Triglyceride–Glucose Index as a Predictor of Sepsis in Urinary Tract Infection: Insights From a Machine Learning‐Driven Cohort Study

Background Early identification of sepsis in ICU patients with urinary tract infection (UTI) may improve outcomes. The triglyceride–glucose (TyG) index has been implicated in systemic inflammation; however, its role in UTI‐related sepsis is unclear. Therefore, this study aimed to investigate the predictive performance of TyG for risk prediction of sepsis in patients with UTIs. Methods In this retrospective study, data were analyzed from adult ICU patients with UTI from the MIMIC‐IV database (2008–2019). Outcomes included sepsis incidence and 28‐day mortality. Baseline characteristics, lab values, and severity scores (SOFA, APSIII, SAPSII, OASIS, and SIRS) were collected. Logistic regression, random forest, and XGBoost models were used to predict sepsis. Results Of 996 patients, 561 (56.3%) developed sepsis. Median TyG was significantly higher in sepsis cases (9.0 vs. 8.8; p < 0.001). Sepsis incidence rose from 51.3% in the lowest TyG quartile to 85.5% in the highest ( p < 0.0001), with a clear dose–response pattern. Elevated TyG levels showed association with in‐hospital mortality in all models (fully adjusted HR: 1.19; 95% CI: 1.11–1.29; p = 0.015). Patients in Q4 had over twice the risk of death compared to Q1 (HR: 2.04; p < 0.001). Subgroup analyses showed no significant interactions with age, sex, BMI, or comorbidities. Adding TyG significantly improved AUCs for SAPSII, OASIS, and SIRS. Adding TyG significantly improved AUCs for SAPSII, OASIS, and SIRS. Among models, XGBoost had the best performance (AUC: 0.90; PR‐AUC: 0.95). Conclusion An elevated TyG index is significantly associated with sepsis and mortality in ICU patients with UTI. TyG enhances early risk stratification and may serve as an effective option for high‐risk patients.