JMIR Publications

ABSTRACT

Background:

The increasing number of attendees at the pediatric emergency departments could negatively affect those children triaged as requiring urgent care, continuous monitoring and analysis of admissions and waiting time of pediatric emergency patients become a necessary. Though the shortage of pediatric medical resources is a big challenge for China’s health care system, few related large-scale studies have performed to analyze the pediatric emergency room visits.

Objective:

This study aims to investigate the characteristics and admission preferences of pediatric emergency department attendances using electronic medical record (EMR) data, construct and evaluate machine learning models to predict the waiting time of the pediatric emergency department visits.

Methods:

This retrospective analysis included patients admitted to the emergency department of Capital Institute of Pediatrics, Affiliated Children’s Hospital from January 1, 2021 to December 31, 2021. Clinical activities of the admissions were extracted from the EMR, and relevant variables of interest, including patient demographics, clinical diagnosis, and timestamps of clinical visits were collected and compared for each indicator. We also constructed and assessed multiple computational models for waiting time prediction.

Results:

In total, 183,024 eligible admissions from 127,368 pediatric patients were included. During the 12-month study period, pediatric emergency department visits were most common in children aged less than 5 years (71.26%, 130,423/183,024), and relatively more male patients (56.90%, 104,147/183,024) were observed than female patients (43.10%, 78,877/183,024). Fever (27.71%, 50,715/183,024), respiratory infection (23.64%, 43,269/183,024), celialgia (5.22%, 9,560/183,024) and emesis (3.77%, 6,898/183,024) were the leading causes of pediatric emergency room visits. The average number of daily admissions was 501.44, and 18.76% (34,339/183,204) PED visits discharged without a prescription or further tests. The median waiting time from registration to seeing a doctor was 27.53 minutes, and prolonged waiting times were observed from April to July with the increased number of arrivals mainly suffering from respiratory diseases. For waiting time prediction, on average, compared with regression methods, Random Forest, LightGBM and Xgboost reduced the root mean-square error by an average of 17.73% and increased the R-square by approximately 29.33%. The SHAP method revealed that the feature “wait.green” and “department” were the most influential factors.

Conclusions:

This study gives an up-to-date investigation of pediatric emergency room visits, it can be sure that the pediatric emergency admission rates highly varied between different time periods and there were certain admission regularities. The machine learning models, especially the ensemble methods, provided more reliable waiting time prediction. The number of patients waiting for consultation or treatment and the triage status were the critical factors for long waiting. Therefore, conducting patient diversion to relieve congestion from emergency departments, or optimizing triage system to reduce average waiting times, should still be effective strategies to improve the quality of pediatric service in China.