Use of serum albumin and activities of daily living to predict mortality in long-lived individuals over 95 years of age: a population-based study.

Both activities of daily living (ADL) and some blood biomarkers (such as albumin) have been associated with mortality in very elderly people, but scarce data is available on the predictive performance of them in isolation or in combination, which is important for clinicians in decision making. Here, based on prospective mortality data over a 6-year follow-up period from 433 long-lived individuals (LLIs) aged 95+ years in the Rugao longevity cohort, we aimed to evaluate Cox proportional hazard ratios (HRs) and discriminative power (ROC curve) of 14 biomarkers and ADL for all-cause mortality. We found that six biomarkers (total triglyceride, albumin, low-density lipoprotein cholesterol, platelet count, lymphocyte count, and neutrophil count) were associated with mortality with a p < .10 in the univariate model. Significant associations of albumin and neutrophil count with mortality were observed when they were simultaneously included in a multivariate model, with HRs of 0.97 (95 % CI 0.94, 0.99; p = .005) and 1.09 (95 % CI 1.00, 1.18; p = .043). With respect to ADL, the corresponding HR was 1.10 (95 % CI 1.07, 1.14; p < .001). Low albumin (<40 g/L) combined with ADL dependent had a significantly increased mortality risk (HR = 2.19; 95 % CI 1.63, 2.95). Albumin and ADL separately showed good discriminative accuracies (area under the curve [AUC] = 0.68 and 0.66, respectively), and their combination had an increased predictive utility (AUC = 0.73). In conclusion, both albumin and ADL are efficient predictors of all-cause mortality in long-lived populations and their combination further increases discriminative power. The preliminary findings, if validated and translated, would help clinicians to identify the elderly people at varying mortality risk.