Multicenter validation of Early Warning Scores for detection of clinical deterioration in COVID-19 hospitalized patients

OBJECTIVE: Investigate the predictive value of NEWS2, NEWS-C, and COVID-19 Severity Index for predicting intensive care unit (ICU) transfer in the next 24 h. DESIGN: Retrospective multicenter study. SETTING: Two third-level hospitals in Argentina. PATIENTS: All adult patients with confirmed COVID-19, admitted on general wards, excluding patients with non-intubated orders. INTERVENTIONS: Patients were divided between those who were admitted to ICU and non-admitted. We calculated the three scores for each day of hospitalization. VARIABLES: We evaluate the calibration and discrimination of the three scores for the outcome ICU admission within 24, 48 h, and at hospital admission. RESULTS: We evaluate 13,768 days of hospitalizations on general medical wards of 1318 patients. Among these, 126 (9.5%) were transferred to ICU. The AUROC of NEWS2 was 0.73 (95%CI 0.68–0.78) 24 h before ICU admission, and 0.52 (95%CI 0.47–0.57) at hospital admission. The AUROC of NEWS-C was 0.73 (95%CI 0.68–0.78) and 0.52 (95%CI 0.47–0.57) respectively, and the AUROC of COVID-19 Severity Index was 0.80 (95%CI 0.77–0.84) and 0.61 (95%CI 0.58–0.66) respectively. COVID-19 Severity Index presented better calibration than NEWS2 and NEWS-C. CONCLUSION: COVID-19 Severity index has better calibration and discrimination than NEWS2 and NEWS-C to predict ICU transfer during hospitalization.