The retrospective analyses on 499 patients hospitalized for COVID-19, the GASS score was associated with both in-hospital death and 30-day death (p<0.001). Dividing the population into 3 categories, according to the GASS score, the association with the COVID-19 outcomes remained (p<0.001 for both variables). The Area Under Curve (AUC) generated by the GASS score towards 30-day death was 0.84, predicting this outcome more accurately than the Charlson Comorbidity Index (CCI, 0.74) and similarly to a LASSO-generated predictive model fitted on 151 variables considered (AUC=0.83).
Bibliography:
Greco, A. Salatiello, N. Fabbri, A. Passaro. "Early Prediction of COVID-19 Outcome: Contrasting Clinical Scores and Computational Intelligence Methods". Book title "Understanding Covid-19: The Role of Computational Intelligence" (Book Chapter).