End‐of‐life health care use among socially isolated and cognitively impaired older adults. Issue 3 (23rd November 2022)

• Record Type: Journal Article
• Authors: Kotwal, Ashwin A.; Cenzer, Irena S.; Yaffe, Kristine; Perissinotto, Carla; Smith, Alexander K.
• Is Part Of: Journal of the American Geriatrics Society. Volume 71:Issue 3(2023)
• Page Start: 880
• Page End: 887
• Publication Date: 2022-11-23
• Subjects: Geriatrics -- Periodicals; 618.97
• DOI: 10.1111/jgs.18131
• Languages: English
• ISSNs: 0002-8614

Abstract

Abstract: Background: Social isolation is common in the last years of life, especially among individuals with cognitive impairment (CI), but it is unknown if social isolation is related to end‐of‐life health care use. Methods: We used nationally‐representative 2006–2018 Health and Retirement Study (HRS) data linked to Medicare claims, including adults age ≥65 interviewed in the last 4 years of life ( N = 2380). We used a validated social isolation scale and three social isolation subscales: (1) household contacts (marital status, household size, nearby children), (2) social network interaction (with children, family, and friends), and (3) community engagement. End‐of‐life health outcomes included 2+ emergency department (ED) visits in the last month of life, hospitalizations or ICU stays in the last 6 months of life, and any hospice use. CI was defined using the validated Langa‐Weir methodology. We used logistic regression to test the association of each social isolation measure with each end‐of‐life outcome, adjusted for sociodemographic covariates, and tested for interaction terms with CI ( p < 0.2). Results: The mean age of our sample of decedents was 81.2 (SD = 9.9), 53% were female, 8% Black, and 4% Hispanic. Overall, social isolation and the community engagement subscale were not associated with end‐of‐life health care use. Fewer household contacts were associated with lower hospice use (aOR = 0.74, p = 0.005). There were significant interaction terms between the social network interaction subscale and CI for ED use ( p = 0.009) and hospitalizations ( p = 0.04), and a trend for ICU stays ( p = 0.15); individuals with both low social network interaction and CI had lower health care use across all three outcomes compared to other groups. Conclusions: Individuals with fewer household contacts had lower hospice use, and cognitively impaired individuals with low social network interaction had fewer end‐of‐life ED visits, hospitalizations, and ICU stays. Clinicians should consider mobilizing external support services to ensure access to goal‐concordant care for older adults with limited end‐of‐life social contact when needed.