Development of an ensemble resource linking MEDications to their Indications (MEDI)

Understanding of medications-disease relationships is critical to distinguish indications from adverse effects, and medication exposures serve as important markers of disease and severity in electronic medical records (EMR). We created a computable medication-indication (MEDI) resource by applying natural language processing and ontology relationships to four public medication resources. Physicians evaluated accuracy of medication-indication relationships. MEDI contained 3,112 medications and 63,343 medication-indication pairs derived from the four resources, whose precisions varied from 56-94%. The MEDI high precision subset (MEDI-HPS) includes indications found within either RxNorm or ≥2 resources and had an estimated precision of 92%. MEDI-HPS contains 13,304 unique indication pairs for 2,136 medications. MEDI is a free, computable resource that links medications with their indications as represented by formal concepts and may assist clinical and research uses of EMR data.