Electronic health record–observed lithium persistence after within-service initiation in bipolar disorder: an incident-initiator cohort study from a multinational tertiary psychiatric service in the United Arab Emirates

Lithium is a first-line maintenance treatment for bipolar disorder, but real-world persistence varies widely across health systems, and no incident-initiator cohort analysis of lithium has been reported from any Gulf Cooperation Council bipolar disorder population. We aimed to characterise lithium initiation, observed within-service treatment persistence, and post-discontinuation treatment pathways in a multinational tertiary psychiatric service in the United Arab Emirates. A retrospective cohort study at a tertiary psychiatric service in the United Arab Emirates from 2018 to 2025. Among adults with a recorded ICD-10 F31 diagnosis, the incident initiator subset comprised adults whose first recorded lithium-bearing encounter occurred more than 90 days after their first service contact. The primary outcome was observed treatment persistence within the health-system record. A patient was classified as having discontinued lithium when at least 90 days elapsed with no lithium-bearing encounter; a 90-day grace period, reflecting routine dispensing intervals, was then added before the discontinuation date. Secondary outcomes included peri-initiation co-prescription and documented next-line treatment within 180 days of discontinuation. Incident initiation was apparent within the study service only; prior or out-of-network lithium exposure could not be excluded. Among 135 incident initiators (median age 35 years, 48.1% female, 31.1% Emirati), lithium was initiated a median of 556 days after first service contact, most commonly during inpatient (43.0%) or outpatient (35.6%) encounters. Only 11.9% received lithium as a single recorded psychotropic; 82.2% had a concurrent antipsychotic, most commonly aripiprazole (40.0%). Estimated observed persistence was 25.0% (17.7 to 32.9) at 6 months, 4.6% (1.7 to 9.8) at 12 months, and 0.0% at 24 months. The documented next-line treatment was most commonly an antipsychotic, with individually reported second-generation antipsychotics accounting for approximately three fifths of discontinuers. In exploratory adjusted analyses, no prespecified covariate was statistically associated with observed discontinuation. Observed health-system-record persistence on lithium in this multinational cohort was very limited and substantially below international benchmarks, consistent with a mixture of true discontinuation, care transfer, out-of-network prescribing, and incomplete within-service medication capture. Strategies maintaining within-service visibility and cross-provider record linkage are warranted.