A novel risk prediction model incorporating triglyceride-glucose-body mass index for in-hospital mortality in hypertensive patients with HFmrEF

BACKGROUND: Patients with heart failure with mildly reduced ejection fraction (HFmrEF) and hypertension represent a high-risk population with limited prognostic tools. The triglyceride-glucose-body mass index (TyG-BMI), a novel metric of insulin resistance and metabolic dysregulation, may enhance risk stratification but has not been evaluated in this specific cohort. METHODS: We retrospectively analyzed 2,550 hospitalized hypertensive patients with HFmrEF from a single tertiary center (May 2012 to October 2023). Participants were randomly assigned to training (n = 1,785) and validation (n = 765) cohorts. Using least absolute shrinkage and selection operator (LASSO) regression and multivariate logistic regression, we developed a nomogram to predict in-hospital mortality. Restricted cubic spline (RCS) analysis was used to characterize the dose-response relation between the TyG-BMI index and mortality risk. Model performance was assessed by area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA). RESULTS: In-hospital mortality occurred in 157 patients (6.16%). Multivariate analysis identified six independent risk factors (age ≥ 75 years [OR 2.79, 95% CI 1.55–5.03], TyG-BMI (per 100-unit increase) [OR 4.47, 95% CI 2.99–6.68], C-reactive protein ≥ 10 mg/L [OR 3.83, 95% CI 1.33–11.03], NT-proBNP elevation [OR 3.86, 95% CI 1.68–8.85], diabetes [OR 2.77, 95% CI 1.64–4.70], and cerebral infarction [OR 4.79, 95% CI 2.81–8.16]) and two factors that were independently associated with a lower risk of in-hospital mortality (ACEI/ARB/ARNI use [OR 0.33, 95% CI 0.20–0.55] and SGLT2i use [OR 0.39, 95% CI 0.18–0.85]). RCS disclosed a non-linear relation between TyG-BMI and in-hospital mortality (P for non-linearity < 0.001), with a steep increase in risk when TyG-BMI was at or above 300. The nomogram demonstrated excellent discrimination in both training (AUC 0.842, 95% CI 0.802–0.882) and validation (AUC 0.844, 95% CI 0.794–0.894) cohorts with good calibration (training P = 0.17; validation P = 0.67) and significant clinical net benefit on DCA. CONCLUSIONS: TyG-BMI displayed a threshold-dependent, non-linear association with in-hospital mortality in hypertensive patients with HFmrEF. A prediction model incorporating TyG-BMI and related variables demonstrated strong discriminative performance, showing potential as a practical tool for early identification of high-risk individuals and supporting risk stratification and clinical vigilance during hospitalization.