Hacker News Reader

That certainly breathes new life into old metaphors like "twilight years".

It sounds like the most immediate takeaway is that anyone analyzing this stuff needs to control for when measure individuals.

View on HN

The potential for systematic bias here is also quite concerning. Imagine your experimental group tends to come in the morning (maybe because they're on-site) but your control group is scheduled to come in the rest of the day...

This problem is VERY real and happens in surprising ways. My biotech has a dataset that was collected over several years and we found this pattern of older people having more Vitamin D in their blood than younger people, which is the opposite of the published literature. It turns out that the sample collection was initially year round but switched to summer only and everyone in the cohort was older, by definition, in the later collection.

And not just this stuff but a lot of health metrics probably have this occult dependence on the time of day. For example MRIs [1].

[1]: https://www.nature.com/articles/s41467-023-42588-6

Reply on news.ycombinator.com