Longitudinal Stress Hyperglycemia Trajectories and Severe Outcome in Patients Hospitalized for Viral Respiratory Infections: A 72-Hour Phenotyping Study

Background/Objectives: Patients with diabetes face increased risks during viral respiratory infections. While chronic glycemic control (HbA1c) is a known risk factor, it may fail to capture acute metabolic failure. This study aimed to evaluate the stress hyperglycemia ratio (SHR) at two time points (admission and 72 h) as a prognostic marker. Methods: We conducted a longitudinal analysis of 430 patients (211 with diabetes and 219 without diabetes) hospitalized with viral respiratory infections. The SHR was calculated at admission (SHR_z1) and at 72 h (SHR_z3). The primary outcome was severe clinical deterioration (defined as oxygen requirement >2–HFNO/OTI or mortality). Multivariate logistic regression and ROC curve analysis were used to determine independent predictors and clinical cut-off points. Results: In the global multivariable model (N = 430), continuous SHR_z3 emerged as an independent predictor of severe outcome (OR = 1.86, 95% CI 1.09–3.20, p = 0.0240) alongside chronic glycemic control (HbA1c: OR = 1.41, p < 0.0001) and age (OR = 1.05, p = 0.0001). After stratification, the independent SHR_z3 effect attenuated below statistical significance in both the non-diabetic (p = 0.6168) and diabetic (p = 0.1954) sub-cohorts, while HbA1c (p = 0.0404) and age (p = 0.0036) remained independent predictors within the diabetic subgroup. An exploratory longitudinal phenotype model classifying patients into Stable Normal, Resolvers, Persistent, and Late Spikers showed the highest in-sample discrimination of all models tested (AUC = 0.797, 95% CI 0.756–0.834). Within this model, the Persistent phenotype (SHR ≥ 1.1 sustained at admission and at 72 h) showed a non-significant trend toward higher adjusted odds of severe outcome compared with Stable Normal patients (OR = 1.81, 95% CI 0.91–3.60, p = 0.088) and was associated with a significantly longer hospital stay (Kruskal–Wallis p = 0.004 across phenotypes). Conclusions: Persistent stress hyperglycemia at 72 h is an independent predictor of severe outcome in the unstratified cohort after adjustment for chronic glycemic control, age, and other covariates. Within metabolic strata, the SHR_z3 signal attenuates, and long-standing suboptimal glycemic control (HbA1c) and age dominate, particularly among diabetic patients. Longitudinal SHR trajectories may therefore provide complementary, hypothesis-generating prognostic information when combined with chronic baseline markers, although these findings require external validation in independent cohorts before clinical use.