Although efficacy trials have been conducted on risperidone long-acting injection (RLAI), its most appropriate utilization in clinical practice remains unclear. This 6-month, follow-up study investigated prognostic indicators for early discontinuation of RLAI. Consecutive sampling was conducted for adult patients with a psychotic disorder commenced on RLAI, whose injection was dispensed by one of three South London psychiatric hospital pharmacies. Prescription data were collected prospectively and clinical data retrospectively. Eightly-one out of 88 (92.0%) eligible patients were included, of whom 29 (35.8%) had treatment refractoriness and 30 (37.0%) discontinued within 6 months. Patients with a preceding oral antipsychotic were more likely to discontinue RLAI than those with a preceding depot; treatment refractoriness weakly confounded this relationship [summary adjusted odds ratio (OR) 2.68, 95% confidence interval (CI) 0.95–7.53, P=0.061]. After adjusting for preceding antipsychotic type, patients with treatment refractoriness were no more likely to discontinue than those without (summary adjusted OR 1.55, 95% CI 0.59–4.11, P=0.376). Sociodemographic factors and other clinical factors were non-predictive of discontinuation. For this first wave of patients commenced on RLAI, many had treatment refractoriness. RLAI discontinuation is high early on but subsequently tapers off. Preceding antipsychotic type (depot versus oral) is a stronger prognostic indicator than treatment refractoriness for RLAI discontinuation.