Assessment and Improvement of Drug Data Structuredness From Electronic Health Records: Algorithm Development and...

Area of Science:

• Health Informatics
• Data Science
• Pharmacology

Background:

• Digitization enables insights from retrospective health data, but unstructured healthcare systems hinder research.
• Real-world data accessibility is crucial for unbiased big data research.
• Drug data often lacks interoperability, posing a significant challenge.

Purpose of the Study:

• To present an approach for identifying and structuring drug data.
• To standardize drug information using Anatomical Therapeutic Chemical (ATC) classification.

Main Methods:

• An initial analysis assessed the baseline structuredness of local drug data.
• Three algorithms were applied to unstructured text to generate ATC codes via string and similarity matching (Levenshtein distance).
• Expert validation and final assessment determined the increased structuredness.

Main Results:

• Initial structuredness was 47.73%; the approach increased this to 85.18% for 1.76 million drug prescriptions.
• Algorithm combinations achieved high correctness rates: 100% (algorithms 1, 2, 3), 99.6% (algorithms 1, 3), and 95.9% (algorithms 1, 2).
• The approach successfully standardized a large volume of medication prescriptions.

Conclusions:

• A product catalog alone is insufficient for generating structured data.
• The 4-step approach effectively and automatically increases data structuredness.
• Similarity matching shows promise for unstructured entries, warranting further research for algorithm enhancement.