Development and internal validation of a diagnostic prediction model for COVID-19 at time of admission to hospital

Abstract

Summary Background Early coronavirus disease 2019 (COVID-19) diagnosis prior to laboratory testing results is crucial for infection control in hospitals. Models exist predicting COVID-19 diagnosis, but significant concerns exist regarding methodology and generalizability. Aim To generate the first COVID-19 diagnosis risk score for use at the time of hospital admission using the TRIPOD (transparent reporting of a multivariable prediction model for individual prognosis or diagnosis) checklist. Design A multivariable diagnostic prediction model for COVID-19 using the TRIPOD checklist applied to a large single-centre retrospective observational study of patients with suspected COVID-19. Methods 581 individuals were admitted with suspected COVID-19; the majority had laboratory-confirmed COVID-19 (420/581, 72.2%). Retrospective collection was performed of electronic clinical records and pathology data. Results The final multivariable model demonstrated AUC 0.8535 (95% confidence interval 0.8121–0.8950). The final model used six clinical variables that are routinely available in most low and high-resource settings. Using a cut-off of 2, the derived risk score has a sensitivity of 78.1% and specificity of 86.8%. At COVID-19 prevalence of 10% the model has a negative predictive value (NPV) of 96.5%. Conclusions Our risk score is intended for diagnosis of COVID-19 in individuals admitted to hospital with suspected COVID-19. The score is the first developed for COVID-19 diagnosis using the TRIPOD checklist. It may be effective as a tool to rule out COVID-19 and function at different pandemic phases of variable COVID-19 prevalence. The simple score could be used by any healthcare worker to support hospital infection control prior to laboratory testing results.