Using electronic health record data to link families: an illustrative example using intergenerational patterns of...

Area of Science:

• Health Informatics
• Population Health Research
• Genetics and Genomics

Background:

• Electronic health record (EHR) data is valuable for population health research.
• EHRs lack critical relational information, limiting comprehensive community representation.
• Emergency contacts offer a novel way to link family members within EHRs.

Purpose of the Study:

• To adapt and validate an algorithm for inferring familial relationships from EHR data.
• To assess the utility of inferred family structures in population health studies.
• To model the association between parent and child obesity using the derived family data.

Main Methods:

• Revised a published algorithm, "relationship inference from the electronic health record" (RIFTEHR), into Pythonic RIFTEHR (P-RIFTEHR).
• P-RIFTEHR identifies emergency contacts, matches them to patients using network graphs, and infers relationships.
• Applied P-RIFTEHR to Northwestern Medicine Electronic Data Warehouse data (approx. 2.95 million individuals) and validated using mother-child pairs.

Main Results:

• The P-RIFTEHR algorithm successfully matched over 1.15 million individuals into 448,278 families.
• Family structures included up to 4+ generations, with a median size of 2.
• Mother-child pair validation achieved 95.1% sensitivity; parental obesity was associated with a 2.30 odds ratio for child obesity.

Conclusions:

• P-RIFTEHR effectively identifies familial relationships in large, diverse health system populations.
• The inferred family structures enable reliable estimation of obesity heritability.
• Findings align with traditional cohort studies, demonstrating the algorithm's validity and utility.