Predicting inpatient pharmacy order interventions using provider action data.

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Clinical Decision Support

Background:

• Electronic health records (EHRs) introduce new medication ordering errors and inefficiencies.
• Current methods for identifying problematic orders rely on patient medical record comparisons.
• New approaches are needed to address EHR-related medication order issues.

Purpose of the Study:

• Develop a machine learning model to identify medication orders requiring pharmacist intervention.
• Utilize provider behavior and contextual EHR features, excluding sensitive patient data.
• Improve the efficiency and accuracy of medication order review processes.

Main Methods:

• Collected provider actions and pharmacy order data from a hospital EHR system over two weeks.
• Developed a classification model to detect orders needing pharmacist intervention.
• Tuned the model to its deployment context and evaluated feature importance.

Main Results:

• The developed machine learning model achieved an area under the receiver-operator characteristic curve of 0.91.
• The model's area under the precision-recall curve was 0.44.
• Provider behavior was identified as a significant predictor for intervention-requiring orders.

Conclusions:

• Provider actions within EHR systems are valuable predictors for identifying medication orders needing pharmacy intervention.
• Context-specific model tuning is crucial for creating effective clinical tools.
• This method offers a way to improve health outcomes by optimizing medication safety without compromising patient privacy.