Many different methods have been used to study the incidence of prescribing errors in hospital inpatients. The objectives of this review were to outline the methods used, highlight their strengths and limitations, and summarise the incidence of prescribing errors reported. Methods used may be retrospective or prospective and based on process or on outcome. Reported prescribing error rates vary widely, ranging from 0.3% to 39.1% of medication orders written and from 1% to 100% of hospital admissions. Unfortunately, there is no standard denominator for use when expressing prescribing error rates. It could be argued that the most meaningful is the number of medication orders written; however, it is also helpful to consider the number of medication orders written per patient stay in order to understand the risk that a given prescribing error rate poses to an individual patient. Because of wide variation in the definitions and methods used, it is difficult to make comparisons between different studies. Each method for identifying prescribing errors has advantages and disadvantages. Process-based studies potentially allow all errors to be identified, giving more scope for the identification of trends and learning opportunities, and it may be easier to collect sufficient data to show statistically significant changes in prescribing error rates following interventions to reduce them. However, studies based on process may be criticised for focusing on many minor errors that are very unlikely to have resulted in patient harm. Focusing instead on harm, as in outcome-based studies, allows efforts to reduce errors to be targeted on those areas that are likely to result in the highest impact. Therefore, the most appropriate method depends on the study's aims. However, using a combination of methods is likely to be the most useful approach if comprehensive data are required.