External validation of the PE-SCORE for predicting early clinical deterioration in acute pulmonary embolism using real-world data.

Acute pulmonary embolism (PE) is associated with substantial short-term morbidity and mortality, particularly within the first few days following presentation. Although existing risk stratification tools primarily predict 30-day mortality, early clinical deterioration may be more relevant for acute management decisions. In this context, tools designed to rule out early deterioration safely-rather than to predict mortality with high specificity-may better inform initial triage and disposition decisions. The Pulmonary Embolism Short-Term Clinical Outcomes Risk Estimation (PE-SCORE) was developed to predict early deterioration and has particular strength as a rule-out tool to identify low-risk patients unlikely to experience early clinical worsening; however, external validation in independent cohorts remains limited. This single-center, retrospective, observational cohort study included consecutive adult patients with acute PE confirmed by computed tomography pulmonary angiography, excluding those with hemodynamic instability, chronic PE, incomplete data, malignancy, alternative diagnoses, or loss to follow-up. PE-SCORE was calculated according to the original algorithm. The primary outcome was a 5-day composite endpoint comprising all-cause mortality or clinical deterioration. A total of 331 patients were analyzed (mean age 69.8 ± 15.8 years; 52.3% female). The 5-day composite outcome occurred in 57 patients (17.2%). The median PE-SCORE was significantly higher among patients with adverse outcomes (3.0 vs. 2.0, p < 0.001). PE-SCORE demonstrated moderate discriminative ability for the 5-day composite outcome (AUC = 0.748; 95% CI: 0.688-0.803). A cutoff value ≥ 3 yielded 77.2% sensitivity and 65.0% specificity. In comparison, binary classification (PE-SCORE ≥ 1) achieved 100% sensitivity and negative predictive value, with a specificity of 12.8% and a positive predictive value of 19.3%, reflecting substantial over-triage with 87.2% of event-free patients classified as at risk. Of note, only 10.6% of patients were classified as low-risk (PE-SCORE = 0). Calibration metrics indicated good model fit. In this external validation cohort, PE-SCORE showed moderate discriminative ability and good calibration for predicting early adverse outcomes in acute PE. The binary classification at PE-SCORE ≥ 1 reliably identified low-risk patients with perfect negative predictive value; however, this threshold was associated with a low positive predictive value and substantial over-triage, and the vast majority of event-free patients were classified as at risk. The low specificity and high rate of over-triage restrict its standalone utility for disposition decisions. Its use in outpatient management decisions may be associated with a high risk of unnecessary hospitalization, and further risk assessment that incorporates clinical judgment and patient preferences remains essential for higher-risk classifications. Prospective implementation studies comparing PE-SCORE-guided strategies with usual care are needed before routine clinical adoption.