Table 1.

Large data classes available in electronic health records (EHRs)

	ICD codes	CPT codes	Laboratory data	Medication records	Clinical documentation
Recall	Medium	Poor	Medium	Inpatient: High Outpatient: Variable	Medium
Precision	Medium	High	High	Inpatient: High Outpatient: Variable	Medium high
Query method	Structured	Structured	Mostly structured	Structured, text queries, and natural language processing	Text queries, natural language processing, rarely structured
Strengths	Easy to query Serves as a good first pass of disease status	Easy to query High precision	Value depends on test High data validity	Can have high validity and marker of severity	Best record of what providers thought
Weaknesses	Disease codes are often used for screening Accuracy is hindered by billing realities and clinic workflow	Most susceptible to missing data errors Influenced by patient and payer factors	Normal ranges and units may change over time Aligning labs can be challenging	Often requires interfacing inpatient and outpatient records Medications prescribed are not necessary taken	Requires natural language processing May suffer from significant cut and paste May be self-contradictory
Has been used in PheWASs to date	Yes (often)	No	Limited	No	In pilot studies
Current challenges in PheWASs	Transitioning from ICD9 to ICD10 in the United States	Missing data	Aligning across sites and identifying normal values	Combining brand and generic names Mapping different forms (ophthalmic versus intravenous)	Extracting robust sets of concepts from notes

With the advent of the meaningful-use requirements and linked incentive payments from the Centers for Medicare and Medicaid Services, adoption of EHRs capable of logging each of the data types shown in this table has increased significantly; US adoption rates of certified EHRs currently exceed 90% for large hospitals (55). Abbreviations: CPT, Current Procedural Terminology; ICD, International Classification of Diseases; ICD9/10, International Classification of Diseases ninth and tenth editions, respectively; PheWAS, phenome-wide association study.