Developing a machine learning model to predict patient need for computed tomography imaging in the emergency
Amirmohammad Shahbandegan1, Vijay Mago1, Amer Alaref2
1Department of Computer Science, Lakehead University, Thunder Bay, Ontario, Canada.
Plos One
|December 15, 2022
Summary
This study developed a machine learning model to predict the need for Computed Tomography (CT) scans in the emergency department (ED) using early triage data. The model helps identify patients requiring CT scans sooner, improving resource allocation and reducing ED overcrowding.
Area of Science:
- Emergency Medicine
- Health Informatics
- Machine Learning
Background:
- Hospital overcrowding, particularly in emergency departments (ED), poses significant challenges to patient care and staff efficiency.
- Timely diagnostic imaging, such as Computed Tomography (CT) scans, is crucial for patient management but can contribute to delays.
- Early identification of patients requiring CT scans is essential for optimizing patient flow and resource allocation within the ED.
Purpose of the Study:
- To develop and validate a machine learning model for the early detection of patients needing a Computed Tomography (CT) exam in the emergency department (ED).
- To utilize readily available administrative triage data for predicting the requirement of CT scans upon patient arrival.
- To assess the model's performance in identifying patients who will ultimately require a CT scan.
Main Methods:
- Data from 81,118 patient visits to the Thunder Bay Regional Health Sciences Centre ED (05/2016-05/2017) were used for model training.
- Various combinations of text embedding, machine learning algorithms, and data resampling techniques were evaluated.
- The optimal model was tested on a separate dataset of 9,013 visits, analyzing factors like chief complaint, treatment area, and triage acuity.
Main Results:
- The best-performing machine learning model achieved a Receiver Operating Characteristic Area Under the Curve (ROC AUC) of 0.86.
- The model demonstrated a sensitivity of 87.3% and a specificity of 70.9% in identifying patients needing a CT scan.
- Chief complaint, treatment area, and triage acuity were identified as the most significant predictors for CT scan orders.
Conclusions:
- A machine learning model effectively predicts the need for CT scans using initial administrative triage data in the ED.
- Early identification of CT scan requirements can facilitate prompt resource allocation and expedite diagnostic imaging.
- Implementing this model has the potential to improve patient flow, reduce diagnostic delays, and mitigate ED overcrowding.
