Objective: To study the effect that voice recognition (VR) has on radiologist reporting efficiency in a clinical setting and to identify variables associated with faster reporting time.
Methods: Five radiologists were observed during the routine reporting of 402 plain radiograph studies using either VR (n = 217)or conventional dictation (CD) (n = 185). Two radiologists were observed reporting 66 computed tomography (CT) studies using either VR (n = 39) or CD (n = 27). The time spent per reporting cycle, defined as the radiologist's time spent on a study from report finalization to the subsequent report finalization, was compared. As well, characteristics about the radiologist and their reporting style were collected and correlated against reporting time.
Results: For plain radiographs, radiologists took 13.4% (P= 0.048) more time to produce reports using VR, but there was significant variability between radiologists. Significant association with faster reporting times using VR included: English as a first language (r = -0.24), use of a template (r = -0.34), use of a headset microphone (r = -0.46), and increased experience with VR (r= -0.43). Experience as a staff radiologist and having a previous study for comparison did not correlate with reporting time. For CT, there was no significant difference in reporting time identified between VR and CD (P = 0.61).
Conclusions: Overall, VR slightly decreases the reporting efficiency of radiologists. However, efficiency may be improved if English is a first language, a headset microphone, and macros and templates are used.