Development and validation of a prediction model for early identification of critically ill elderly COVID-19 patients

In this study, we established a simple and practical tool for early identification of potentially high-risk individuals among elderly COVID-19 patients. Included were 2106 laboratory-confirmed COVID-19 patients aged 60 years and above in 30 provinces of mainland China. Using discrimination (the area under the receiver-operator characteristic curve [AUC]) and calibration (Hosmer-Lemeshow goodness-of-fit test and calibration plots), a nomogram for predicting critically ill cases was developed, and its performance was examined using an internal validation cohort (444 patients) and external cohort (770 patients). The proportion of critically ill patients was 11.8% (248/2106). The most common symptoms at the onset of illness were fever (66.6%), cough (34.1%), fatigue (23.3%), and expectoration (23.6%). Older age, history of chronic obstructive pulmonary disease, fever, fatigue, shortness of breath, and lymphocyte percentage lower than 20% at admission were associated with increased risk of becoming critically ill. The AUCs for the six-variable-based nomogram were 0.77 (95% CI: 0.73-0.82), 0.73 (95% CI: 0.67-0.79), and 0.77 (95% CI: 0.71-0.83) in the development, internal validation, and external validation cohorts, respectively. This six-variable-based nomogram could potentially serve as a practical and reliable tool for early identification of elderly COVID-19 patients at high risk of becoming critically ill.