Objective To investigate whether cannabis use is associated with increased risk of relapse, as indexed by number of hospital admissions, and whether antipsychotic treatment failure, as indexed by number of unique antipsychotics prescribed, may mediate this effect in a large data set of patients with first episode psychosis (FEP).

Design Observational study with exploratory mediation analysis.

Setting Anonymised electronic mental health record data from the South London and Maudsley NHS Foundation Trust.

Participants 2026 people presenting to early intervention services with FEP.

Exposure Cannabis use at presentation, identified using natural language processing.

Main outcome measures admission to psychiatric hospital and clozapine prescription up to 5 years following presentation.

Mediator Number of unique antipsychotics prescribed.

Results Cannabis use was present in 46.3% of the sample at first presentation and was particularly common in patients who were 16–25, male and single. It was associated with increased frequency of hospital admission (incidence rate ratio 1.50, 95% CI 1.25 to 1.80), increased likelihood of compulsory admission (OR 1.55, 1.16 to 2.08) and greater number of days spent in hospital (β coefficient 35.1 days, 12.1 to 58.1). The number of unique antipsychotics prescribed, mediated increased frequency of hospital admission (natural indirect effect 1.09, 95% CI 1.01 to 1.18; total effect 1.50, 1.21 to 1.87), increased likelihood of compulsory admission (natural indirect effect (NIE) 1.27, 1.03 to 1.58; total effect (TE) 1.76, 0.81 to 3.84) and greater number of days spent in hospital (NIE 17.9, 2.4 to 33.4; TE 34.8, 11.6 to 58.1).

Conclusions Cannabis use in patients with FEP was associated with an increased likelihood of hospital admission. This was linked to the prescription of several different antipsychotic drugs, indicating clinical judgement of antipsychotic treatment failure. Together, this suggests that cannabis use might be associated with worse clinical outcomes in psychosis by contributing towards failure of antipsychotic treatment.