DSI Medical Voice Recognition for
Computer Assisted Medical Transcription (CAMTran)
By Kevin Goldstein, CEO, DSI Medical
DSI Medical calls its voice recognition system CAMTran, for Computer Assisted Medical Transcription. Why the distinction? Why not just call it voice recognition?
Having been active in all phases of the computer industry virtually since its inception (over forty years -- since 1967, to be exact), I am keenly -- and at times painfully -- aware not only of the strengths of most computer technologies, but also their weaknesses. In the case of voice recognition, the state of the art has reached a point where it can, when used appropriately, offer certain advantages to the Medical Records Director and her (or his) institution. It has not, however, yet reached either the point where it is the best solution in every situation, nor where it can be used solely by itself in lieu of a medical transcriptionist or editor. That is because, while the best voice rec technology (such as the one DSI Medical uses) can reach real-life accuracies of 95% (and occasionally better), they can do so only in the best of circumstances, and to do so they will require some changes in the way your doctors work. Those changes will often take additional time out of your doctors’ busy schedules, leaving them less time for patient care.
Equally important, while an accuracy of 95% may at first hearing sound good, when you start digging a little deeper, you realize that for medical transcription, it is simply unacceptable. For the purposes of this discussion, though, let's instead assume a best case scenario: an excellent dictator in a quiet environment, able to maintain 98% accuracy. Consider that a typical report that is 98% accurate will have on average over 6 mistakes per report, less on some, more on others. Any one of those mistakes could have dire consequences in the treatment of the patient.
The technology is already useable in certain areas (when was the last time an operator answered your “411” call?), but it has yet to reach the point where it can be used stand-alone in critical applications or those requiring exceptionally high accuracy. (You probably don’t have to think back very far to recall the last time you made a call and the voice recognition at the other end either couldn’t understand you at all, or misinterpreted you.) Back-end speech recognition is the technology currently showing the most promise for medical records. While by itself it is not yet able to offer the kind of fluent recognition many were expecting of it (for the past, oh, ten years or more), when combined with a well-designed methodology that interlocks speech recognition with oversight and editing by expert medical editors, it may meet the expectations sought of it. It then may have the additional advantage of being transparent to the dictator, requiring no change in the way he or she works.
CAMTran is based not on what voice recognition can offer next year, next week, or even tomorrow, but rather on a realistic assessment of what the technology can offer right now: it is a balanced blending that maximizes the best technology available now by combining it with human intervention and careful oversight. While this way of looking at technology guides many important technology-related decisions at DSI, it is of particular importance here, because our long experience and deep understanding of computer and software development causes us to approach the implementation of leading-edge voice recognition differently and -- at times -- more cautiously, than most other companies. It is a policy that is serving our clients well, offering them the advantages of new technology, without the time, cost, and heartache of being the guinea pig for someone else's research project.