Development of a Preliminary Age-Friendly Index in the Health and Retirement Survey: A Calibration-Validation Approach

Background/Purpose: The World Health Organization’s (WHO) age-friendly community program has drawn increased attention as a public health and community capacity building model to make place-based communities more conducive to long and healthy lives (Greenfield et al., 2015). To-date, most quantitative research has explored this topic through primary data collection or age-friendly specific secondary data sets such as the AARP Age-Friendly Community Survey and the AARP Livability Index (e.g., Kim et al., 2022; Zhang et al., 2020). However, there has been limited use of more general aging-related secondary data sets—such as the Health and Retirement Survey (HRS)—that explore the potential population health implications of the age-friendly model.

This project therefore aimed to develop a preliminary age-friendly index within the HRS. Such an index offers the promise of integrating the multi-dimensional WHO age-friendly concept with the methodological strengths of the HRS. Specifically, this project explored the extent to which the dimensions underlying a section of variables in the 2016 wave of the HRS aligned with the WHO age-friendly conceptual framework, which spans eight domains of livability across three dimensions (built, social, and service environments) (John & Gunter, 2016; WHO, 2007).

Methods: A split-sample exploratory/confirmatory factor analysis was performed to explore the underlying dimensionality within the theoretically selected HRS variables.

Data and sample: The final analytic sample for this study was restricted to a maximum sample of 982 participants of the 2016 RAND flat file (biennial wave) of the HRS. Inclusion criteria included no missing variables on the main index or demographic variables. Individuals were excluded if they were less than 45 years old and if they were not the primary or spousal respondent. The EFA and CFA split sample halves consisted of 491 participants each.

Measures: The final model consisted of ten variables, including two variables representing the built environment (neighborhood safety, physical condition of home), six variables for the social environment (time spent helping friends, modes of communication with friends, frequency of social activities, time spent volunteering, and social inclusion/exclusion scales), and 2 service environment variables (satisfaction with healthcare and frequency preferences were taken into account by healthcare system).

Results: The EFA identified four factors among 10 variables in the final model, representing six of the eight age-friendly domains and all three dimensions (transportation and communication & information were not represented). The total variance accounted for by each factor ranged from 23.3% to 10.52%. Specifically, the factors aligned with the built environment dimension (outdoor spaces & buildings and housing domains), a factor combining the social and civic participation domains, the community support & health services domain, and the social respect & inclusion domain. CFA findings suggested that the EFA factor structure fit the data well and that a second-order model provided the most parsimonious model-to-data fit.

Conclusions/Implications: The results of preliminary criterion-validation approach offered support for the continued development of an age-friendly index in the HRS. Future research should continue model refinement, including testing for convergent and discriminant validity and developing more robust theoretical methods for variable identification.