Development and Evaluation of a Brief Social Isolation Measure in the Nationally-Representative Health and Retirement Study.

Background: The Health and Retirement Study (HRS) has been central to identifying social isolation as a key determinant of health among older adults. However, prior HRS social isolation measures have relied on a subset of participants responding to a leave-behind questionnaire (LBQ), which limits longitudinal analyses, sample size, and inclusion of individuals with cognitive impairment who require a proxy. This study developed a brief social isolation scale that can be used in the full HRS cohort.

Methods: We used 2016 and 2018 HRS data, including 6122 adults ≥ 65 years old. The brief 5-Item "Core" Social Isolation Measure (Range: 0-8 Points) Included Items From the HRS core interview administered to all respondents: marital status, household size, proximity to children, religious participation, and volunteering. We compared three cutoffs (≤ 1, ≤ 2, ≤ 3) to cutoffs for previously established scales (Kotwal and Crowe) Using Sensitivity, specificity, and overall classification accuracy. We also compared the prevalence across demographic and health subgroups, and construct validity through associations with loneliness, depressive, symptoms, and life satisfaction.

Results: Participants were on average 75.4 years old (SD 7.2), 60% women, and 10% socially isolated using the Kotwal measure and 27% using the Crowe measure. A core measure cutoff of ≤ 2 correctly classified 83.1% vs. Kotwal (Sensitivity: 81.0%, Specificity: 83.2%) and 77.4% vs. Crowe (51.3%, 86.8%). At cutoff ≤ 3, accuracy was 66.6% vs. Kotwal (95.6%, 63.2%) and 70.3% vs. Crowe (74.6%, 68.8%). All three measures showed strong construct validity with loneliness, depressive symptoms, and life satisfaction, and relatively consistent prevalence estimates across sociodemographic and health groups.

Conclusions: A Brief Social Isolation Measure Aligns Well With Longer, Previously-Established Measures While Addressing Key Limitations in Data Availability and Sample Representativeness. Researchers Should Consider the Appropriate Cutoff Depending on the Research Context, and the Relative Importance of Sensitivity, Specificity, and Sample Size.