Development and Validation of a Novel Risk Score for Predicting Clinical Deterioration in COVID-19 Patients: The ABCD Risk Score

Abstract IntroductionFor the appropriate allocation of medical resources to at-risk COVID-19 patients, a simple risk scoring model for the risk stratification of clinical deterioration in the early stage of symptom onset would be invaluable. Here, we report the development and validation of such a model for clinical deterioration in COVID-19.MethodsThis multicenter retrospective cohort study conducted in Japan included adult patients (≥18 years old) with confirmed COVID-19 according to molecular diagnostic methods and hospitalized in two hospitals. Patients who did not undergo laboratory tests within 10 days of initial symptom onset and those who were treated with oxygen therapy before hospitalization were excluded. Patients were divided into derivation (n=446) and temporal validation (n=305) datasets based on hospitalization period. The primary outcome was need for oxygen therapy within 14 days of hospitalization.ResultsA novel ABCD Risk Score (range, 0–12) comprising age, body mass index, C-reactive protein, and lactate dehydrogenase levels at admission was developed in the derivation dataset. This risk score showed good discrimination for clinical deterioration (concordance statistics, 0.86; 95% confidence interval [CI]: 0.72–0.90) with good calibration (intercept, 0.01; slope, 0.99) in the temporal validation dataset. Three risk groups were defined: low risk (≤3 points), intermediate risk (4–6 points), and high risk (≥7 points). In the validation dataset, the clinical deterioration rates for these three groups were 7.1% (95% CI: 3.1%–13.6%), 32.9% (95% CI: 22.3%–44.9%), and 73.3% (95% CI: 64.5%–81.0%), respectively. The risk score showed better discrimination and calibration performance than four previously reported risk scoring models.ConclusionOur novel ABCD Risk Score can be used for the risk stratification of clinical deterioration in COVID-19 patients at an early stage of symptom onset.