Fig 1.
Ontology-driven data transformation in i2b2.
The ontology, which defines concept metadata, drives the transformation from i2b2 to OMOP. Data are retrieved from the i2b2 fact table, converted to OMOP codes via ontology lookups, and then written to the OMOP tables specified through the ontology concept path.
Table 1.
ARCH ontologies and terminologies vs OMOP.
Table 2.
ARCH Ontology modifiers vs. those in OMOP.
Table 3.
Codes that could not be found in the OMOP concept dictionary.
Table 4.
Transformation source to target table.
Fig 2.
Mapping distribution from ARCH terminologies to OMOP.
ICD and CPT codes map to six different tables in OMOP. This is just one (easily visualizable) aspect of the many complexities encountered in mapping. Boxes in the treemap are sized in a logarithmic scale.
Fig 3.
Achilles results on our “10% of Partners’ data” dataset.
From top left to bottom right: (a) data density (notice all are in the same magnitude); (b) age at first observation (notice the expected peak in 20s followed by decrease, with a spike at age 0 representing babies born in the hospital but not receiving follow-up care); (c) population distribution by race.