Study on frailty trajectories and their association with mortality risk in elderly hemodialysis patients: an analysis based on a multi-state Markov model

Abstract

Background Frailty is common in older adults receiving maintenance hemodialysis (MHD) and predicts death, yet it changes in both directions over time. Most studies treat frailty as a single baseline or time-updated covariate and do not jointly estimate worsening, recovery, and mortality. Objectives To describe frailty-state transitions in elderly MHD patients and to quantify their association with all-cause mortality using a continuous-time multi-state Markov model. Methods We enrolled 426 patients aged ≥65 years on thrice-weekly in-center MHD. The Fried phenotype was assessed at baseline and at scheduled visits (median 3 assessments). States were robust (0 criteria), prefrail (1–2), and frail (≥3); death was absorbing. A time-homogeneous multi-state model with interval-censored living-state transitions was fitted. Covariate effects were expressed as intensity ratios (IRs). Secondary analyses used time-updated Cox models and two-wave trajectory groups. Results At baseline, 18.5% of patients were robust, 39.4% prefrail, and 42.0% frail. Over a median 28.4 months (912 person-years), 168 patients (39.4%) died. Among consecutive visit pairs, 26.5% showed worsening and 19.1% showed improvement. Mean sojourn times were 14.8, 11.3, and 13.1 months in the robust, prefrail, and frail states, respectively. Three-year probabilities of death were 0.22 from robust,0.36 from prefrail, and 0.54 from frail.Relative to robust occupan cy, death intensities were higher from prefrail (IR 2.41, 95% CI 1.09–5.33) and frail (IR 5.82, 2.71–12.51) states. Older age and higher Charlson index increased intensities of deterioration and death; higher albumin favored recovery from frailty and lowered the frail-to-death intensity.Patients who remained frail or worsened had substantially higher Cox hazards than those who remained robust; improvement did not differ significantly from stable robust status. Conclusions In elderly MHD patients, frailty is dynamic but strongly prognostic. Persistent or incident frailty confers the highest intensity of death, whereas recovery is associated with a more favorable outlook. Serial frailty assessment and interventions aimed at preventing deterioration and supporting recovery are warranted.