Aging
Triangulating Evidence on Social Isolation, Cognitive Function, and Dementia Risk Among Adults Aged 90 Years and Older Andrew Philips* Andrew Philips Philips Philips Philips Philips Philips Philips Philips Boston University
Social relationships change with age, and the impact of social isolation may differ in those aged ≥90 years. However, evidence on associations of social isolation with cognitive aging in this age group is limited.
We used the 2012-2020 Health and Retirement Study (HRS) and the National Health and Aging Trends Study (NHATS), restricted to participants ≥90 years (HRS n=416, NHATS n=396). In both cohorts, social isolation was operationalized using items conveying marital status, living arrangement, frequency of contact with network members, and social participation (HRS: 0-5, NHATS: 0-4). Social isolation was modeled both as a continuous and binary measure (top quintile in HRS, ≥2 in NHATS). Cognitive function in HRS (range 0–27) and NHATS (range 0–33) was assessed using measures capturing memory and orientation, with scores standardized to baseline mean and SD. Dementia was defined using the Langa-Weir algorithm for HRS and AD-8 for NHATS. We fit linear mixed-effects models for cognitive level and decline, and Cox proportional hazards models for dementia risk for those without prevalent dementia at baseline, adjusting for baseline age, sex, race, education, and depression. Inverse probability weights accounted for attrition due to death and loss to follow-up.
In NHATS, greater social isolation was associated with lower cognitive levels (-0.21, 95% CI: -0.3, -0.12) and faster annual rate of decline (-0.03, 95% CI: -0.05, -0.01), with similar findings for the binary measure. However, no such associations were observed in HRS. 33% of HRS participants and 53% of NHATS participants developed dementia during follow-up, but social isolation was not associated with dementia risk in either cohort.
Social isolation may be associated with lower cognitive function and faster cognitive decline among those aged ≥90 years, although findings were inconsistent across cohorts. More data and a harmonized measurement approach is needed to strengthen evidence in this population.
