Predictive Analytics
The Challenge
Configure CV for Sepsis product on the hospital’s retrospective data and run simulations to test the correct behavior of the rules engine.
The Approach
- Analyzed and investigated the raw-data files and made sure all the required elements by CV are present i.e., verify for the completeness of the received flat-files. Focus only on the fields that CV requires and ignore other fields from hospital data files
- Converted the raw flat-files into intermediary XML format using the tool (script transform-flatfile.sh) provided by client that uses XSLT transforms to produce client’s standard XML
- Analyzed the CV rules engine to ensure that it handles the data correctly. i.e., perform data normalization (review datanormalization.clp and update the rule, if needed). For the pilot effort, most of the data normalization will be performed in the lab data fields, for instance – CBC, urinalysis, etc.
- Redefined the rules so that the rules are asserted and data normalization works properly
- Applied data transformation wherever needed
- Ensured CV rules engine is ready to perform simulation test
- Investigated alerts for validity: Verified that triggering conditions are correct, check whether any alerted patients had ICD-9 coded discharge diagnoses for septicemia, sepsis, severe sepsis, or septic shock
- Prepared test results report
Accomplishments
- Flat-files can be analyzed and converted into XML format using client provided tool
- Successfully configured the CV rules engine
- Generated report of simulation test results
- Fully HIPAA-compliant