NegEx Algorithm

Description: The NegEx algorithm is a simple one for identifying negatives in textual medical records. It was created by Will Bridewell in 2002 and refined in 2003 with 291 new phrases added to the “negation phrase list”. You input a text document to try and see if the NegEx algorithm will find a phrase or phrases within the document.

NegEx is different from the algorithms that they already had out to scan documents. The baseline algorithm that was already out “negates everything from the occurrence of the negation until the end of the sentence,”2 while “NegEx differs between two basic negation types…”3. NegEx catches things they call double negatives (e.g., “not ruled out”) that the original algorithm would miss. It is important for the software to catch these things so that they don’t miss anything during their diagnosis and can treat the patient as safe and as quickly as possible. In the medical field, clerical errors contribute to a lot of things that go bad, so not missing these important phrases that the original algorithm would miss is very helpful to the medical community.

Here are some numbers to show you how the NegEx algorithm differs from the original baseline one.
“NegEx had a specificity of 94.5% (versus 85.3% for the baseline), a positive predictive value of 84.5% (versus 68.4% for the baseline) while maintaining a reasonable sensitivity of 77.8% (versus 88.3% for the baseline).”4

negex_chart.JPG

Applications:

· Discharging patients – NegEx is used to scan the documents and charts of an outgoing patient to make sure that the doctors haven’t missed anything important. They may have looked over a symptom and putting the document in to be read by NegEx saves the doctors’ time without having to go through the records and look everything over with their own eyes.

Web Resources:

· http://web.cbmi.pitt.edu/~chapman/NegEx.html
· http://www.ncbi.nlm.nih.gov/sites/entrez?cmd=Retrieve&db=PubMed&dopt=AbstractPlus&list_uids=12123149
· http://rods.health.pitt.edu/LIBRARY/NLP%20Papers/chapman_JBI_2001_negation.pdf

Related Terminology:

· ConText
· MedLEE – A Medical Language Extraction and Encoding System
· UMLS – Unified Medical Language System (or here)

Citations/References:

1,2,3 Gindl, Stephan. Negation Detection in Automated Medical Applications. Vienna University of Technology. ©October 2006.

4 Chapman, Wendy W.; Bridewell, Will; Hanbury, Paul; Cooper, Gregory F.; Buchanan, Bruce G. A Simple Algorithm for Identifying Negated Findings and Diseases in Discharge Summaries. University of Pittsburgh. ©May 9th, 2002.

Graphics:

negex_code.JPGnegex_screenshot.JPG