Amilcare Gentili MD, Presenter: Nothing to Disclose

Brian E Chapman PhD, Abstract Co-Author: Nothing to Disclose

Using pyConText is possible to correctly classify reports containing critical results, simplifying the task of auditing reports for documentation of communication of critical test results.

The Joint Commission requires hospitals to define critical test results and the acceptable length of time between the availability of critical results and receipt of the results by the responsible licensed caregiver. Our hospital policy requires documentation of communication of critical results in the radiology reports. To audit compliance with this requirement we used pyConText to select reports containing critical results. ConText is a simple text-processing algorithm that uses simple lexical cues to relate modifying phrases, such as expressions of uncertainty, temporality, or negation, to findings described in text. For this study we used pyConText, a Python implementation of the ConText. pyConText has previously been used to identify and characterize pulmonary embolism findings in radiology reports, and to characterize personal and family history of ancillary cancers from the history and physical reports for patients with mesothelioma.  

All radiology reports obtained over a 2 month period were included in the study. We compared the reports classified as containing critical results by a radiologist reviewing the reports with pyConText classification.

Out of 18424 reports pyConText found 989 reports containing 1260 mentions of possibly critical diagnosis. Based on modifying phrases it classified these possibly critical diagnosis as positive in 157 cases and negative in 1193 cases, including 14 false negative and 45 false positive for a sensitivity of 88% and a specificity of 96%.

Gentili, A, Chapman, B, Use of pyConText to Classify Reports Containing Critical Results.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11014534.html