BACKGROUND: A study was conducted to explore the value and limitations of voluntary medical error reports and to learn about common errors in warfarin use. METHODS: Voluntary reports of 8,837 inpatient errors and 820 outpatient errors in warfarin use submitted by 445 hospitals and 192 outpatient facilities participating in MEDMARX, a voluntary medication error reporting system, from 2002 to 2004, were gathered. RESULTS: Overall, errors occurred most often during transcription/documentation (35%) and administration (30%) in hospitals, and during prescribing (31%) and dispensing (39%) in outpatient settings. Dosing errors were the most common type. In hospitals, more than 50% of reported errors were initiated by nurses, and 50% were intercepted by nurses, whereas in outpatient settings, about 50% of reported errors occurred in pharmacies and 50% were intercepted by pharmacists. About 17% of inpatient and 13% of outpatient warfarin errors resulted in changes in patient care, and 42% of inpatient and 62% of outpatient errors resulted in procedural changes. Cascade analysis and textual descriptions further located specific, correctible safety lapses. DISCUSSION: Voluntary medical error reporting systems can, to some extent, provide meaningful and actionable information to guide patient safety improvement, but their usefulness is limited because of a lack of details, incomplete reporting, underreporting, and various reporting biases.