Why does purpose in life predict mortality in older adults?

Background Previous work documents a strong association between a higher sense of life purpose and lower all-cause mortality risk even when controlling for baseline health and proposes that life purpose intervention may provide a low-cost lever to improve health and longevity. Causation, however, is less clear: lower purpose may cause poorer health and decreased longevity, or poorer health may cause decreased longevity and lower purpose. We examine the extent that (1) more comprehensive health metrics and (2) horizon mitigate or strengthen the relation between purpose and mortality risk to better understand causation. Methods Prospective cohort sample of 8 425 individuals aged 50 and older who were eligible to participate in the 2006 Health and Retirement Study Psychosocial and Lifestyle questionnaire. Individuals were followed for three subsequent four-year periods: 2006-2010, 2010-2014, and 2014-2018. A total of 1 597 individuals were excluded in the initial four-year period due to lack of follow up, sample weights, or covariates leaving an initial sample of 6 828 individuals. For the second and third four-year periods, an additional 168 and 349 respondents were lost to follow up, respectively. Cox models were estimated to examine the relation between life purpose and mortality for three horizons (years 1-4, 5-8, and 9-12) with more comprehensive measures of current health. Covariates included age, sex, education, race, marital status, smoking status, exercise, alcohol, BMI, and functional health score. Findings The relation between life purpose and mortality was substantially attenuated or disappeared at longer horizons or when using more comprehensive measures of current health. Interpretation Much of the documented relation between life purpose and longevity arises from poor health causing higher mortality risk and lower purpose (i.e., reverse causation). As a result, life purpose intervention is likely to be less effective than the previous evidence suggests.