Seven routine blood markers predict five-year mortality in older Indians: a risk score from the LASI-DAD cohort

Validated, blood-based tools to estimate short-to-medium-term mortality in older adults are derived almost entirely from European-ancestry or East Asian populations and transport poorly to South Asians, who present a distinct cardiometabolic and body-composition phenotype. No mortality risk score using routine blood markers exists for community-dwelling older Indians.

We used Wave-1 data from the Longitudinal Aging Study in India - Diagnostic Assessment of Dementia (LASI-DAD), a nationally representative cohort of adults aged ≥ 60 years with venous biomarker phenotyping. Vital status was ascertained through an End-of-Life informant module to Wave 2. From the available venous panel, we pre-selected seven markers on biological grounds - N-terminal pro-B-type natriuretic peptide (NT-proBNP), serum albumin, cystatin C, high-sensitivity C-reactive protein (hs-CRP), the neutrophil-to-lymphocyte ratio (NLR), glycated haemoglobin (HbA1c) and alanine aminotransferase (ALT) - and fitted a Cox proportional-hazards model with age and sex. Discrimination (Harrell's C), bootstrap optimism correction, calibration at three years, and risk-tertile stratification were assessed; decision-curve analysis quantified clinical utility.

Among 2144 participants (540 deaths over a median 4.4 years of survivor follow-up; 25% mortality), the seven-marker score plus age and sex achieved an apparent C of 0.74 (optimism-corrected 0.74), versus 0.68 for age and sex alone (Δ ≈ 0.06); the blood panel alone reached C 0.72. NT-proBNP (hazard ratio [HR] per 1 SD 1.31), HbA1c (1.24) and albumin (0.81) carried most of the signal, with smaller contributions from cystatin C, ALT, NLR and hs-CRP. Calibration was excellent at three years. Participants in the highest score tertile had 8.7-fold the mortality hazard of the lowest (95% CI 6.5-11.6; log-rank p < 0.001), with three-year mortality of 34% versus 6%. The score provided positive net benefit across clinically relevant decision thresholds. In sensitivity analyses, all associations persisted after adjustment for comorbidity, cognition and function, and the panel out-performed a clinical-plus-cognitive model.

A seven-marker panel obtainable from a single venous sample, combined with age and sex, stratifies five-year mortality in older Indians with good discrimination and calibration. Pending external validation, it offers a clinically translatable, population-specific tool for prognostic risk stratification in a large and under-studied population.

Not applicable (observational cohort analysis).