Sex-specific multimorbidity clusters and all-cause mortality among community-dwelling older adults: findings from the ASPREE cohort

Introduction

Multimorbidity, commonly defined as the coexistence of 2 or more chronic conditions, is becoming increasingly prevalent with the aging population, posing a major challenge for public health and healthcare systems.^1,2 The prevalence of multimorbidity rises sharply with age, affecting approximately 65% of adults aged 65 years or older and 80% of those aged 85 years or older.² This growing burden is associated with poorer quality of life, functional decline, psychological distress, greater healthcare use, polypharmacy, and increased mortality.³ However, its management remains complex because clinical guidelines and healthcare delivery models often focus on single diseases, making it difficult to address the needs of older adults with diverse combinations of coexisting chronic conditions.^3,4 Multimorbidity is often measured using disease count, which provides an intuitive summary of overall disease burden.⁵ Beyond disease count, cluster-based approaches can clarify the nonrandom co-occurrence of chronic conditions and identify groups with distinct condition distributions.^6–8 Clusters may be defined using predetermined clinical groupings or identified through data-driven approaches.^7,9 Examining their associations with subsequent health outcomes can establish their prognostic relevance and provide a basis for future research evaluating potential clinical applications.^10,11

Existing studies of multimorbidity clusters show substantial heterogeneity.^6,8 Nevertheless, large systematic reviews have consistently identified cardiometabolic or cardiovascular-related clusters, often mixing chronic conditions such as hypertension and diabetes, with major health events like ischemic heart disease, heart failure, and stroke.^6,8 Less is known about multimorbidity clustering in community-dwelling older adults without major clinical cardiovascular disease, despite a large proportion of the population living to older age in relatively good health.^12–14 Characterizing multimorbidity in this relatively healthy population may provide insight into patterns of co-occurring conditions which can be different from those among individuals with established cardiovascular disease. Moreover, among the few sex-stratified analyses conducted to date, most have been conducted in populations from Europe, Asia, and North America, and their largely cross-sectional designs have limited the examination of associations between multimorbidity clusters and subsequent health outcomes.⁸ Using data from a large cohort of relatively healthy community-dwelling adults aged 70 years or older in Australia, this study aimed to characterize sex-specific multimorbidity clusters and their associations with all-cause mortality.

Methods

Results

Discussion

In this large cohort of relatively healthy community-dwelling older adults aged 70 years and over, we applied LCA to identify probabilistic patterns of co-occurring baseline health conditions and examined their associations with mortality.^8,21 We found that multimorbidity was highly prevalent, and the composition and mortality associations of the identified patterns varied descriptively between males and females. These sex-stratified patterns may provide prognostic information beyond the individual-level number of chronic conditions.

A key difference between our study and many previous multimorbidity studies is that the ASPREE participants were free of major cardiovascular disease at enrollment, including stroke, coronary artery disease, heart failure, and peripheral artery disease.¹⁷ Although cardiovascular disease remains a major cause of mortality in later life, declining age-standardized cardiovascular mortality means that an increasing number of people are reaching older age without established clinical cardiovascular disease. Previous studies have identified cardiovascular-dominant multimorbidity clusters and their strong associations with mortality.^7–9 When established cardiovascular diseases are included in clustering analyses, they may strongly influence cluster formation and limit the characterization of less prominent clusters involving other health conditions.^22,23 The ASPREE eligibility criteria therefore provided an opportunity to characterize multimorbidity clusters and their associations with mortality among relatively healthy older adults without established major cardiovascular disease.

