High alert drugs screening using gradient boosting classifier.

Area of Science:

• Medical Informatics
• Pharmacovigilance
• Machine Learning in Healthcare

Background:

• Prescription errors involving high alert drugs (HAD) pose significant risks in healthcare.
• Existing hospital interventions and guidelines have shown limited effectiveness in reducing these errors.

Purpose of the Study:

• To develop and evaluate a machine learning-driven clinical decision support system (CDSS) for identifying high alert drug prescription errors.
• To assess the efficacy of a Gradient Boosting Classifier model in screening for inappropriate HAD use.

Main Methods:

• A machine learning model utilizing Gradient Boosting Classifier was developed.
• The model was trained and tested on outpatient and inpatient drug prescriptions from Maharaj Nakhon Chiang Mai hospital in 2018.
• Screening parameters were established to identify potential HAD prescription errors.

Main Results:

• The machine learning algorithm demonstrated high performance in screening for HAD prescription errors.
• The model achieved over 98% accuracy in identifying actual HAD mismatches in the test set.
• An accuracy of 99% was observed in the evaluation set for detecting HAD prescription errors.

Conclusions:

• Machine learning, particularly through CDSS, shows significant potential for reducing errors in high alert drug prescriptions.
• The developed HAD screening protocol effectively identifies at-risk prescriptions, enhancing patient safety.
• This study highlights the crucial role of machine learning in improving medication safety and preventing adverse drug events.