Social isolation is a better predictor of mortality than traditional clinical risk factors in certain populations

Background: Social isolation has been linked to mortality across a range of middle-aged and older populations, although it has not been compared with other clinical risk factors in a national sample. Methods: Data came 12,403 adults from the National Health and Nutrition Examination Survey III and the National Death Index. The outcome was mortality after eight years. Predictor variables included individual social isolation factors and a composite score. Clinical predictors used for comparison included smoking, BMI, high blood pressure, and hypercholesterolemia. Logistic regressions and recursive partitioning - which classifies populations into low- and high- risk groups by choosing the best predictor variables – were used to rank the importance of these predictors. Models were nationally weighted. Results: Male mortality was significantly predicted by current smoking (OR 1.81, CI 1.31-2.51), high blood pressure (OR 1.40, CI 1.06-1.85), and, most strongly, by social isolation (OR 2.25, CI 1.51-3.35). Only smoking (OR 2.33, CI 1.60-3.40) predicted mortality among females, although social isolation approached significance (OR 1.55, CI 0.99-2.44). For males, significant individual social isolation factors included being unmarried (OR 1.67, CI 1.20-2.31), not participating in religious activities (OR 1.31, 1.04-1.63), and lacking club/organization associations (OR 1.29, CI 1.00-1.67). Recursive partitioning identified social isolation factors as the most important predictors of mortality for young men. Conclusions: Social isolation is an equal if not better predictor of mortality than traditional clinical risk factors among males. It suggests that evaluating patients' social isolation in a clinical setting to identify individuals at high risk for mortality could be valuable.