EHR phenotyping

What does EHR phenotyping mean in clinical research?

Phenotyping is the task of finding patients with specific conditions or outcomes in EHR data. A 2018 review in the Annual Review of Biomedical Data Science calls it one of the most fundamental research problems in using EHR data, and the basis of translational research, comparative effectiveness studies, clinical decision support and population health analyses.^⁠[1] Methods range from rule-based definitions, such as diagnosis codes, lab thresholds and medication lists, to supervised and unsupervised machine learning models.^⁠[1]

Trial pre-screening is phenotyping with a protocol attached: each inclusion and exclusion criterion becomes a definition to test against the chart. Coded fields alone often miss what decides eligibility, such as disease stage, prior lines of therapy or a documented contraindication, because most digital data in healthcare are unstructured and need significant processing before research use.^⁠[2]

Practical note: ask any phenotyping tool to show the evidence behind each criterion decision, so a coordinator can check it in seconds instead of re-reading the chart. Identify reads structured and unstructured records against the protocol and shows criterion-to-evidence rationale for each ranked candidate.