The dominant clusters identified in the overall population were broadly consistent with previously reported multimorbidity clusters, including metabolic, mental health, and frailty- or physical function-related clusters.^{6,8,21,22,24,25} In the overall analysis, 5 clusters were identified, which differed in the relative representation of the included conditions. Cluster labels were used as concise summaries of relatively prominent or over-represented conditions and were not intended as exhaustive descriptions or discrete clinical phenotypes. The sex-stratified analyses further characterized the multimorbidity patterns identified separately within males and females. Consistent with previous studies, we observed clusters characterized by hypertension and dyslipidemia and by metabolic conditions, including diabetes and obesity.^8,23 A cluster characterized by depressive symptoms, osteoporosis, and frailty was identified in the overall population and in both sex-stratified analyses, although the relative representation of its component conditions differed between males and females. Gout and kidney disease were also grouped within clusters identified separately among males and females. Among males, these conditions occurred within a broader metabolic cluster that included diabetes and obesity, whereas among females, they clustered with anemia and diabetes. These findings may partly reflect sex-related variation in the prevalence and co-occurrence of metabolic, renal, and hematological conditions.²⁶

Distinct multimorbidity clusters exhibited different associations with all-cause mortality in the overall and sex-stratified analyses. In the overall population, 3 clusters showed similarly strong associations with higher mortality compared with the least distinctive reference cluster, whereas the hypotension, thyroid disorder, and cancer history cluster showed a more modest association. This may partly reflect the relatively stable or well-controlled nature of these conditions in this relatively healthy older population, as well as the presence of cancer history and thyroid disorder across multiple clusters, which may have attenuated the mortality contrast between clusters.^27,28 Sex-stratified analyses provided further insight into cluster-specific mortality associations. In males, both non-reference clusters were associated with higher mortality, with a numerically stronger association for the metabolic, gout, and kidney disease cluster. In females, the metabolic cluster showed a more modest association, potentially reflecting the complex relationship between BMI and mortality in older adults, including reverse causation, illness-related weight loss, selective survival, and differences in body composition.²⁹ The depressive symptoms, osteoporosis, and frailty cluster was associated with higher mortality in both the male and female analyses.²⁵ Its composition and association with mortality may be shaped by age, hormonal changes, medication use, and the combined psychological, skeletal, and physical vulnerability represented by these conditions.^30–33 Overall, these findings highlight the value of sex-stratified analyses for characterizing cluster–outcome associations separately within males and females, which may be less apparent in pooled analyses.⁷ The observed associations may be shaped by biological, behavioral, healthcare-related, and survival-related factors.^26,34–36

The clusters identified by LCA represent statistical groupings of co-occurring conditions rather than distinct biological entities. These groupings may be interpreted using a framework similar to that applied to metabolic syndrome, in which related abnormalities co-occur and jointly contribute to cardiovascular risk.^37,38 From this perspective, the observed clusters may provide clues to potential interrelationships among conditions and cumulative physiological dysregulation. The clustering of kidney disease with anemia may reflect their established clinical relationship, as impaired kidney function can contribute to anemia through reduced erythropoietin production, inflammation, and disturbances in iron metabolism.^8,39–42 The depressive symptoms, osteoporosis, and frailty cluster may represent the convergence of mental health burden and physical vulnerability. From a clinical practice perspective, frailty present at baseline warrants assessment and management as a clinically relevant condition that can influence treatment decisions and care needs. When combined with depression and other somatic conditions, frailty may contribute to poorer outcomes, including mortality.³¹ The cluster characterized by hypotension, thyroid disorder, and cancer history has been less commonly reported in previous studies. Although hypotension is less frequently included in clustering studies, among older adults it may reflect impaired hemodynamic regulation, medication effects, or underlying disease burden, while thyroid disorder and cancer history may add further clinical complexity.^43–45 Taken together, this cluster may represent a heterogeneous grouping of health conditions rather than a distinct clinical entity. Further research is needed to validate the reproducibility and prospective predictive value of these clusters before they can be considered for clinical screening or intervention.

Beyond their associations with mortality, multimorbidity clusters may have implications for research and, with further validation, clinical care. Identifying frequently co-occurring conditions can generate hypotheses about shared disease mechanisms, common risk pathways, and the natural history of multimorbidity.²³ Clinically, validated cluster information could help identify coexisting conditions that might otherwise be overlooked and support more person-centered assessment of treatment goals, medication burden, functional status, and patient priorities.^10,21,36 If shown to be reproducible and prospectively predictive, these clusters may eventually inform screening approaches, intervention strategies, clinical guidelines, and trial designs that better reflect the complexity of multimorbidity in older adults.³⁶ The clusters identified separately within males and females also suggest that future validation should consider sex-stratified analyses.

