MEDI-2

MEDI-2 provides comprehensive medication–indication mappings for analysis of electronic health records to support drug utilization studies, off-label use detection, and pharmacovigilance.

Key Features:

• Expanded dataset: Contains 3,031 prescribable medications and 186,064 indications, representing a 13% increase in medications and more than a threefold increase in indications compared to MEDI-1.
• High Precision Subset (HPS): MEDI-2 HPS comprises indications verified within RxNorm or corroborated by at least three additional resources to improve reliability.
• Enhanced natural language processing extraction: Indications were extracted using natural language processing from six publicly available resources: RxNorm, Side Effect Resource 4.1, Mayo Clinic, WebMD, MedlinePlus, and Wikipedia.
• Performance and evaluation: Overall precision is 0.60 and HPS precision is 0.92; overall recall is 0.89 (up from 0.79 in MEDI-1) and HPS recall is 0.65 (up from 0.55), with combined MEDI-1 and MEDI-2 recall of 0.95.

Scientific Applications:

• Drug utilization analysis: Enables in-depth analyses of prescribing and medication use patterns within EHR data.
• Off-label and secondary use investigation: Supports identification and study of off-label or secondary medication uses.
• Pharmacovigilance and adverse event monitoring: Improves contextualization of adverse events by providing indication mappings for medications.

Methodology:

Indications were systematically extracted using natural language processing and ontology relationships across RxNorm, Side Effect Resource 4.1, Mayo Clinic, WebMD, MedlinePlus, and Wikipedia.