Abstract
OBJECTIVES: To develop and validate a clinical model to predict 30-day mortality in high-risk COVID-19 patients evaluated in the Emergency Department, and to evaluate outcomes in adults with COVID-19 pneumonia classified as low-risk by the model and discharged home. METHODS: Secondary analysis of the COVID-19\_URG-HCSC registry at Hospital Clínico San Carlos (Madrid). Phase 1 (Feb-Apr 2020): derivation of a model in patients aged ≥18 years with probable or confirmed COVID-19 who were classified as high-risk at triage (age ≥55 years, baseline oxygen saturation <96% on arrival, or pulmonary infiltrates on chest radiograph). The primary outcome was 30-day all-cause mortality. Phase 2 (Sep 2020-May 2021): validation in adults with low-risk pneumonia discharged home and monitored by telephone. Primary outcomes were 30-day all-cause mortality, revisits, and hospitalization. RESULTS: Of 2,436 eligible patients in phase 1, age, oxygen saturation, renal function, LDH, CRP, platelet count and dementia were independent mortality predictors. The model showed excellent discrimination (AUC 0.918) and good calibration, and was implemented in a web-based tool. In phase 2, 565 patients with low-risk pneumonia were monitored; no deaths occurred at 30 days. Re-attendance was 8.0% and hospitalisation 12.6%, mainly for respiratory failure. All hospitalised patients were ultimately discharged home. CONCLUSIONS: The PrediCOVID model provides accurate risk stratification and, when combined with telemonitoring, enables safe early discharges while reducing hospital burden.