Instructors:
Werner Ceusters
William R. Hogan
Electronic health records (EHR) consist in part of structured data such as demographics, coded diagnoses and procedures, medications, problem list, adverse events, results from lab tests and so forth. This list will most likely expand with the move towards 'meaningful use'. To make these data useful for translational research, they are often exported to data warehouses and pooled with similar data from other systems (e.g. the i2b2 initiative). This is most typically achieved by converting data to a common format and applying data mapping procedures using reference terminologies. These techniques often lead to some form of data loss in terms of granularity or terminological detail.
In this tutorial we will explain and provide practical examples of how Ontological Realism (OR) and Referent Tracking (RT) can be used to achieve data enrichment in data warehouses rather than data loss. This will be done by demonstrating how the principles of OR and RT, thereby further exploiting how the entities represented in the Ontology of General Medical Science (OGMS) relate to each other, can be used to improve the transform component of traditional Extract-Transform-Load (ETL) procedures in data warehousing. For example, explicit representation of entities referred to by ICD-9-CM diagnosis codes can simplify querys and help track individual diseases over time. We will also demonstrate that our methods will not lead to a completely faithful representation of the reality expressed by means of the EHR data because of underspecification in prevailing EHR paradigms. It will however be a means to provide recommendations for future meaningful use criteria.
This half-day tutorial will be of interest for anyone involved in:
Attendees may submit until one month prior to the tutorial examples of what they are practically struggling with in their environment. These examples will be worked out during the tutorial.