Mediation analysis of social isolation and mortality by health behaviors

Social isolation associated with greater risk for mortality is well documented, however, it is unclear whether and the extent to which social isolation influences mortality through health behaviors. We therefore assessed the mediation between social isolation and mortality by health behaviors. 15,000 adults ages 18-89 years in the Third National Health and Nutrition Examination Survey in the US (NHANES III) (1988-1994) were followed through 2015. The outcomes of overall and cause-specific mortality were assessed by linking with the National Death Index. Social isolation was based on a modified Social Network Index. Mediation analyses were conducted under the counterfactual framework. Compared with participants who were not socially isolated, those who were isolated had significantly increased risk for overall mortality (HR = 1.33, 95% CI: 1.15-1.54), cardiovascular disease (CVD)-specific mortality (HR = 1.28, 95% CI: 1.01-1.62) and mortality from causes other than CVD or cancer (HR = 1.53, 95% CI: 1.22-1.90). Social isolation was not independently associated with risk for cancer-specific mortality. The association between social isolation and cancer-specific mortality was largely (74%) mediated by health behaviors, including smoking, physical activity and quality of diet. The associations between social isolation and CVD or other specific mortality were partly (about one-fifth) explained by unhealthy behaviors. In conclusion, the extent to which social isolation and mortality was mediated by healthy behaviors varies by causes of death. Given mounting evidence that social isolation is increasing in society, it is important to add social isolation to the list of public health concerns in addition to conventional risk factors.