Â鶹´«Ã½

I recently had an article accepted for publication in Academic Pediatrics. I collaborated with emergency department (ED) pediatricians at Yale (Kirsten Bechtel, MD) and the University of Connecticut (UConn) (Nina Livingston, MD, now at the University of New Mexico), and a public health researcher at UConn (Amy Hunter, MPH, Ph.D.).

We began working together in 2020, when there were questions about child maltreatment (abuse and neglect) cases decreasing in the ED. We weren’t sure if maltreatment was occurring less often, or if families were not seeking care at a time when there was uncertainty and fear about coming to a hospital or ED. In our first study, we described the change in number of cases at Yale-New Haven Children’s Hospital and Connecticut Children’s Medical Center, showing that there was a substantial, quick decrease in cases when schools and businesses shutdown in March 2020. There was then a gradual increase over the summer of 2020.

A question that we encountered during the first study was: What is the best approach to identify cases of maltreatment in the ED? Many studies use ICD-10 codes, which is a system to classify diagnoses in a healthcare encounter. In the 2015 update to the ICD system, the codes documenting child maltreatment were updated, including separate codes for confirmed vs. suspected maltreatment.

When we began this work in 2020, there was limited evidence regarding how these codes were being used in practice, although studies consistently show that maltreatment is underdiagnosed. We decided to use a broader search strategy to minimize the number of cases we missed. The strategy included ICD-10 codes, keywords in medical notes, and terms in chief complaints (reason(s) for the visit documented in triage). Because this strategy increased the risk of including cases that were not related to child maltreatment, two physicians and two medical students reviewed each case to determine if the attending provider documented any concern related to maltreatment. This provided us with the data we used in our recently accepted paper, in which we compared the three approaches to flag cases (ICD-10, provider notes, chief complaints) in terms of how often they included cases that were not related to child maltreatment and how often they missed cases that at least one of the other methods did not identify.

Our major finding was that using ICD-10 codes alone missed many cases (53%) that were identified by the other approaches. This is important because many studies use only ICD-10 codes to identify cases. However, the manual review of all cases was very time-consuming and may not be feasible for all studies.

In our study of 5-1/2 months of data from two hospitals, we reviewed 3,851 cases and found that there was documented concern for maltreatment in 1,248 of them. Note that conclusions regarding child maltreatment cannot always be made in the ED, so we do not know what percentage of these cases would be substantiated as maltreatment. We also are depending on what the attending provider documented, meaning there may be other cases in which maltreatment occurred but was not noted.

This collaboration began as a way to respond to the disruptions of COVID-19 and to contribute in some way to documenting and understanding the ripple effects of the pandemic. Discussions for the first study began less than a month into the pandemic, starting with an email to a colleague asking what new issues were arising that we could study.

It was a very interesting time to form collaborations and determine what questions would be meaningful by the time we completed the work. I was fortunate to already have connections with the child abuse and neglect team at Yale-New Haven Children’s Hospital, primarily through the director of that team at the time, John Leventhal, MD, who contributed to the first study. Through those connections, we expanded into a larger group, before settling on the group involved in these two articles.

The second article was not planned when we obtained the data. Instead, working through the first article led us to new questions, some of which we examined in the second paper. An additional side-effect of the study beginning in 2020 is that, even though we were all in Connecticut, I have still never met one of my co-authors in person, and another colleague I’ve only met once (at a national meeting in Atlanta, GA!). However, through our shared interests and complementary skills, we formed a productive team.

In the early months of the pandemic, it was not clear what most of us could do to help, beyond staying home and away from other people. With training in health services research, I felt that I had skills to contribute, even though I do not have expertise in infectious diseases. When I emailed my colleague, I did not know if anything would come out of it. I’m so glad I took a chance and reached out at a time when there was so much uncertainty. This project gave me something new to work on that was unique to the moment, increased my connections at a time when connections were challenging, and led to findings that have the potential for longer-term impact.

Holland, M. L., Hunter, A. A., Livingston, N., & Bechtel, K. (2025). Identification of child-maltreatment-related emergency department visits from electronic health records. Academic Pediatrics, 103135. https://doi.org/10.1016/j.acap.2025.103135