IntroductionWhile randomized-controlled trials have shown that heroin-assisted treatment (HAT) is superior to methadone maintenance alone in treatment of refractory clients, little is known about client factors associated with retention in HAT in routine care. MethodsThis retrospective cohort study assessed predictors of retention in first treatment episode among a consecutive cohort of clients admitted to HAT in Denmark from 2010 to 2018, who could be matched to the Danish population register and for whom a Short Form Health Survey (SF-36) was available at admission (N = 432). The study derived predictors from client self-reports at intake and administrative data available in national registers. Cox proportional hazards regression modelled retention in treatment. ResultsThe one-year retention rate was 69.63 % (95 % CI 65.06 %–73.74 %), and the median time in treatment was 2.45 years (95 % CI, 1.83–3.12). Bivariate analyses showed that retention was lower for clients who had recent cocaine or benzodiazepine use and among those who had experienced an overdose in the year prior to enrollment in HAT. Age below 40, recent illegal activity, poorer emotional wellbeing, previous residential treatment experience, and previous intensive outpatient treatment were also predictors of dropout from HAT. ConclusionsThis observational study found that retention in HAT in routine care was similar to rates observed in randomized-controlled trials conducted in other countries. The results suggest that addressing polysubstance use as part of the HAT program may promote long-term retention, as may directing resources to certain subgroups identified at intake, including clients under 40 years and those who report recent criminal activity, emotional problems, or overdoses. The findings that previous residential treatment and intensive outpatient treatment were associated with dropout were unexpected.