OBJECTIVES: Most medication errors occur in primary and long-term care, and a wide range of medication safety interventions have been implemented, but these are often expensive, with little evidence around cost-effectiveness. We report a systematic review of economic evaluations of these interventions within primary and long-term healthcare settings. METHODS: A comprehensive search was conducted in databases (Medline, Embase, Econlit and PsycINFO) for full economic evaluations of primary care interventions targeting all errors in the medication use process (January 2004 to September 2025). Methodological and reporting qualities were assessed using standard tools. RESULTS: From 8523 records, 44 studies evaluating interventions in general/family practice (22), community pharmacy (11) and nursing/care/residential homes (11) met the inclusion criteria, 24 of which were either pharmacy led (19) or multidisciplinary medication reviews (5). All but one study looked at prescribing or monitoring interventions only. A total of 12 studies included all patients, with 24 focusing on older adults (> 65 years) and 3 focusing on condition-specific groups. Most studies only included costs from a healthcare perspective (39). Outcomes ranged from prescribing errors (9), hospital utilisation (13) and health-related quality of life (15) to falls (6) and adverse drug events (6). In total, 21 studies carried out an incremental cost-effectiveness analysis (16 including the incremental cost per quality-adjusted life year gained), and 14 reported the intervention cost-effectiveness. Remaining studies were cost-consequence (18) and cost-benefit analyses (5). Study reporting quality varied considerably, with lack of transparency in the design of the decision-analytic model, varied reporting of costs, little consideration of indirect costs or the impact of loss of trust on future use of healthcare, limitations in handling of uncertainty or discounting and very little patient involvement around targeting patients or designing interventions. Of the ten studies using decision models, all scored poorly for model validation. The quality of studies has not improved over time. CONCLUSIONS: While some interventions demonstrated cost-effectiveness, study quality was variable, with generally poorly validated models. Study heterogeneity precluded meaningful direct comparison between studies. Significant research gaps remain as studies focused mainly on prescribing and monitoring errors, there was little or no investigation of technology-based interventions and there was inadequate targeting of patients most vulnerable to harm.