Frailty index trajectories and health behaviour transitions among middle-aged and older adults with cardiovascular disease: a longitudinal CHARLS study

Frailty is common in adults with cardiovascular disease (CVD). It is also linked to worse outcomes. Less is known about Frailty Index (FI) trajectories in CVD populations. The relation between these trajectories and changes in health behaviours remains unclear. This study identified FI trajectory subgroups in middle-aged and older adults with CVD. It also examined whether health-behaviour transitions were associated with unfavorable FI trajectories and FI accumulation. We used four waves of the China Health and Retirement Longitudinal Study (CHARLS, 2011–2018). Participants were adults aged at least 45 years with baseline CVD. CVD was defined as self-reported physician-diagnosed heart disease or stroke. A 26-item FI was built using the deficit-accumulation approach. Behavioural exposure variables and CVD-defining conditions were excluded. Latent class mixed models (LCMM) identified FI trajectory subgroups. Wave 1 to Wave 4 health-behaviour transitions were classified as persistently favorable, improved, deteriorated, or persistently unfavorable. Logistic regression and linear mixed-effects models tested the associations. Sensitivity analyses examined temporal ordering, behaviour scores, trajectory outcomes, posterior probabilities, IPW, and multiple imputation. The baseline CVD cohort included 2,427 participants. A total of 2,077 participants contributed 7,257 repeated FI observations. Three trajectory subgroups were identified. They were low-level slowly increasing ( n = 685, 33.0%), moderate-level increasing ( n = 1,180, 56.8%), and high-level rapidly increasing ( n = 212, 10.2%). In the fully adjusted model, deteriorated behaviour was associated with unfavorable FI trajectory membership. The OR was 1.92 (95% CI: 1.22–3.04; P = 0.005). Persistently unfavorable behaviour showed a similar association. The OR was 1.78 (95% CI: 1.16–2.73; P = 0.008). FI increased over time (β = 0.0154 per year; P < 0.001). The persistently favorable group accumulated FI more slowly than the deteriorated group. The interaction was β = −0.0025 ( P = 0.004). In lagged modelling, each additional current-wave favorable behaviour predicted lower next-wave FI. The estimate was β = −0.0047 (95% CI: −0.0074 to − 0.0020; P < 0.001). IPW and multiple-imputation analyses gave similar directions. Among middle-aged and older Chinese adults with CVD, FI progression followed three exploratory trajectories. Deteriorated and persistently unfavorable behaviour patterns were associated with unfavorable FI trajectory membership. Persistently favorable behaviour was associated with slower FI accumulation than behavioural deterioration. Temporally stricter and probability-based analyses supported caution. Longitudinal behaviour patterns may help identify concurrent frailty risk. These observational data cannot show causal effects of behaviour change on FI trajectories.