Hacker News

Yes, but it's still not a good separation. From the article, there are a number of issues with their method:

The study has several limitations. First, the database includes countries that require a prescription for melatonin (such as the United Kingdom) and countries that don't (such as the United States), and patient locations were not part of the de-identified data available to the researchers. Since melatonin use in the study was based only on those identified from medication entries in the electronic health record, everyone taking it as an over-the-counter supplement in the U.S. or other countries that don't require a prescription would have been in the non-melatonin group; therefore, the analyses may not accurately reflect this. Hospitalization figures were also higher than those for initial diagnosis of heart failure because a range of related diagnostic codes may be entered for the hospitalization, and they may not always include the code for a new diagnosis of heart failure. The researchers also lacked information on the severity of insomnia and the presence of other psychiatric disorders.

An additional confounding variable is the fact that the USDA doesn't regulate melatonin sold in the United States, so it could be pill capsules full of actual bull shit for all the customers know.