This study has several strengths. The large, community-dwelling ASPREE cohort and long follow-up enabled robust examination of multimorbidity clusters and their associations with mortality among adults aged 70 years and over. The ASPREE eligibility criteria also enabled examination of these associations among relatively healthy individuals who had reached later life without established major cardiovascular disease. The use of LCA supported the data-driven identification of probabilistic clusters, while analyses conducted separately within males and females characterized the clusters and their mortality associations within each sex.

Several limitations should be noted. First, the selected ASPREE cohort comprised relatively healthy older adults, limiting generalizability to those with established cardiovascular disease, dementia, major physical disability, or greater clinical burden. Second, only 15 indicators were included, and alternative indicators may have resulted in a different number or composition of clusters. Third, disease severity was not captured, and class assignment showed some uncertainty and overlap; the clusters should therefore be interpreted as probabilistic groupings rather than discrete clinical phenotypes. Fourth, clusters were based only on baseline conditions and did not capture incident conditions, disease progression, treatment changes, or transitions between clusters over time. Finally, residual confounding from unmeasured factors may have influenced the observed associations.

In conclusion, distinct multimorbidity clusters were identified among relatively healthy older adults, both overall and separately in males and females. Several clusters, including those characterized by frailty, depressive symptoms, osteoporosis, or kidney disease, were associated with higher all-cause mortality. These findings are hypothesis-generating and support further investigation of the relationship between multimorbidity clustering and mortality in older adults.

Supplementary Material

Supplementary material

Supplementary material is available at The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences online.

Funding

This work was supported by the National Institute on Aging and the National Cancer Institute at the National Institutes of Health (U01AG029824 and U19AG062682); the National Health and Medical Research Council (NHMRC) of Australia (334047 and 1127060); Monash University (Australia) and the Victorian Cancer Agency (Australia). Rong Du is supported by an Australian Government Research Training Program (RTP) Scholarship and a Monash International Tuition Scholarship (MITS). Aung Zaw Zaw Phyo is supported by the National Heart Foundation of Australia—Postdoctoral Fellowship (110205-2025_PDF). Joanne Ryan is supported by a National Health and Medical Research Council Research Leadership Investigator Grant (2016438). The sponsors were not involved in the design, methods, subject recruitment, data collection, analysis, or preparation of the paper.

Conflicts of interest

K.S. reports grants from the American Association of Retired Persons and Optum Labs, National Institute on Ageing, and Agency for Healthcare Research and Quality, as well as honoraria from the International Antiviral Society USA. J.R. reports funding from the National Institute of Health, the National Health and Medical Research Council, the Medical Research Future Fund, and the Australian Research Council, honoraria from the Victorian Coronial Council–Department of Justice and Community Service, and serving as a non-paid committee member and a scientific board member for Dementia Australia. No other competing interests were reported.

Data availability

The data used in this study are not publicly available owing to participant privacy and ethical restrictions. Information regarding requests for access to ASPREE data is available through the ASPREE website.

Author contributions

Rong Du (Conceptualization [equal], Formal analysis [lead], Methodology [equal], Visualization [equal], Writing—original draft [equal], Writing—review editing [equal]), Swarna Vishwanath (Conceptualization [equal], Formal analysis [equal], Methodology [equal], Supervision [equal], Validation [equal], Writing—review editing [equal]), Zimu Wu (Conceptualization [equal], Methodology [equal], Writing—review editing [equal]), Aung Zaw Zaw Phyo (Conceptualization [equal], Methodology [equal], Writing—review editing [equal]), Suzanne G. Orchard (Conceptualization [equal], Methodology [equal], Writing—review editing [equal]), Kerry Sheets (Conceptualization [equal], Methodology [equal], Writing—review editing [equal]), Michael E. Ernst (Conceptualization [equal], Methodology [equal], Writing—review editing [equal]), and Joanne Ryan (Conceptualization [equal], Funding acquisition [equal], Methodology [equal], Supervision [equal], Validation [equal], Writing—review editing [equal])