Validating OMOP results with reproducible detailed data investigation to address false positive DOI-HOI pairs
The primary research objective of this work is to investigate three of the false positive DOI-HOI pairs that arose during the initial OMOP research. We are focusing on the false positive pairs (vs. false negative pairs) because they represent a particularly strong barrier to the adoption of automated surveillance methods. Three DOI-HOI pairs produced false positive associations consistently across the OMOP databases. The three false positives being studied include: 1) Antibiotics & Acute Renal Failure; 2) Typical Antipsychotics & GI Ulcer Hospitalization; and 3) Warfarin & Hip Fracture. The Informatics for Integrating Biology and the Bedside (i2b2) workbench will be used to explore these three false positive results. To start, we will develop a standardized approach for looking deeply into the data in order to understand better how to correctly identify true associations in the coded data, which, importantly, will be enriched by the addition of text notes from the patient record.
The secondary research objective is to build a process that will transform the OMOP data model and Ontology to the i2b2 platform. This will allow the extensive tool set in i2b2 to be utilized in the investigation of the three DOI-HOI false positive pairs. This will allow us to take a more in-depth look at these associations to assess the underlying problems that may be causing the pairs to be misrepresented. Overall, we plan to perform a transformation from the OMOP data model to the i2b2 data model, to use the resulting data on the i2b2 platform to investigate the false positive results, adding and transforming, (such as with natural language processing,) the data as necessary from our Electronic Medical Record system to investigate the results, and to transform some of the data back into the OMOP data model in order to recalculate the results using the derived data.
Deliverables of this research will be posted below as available.