Patterns, Trends and Profiles of Acute Poisoning and Predictors of Hospitalization and Mortality in Prehospital EMS Records in a U.S. Based Dataset

Abstract

Objectives: To describe the distribution of poisoning cases reported in the ESO Collaborative dataset by sociodemographic factors and mechanisms of poisoning. To investigate predictors associated with hospitalization and mortality among poisoned patients treated in the pre-hospital EMS care. Methods: A cross-sectional design, secondary data with a multimethod analysis strategy. The ESO Data Collaborative records of the 2022 calendar year were used. It is a de-identified prehospital EMS electronic health records containing more than 1,300 EMS agencies across the U.S. A sample of 5,278,986 poisoning cases was obtained. Data management, cleaning, and analysis were performed using RStudio version 2025.05.0+496. Results: The annual incidence of poisoning encounters during 2022 was 0.41 based on the ESO dataset. The majority of encounters were adult males aged between 20 and 40 years old. There is an even distribution of poisoning cases over the day, with 3 peak times at mid-day, afternoon, and late afternoon hours. 67.3% of poisoning encounters were among the white race, and more than half of the cases (54.77%) occurred in the southern region of the United States. The majority of cases (67.5%) were in urban areas; however, rural areas had significantly more fatal poisoning, and super rural areas had significantly higher severity of poisoning cases. Thematic content-analysis showed that the majority of poisoning encounters were due to substance use, among substances, opioids and alcohol were the most commonly used. Significant predictors of hospitalization were an increase in weight, the Midwest and western geographical regions, a super rural area, and having at least 1 impairment. Significant predictors of mortality were older age, an increase in weight, female gender, the Midwest and south geographical regions, rural area and having at least 1 impairment, and the patient being unconscious. Conclusion: Identified significant predictors in this study help in preventing future poisoning-related hospitalizations and deaths by rapid prediction of severity (triage at the prehospital level). In the aim of reducing the burden of poisoning, the integration of prehospital EMS records with hospital records provides a comprehensive understanding of patient disposition and clinical outcomes. Predictive modeling and machine learning techniques can be used to facilitate the identification of high-risk patients using real-time prehospital data.