A multistate model of health transitions in older people: a secondary analysis of ASPREE clinical trial data

Abstract

Background: Understanding the nature of transitions from a healthy state to chronic diseases and death is important for planning health-care system requirements and interventions. We aimed to quantify the trajectories of disease and disability in a population of healthy older people. Methods: We conducted a secondary analysis of data from the ASPREE trial, which was done in 50 sites in Australia and the USA and recruited community-dwelling, healthy individuals who were aged 70 years or older (≥65 years for Black and Hispanic people in the USA) between March 10, 2010, and Dec 24, 2014. Participants were followed up with annual face-to-face visits, biennial assessments of cognitive function, and biannual visits for physical function until death or June 12, 2017, whichever occurred first. We used multistate models to examine transitions from a healthy state to first intermediate disease events (ie, cancer events, stroke events, cardiac events, and physical disability or dementia) and, ultimately, to death. We also examined the effects of age and sex on transition rates using Cox proportional hazards regression models. Findings: 19 114 participants with a median age of 74·0 years (IQR 71·6–77·7) were included in our analyses. During a median follow-up of 4·7 years (IQR 3·6–5·7), 1933 (10·1%) of 19 114 participants had an incident cancer event, 487 (2·5%) had an incident cardiac event, 398 (2·1%) had an incident stroke event, 924 (4·8%) developed persistent physical disability or dementia, and 1052 (5·5%) died. 15 398 (80·6%) individuals did not have any of these events during follow-up. The highest proportion of deaths followed incident cancer (501 [47·6%] of 1052) and 129 (12·3%) participants transitioned from disability or dementia to death. Among 12 postulated transitions, transitions from the intermediate states to death had much higher rates than transitions from a healthy state to death. The progression rates to death were 158 events per 1000 person-years (95% CI 144–172) from cancer, 112 events per 1000 person-years (86–145) from stroke, 88 events per 1000 person-years (68–111) from cardiac disease, 69 events per 1000 person-years (58–82) from disability or dementia, and four events per 1000 person-years (4–5) from a healthy state. Age was significantly associated with an accelerated rate for most transitions. Male sex (vs female sex) was significantly associated with an accelerate rate for five of 12 transitions. Interpretation: We describe a multistate model in a healthy older population in whom the most common transition was from a healthy state to cancer. Our findings provide unique insights into the frequency of events, their transition rates, and the impact of age and sex. These results have implications for preventive health interventions and planning for appropriate levels of residential care in healthy ageing populations. Funding: The National Institutes of Health.