Volunteering and mortality risk: a partner-controlled quasi-experimental design

The consensus that volunteering is associated with a lower mortality risk is derived from a body of observational studies and therefore vulnerable to uncontrolled or residual confounding. This potential limitation is likely to be particularly problematic for volunteers who, by definition, are self-selected and known to be significantly different from non-volunteers across a range of factors associated with better survival.

This is a census-based record-linkage study of 308 733 married couples aged 25 and over, including 100 571 volunteers, with mortality follow-up for 33 months. We used a standard Cox model to examine whether mortality risk in the partners of volunteers was influenced by partner volunteering status-something expected if the effects of volunteering on mortality risk were due to shared household or behavioural characteristics.

Volunteers were general more affluent, better educated and more religious than their non-volunteering peers; they also had a lower mortality risk [hazard ratio (HR)adj = 0.78: 95% confidence interval (CI) = 0.71, 0.85 for males and HRadj = 0.77: 95% CI = 0.68, 0.88 for females]. However, amongst cohort members who were not volunteers, having a partner who was a volunteer was not associated with a mortality advantage (HRadj = 1.01: 95% CI = 0.92, 1.11 for men and HRadj = 1.00: 95% CI = 0.88, 1.13 women).

This study provides further evidence that the lower mortality associated with volunteering is unlikely to be due to health selection or to residual confounding arising from unmeasured selection effects within households. It therefore increases the plausibility of a direct causal effect.