PREDICTING DEVELOPMENT OF CRITICAL ILLNESS IN PATIENTS WITH COVID-19 PRESENTING TO A TERTIARY CARE HOSPITAL IN A DEVELOPING COUNTRY

Background: With the increasing number of COVID-19 patients and limited resources available to accommodate them, there is a need for risk stratification tools to ensure better utilization of resources. Methods: We conducted a retrospective observational cohort study in patients discharged from the COVID designated areas of a large tertiary care hospital in Karachi, Pakistan from the 1^stof May to the 31^stof July, 2020. 581 patients were included and the COVID GRAM score was calculated at the time of admission and patients developing critical disease as per COVID GRAM study criteria (need of intensive care unit admission, invasive ventilation or death) after 24 hours of admission were noted. Results: The mean age of the study population was 56.3±14.8 years. Patients that developed critical illness (as per COVID GRAM study criteria) beyond 24 hours after admission had higher COVID GRAM scores at admission versus those that did not (183.2±80.7 versus 130.3±42.6). The Area under the Receiver Operator Curve for the COVID gram score to predict critical illness in the study population was 0.802 (95% confidence interval, 0.753-0.850). On binary logistic multivariable regression analysis, the COVID GRAM and SOFA scores on admission and need of ICU admission during hospitalization were significant predictors of mortality 24 hours after admission. Conclusion: The COVID GRAM score is a useful risk assessment tool and can be used for appropriate allocation and prioritization of resources where they are most needed.