Study population and setting
Data on 372 patients classified with non-severe COVID-19 at admission from January 20, 2020 to March 2, 2020 were collected retrospectively from two hospitals in Guangzhou, China and one hospital in Wuhan, China. Patients were followed for at least 15 days after admission. Patients were divided into a training cohort (n=189, median age 49, 53% female) and two validation cohorts of 165 and 18 subjects. 72 patients overall progressed to severe disease.
Summary of Main Findings
Authors developed a nomogram, which is a prediction tool to predict progression to severe COVID-19 disease, based on seven patient covariates (age and six laboratory values) chosen via LASSO regression. The nomogram performed well in both training cohort and validation cohorts (validation sensitivity: 78%; validation specificity: 78%).
Nomograms are intuitive and amenable to utilization by clinicians. The statistical approach in this study is straightforward and the model performed well on the validation cohorts.
The study population is small: the training cohort n=189, and one of the validation cohorts consisted of only 18 patients. Results may not be generalizable, since they are taken from two cities in China during the early phase of the pandemic. Biomarker concentrations and their relationship to clinical progression may depend on unique features of the study population, including unmeasured comorbidities. Furthermore, the composite outcome of severe disease includes tachypnea which can result from a number of comorbidities and does not include severe dispositions such as ICU admission or death. This study is not yet published and has not undergone peer-review.
The study is one of the first efforts to construct a prediction tool for severe COVID-19.