Evaluation of a voice recognition system for generation of therapeutic ERCP reports

We evaluated a computerized report-generating system using voice recognition technology (EndoSpeak) for producing therapeutic ERCP reports. For 30 consecutive ERCP cases, reports using both EndoSpeak and standard dictation were generated at the end of the procedure. The cases were scored for overall difficulty and the number of component procedures performed. The time to generate a report with EndoSpeak was significantly longer than with dictation (10.0 +/- 2.9 (SD) min vs. 6.5 +/- 1.6 min, p less than 0.0001), and this difference was greater for more lengthy reports (r = 0.85, p less than 0.001). The procedure difficulty score and the number of component procedures were both significant predictors of the physician's report generation time for EndoSpeak and dictation. Dictated reports were judged to have a higher information content than EndoSpeak reports for 90% of the cases (p less than 0.00001). Several specific difficulties were encountered with the EndoSpeak software. Although EndoSpeak offers promise, substantial software modifications will be required before this system can become competitive with dictation for therapeutic ERCP reports.