Do health records data accurately identify repeat self-harm after emergency department visits?

ObjectiveEvaluate how often encounter diagnoses of self-harm soon after an emergency department visit for self-harm represent new self-harm events. MethodsElectronic health records (EHR) and insurance claims data from a large integrated health system identified emergency department encounters for injury or poisoning coded as self-harm and then selected those with another encounter diagnosis of self-harm occurring within 91 days. Review of clinical text for these pairs of encounters examined whether the subsequent self-harm diagnosis represented a distinct new self-harm event vs. a repeat ResultsOf 121 pairs of encounters with relevant clinical text available for review, records indicated a distinct repeat self-harm event for 50 (41%, 95% CI 33%-49%). The proportion confirmed as distinct new events ranged from 3% (95% CI 0%-10%) for self-harm diagnoses the following day to 50% (95% CI 19%-81%) for diagnoses 2 to 7 days later to 100% (95% CI 92%-100%) for diagnoses 8 to 91 days later. The proportion confirmed as distinct, new events did not vary by healthcare setting of recording for the subsequent diagnosis (inpatient, emergency department, other outpatient) or similarity of injury or poisoning type between the two events. ConclusionsHealth systems, researchers, and public health agencies using insurance claims or EHR diagnoses to identify early recurrence of self-harm should be cautious regarding diagnoses that appear to represent early repetition of self-harm. Self-harm diagnoses the day after an emergency department self-harm visit rarely represent a distinct new event, while diagnoses recorded more than a week later usually do. HIGHLIGHTS{blacktriangleright} Health records may over-estimate repeat self-harm after emergency department visit. {blacktriangleright}Self-harm diagnoses the day after an ED visit rarely represent distinct new events. {blacktriangleright}Self-harm diagnoses a week or more after an ED visit usually indicate new events.