The article doesn’t mention if they controlled for income, and a lot of these studies don’t. This was famously what debunked the “one to two glasses of wine a week improves your health” since people who drink zero are possibly either too poor to afford wine or alcoholics who are sober, and people who are either poor or who drink more than three per night will have worse health outcomes in general. The people who drink 1-2 have disposable income and are generally otherwise healthy. I wonder if they controlled for caffeine in a similar way, as I could see people who can afford the time and money for 2-3 cups per day are just generally healthier anyway because they might tend to have more disposable income.