Using administrative data as a means of detecting safety problems has inherent appeal, as these data are already routinely collected. However, many prior studies have shown that these data have limited ability to identify adverse events, in large part because adverse events are not specifically delineated in these databases, which are assembled primarily for billing purposes. This systematic review found that this problem is likely to persist with the new ICD-10 coding system. The review searched for studies that attempted to identify adverse drug events by searching administrative data and found that more than 800 different ICD-10 codes could be used to detect medication errors, with the combination of codes varying widely across studies. These findings support the concept that complementary data sources must be used to comprehensively assess patient safety within an organization.