Introduction:

Accurate risk assessment is essential for sound decision-making in patients hospitalized with COVID-19, and several prediction rules have been developed to estimate their risk of death or admission to intensive care.

Objective:

To validate the CALL, COVID-GRAM, and ISARIC-4C scores for predicting intensive care unit admission or death among patients hospitalized with COVID-19 in a cohort from three Bogotá hospitals during 2020.

Methods:

A prospective multicenter analytical observational study that included adults hospitalized for COVID-19 between April and November 2020 in three fourth-level hospitals in Bogotá. The scores from the three scales were calculated for each participant. Receiver-operating-characteristic curves were plotted and areas under the curve with 95 % confidence intervals were estimated; calibration was assessed with the Brier statistic and LOESS smoothing, and clinical utility was examined using decision-curve analysis.

Results:

Of 793 patients, 358 had data to calculate the three prediction rules. 186 (26.2 %) were admitted to intensive care and 120 (16,9 %) died. The areas under the curve for intensive care admission were: COVID-GRAM 0,67 (95 % CI: 0,61-0,73), CALL score 0,54 (95 % CI: 0,48-0,60), ISARIC-4C 0,61 (95 % CI: 0,55-0,67). For mortality they were: COVID-GRAM 0,76 (95 % CI: 0,70-0,81), CALL score 0,65 (95 % CI: 0,58-0,73), ISARIC-4C 0,75 (95 % CI: 0,6-0,8).

Conclusions:

the clinical prediction rules evaluated showed poor to moderate performance for intensive care admission and good performance for mortality. The ISARIC-4C and COVID-GRAM could be considered to evaluate the risk of death.

        · abstract in Spanish     · text in English     · English ( pdf )