Johns Hopkins Paper Links Legal Online Sports Betting to Small Rise in Depression and Anxiety Care

Legalising online sports betting was associated with modest increases in outpatient depression and anxiety diagnoses and related prescriptions, Johns Hopkins Bloomberg School of Public Health researchers reported on 14 October 2026. Using employer-sponsored insurance claims from January 2016 to December 2023, the authors compared 22 states that legalised online sports betting between 2018 and 2022 with 13 states that had not by the end of 2023 and found relative increases in several behavioural-health outcomes.

The headline results on online sports betting and behavioural health

The study, published online in JAMA Psychiatry on 14 October 2026 and titled "Online Sports Betting Legalization and Behavioral Health Care Use", reports relative increases in outpatient diagnoses and prescription fills against the sample averages. The team — Benjamin Westermeyer, Matthew D. Eisenberg and senior author Mark K. Meiselbach of the Bloomberg School — measured outcomes at the state-year level across 61,951,305 enrollee-years drawn from a national convenience sample of people aged 18 to 64 with employer-sponsored insurance.

Relative to sample averages the authors report:

• probability of an outpatient depression diagnosis: +6.73% for men and +5.47% for women;

• probability of an antidepressant fill: +5.44% for men and +5.00% for women;

• probability of an outpatient anxiety diagnosis: +7.20% for men and +6.50% for women;

• probability of an anti-anxiety medication fill: +4.99% for men and +5.17% for women.

The paper also gives absolute changes in percentage points: antidepressant fills rose by 0.57 percentage points for men and 1.08 points for women, and outpatient anxiety diagnoses rose by 0.67 points for men and 1.15 points for women. The authors note that the antidepressant estimate for men equates to roughly six additional insured men per 1,000 filling at least one prescription in a year.

Methods, sample and what the numbers mean for operators and regulators

The design is a staggered difference-in-differences comparison of 22 treated states (legalised online sports betting between 2018 and 2022) against 13 control states that had not legalised by the end of 2023. States that legalised in 2023 or before 2018 were excluded. Outcomes are the share of enrollees in each state-year with at least one prescription fill or at least one outpatient visit for depression, anxiety or substance use disorder.

The dataset covers claims from January 2016 through December 2023 and the sample is 52% women with a mean age of 41.9. The authors stress the outcome measures are at the population level; they cannot link claims to individual betting behaviour because the claims data do not record who placed bets.

For operators and regulators, the practical implication is that the effect appears in routine outpatient care and prescription fills rather than emergency or inpatient services. That means common industry harm metrics — helpline calls, self-exclusions, and screened problem-gambling rates — may not capture this set of harms. The paper contrasts this pattern with an Ontario study of emergency visits coded as gambling disorder, which recorded a doubling of gambling-disorder emergency visits after Ontario’s online market opened.

Limits, precision and subgroup notes

The authors list two principal limits in their announcement: the results apply only to people with employer-sponsored insurance, and claims data cannot identify who bets. The JAMA Psychiatry abstract does not name the claims database or the estimator; the full text, including funding and conflict-of-interest statements, is paywalled.

Confidence intervals are wider for women; the paper says associations were “more precise and robust for men than women.” For example, women’s confidence interval for the antidepressant change runs from 0.23 to 1.94 percentage points, while men’s runs from 0.26 to 0.89 points. There was no detectable change in substance use disorder treatment and no association with emergency department or inpatient visits.

"I hope the results would make doctors and online sports bettors aware that these online apps can lead to harms beyond financial risks," said Mark K. Meiselbach, assistant professor of health policy and management at the Bloomberg School.

How this study fits alongside financial and population research

Population-level evidence on legal sports betting has largely been financial to date. The Johns Hopkins study adds a clinical cost that lands on employer health plans. The article cites working research by Brett Hollenbeck, Poet Larsen and Davide Proserpio suggesting personal bankruptcies rose 25–30% in states with online sports betting, and a Journal of Financial Economics paper by Scott Baker and colleagues that found online betting reduced savings and raised credit-card debt among financially constrained households.

A 2024 Economics Letters paper from Hamilton College economists reported mixed mental-health effects by age and marital status, underscoring that the literature is not uniform. The Johns Hopkins authors explicitly do not claim causation; they present an association at the population level and discuss mechanisms and limitations in the paywalled full text.

Policy context and public funding

The paper arrives while states wrestle with funding for problem-gambling services. Fiscal-year 2025 data show states spent $145m of the $188m set aside for problem gambling services. Advocacy group GamblingHarm.org noted the Johns Hopkins data end in December 2023 and therefore predate the expansion of sports contracts on prediction markets. By 16 October 2026, iGaming Times had not found a public response from the American Gaming Association.

For regulators, employers and insurers the central takeaways are empirical and operational: the measured increase is population-level, concentrated in outpatient settings, and observable within employer-sponsored plans. Whether the pathway is betting or another correlated mechanism cannot be determined from the claims data used in this analysis.