Identification of child-maltreatment-related emergency department visits from electronic health records

OBJECTIVES: Since the implementation of ICD-10-CM in 2015, emergency department (ED) records using expanded child maltreatment codes have been used to estimate the prevalence of child maltreatment in the US. This study compares the efficacy of ICD-10-CM codes with other search strategies in identifying child maltreatment. 

MATERIALS AND METHODS: This study was conducted at Yale-New Haven Children's Hospital (YNHCH) and Connecticut Children's Medical Center (CCMC). We examined ED visits for confirmed or suspected child maltreatment concerns from 1/1/2019 to 8/31/2020 among children under 18 years old. Cases were identified using three search strategies: (1) ICD-10-CM codes, (2) keywords in provider notes, and (3) chief complaints. After cases were identified, ED provider notes were reviewed by clinicians to determine if a concern for maltreatment was indicated. Sensitivity and positive predictive values were calculated for each search strategy. 

RESULTS: Of 3841 ED records identified by search strategies, 1248 were suspected to involve maltreatment. Using ICD-10-CM codes alone had a sensitivity of 47%; the sensitivity of keywords in provider notes and chief complaints were 65% and 58%, respectively. When ICD-10-CM codes were combined with keywords in provider notes and chief complaints, the sensitivity increased to 82% and 77%, respectively. 

PRACTICE IMPLICATIONS: The use of any single method could miss 35 to 53% of cases, although time-consuming chart review was required to remove identified cases that were not associated with maltreatment. Future research should refine the search strategies and consider utilizing multiple approaches to identify cases.