For two decades, national surveys have shown that overall rates of teenage cannabis use in the United States have stayed relatively flat. That figure has been cited, constantly and reasonably, as evidence that legalisation has not affected young people.
A study published in Addiction this week, led by Alisa Padon of the Public Health Institute with investigators at Kaiser Permanente, looked at something else. It tracked new diagnoses of cannabis use disorder across 637,461 Kaiser Permanente Northern California adolescents. New diagnoses were falling by around six percent a year before California’s 2016 legalisation vote, and rising by around eight percent a year after it. By early 2020 they had returned to 2014 levels, erasing a three-year decline.
Padon’s framing of the problem is the sentence worth keeping. Most surveys of adolescent cannabis use tell us how many teens have tried cannabis, but not how many are developing a disorder that brings them into clinical care. The public health goal is to reduce harm, and we cannot reduce what we do not measure.
There is a further detail that sharpens it. The shift began at the passage of legalisation rather than at the opening of retail stores, which the authors suggest may reflect changes in the legal and social environment around the vote itself, including reduced juvenile penalties and a normative signal amplified by the ballot campaign. A companion study in JAMA Network Open found self-reported use following a similar shape, falling before legalisation, climbing after passage and before retail, peaking at 9.5 percent in early 2020 and sitting at 6.5 percent in 2024.
Now the part that belongs on a business page.
Prevalence and severity are different measurements and they can move in opposite directions. The number of people who do a thing can hold perfectly steady while the intensity with which some of them do it changes completely, and any instrument that only counts heads will report calm throughout.
Every cafe owner has been fooled by exactly this. Your door counter says four hundred and twenty people a day, same as last year, and you conclude nothing has changed. What the counter cannot see is that you have lost eighty regulars who came four times a week and gained eighty occasional visitors who come once a month, and that the eighty who left were the ones who bought a pastry and the ones who arrived are the ones who sit for two hours with a filter. Same footfall, different business, and by the time the revenue line catches up you have spent a year believing your own dashboard.
The general rule the researchers are pointing at is that you have to measure the thing you actually care about, not the thing that is easy to count. Trying cannabis once and developing a disorder are not the same outcome and one is not a proxy for the other, particularly since disorder risk rises sharply with frequency and with earlier first use. Visits and value are not the same outcome either.
The methodological point is that surveys asking about prevalence were doing their job correctly. They simply were not designed to answer the question they were being used to answer, and nobody noticed because the answer they gave was reassuring and complete-looking.
What this study did was go and find a different instrument. Clinical records, tracked over time rather than sampled at a point.
The line was flat. Nobody had asked the line about severity.
Source: Kaiser Permanente Division of Research, “Teen cannabis surveys may miss rise in use disorder”
