Background: External causes of injury codes (E-codes) are used in administrative and claims databases for billing and often employed to estimate the number of self-injury visits to emergency departments (EDs). Aims: This study assessed the accuracy of E-codes using standardized, independently administered research assessments at the time of ED visits. Method: We recruited 254 patients at three psychiatric emergency departments in the United States between 2007 and 2011, who completed research assessments after presenting for suicide-related concerns and were classified as suicide attempters (50.4%, n = 128), nonsuicidal self-injurers (11.8%, n = 30), psychiatric controls (29.9%, n = 76), or interrupted suicide attempters (7.8%, n = 20). These classifications were compared with their E-code classifications. Results: Of the participants, 21.7% (55/254) received an E-code. In all, 36.7% of research-classified suicide attempters and 26.7% of research-classified nonsuicidal self-injurers received self-inflicted injury E-codes. Those who did not receive an E-code but should have based on the research assessments had more severe psychopathology, more Axis I diagnoses, more suicide attempts, and greater suicidal ideation. Limitations: The sample came from three large academic medical centers and these findings may not be generalizable to all EDs. Conclusion: The frequency of ED visits for self-inflicted injury is much greater than current figures indicate and should be increased threefold.