The identification of outcomes within EHRs may be easier when computable phenotypes have been created for the conditions of interest.
A computable phenotype is defined as “a clinical condition, characteristic, or set of clinical features that can be determined solely from the data in EHRs and ancillary data sources and does not require chart review or interpretation by a clinician… For a phenotype definition to be valid, it must identify the condition for which it was developed and claims to identify and meet the desired degree of sensitivity and specificity" (Richesson and Smerek 2014).
The NIH Collaboratory’s Phenotypes, Data Standards, and Data Quality Core, which includes representatives from the PCT Demonstration Projects, has published an article describing approaches for using “phenotypes” data to identify clinically equivalent populations across multiple sites. The Core describes how to assess whether the data collected from health systems are comparable, valid, and reliable (Richesson et al. 2013).
Some key questions and considerations related to outcomes within EHRs include the following:
Is the outcome a medically significant such that a patient would seek care?
- Will the endpoint be medically attended?
- Does it require hospitalization?
- Is the treatment for the outcome generally provided in inpatient or outpatient settings?
- Outpatient events may include diagnoses that justify a specific test order, also called a “rule out”, and rule-out diagnoses might not indicate true outcomes.
What is the intensity of medical care?
- If high, as with a myocardial infarction, then there will be a clear record in claims and/or EHR data.
- If low, as with the gout example, there may or may not be a record of the event. A solution to this problem is to use a PRO, and reach out to the participant at specified intervals.
Where would the signal show up?
- EHR (laboratory values, treatments, etc.)
- Claims data (does the event generate a bill?)
What sensitivity is required?
- Conditions may not be consistently and reliably recorded.
Will the data be structured or unstructured?
- If structured, the data may be usable as is.
- If unstructured, some work will need to be done to ensure the data are captured in a uniform way. (Prompts can be added to the EHR system, clinicians can be given special training, etc.)
Are there overlapping conditions (e.g., chest pain and unstable angina)?
- The data may need adjudication, especially if they are claims data.
An additional consideration, discussed in greater detail in the chapter Using EHR Data, is the ability to capture outcomes from a number of different sources over time. For example, in the Aspirin Dosing: A Patient-Centric Trial Assessing Benefits and Long-term Effectiveness (ADAPTABLE) trial, patients enroll through an online portal, which also provides online consent and randomization. Data collected during routine care are used in the study, however, capture complete information about all outcomes—which include myocardial infarction, mortality, and hospitalizations—participants are asked to login to the portal and report hospitalizations. If the patient does not go back to the portal, the call center at the Duke Clinical Research Institute calls for follow up.
(See the chapter Using EHR Data for more on how claims data are linked to study data.)
Electronic health records based phenotyping in next-generation clinical trials: a perspective from the NIH Health Care Systems Collaboratory (Richesson R, et al. J Am Med Inform Assoc 2012)
Living Textbook resource chapter: Electronic Health Records-Based Phenotyping
Richesson RL, Hammond WE, Nahm M, et al. 2013. Electronic health records based phenotyping in next-generation clinical trials: a perspective from the NIH Health Care Systems Collaboratory. J Am Med Inform Assoc. 20:e226–e231. doi:10.1136/amiajnl-2013-001926.
Richesson RL, Smerek M. Electronic Health Records-Based Phenotyping. In Rethinking Clinical Trials: A Living Textbook in Pragmatic Clinical Trials. NIH Health Care Systems Research Collaboratory. Published June 27, 2014.