Abstract Body (Do not enter title and authors here): Background:
Evidence linking social isolation to cardiovascular disease morbidity and mortality has grown in recent years. Still, information on how this may manifest in real world settings and its implications for screening practices is limited. In 2019, our large national integrated health care system implemented screening for social isolation as part of a broader universal social risk assessment. This repository of screening data was joined to administrative claims to test these associations in real world data and explore differences by demographic and medical factors.
Methods:
Social isolation responses recorded from 2019-2022 were included for a cohort of adult health plan members with documented atherosclerotic cardiovascular disease (ASCVD). We selected a single random assessment for each member and retained any other responses for sensitivity analyses. Cohort members had at least 10 months of enrollment surrounding assessment date for use as the baseline period and were followed for 365 days. We used cox proportional hazards regression with right censoring for coverage gaps to estimate the risk of all-cause mortality conferred by social isolation. We used Poisson regression to model the rate of inpatient stays.
Results:
There were 881 deaths among 7,484 members (18% of those with social isolation; 11% of those without). The isolated group skewed less male (54% vs. 65%, p <0.001) and older (68.4 [SD 12.7] vs. 67.5 [SD 12.6] years, p <0.05) than those not isolated. At baseline, the isolated group had higher rates of mean inpatient stays (2.3 [SD 1.9] vs. 1.9 [SD 1.6], p <0.001), ED visits (1.2 [SD 1.9] vs. 0.9 [SD 1.6], p < 0.001), and smoking (50% vs. 44%, p < 0.05), and a higher mean Charlson Comorbidity Index (CCI) (3.6 [SD 2.6] vs. 3.0 [SD 2.4], p < 0.001). Adjusted survival models suggested an independent 33% increase in all-cause mortality risk among those screening positive for social isolation (1.33 HR 95% CI 1.10, 1.62). Adjusted Poisson models found a similar increase in inpatient stays (1.36 IRR 95% CI 1.14, 1.62).
Conclusions:
Our findings corroborate the association between social isolation and morbidity and mortality among those with known ASCVD. The effect for social isolation was similar in magnitude to clinical comorbidities and smoking, suggesting that routine screening for social isolation may provide valuable information for assessing and managing the risk of death among patients with ASCVD.