CHICAGO--The widespread use of speech recognition (SR) has decreased report turnaround time, but also has introduced errors into the radiology report, said Zombor Zoltani, BA, of the University of Maryland, at the Radiological Society of North America (RSNA) annual conference on Tuesday.
“The radiology report is the primary means by which a radiologist communicates,” explained Zoltani, who identified three critical functions which the report serves, including the conveyance of critical information for diagnosis, serving as a record of medical information for patients and acting as a medical legal document.
A study by Zoltani and colleagues conducted at their facility found that 68 percent of more than 17,000 finalized reports contained errors, 15 percent of which could potentially change the meaning of the report. A radiologist's experience, sex and caseload were not associated with significant differences in error rates.
“We did not see an association of demographic factors with an increase of error reports, except for English as a second language, which was clinically significant, although we did not have many non-native speakers in this study,” Zoltani said.
The errors that the study identified included word substitution and omission, spelling, unintelligible phrasing, wrong patient or report, and formatting and punctuation errors.
“Potentially confusing errors that could alter the meaning or the interpretation of the report” included word omission and substitution, misspellings and unintelligible phrasing.
Approximately 12,000 of the reports sample contained at least one error. The reports that were classified as having potentially confusing errors totaled 2,709.
Of the 2,709 reports, misspelling errors accounted for 5.8 percent, word substitution 4.8 percent, unintelligible errors 1.9 percent and word omission 0.8 percent.
“Radiologists must vigilantly edit and read each report before sign-off. Further work is needed to reduce errors in SR from both radiologists and SR vendors,” concluded Zoltani.