Fig. 1. Mapping, propagation, and binning as a three-step process for clinical data harmonization from electronic medical records.

We present the hypothetical example of two individuals seen for an outpatient encounter or admission at 6 months or 7.5 months respectively. By (a) mapping clinical diagnosis terms to Human Phenotype Ontology (HPO) terms, the clinical data are harmonized to a joint ontological framework. By adding all higher-level (ancestral) terms, it is now possible to identify common shared higher-level terms, a method we refer to as (b) propagation. Finally, by (c) binning the propagated unique phenotype terms (removing duplicated terms) into discrete time intervals (3 months), phenotypes can be compared across both individuals. Shared HPO terms (e.g., “Seizures,” HP:0001250) are highlighted in bold. IMO Intelligent Medical Objects terminology.