Nima Momenin MD, Presenter: Nothing to Disclose

Fatemeh Abdollahi Mofakham MD, Abstract Co-Author: Nothing to Disclose

Andrew Berwick, Abstract Co-Author: Nothing to Disclose

Wilbur L. Smith MD, Abstract Co-Author: Nothing to Disclose

The aim of this study is to compare the frequency and spectrum of document errors in finalized chest radiology reports generated in pre- Voice Recognition with the reports made in 3 discrete intervals after implementation of the voice recognition (VR) in order to assess the effect of VR technique on the quality of the reports. We also evaluated progress in report quality in using voice recognition software over several years.

100 chest X-Ray reports from 5 radiologists were randomly sampled retrospectively for each of four categories including: 1.Traditional dictation and transcription method (before VR), 2.VR with transcriptionist correction, 3. VR with self-correction three months after implementation, and 4. VR two years after the implementation of the continuous speech recognition system. 400 reports were scrutinized for errors using a validated taxonomy of error to study reproducibility and quality improvement. Errors were divided into two categories, minor and major.

Error rate using VR with correctionists was slightly better than the pre-VR rate; however when the correctionist was removed, the error rate rose to 2.5 times higher in reports generated by VR with self correction. The rate of errors after 2 years of using VR rose to 1.5 times higher than 3 months after implementation of VR. Fortunately only 2 and 1 reports in third and fourth category contained major errors.

Using voice recognition technology results in a high frequency of transcription errors compared to traditional transcription method, jeopardizing the quality of radiology reports. Most of the errors did not affect the core meaning of the report and are unlikely to alter patient management; however many made strangely worded reports. Comparison of the reports made 2 years after using VR with those generated 3 months after its implementation showed significant increase of the error rates, which questions the value of the ‘continuous learning’ ability of the software. Recognition of deterioration in performance suggests that VR must be carefully monitored after installation to maintain report quality and that prolonged familiarity with the device is not sufficient to sustain report quality.

" VR must be carefully monitored after installation to maintain report quality and that prolonged familiarity with the device is not sufficient to sustain report quality "

Momenin, N, Abdollahi Mofakham, F, Berwick, A, Smith, W, Voice Recognition vs Transcriptionist for Radiology Reporting and Progress in Report Quality Over Time.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12030382.html