Fig. 1

Demonstration of the overall study design. (A) The number of patients and time points collected and excluded during data preprocessing across the three cohorts. (B) The conceptual depiction of CovSF predicting one’s severity level (green box) based on the given clinical input (blue box). (C) CovSF scores, feature values, and feature importance for deteriorating and recovering patients in the KNIH cohort. The grey area beside the CovSF score represents the output variability (i.e. standard deviation) arising from multiple predictions on the same day. Feature values closer to severe than the baseline are shaded in red, and those closer to the mild are shaded in blue. Feature importance (F.i.) is shown by the red line.