Return To Abstract Listing

Efficacy, Education, Administration, Informatics

E2315. Frequency and Analysis of Nonclinical Errors in Radiology Reports Using Voice Recognition Dictation

Motyer R,  Liddy S,  Torreggiani W,  Buckley O. The Adelaide and Meath Hospital, Dublin, Ireland

Address correspondence to R. Motyer (ronanmotyer@gmail.com)

Objective: Voice recognition (VR) dictation of radiology reports has become the mainstay of reporting in many institutions worldwide. Despite benefit, such software is not without limitations, and transcription errors have been widely reported. Our aim is to evaluate the frequency and nature of nonclinical transcription error using VR dictation software.

Materials and Methods: A retrospective audit of 378 finalized radiology reports was conducted. Errors were counted and categorized by significance, error type, and subtype. Data regarding imaging modality, report length, and dictation time were collected.

Results: Sixty-seven (17.72%) reports contained 1 error, with seven (1.85%) containing “significant” and nine (2.38%) containing “very significant” errors. A total of 90 errors were identified from the 378 reports analyzed, with 74 (82.22%) classified as “insignificant,” seven (7.78%) as “significant” and nine (10%) as “very significant.” Sixty-eight (75.56%) errors were “spelling and grammar,” 20 (22.22%) “missense,” and two (2.22%) “nonsense.” “Punctuation” error was the most common subtype, accounting for 27 errors (30%). On average, complex imaging modalities had higher error rates, with CT containing 0.57 errors per report compared to plain film with 0.08. Longer reports had a higher error rate, with reports longer than 25 sentences containing an average of 1.23 errors per report compared to 0–5 sentences containing 0.09.

Conclusion: This retrospective audit highlights the limitations of the use of VR dictation software and places emphasis on the importance of a review process prior to the completion of the report. While a large proportion of errors were made in spelling and grammar and deemed insignificant, errors were identified that may interfere with the interpretation of the report and, most significantly, influence patient management and safety. Reports of greater length and on more complex imaging modalities had a higher error rate and this should be taken into account by the reporting radiologist. Furthermore, institutions should ensure appropriate and effective risk management systems are put in place to adequately quantify and qualify error and implement measures aimed at improving the use of VR software.