361 - Preliminary Development of an Algorithm to Detect Child Physical Abuse Using Emergency Department Records.

Publication Number: 361.6208

Amy Hunter, UCONN Health, Farmington, CT, United States; Zhenyu Xu, University of Connecticut School of Medicine, Vernon, CT, United States; Kun Chen, University of Connecticut, Storrs, CT, United States; Jun Yan, University of Connecticut, Storrs, CT, United States; Rob H. Aseltine, UCONN Health Center, Farmington, CT, United States; Shane J. Sacco, University of Connecticut School of Medicine, Farmington, CT, United States

Background: Victims of child physical abuse (CPA) disproportionately utilize the emergency department, yet identifying CPA remains challenging in this setting. There is an urgent need to develop tools supporting providers in diagnosing violence-related injuries in children. Developing a CPA predictive algorithm could improve identification and clinical familiarity with associated injuries and symptoms.

Objective: To develop a CPA predictive algorithm and assess its performance.

Design/Methods: Our study population included patients aged 0-17 captured in EPIC at a single pediatric hospital from May 2017 to March 2022. We developed a CPA predictive algorithm using statistical and machine learning techniques including XGBoost, lasso regression and deep neural network models. Model performance was evaluated using out-of-sample area under the receiver operating characteristics curve (AUROC), sensitivity, and positive predictive value (PPV). Performance metrics were stratified by age group and gender.

Results: Among 138,234 patients, 298 experienced CPA ( < 1%). Victims were most often 0-9 years, on Medicaid, and of Hispanic/Latino ethnicity. XGBoost demonstrated the best performance. Based on the random splitting procedure, the out of sample AUROC was 82.6% (SE= 1.6%). At 90% specificity, sensitivity was 57.7% (SE=3.1%) and PPV was 1.2% (SE=0.1%). Among high-risk patients, no gender differences in diagnoses were identified. Older children presented with behavioral health diagnoses, while young children presented with non-specific symptoms.

Conclusion(s): Our findings underscore the unique role of behavioral health professionals in recognizing CPA. Further study examining the intersectionality of gender and age among children at high risk for CPA may reveal opportunities for clinical intervention.