Extracting antipsychotic polypharmacy data from electronic health records: developing and evaluating a novel process

Area of Science:

• Psychiatry
• Clinical Informatics
• Pharmacology

Background:

• Antipsychotic prescription data is often limited to structured fields.
• Investigating antipsychotic polypharmacy (APP) requires comprehensive data sources.
• This study presents a novel method for extracting APP data from electronic health records (EHRs).

Purpose of the Study:

• To describe and evaluate a new method for extracting APP data from EHRs.
• To assess the precision and recall of the novel extraction method.
• To estimate the prevalence and patterns of APP in patients with serious mental illness (SMI).

Main Methods:

• Utilized anonymised EHRs from South London and Maudsley NHS Foundation Trust.
• Employed natural language processing and a bespoke algorithm to extract antipsychotic co-prescribing information.
• Validated extracted data against a manually coded gold standard.

Main Results:

• High precision (0.94-0.97) and moderate recall (0.57-0.77) for individual antipsychotic prescriptions.
• Baseline APP detected with 0.92 precision and 0.74 recall.
• Long-term APP identified in 4.7% of SMI patients, with specific drug combinations noted.

Conclusions:

• The novel method is a practical tool for large-scale identification of polypharmacy in mental health EHRs.
• Extracted data facilitates research into polypharmacy patterns, predictors, and patient outcomes.
• This approach enhances the understanding of complex medication prescribing in SMI.