Temporal risk stratification using inflammation–nutrition indices in adults with diabetes

This study aimed to characterize the temporal prognostic roles of composite inflammation-nutrition indices in mortality risk stratification among adults with diabetes. We conducted a retrospective cohort study in a large population-based sample of adults with diabetes derived from a nationally representative health survey conducted between 1999 and 2018. Seven composite inflammation-nutrition indices were evaluated, including nutritional indices (PNI, GNRI), albumin-based ratios (MAR, NAR, RAR), and integrated scores (ALI and HALP). Associations with all-cause and cardiovascular mortality were assessed using Kaplan-Meier analysis and multivariable Cox regression. Nonlinear relationships were explored using restricted cubic splines, and time-dependent discriminative performance was evaluated across different follow-up periods. Kaplan-Meier analysis showed that higher levels of MAR, NAR, and RAR were significantly associated with increased all-cause mortality, whereas higher ALI, PNI, GNRI, and HALP scores were associated with reduced mortality risk. Multivariable Cox regression analysis indicated that, except for GNRI and HALP, all other indicators were independent predictors of cardiovascular mortality risk; Most evaluated indicators showed significant prognostic value for all-cause mortality, whereas the association for HALP was attenuated and no longer statistically significant after FDR correction. Marked temporal heterogeneity in prognostic performance was observed. RAR demonstrated stronger discrimination for short-term (1-year) mortality, whereas ALI provided more stable prognostic stratification across mid- to long-term follow-up. The overall prognostic importance of ALI was further supported by the Random Survival Forest model, in which ALI consistently ranked among the most influential predictors. This study highlights the complementary, time-dependent prognostic roles of inflammation-nutrition indices in diabetes, underscoring their potential utility for integrated and temporally informed risk assessment.