Evaluating the outcome of Ambulances Collisions Review Panel in Middle Eastern environment using epidemiological supervised and unsupervised machine learning analyses.

Background: Ambulance collisions pose a significant occupational risk to personnel, patients, and the public. Despite ongoing efforts to improve safety measures, the complex nature of emergency response operations continues to pose challenges in reducing collision risks.

Objective: This study investigates the role of the dedicated Vehicle Collisions Review Panel at Hamad Medical Corporation Ambulance Service (HMCAS) in identifying, understanding, and managing risks associated with ambulance collisions.

Methods: A retrospective quantitative analysis of HMCAS ambulance collision records from 2023 was conducted using descriptive and bivariate analyses, along with supervised and unsupervised machine learning (ML) techniques - including multinomial logistic regression (MLR), decision tree (DT) analysis, association rule mining (ARM), and time series forecasting - to uncover hidden patterns, predictive insights, and future projections.

Results: A total of 131 ambulance collisions were analyzed. The majority of incidents involved emergency urban ambulances. MLR and DT achieved prediction accuracies of 41% and 35%, respectively. ARM revealed significant association between daytime incidents, normal road conditions, and the absence of patient involvement. Time series forecasting predicted a gradual increase followed by stabilization in collision incidents.

Conclusion: This study highlights the crucial role of a dedicated collision review panel in managing and mitigating ambulance collision risks. ML techniques provided evidence-based support for decision-making. Future research is needed to evaluate the long-term impacts of targeted training programs and safety protocols.