Derivation and Validation of a Novel Prediction Model to Identify Low-Risk Patients With Acute Pulmonary Embolism

Accurate identification of low-risk patients with acute pulmonary embolism (PE) who may be eligible for outpatient treatment or early discharge can have substantial cost-saving benefit. The purpose of this study was to derive and validate a prediction model to effectively identify patients with PE at low risk of short-term mortality, right ventricular dysfunction, and other nonfatal outcomes. This study analyzed data from 400 consecutive patients with acute PE. We derived and internally validated our prediction rule based on clinically significant variables that are routinely available at initial examination and that were categorized and weighted using coefficients in the multivariate logistic regression. The model was externally validated in an independent cohort of 82 patients. The final model (HOPPE score) consisted of 5 categorized patient variables (1, 2, or 3 points, respectively): systolic blood pressure (>120, 100 to 119, <99 mm Hg), diastolic blood pressure (>80, 65 to 79, <64 mm Hg), heart rate (<80, 81 to 100, >101 beats/min), arterial partial pressure of oxygen (>80, 60 to 79, <59 mm Hg), and modified electrocardiographic score (<2, 2 to 4, >4). The 30-day mortality rates were 0% in low risk (0 to 6 points), 7.5% to 8.5% in intermediate risk (7 to 10), and 18.2% to 18.8% in high-risk patients (≥11) across the derivation and validation cohorts. In comparison with the previously validated PESI score, the HOPPE score had a higher discriminatory power (area under the curve 0.74 vs 0.85, p = 0.033) and significantly improved both the discrimination (integrated discrimination improvement, p = 0.002) and reclassification (net reclassification improvement, p = 0.003) of the model for short-term mortality. In conclusion, the HOPPE score accurately identifies acute patients with PE at low risk of short-term mortality, right ventricular dysfunction, and other nonfatal outcomes. Prospective validation of the prediction model is necessary before implementation in clinical practice.