Back

Building Population Phenotypic Journeys from Laboratory Tests in Electronic Health Records for Translational Research

Zhang, X. A.; Lee, K.; Jin, L.; Liu, Z.; Ai, L.; Jun, T.; Oh, W.; Higashi, M.; Pan, Q.; Stolovitzky, G.; Schadt, E.; Robinson, P. N.; Wang, X.

2022-10-11 health informatics
10.1101/2022.10.10.22280880 medRxiv
Show abstract

Abundant volumes of clinical laboratory test results available within Electronic health records (EHRs) are essential for differential diagnosis, treatment monitoring, and outcome evaluation. LOINC2HPO is a recently developed deep phenotyping approach to transform laboratory test results into the Human Phenotype Ontology (HPO) terms. Here, we deployed the approach on a large EHR dataset from the Sema4 Data Warehouse to build patient phenotypic journeys at scale. Among 1.07 billion laboratory test results, we successfully transformed 774 million (72.5%) into HPO-coded phenotypes and built a patient phenotypic journey for over 2.2 million patients. First, a global analysis of patient phenotypic journeys revealed a longitudinal increase in patients with genitourinary system abnormality. The analysis also revealed abnormal phenotypes with strong racial patterns. Second, using severe asthma as an example case, we identified abnormal phenotypes in the past three years that were correlated with asthma progression to severe state. Lastly, we demonstrated that converting laboratory test results into HPO terms resulted in limited information loss. Our study demonstrated that the phenotypic journey framework opens the way to characterize phenotypic trajectories in population level and screen biomarkers for translational research.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.