Radiologists Say Speech Recognition Does Work
Like handwriting recognition, software-based speech recognition seems to have been three years away from being really effective for the last 15 years. Industry leaders such as Microsoft’s Bill Gates have touted its benefits for a while but error rates have remained stubbornly too high to be practical for a long time. (HospitalBuyer resisted the temptation to publish a typo-laden article to mimic how such software is perceived to perform.) According to Health Imaging & IT, actual performance is now good enough for radiology departments that say they have been able to implement successfully technology such as MedQuist’s SpeechQ or Nuance’s PowerScribe.
As an example, 109-bed Northeast Regional Medical Center (NRMC) in Kirksville, MO (one of 73 hospitals in 22 states operated by Community Health Systems Inc), started implementing SpeechQ in September 2005. Two administrators were trained with it so they could then transfer that knowledge to radiologists on the fly. It only took a couple of weeks before the old portable dictation devices could be taken away. Report turnaround time, a critical metric, is now less than 30 minutes during the week. Doctors at first were lukewarm to the idea of having to self-edit their reports but it turns out hospitals that introduce speech recognition can eventually phase out transcription services. Yet some facilities actually keep using transcriptionists and use recognition as a way to increase their output.
January 7, 2007 Related topics: Radiology, IT & software