Prediction of 30-Day Complicated Clinical Course in Acute Pulmonary Embolism: A Prospective Cohort Comparison of the ESC, Bova, and Modified FAST Models

Background/Objectives: Risk stratification of hemodynamically stable acute pulmonary embolism guides the intensity of monitoring and the site of care, but the available tools use different variables and may classify the same patients differently. We aimed to compare the prognostic performance of three established models for predicting a complicated clinical course. Methods: In a prospective single-center cohort, consecutive hemodynamically stable adults with computed tomography-confirmed acute pulmonary embolism in the emergency department were enrolled. The European Society of Cardiology (ESC) algorithm, the Bova score, and the modified FAST score were subsequently calculated from prospectively collected baseline data. The primary outcome was a complicated clinical course within 30 days (a composite of death, hemodynamic collapse, cardiac arrest, mechanical ventilation, or rescue reperfusion). Discrimination was assessed by the area under the curve (AUC) with DeLong confidence intervals (CI) and pairwise comparison. Prognostic classification performance was assessed at each model’s prespecified intermediate–high-risk threshold. Results: Of 383 patients (median age 72 years, 51.7% women), 56 (14.6%) developed a complicated course. Discrimination was modest across the three tools: modified FAST AUC 0.680 (95% CI 0.608–0.752), Bova 0.663 (0.587–0.740), and ESC 0.617 (0.560–0.673). No statistically significant pairwise difference in AUC was detected (all Holm-adjusted p > 0.3). Classification characteristics differed substantially: the modified FAST score was the most sensitive (sensitivity 53.6%, specificity 74.0%), whereas Bova Stage III was the most specific (specificity 95.7%, sensitivity 14.3%), with the ESC algorithm intermediate (sensitivity 26.8%, specificity 86.2%). Agreement between the tools for intermediate–high-risk classifications was poor to fair (Cohen’s kappa 0.12–0.41). Conclusions: The three tools showed modest discrimination for a 30-day complicated course, and no statistically significant pairwise differences in AUC were detected. They nevertheless classified markedly different proportions of patients as intermediate–high risk and showed distinct sensitivity–specificity profiles. Because the tools carry distinct sensitivity–specificity trade-offs and are not interchangeable, the choice among them should match the clinical question, and no single score should determine treatment alone.