374 - Identifying Law Enforcement Presence in Pediatric Emergency Department Encounters: A Novel Key-Word Algorithm

Poster Number: 374
Publication Number: 374.148

Background: The role of law enforcement officers (LEOs) in the pediatric emergency department (ED) is vague and variable. It is unclear how LEO presence may delay or otherwise affect patient care. Quantitative analysis of LEO prevalence in EDs and potential impact on health outcomes has been limited by challenges identifying LEO presence via electronic health record (EHR) documentation.

Objective: To develop an EHR based key-word algorithm that accurately identifies LEO presence in a pediatric ED.

Design/Methods: The study included ED encounters from 10/1/2021-1/31/2022 at a high-volume children’s hospital ED. We identified a retrospective “high prevalence cohort” with higher likelihood of LEO presence for algorithm development and testing (figure 1). This cohort included patients ≤22 years of age who arrived via emergency medical services or public (e.g. LEO) vehicle; walk-in, private car, and air arrivals were excluded for algorithm development. LEO presence was identified by one discrete EHR data field and key-word algorithm. The initial algorithm included 14 unique words chosen by pediatric ED clinicians located in two note types (ED notes, ED provider notes). We manually reviewed 406 random encounters to determine LEO presence and create a reference standard set. We then tested the algorithm on the reference set and excluded words with poor specificity and adjusted certain words to be more inclusive (figure 1). We then reapplied the adjusted algorithm to the reference standard to reevaluate specificity and sensitivity. The final algorithm was then applied to the entire sample. LEO prevalence and algorithm sensitivity, specificity, and positive predictive value are reported.

Results: 3,884 encounters were included in the high prevalence cohort for algorithm development. LEO presence was detected with 95% sensitivity, 94% specificity, and 94% accuracy (PPV 64%). LEO presence was identified in 566 of encounters (15%). LEO presence was higher among those who identified as Black (19.7%, p=0.002) and lower among those who identified as Asian (6.9%, p=0.048) when compared to those identifying as White (13.8%) (table 1). There was no significant difference in LEO presence between Hispanic/Latino patients and Not Hispanic/Latino patients (p=0.061).

Conclusion(s): Our novel algorithm identified LEO presence in EHR records in an enriched population, overcoming the largest hurdle to quantitative analysis. We noted significant racial disparities in LEO presence. The algorithm and results require further external validation in the general population to facilitate examination of LEO prevalence and impact on outcomes.