Although better diet quality is inversely associated with mortality risk, the association between diet quality and mortality remains unclear in frail and non-frail older adults. Thus, we aimed to examine this association in older Japanese adults.

A prospective cohort study.

We used the data of 8,051 Japanese older adults aged ≥65 years in the Kyoto-Kameoka study

Dietary intake was estimated using a validated food frequency questionnaire. Diet quality was evaluated by calculating the adherence scores to the Japanese Food Guide Spinning Top (range, 0 [worst] to 80 [best]), which were stratified into quartiles. Frailty status was assessed using the validated self-administered Kihon Checklist (KCL) and the Fried phenotype (FP) model. Survival data were collected between February 15, 2012 and November 30, 2016. Statistical analysis was performed using the multivariate Cox proportional hazard analysis and the spline model.

During the median 4.75-year follow-up (36,552 person-years), we recorded 661 deaths. After adjusting for confounders, compared with the bottom adherence score quartile, the top quartile was associated with lower hazard ratio (HR) of mortality in frailty (HR, 0.73; 95% confidence interval [CI], 0.54–1.00) and non-frailty, as defined by the KCL (HR, 0.72; 95% CI, 0.52–1.01). In the spline model, regardless of frailty status defined by the KCL and FP model, adherence score showed a strongly dose-dependent inverse association with mortality up to approximately 55 points; however, no significant differences were observed thereafter. This association was similar to the results obtained in individuals with physical, cognitive, and depression as domains of KCL in the spline model.

Our findings demonstrate an L-shaped association between diet quality and mortality in both frail and non-frail individuals. This study may provide important knowledge for improving poor diet quality in older individuals with frailty or domains of frailty.