Automated identification of adverse events related to central venous catheters
Journal
J Biomed Inform
Publication Date
2007 Apr
Volume
40
Issue
2
Pages
174-82
Summary:
- HIT Description: natural language processing program and phrase-matching algorithm to identify adverse events related to central venous catheter placement documented in an electronic medical record More info...
- Purpose of Study: evaluate the sensitivity and specificity of two different semi-automated processes for identifying adverse events documented in an electronic medical record
- Years of study: 1999-2004
- Study Design: other predictive analysis
- Outcomes: impact on patient safety
- Settings: electronic records of patients admitted to the surgical service in the Surgical or Medical intensive care units of the Salt Lake City VA Medical Center who had a central venous catheter (CVC) placed
- Intervention: Compare the identification of adverse events (AE) related to CVC placement using a natural language processing system, word and phrase-matching algorithms and manual chart review
- Evaluation Method: chart review reference standard
- Description: Used Medical Language Extraction and Encoding system (MedLEE) (a natural language processing system) on a partially simulated data set to optimize MedLEE and its ability to identify adverse events. Developed phrase-matching system using common phrases related to CVC use, suggestions from three general surgeons and synonymous phrases for CVCs and AEs from the Unified Medical Language System Metathesauras
- HIT System Sustainability: pselling errors and local abbreviations
- Quality of Care and Patient Safety Outcome: The natural language processing system had relatively low sensitivity (50%) but high specificity (90%) while the phrase-matching system had higher sensitivity (70%) and lower specificity (55%). Using the two as a combined instrument resulted in ÒacceptableÓ sensitivity and specificity (72% and 80%)