Frailty and Mortality Among Older Americans: A 20-Year Follow-up from the Health and Retirement Study

Background and Purpose: Approximately 9% to 34% of adults in the US are considered frail, with prevalence rising significantly with age. Frailty is central to geriatric social work and healthcare as it represents cumulative physiological and functional vulnerabilities associated with elevated risks of hospitalization, institutionalization, and adverse health outcomes such as multimorbidity and mortality. Understanding frailty and its health impacts thus has implications for preventing premature death and extending healthy life expectancy. However, the current evidence remains mixed due to inconsistent variable selection, varied frailty model construction, diverse age-sampling without focusing on older adults, and inconsistent mortality outcomes. To address these gaps, this study constructs the frailty index using the standardized cumulative deficit model and a model-based principal component analysis served as a sensitivity test, and investigates their predictive power for mortality among adults aged 50+ using a US population-based panel survey.

Methods: We used the 1998-2022 RAND Health and Retirement Study (HRS) longitudinal file and selected participants aged 50+ across all waves, yielding 35,403 respondents and 192,578 person-waves. Frailty was constructed using 35 variables (e.g., health status, chronic conditions, mobility, ADLs, IADLs, sensory impairment, etc.). Two mortality outcomes, all-cause and cardiovascular disease (CVD) mortality, were selected from the HRS Tracker file. Two approaches were used to construct the frailty indices: a standardized method that combined all 35 variables into a continuous cumulative deficit score (FI-CD; range: 0-1) and a PCA-based analysis to reduce the multidimensionality of frailty (FI-PCA); missing values in frailty were handled using multiple imputation (n=5). We used a lagged design to model the influence of frailty (t-1) on mortality (t), controlling for age, gender, race, education, smoking, alcohol use, and wave-time, using discrete-time survival analysis with robust standard errors that account for the complex survey design. For easier interpretation, frailty was divided by 100, making the percentage changes in mortality correspond to a 0.01-unit change in frailty.

Results: A strong correlation (Pearson r=0.984; Spearman ρ=0.990) was observed between FI-CD and FI-PCA. Overall, both frailty measures showed similar distributions (Mean/SD: FI-CD [0.203/0.149]; FI-PCA [0.179/0.153]), and they increase with age over time. The discrete-time survival analysis showed that frailty was significantly associated with 4.0-4.7% increases in all-cause mortality (FI-CD HRs: 1.047; FI-PCA HRs: 1.040) and 3.8-4.6% increases in CVD mortality (FI-CD HRs: 1.046; FI-PCA HRs: 1.038).

Conclusions and Implications: Findings suggest that frailty is predictive of and positively associated with both all-cause and CVD mortality. Our findings have implications for one of the Grand Challenges in Social Work: closing health gaps, particularly through improved frailty measurement and timely health interventions. As frailty reflects a risk continuum and a potentially reversible process that can be modified, we urge further research using varied algorithms to develop an accurate measurement protocol for frailty assessment and detection. Additionally, since frailty integrates multilevel declines in physiological and functional health, health-promoting actions (e.g., healthy diet, exercise and rehabilitation, health monitoring, social care services, and aging-friendly community programs) should be implemented to reduce age-related declines in health and support healthy longevity across the life course.