Estimating Patient Condition Codes Using Data Mining Techniques
Abstract
The PROBLEM: TBE wanted to improve our TML + mortality algorithm by factoring in the chance of death as a function of time based on real-world data. The current DOW mortality algorithm is based on responses from a panel of medical SMEs rather than empirical data. Due to time limitations and the amount of data, we decided to pursue an automated method for assigning patient condition codes. The Combat Trauma Registry (CTR) database maintained by the Naval Health Research Center records medical data from patients treated at Navy and Marine corps medical facilities. Each CTR record documents a single visit (encounter) by a single patient. Each record contains over 50 individual fields of data, though often the data is incomplete. The most relevant data (such as injury descriptions) are primarily free text containing numerous abbreviations, medical terms and misspellings.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2007
- Accession Number
- ADA488437
Entities
People
- Douglas Lowe
- Joseph Parker
- Martin Hill
- Michael Galarneau
- Peggy Fridlund
- Ray Mitchell
Organizations
- Teledyne Technologies