Illness Severity Modifies the Mortality Burden of Sepsis in Critically Ill Adults: Evidence of Severity-Dependent Shock Heterogeneity in 64,703 ICU Admissions

Background: Illness severity scores such as SAPS II implicitly assume that condition-specific mortality contributions — including that of sepsis — remain constant across the severity spectrum. From a shock pathophysiology perspective, distributive shock from sepsis may carry different reversibility profiles than cardiogenic or neurogenic shock at equivalent severity. Whether illness severity modifies the mortality burden of sepsis remains unevaluated. Methods: We conducted a retrospective cohort study of 64,703 ICU admissions from MIMIC-IV (2008–2019). Sepsis was defined by Sepsis-3 criteria; baseline severity by SAPS II with fractional polynomial modeling. Multivariable logistic regression with interaction terms assessed effect modification; average marginal effects (AMEs) were estimated across the SAPS II spectrum. ICU-type interaction heterogeneity was tested. Results: Sepsis occurred in 42.7% of admissions; 30-day mortality was 12.9%. The sepsis–mortality association varied significantly by severity (interaction p<0.001). At SAPS II 30, sepsis was associated with higher mortality (AME +3.8%, 95%CI 3.2–4.3%); this attenuated and became non-significant at SAPS II 50–55. At SAPS II 80, septic patients had lower adjusted mortality (AME −7.3%, 95%CI −9.6 to −5.0%). This crossover was present in unadjusted data (SAPS II ≥70: 61.5% vs 69.2%) and consistent across eight sensitivity analyses. Conclusions: The mortality burden of sepsis is not constant across the severity spectrum: it confers additional risk at low-to-moderate severity, but a crossover pattern emerges with lower mortality at high severity — robust in crude and adjusted analyses and consistent across ICU types. These findings implicate severity-dependent shock reversibility and have implications for prognostication, clinical trial design, and quality benchmarking.