Comparison of easy-to-use clinical prognostic models to identify low-risk normotensive patients with pulmonary embolism.

Background Lack of practicality is a common barrier to the use of clinical prognostic models for acute pulmonary embolism (PE).Objectives To compare easy-to-use prognostic models with <10 readily available variables to identify PE low-risk patients who are candidates for home care.Methods Based on prospective cohort data from 677 patients with acute PE from 9 Swiss hospitals, we calculated 11 easy-to-use prognostic models (sPESI, Geneva Prognostic Score, shock index [SI], BOVA, modified FAST, eStiMaTe, PATHOS, Uresandi, Agterof, Huang, ESC-2019) and classified patients as low vs. higher risk. The primary outcome was 30-day overall mortality. For each model, we examined test characteristics and the weighted clinical net benefit, i.e., the percentage of avoided hospitalizations in low-risk patients minus the percentage of deceased low-risk patients, assuming that death is 100 times worse than a not avoided hospitalization.Results Overall, 21/677 patients (3.1%) died within 30 days. The proportion of patients classified as low risk varied widely from 18.9% (eStiMaTe) to 95.8% (SI). Mortality among low-risk patients ranged from 0% (sPESI, eStiMaTe, ESC-2019) to 3.0% (modified FAST), with sensitivities varying from 100% (sPESI, eStiMaTe, ESC-2019) to 9.5% (SI). Similarly, the negative likelihood ratios for mortality ranged from 0.06 (sPESI) to 0.95 (modified FAST). The weighted net benefit was highest for the sPESI (35.9%) and lowest for the SI (-187.6%).Conclusion The sPESI, the eStimMaTe score, and the ESC-2019 model most accurately identified patients with PE who are at low risk of mortality. The sPESI showed the highest clinical net benefit.