Actual paper (paywall but abstract available): https://www.nejm.org/doi/full/10.1056/NEJMoa1806828?query=featured_home https://www.nejm.org/doi/full/10.1056/NEJ

Actual paper (paywall but abstract available): https://www.nejm.org/doi/full/10.1056/NEJMoa1806828?query=featured_home https://www.nejm.org/doi/full/10.1056/NEJMoa1806828?query=fe... One part out of many that interested me in the discussion: "data from insurance claims do not allow us to determine when a child starts school. Parents may delay school entry for children born in August, which would mean that those children start school at 6 years of age rather than 5 years of age. Because we did not directly observe children’s ages when they entered school, we cannot know how often this occurred" There are other confounds, but suffice to say that this is not a very well controlled cohort study. It is however interesting for guiding further research.