Common Data Model Harmonization (CDMH) - FHIR Implementation Guide

The nation is reaching a critical mass of Health IT systems (EHRs, Data Warehouses etc.) that comply with data and vocabulary standards. The wide deployment of Health IT systems has created unique opportunities for providers, provider support teams, healthcare professionals, and researchers etc. to access and use the patient data that is already collected during clinical workflows. However in order to use the data collected during clinical workflows by researchers, typically data is extracted, curated and then populated into data marts from which researchers access this data.

However this data is typically curated, managed and stored based on different use cases and different data models are used. Some of the data models used for research are PCORnet CDM (Common Data Model), OMOP, i2b2 and Sentinel. The usage of different data models make it hard for researchers to write queries and extract data from multiple data marts without going through a lot of intensive mapping and translation processes.

This project intends to harmonize these different models to the BRIDG conceptual model and create mappings. The gaps found during the mappings will be added to the BRIDG model. BRIDG 5.1 which is in September 2018 ballot cycle has the CDMH project semantics additions. The CDMH subset of the BRIDG 5.1 model will inform the database design of the CDMH physical model. The project further provides mappings from the CDMH subset of the BRIDG model to FHIR. These mappings are expected to be used for enabling data access by composing queries in a common format, translating these queries to the appropriate local format, extracting the results and then translating the results back to a common format. For the purposes of the project the extracted results are planned to be translated into two different formats namely HL7 FHIR and CDISC SDTM to facilitate common APIs for accessing data using FHIR and then providing regulatory compliance with FDA requirements using SDTM.

From a standards perspective, the project will specifically perform the following

- Create and update FHIR profiles which can be used to represent the data that is being accessed by the researchers.

- Update the DAF-Research IG which has already created some of these profiles.

- Update the Data Extraction and transformation processes within DAF-Research to use the various profiles.

The project will use the entire inventory of FHIR resources and propose extensions as needed. The profiles created for this effort will be for US Realm and will re-use the US-Core Profiles.

Note: This is currently being scoped to US Realm and the effort will leverage US Core Profiles. There may be a need for a follow-up effort to this since EU is also interested in similar effort in relation to I2B2, OMOP, PCORnet and Sentinel.