Machine Learning Tool to Identify Patients at Risk of Medication Errors
Author Information
Author(s): Abdo Ammar, Gallay Lyse, Vallecillo Thibault, Clarenne Justine, Quillet Pauline, Vuiblet Vincent, Merieux Rudy
Primary Institution: Institut d’Intelligence Artificielle en Santé, CHU de Reims, Université de Reims Champagne-Ardenne
Hypothesis
Can a machine learning-based tool effectively identify patients at high risk of medication errors upon hospital admission?
Conclusion
The machine learning tool significantly improved the identification of patients at risk of medication errors, outperforming existing methods.
Supporting Evidence
- The tool identified 45% of patients with medication discrepancies compared to 21% with the existing tool.
- The machine learning models were trained on data from 7200 patients.
- The voting classifier model showed a recall of 0.75 and an F1 score of 0.70.
Takeaway
Researchers created a smart computer program to help doctors find patients who might get their medications mixed up when they go to the hospital.
Methodology
Data from 7200 patients was analyzed using four machine learning models to predict medication errors based on 52 variables.
Potential Biases
The model's performance may not generalize to other healthcare settings due to its single-center design.
Limitations
The study was conducted at a single center and did not include patients from neonatology and intensive care units.
Participant Demographics
Patients admitted to various hospital departments, with a mean age of 73 years.
Statistical Information
P-Value
<0.0001
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website