[2716057]

Anecdotally I agree with the message, but the research looks weak indeed.

A simple snapshot assessment and some scoring of an individual's (entire, self-reported!) social life is too simplistic. The measurements would have to be performed throughout the life of each participant with sufficiently high frequency.

[derbOac]

For what it's worth, the dataset they used does in fact have measurements spanning 30 years of adulthood, and similar papers from that dataset leveraging the longitudinal data have found similar conclusions.

Why it happens is less clear. It could be stress effects, or it could be something like people with more social support are more likely to get help going in for preventative care etc.

[nicce]

If someone starts now, we may see results in 50 years?

[psychoslave]

You're the lucky guy today: https://pmc.ncbi.nlm.nih.gov/articles/PMC11852352/