Table 3.

DILI EHR phenotyping algorithm implementation: data access in use

	CU	Mayo	MSSM
Diagnoses	ICD-9 codes and UMLS codes for NLP of discharge summary notes only	ICD-9 codes and NLP of clinical notes	ICD-9 codes
Medications	MED codes and UMLS codes for NLP of discharge summary notes only	RxNorm codes (data derived from a structured registry and NLP of drug orders)	Medication codes and text terms of medication fields
Laboratory values	MED codes	Laboratory information system	Test codes and text terms of procedure fields
Data access	CDW. 1 045 125 patients. 2004–10. Limited to inpatients, >1 year old	Mayo Clinic eMERGE cohort. 6916 patients. 2007–13	Biobank clinical datamart. 23 200 patients. 2006–13

CDW, clinical data warehouse; CU, Columbia University; DILI, drug-induced liver injury; EHR, electronic health record; eMERGE, Electronic Medical Records and Genomics; ICD-9, International Classification of Diseases, revision 9; Mayo, Mayo Clinic; MED, medical entities dictionary; MSSM, Mount Sinai School of Medicine; NLP, natural language processing; UMLS, unified medical language system.