Developed a continuous sepsis severity index using 43 routinely charted variables over a 72-hour treatment window from two hospital systems (total 36,807 patients). Unlike previous work, they used mortality as a treatment-level ranking signal rather than a per-state target, allowing non-uniform credit redistribution across timesteps.
Non-survivors scored 1.19-1.64 points higher on a 0-10 scale within all SOFA-2 strata. Within-patient changes correlated with lactate (Spearman rho=0.39) and more weakly with MAP and creatinine. Cross-institutional agreement was 70-77% of same-site correlation.
The index provides hourly prognostic information that separates patient outcomes and aligns with clinical expectations, suggesting potential as a decision support tool complementing clinical judgment.