Week 11: What it costs

The mortality evidence, read properly — what the effect size means, and what a meta-analysis of observational studies cannot establish.

10 May 2027

148 studies, about 308,000 people, and a 50% greater likelihood of survival for those with stronger social relationships. The comparison that made the paper famous sits in the discussion: an effect comparable to giving up smoking, and larger than obesity or physical inactivity. That sentence is why health departments started using the word loneliness.

This lecture reads the paper properly, which means spending as long on what it does not establish.

These are observational studies. People with more social connection differ from people with less in many ways, and several of those ways — money, health at baseline, mobility, mental illness — cause both the connection and the survival. The arrow may also run backwards: being unwell shrinks your social world before it kills you.

“Social relationship” is measured wildly differently across 148 studies, from marital status to a count of contacts to a composite index. The meta-analysis pools them, and the heterogeneity is not a footnote.

And the comparison to smoking is a comparison of effect sizes, not of interventions. We know how to make someone stop smoking. Nobody knows how to give someone a friend.

Cacioppo and Hawkley are the secondary reading and they supply the mechanism — perceived isolation putting the brain into hypervigilance for social threat, which then makes connection harder. That is the plausible pathway, and a plausible pathway is not proof.

You should leave able to state the finding accurately and able to say what it does not show. Both halves are marked.