O2175: Smarter Transfers: The Potential of Artificial Intelligence Machine Learning to Dramatically Reduce Hospital Costs for Unnecessary Pediatric Appendicitis Transfers

Background: Community diagnosis of pediatric appendicitis can be challenging, with high risks associated with misdiagnosis. Often, patients are transferred to children’s hospitals for diagnostic confirmation. Unnecessary transfers contribute to healthcare overutilization and impose a burden on both the patients and their families. Using artificial intelligence, we created a transfer machine learning model (MLM) that predicts appendicitis diagnosis with 96% accuracy. We aim to outline the overall estimated costs (assessed from the hospital perspective) and the potential savings associated with the application of the MLM.

Methods: A retrospective cohort study was performed of all pediatric patients ( < 18 years) who were transferred to a children’s hospital within a large major metropolitan health system for appendicitis care from July 2021 to September 2023. Total cost was obtained from the hospital cost accounting system and inflated to 2025 US$ using the consumer price index for medical services. Private ambulance costs were imputed using the mean cost for transfers obtained from the hospital’s in-house ambulance service. Missing cost data (2%) was imputed with mean imputation by the disposition group. The previously developed MLM was applied to estimate the possible cost savings with implementation. Descriptive analyses were performed. IRB#HSC-MS-24-0073.

Results: Over the study period, 1,864 patients were transferred for appendicitis care. The median age was 10 years (IQR 7-13), and 58% were male. Governmental insurance was identified in 62% (1,153) and 8% were self-pay (156). The majority, 84% (1,562), were transferred from a hospital within the same health system as the children’s hospital. One-third (30%, 554) were discharged from the receiving emergency department (ED) without need for further treatment. Two-thirds (60%, 1,128) of all transferred patients underwent appendectomy, and 10% (182) were admitted for diagnosis, observation, or treatment with antibiotics. Ambulance was utilized in 86% (474) of the transferred patients discharged from the receiving ED. The total cost for the entire transferred cohort was $20,895,910 (95% CI $20,895,460- $20,896,360) with a mean cost per patient of $11,210 (95% CI $10,760- $11,660). The total cost for the ED discharged group was $1,038,653 (95% CI $1,038,480-$1,038,826) with a mean cost per patient of $2,120 (95% CI $1,946-$2,294). Applying the MLM, transfer would be prevented in an estimated 532 patients, for a projected overall total cost savings of $997,107 (95% CI $996,941- $997,273) or $2,035 (95% CI $1,868-$2,202) per transferred patient discharged from the receiving ED (TABLE).

Conclusion: The application of an artificial intelligence machine learning model to predict appendicitis at the time of transfer request stands to dramatically reduce overutilization. Over a three-year study period this represents a potential cost savings of one million U.S. dollars within a single health system. Further validation with multicenter trial and prospective implementation is required.