Development and validation of a predictive model to determine the level of care in patients confirmed with COVID-19

Background
The COVID-19 pandemic has imposed significant challenges on hospital capacity. While mitigating unnecessary crowding in hospitals is favorable to reduce viral transmission, it is more important to prevent readmissions with impaired clinical status due to initially inappropriate level of care. A validated predictive tool to assist clinical decisions for patient triage and facilitate remote stratification is of critical importance.
Methods
We conducted a retrospective study in patients with confirmed COVID-19 stratified into two levels of care, namely ambulatory care and hospitalization. Data on socio-demographics, clinical symptoms, and comorbidities was collected during the first (N=571) and second waves (N=174) of the pandemic in Belgium (March 2 to December 6, 2020). Univariate and multivariate logistic regressions were performed to build and validate the prediction model.
Results
Significant predictors of hospitalization were old age (OR=1.08, 95%CI:1.06-1.10), male gender (OR=4.41, 95%CI: 2.58-7.52), dyspnea (OR 6.11, 95%CI: 3.58-10.45), dry cough (OR 2.89, 95%CI: 1.54-5.41), wet cough (OR 4.62, 95%CI: 1.93-11.06), hypertension (OR 2.20, 95%CI: 1.17-4.16) and renal failure (OR 5.39, 95%CI: 1.00-29.00). Rhinorrhea (OR 0.43, 95%CI: 0.24-0.79) and headache (OR 0.36, 95%CI: 0.20-0.65) were negatively associated with hospitalization. A receiver operating characteristic (ROC) curve was constructed and the area under the ROC-curve was 0.931 (95% CI: 0.910-0.953) for the prediction model (first wave) and 0.895 (95% CI: 0.833-0.957) for the validated data set (second wave).
Conclusion
With a good discriminating power, the prediction model might identify patients who require ambulatory care or hospitalization, and support clinical decisions by Emergency Department staff and general practitioners.