🧠 What Inspired This Study?
Frailty, characterized by age-related declines in physiological systems, poses significant challenges to the well-being of older adults and the sustainability of healthcare systems. While it's well-established that social determinants of health (SDoH)—such as socioeconomic status, social connections, and health behaviors—play crucial roles in influencing frailty, the relative importance of these factors across different countries remains underexplored.
Our motivation stemmed from the need to understand how various SDoH contribute to frailty in diverse sociocultural contexts. By examining data from the USA, England, and China, we aimed to uncover both universal and country-specific determinants of frailty, providing insights that could inform targeted public health interventions.
🔍 Our Approach
We analyzed data from participants aged 45 and above, encompassing:
- USA: 5,792 individuals
- England: 3,773 individuals
- China: 5,016 individuals
Each country's dataset included a comprehensive set of SDoH variables:
- USA: 121 variables
- England: 125 variables
- China: 94 variables
These variables spanned seven domains, including adverse childhood experiences (ACEs), socioeconomic status (SES), material circumstances, social connections, social stressors, health behaviors, and healthcare systems.
To predict frailty at a 4-year follow-up, we employed Extreme Gradient Boosting (XGBoost), a powerful machine learning algorithm known for its predictive accuracy. To interpret the model's outputs and understand the contribution of each SDoH variable, we utilized SHapley Additive exPlanations (SHAP). This approach allowed us to quantify the impact of individual variables and domains on frailty.
🚧 Challenges Encountered
One of the primary challenges was harmonizing data across three countries with distinct cultural, economic, and healthcare landscapes. Ensuring consistency in variable definitions and measurement scales was crucial to maintain the integrity of our cross-national comparisons.
To address this challenge, we grounded our variable classification in the World Health Organization’s framework on SDoH. This provided a conceptual foundation to organize diverse variables into coherent domains. We also drew extensively from prior studies that used the HRS (USA), ELSA (England), and CHARLS (China) datasets to guide variable selection and domain alignment. Key references that informed this process included:
- Liu, Z. et al. Associations of genetics, behaviors, and life course circumstances with a novel aging and healthspan measure: Evidence from the Health and Retirement Study. PLOS Medicine 16, e1002827 (2019)
- Shah, S. J. et al. Social Frailty Index: Development and validation of an index of social attributes predictive of mortality in older adults. Proc. Natl. Acad. Sci. 120, e2209414120 (2023)
- Steptoe, A., Breeze, E., Banks, J. & Nazroo, J. Cohort Profile: The English Longitudinal Study of Ageing. Int. J. Epidemiol. 42, 1640–1648 (2013)
- Cao, X. et al. Contribution of life course circumstances to the acceleration of phenotypic and functional aging: A retrospective study. eClinicalMedicine 51, 101548 (2022)
These resources helped us to ensure theoretical consistency and empirical relevance across datasets, allowing for more robust cross-national comparisons of frailty predictors.
📈 Key Findings
Our models explained a significant portion of the variance in frailty index:
- USA: 24.2% (95% CI: 20.3%–28.1%)
- England: 25.8% (95% CI: 19.0%–32.4%)
- China: 17.2% (95% CI: 12.6%–21.5%)
Notably, the most influential SDoH domains varied by country:
- USA & England: Health behaviors and social connections/stressors were predominant.
- China: Material circumstances, such as housing conditions, had the most significant impact.
Common predictors across all countries included body mass index (BMI) and sleep duration. However, the nature of their relationships with frailty differed, highlighting the importance of context-specific analyses.
🌍 Implications for Public Health
Our findings underscore the necessity of tailoring public health strategies to the unique social determinants prevalent in each country. While some factors like BMI and sleep duration are universally relevant, others require localized interventions. For instance, addressing material deprivation may be more critical in China, whereas promoting healthy behaviors and social engagement might be more effective in the USA and England.
By identifying and prioritizing these determinants, policymakers and healthcare providers can develop targeted interventions to mitigate frailty and promote healthy aging.
🔮 Future Directions
Building upon our findings, future research could explore:
-
Expanded Data Collection: Future studies would benefit from more comprehensive data on SDoH, such as dietary patterns, which were not included in our current analysis but are known to influence frailty and aging.
-
Causal Inference and Interventional Research: To move beyond associations, it is crucial to explore the mechanisms underlying the identified relationships. This could be achieved through causal inference methods and interventional studies, which would help determine whether modifying specific social determinants can effectively reduce frailty risk.
-
Broader Populations: Expanding analyses to include other countries and diverse demographic groups to enhance generalizability and better understand global patterns of social influence on frailty.
🙏 Acknowledgments
We extend our gratitude to the participants and data providers from the HRS (USA), ELSA (England), and CHARLS (China). We appreciate the Gateway to Global Aging team for providing harmonized datasets and codebooks for the HRS, ELSA, and CHARLS, which made this cross-national analysis possible.
📄 Read the Full Study
For a comprehensive understanding of our methodology and findings, please refer to the full article: