.png){kind=icon}
.png){kind=icon}
Return
A multivariable explanation of emergency response time for highway crash incidents
.png){kind=icon}
PRE
AI
Y
H
L
DOI：10.1080/12265934.2026.2679511.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Emergency Response Time (ERT), defined as the time interval from the occurrence of a traffic accident to the arrival of Emergency Medical Services (EMS) at the accident location, is a fundamental factor that affects survival outcomes and the severity of injuries. It is indeed a daunting task to skillfully map and explain different ERT scenarios, given that the cause factors are complicated, nonlinear, and heavily intertwined. Prior research has addressed the matter in a relatively simple or isolated manner. They fail, however, to offer insights into the integrative impact of spatial, temporal, operational, and crash-related variables through a universal model, let alone various road accident situations under the real-world environment. This study presents a multivariable analysis aimed at explaining variations in ERTs across diverse highway crash incidents in the United States. Using a real-world dataset comprising categorical and continuous variables, latent patterns influencing response efficiency were explored. By employing clustering techniques such as K-Means and Hierarchical Density-Based Clustering, it was possible to unveil the data’s hidden patterns and structure. After that, using internal validation with a few metrics like the Calinski-Harabasz Index (CHI), Davies-Bouldin Index (DBI), and Silhouette Score was carried out. Clusters were interpreted as distinct incident profiles, revealing associations between contextual crash features and emergency response delays. Subsequently, the dataset was discretized into separate datasets based on clustering labels to facilitate supervised prediction modelling. It turned out from the prediction results that the Random Forest (RF) fine-tuned by Lung’s Performance-based Optimization (LPO) reached the top performance in Clusters 1, 2, and 3. During the testing phase, it scored R2 figures of 0.954, 0.876, and 0.845, RMSE figures of 1.613, 2.602, and 2.909, and NSE figures of 0.954, 0.876, and 0.845, each in turn.
Keywords:
Emergency reaction interval
unregulated grouping
regulated modeling
data engineering
analysis of sensitivity
Journal
I
IF:
3
Papers:
463
Citations:
1.1K
Organization
Cited Papers