low risk versus 19.1% by Hestia. All models demonstrated 100% sensitivity and negative predictive value for 30-day mortality, but modest discriminatory power (AUC range: 59.2-67.1). sPESI consistently outperformed other models in both timeframes. Including RVD with sPESI or Hestia did not enhance accuracy and slightly reduced performance. The net reclassification index indicated minor improvement in non-event classification with RVD, but no benefit for identifying deaths. sPESI remains a modest yet effective predictor of mortality risk within 90 days following acute PE, consistently outperforming sPESI + RVD, Hestia alone, and Hestia + RVD at both 30 and 90 days. Adding RVD minimally improved predictive accuracy.