D93-05 Sepsis Subphenotypes Have Distinct Premorbid Medical Disease Profiles With Increased Long-term Morbidity Burden and Late Mortality in Survivors

Abstract Background Latent class analysis has revealed two molecular subphenotypes of sepsis (Hyperinflammatory and Hypoinflammatory) with markedly divergent outcomes and responses to therapeutic interventions. While extrinsic factors including infection with specific pathogens can influence which subphenotype is manifested, it remains uncertain how intrinsic host features such as premorbid medical disease burden also impact manifested subphenotype, or if long-term health outcomes differ between subphenotypes. To address this, we used longitudinal electronic health records (EHR) to investigate the premorbid host medical disease burden and late consequences of sepsis subphenotypes. Methods Retrospective observational cohort study of adults admitted with sepsis to any Vanderbilt ICU from 2004-2024. Sepsis was identified using the Rhee EHR algorithm (PMC5710396). Subphenotypes were ascertained with a validated Clinical Classifier Model using vital signs and clinical laboratory data (PMC10841178). We used phenome-wide association study (PheWAS) to identify premorbid diagnoses associated with each subphenotype and new outpatient diagnoses arising among survivors after discharge. Diagnostic billing codes were mapped to curated PheWAS codes (phecodes) capturing 1,861 clinically meaningful medical diagnoses. The association between the sepsis subphenotypes and individual phecodes in each patient’s EHR was assessed by logistic regression, adjusting for standard PheWAS covariates age, sex, race/ethnicity, and EHR record duration. Post-discharge all-cause mortality in survivors was assessed using Cox proportional hazards regression. We validated our premorbid analyses using the MIMIC-IV ICU dataset (2008-2022). Results We included 30,626 adults admitted with sepsis to a Vanderbilt ICU and 11,695 in MIMIC-IV. Demographics and subphenotype distributions were similar between cohorts. Using a Bonferroni-adjusted cutoff (Z-Score:4.09, p 4.24 × 10-5), PheWAS identified 105 premorbid medical diagnoses associated with increased risk for Hyperinflammatory sepsis (Figure-A) including cirrhosis (Z-Score:22.47), chronic renal disease (Z-Score:13.28), leukemias (Z-Score:11.67), and hypertension (Z-Score:5.51). Only 10 premorbid diagnoses were associated with Hypoinflammatory sepsis including chronic airway obstruction (Z-Score:6.36) and oropharyngeal cancers (Z-Score:4.78). Post-survival PheWAS (Figure-B) revealed increased risk for 10 new diagnoses in Hyperinflammatory survivors including valvular heart disease (Z-Score:5.82) immunodeficiency (Z-Score:4.67); dyschromia/vitiligo (Z-Score:4.66); and surgical complications (Z-Score:4.37). Hyperinflammatory sepsis survivors also had worse all-cause mortality in long-term EHR follow up (Hazard ratio: 1.32, 95% CI: 1.23-1.42, p 0.001, Figure-C). Analysis using MIMIC-IV replicated 65 premorbid associations observed in Vanderbilt. Conclusions Hyperinflammatory sepsis was associated with a spectrum of premorbid disease burden, new medical problems in survivors, and increased all-cause mortality after discharge. Elucidating these associations’ biological drivers and the interplay between medical disease burden with extrinsic factors may reveal new opportunities to improve longitudinal care of sepsis survivors. This abstract is funded by: K01 HL157755 (VEK); R01 HL171809, R01 AG069900, R01 AG084550 (WQW); R35 GM145330 (MMC); R01HL173531, R35GM142992 (PS); U01 HL168412, R01 HL164937 (LBW); UL1 TR002243 (VUMC)