Computer-based speech recognition as a replacement for medical transcription

We have implemented computerized speech recognition in a high-volume clinical setting using a newly developed commercial software system. This paper compares the performance of the computerized system with conventional transcription during a trial week followed by 3 months of routine clinical use.

The speech recognition system was used for 97% of the cases during the trial week and 87% of cases during the months of continuous use. For a similar mixture of cases, no change occurred in the length of reports after introduction of the computerized system. Speech recognition decreased the time until the report became available by 99% and resulted in a substantial savings in transcription costs. We conclude that speech recognition by computer is practical in a high-volume clinical implementation.