Observational Health Data Sciences and Informatics

Meeting Notes June 25, 2019

Topics

• Rimma Belenkaya went over the ETL instructions for ETLing oncology diagnosis modifiers from NAACCR formatted tumor registry date into the OMOP Oncology CDM extension. See draft document here: https://github.com/OHDSI/OncologyWG/blob/master/documentation/NAACCR.Ingestion.into.OMOP.Vocabulary.-.Diagnostic.Modifiers.docx
• Maxim Moinat raised the issue of whether he should map European tumor registry data to NAACCR or go directly to SNOMED and/or LOINC.
• It was argued that even though it is less than ideal to map to NAACCR, this is the current right choice. We need to conventionalize our representation of oncology diagnosis modifier structure and semantics to make analyses comparable across institutions.
• It was stated that making NAACCR the standard oncology diagnosis modifier vocabulary is just a bootstrap towards agreeing upon canonical representations in SNOMED. Hopefully via the Nebraska Lexicon.
• Maxim Moinat agreed to attempt to map his European tumor registry data to NAACCR. If the overlap with NAACCR is not possible, the advice to map to NAACCR will need to be reconsidered.
• The group was asked for volunteers to write SQL to derive Hemonc.org oncology drug regimens from low-level DRUG_EXPOSURE and EPISODE oncology diagnosis entries. Michael Gurley pointed to the following paper as a starting point: https://ascopubs.org/doi/pdf/10.1200/CCI.17.00002
• Dmytry Dymshyts expressed doubt about the possibility of derivation of oncology drug regimens from discrete DRUG_EXPOSURE data and EPISODE oncology diagnosis entries.
• The group agreed that validation of derivation will need to be compared to gold standard manually chart abstracted oncology drug regimens. Possibly from clinical trials or registry patients.
• Michael Gurley pointed out that some institutions' source systems will have pre-derived oncology drug regimens within their oncology drug regimen order template systems. These folks are the “lucky ones”.
• Michael Gurley mentioned that Jeremy Warner (one of the principal architects of Hemonc.org) thinks that the use of NLP against oncologist progress notes will be necessary to derive oncology drug regimens.
• Andrew Williams raised the need for the group to draft concrete oncology informatics use cases that the beta testing will be able to answer not previously answerable within the OMOP CDM.
• Michael Gurley pointed out that any person engaged in the ongoing support of oncology informatics requests (posted to an EDW, a research registry or quality tracking/improvement program) should be able to generate examples.
• The group decided to collect oncology use cases within a central location. An existing forum post was suggested as a good central location: http://forums.ohdsi.org/t/oncology-data-use-cases/4382