Methods for managing variation in clinical drug names.

Area of Science:

• Pharmacoinformatics
• Natural Language Processing in Healthcare
• Clinical Data Management

Background:

• Clinical drug names exhibit significant variation due to abbreviations, formatting, and salt forms.
• Managing this variation is crucial for accurate data matching in electronic health records and formularies.
• Existing generic normalization methods show limited effectiveness in handling drug-specific nomenclature.

Purpose of the Study:

• To develop and evaluate drug-specific normalization methods for clinical drug names.
• To improve the recall of drug names from local formularies against standard terminologies like RxNorm.
• To assess the impact of normalization on data ambiguity.

Main Methods:

• Manual analysis of drug names from RxNorm and local formulary data.
• Identification of three key normalization rule types: abbreviation expansion (e.g., 'tab' to 'tablet'), element reformatting (e.g., number-unit spacing), and salt variant removal (e.g., 'succinate').
• Application of these drug-specific rules to a dataset of non-matching drug names.

Main Results:

• Drug-specific normalization achieved an overall recall of 45% for previously non-matching names, with some subsets reaching 70%.
• This represents a substantial improvement compared to 10-20% recall achieved with generic normalization.
• The normalization process did not significantly increase ambiguity within the RxNorm dataset.

Conclusions:

• A targeted set of drug-specific normalization operations offers superior performance over general language normalization techniques.
• These methods effectively address common variations in clinical drug nomenclature.
• The findings support the implementation of drug-specific rules for enhanced clinical data interoperability.