Intended for healthcare professionals

Information In Practice Infopoints

CyberTranscriber—the virtual medical secretary on your desk

BMJ 2000; 321 doi: (Published 09 September 2000) Cite this as: BMJ 2000;321:618
  1. A Depuyt, house officer,
  2. E Carty, specialist registrar,
  3. I Mason, gastroenterology nurse specialist,
  4. C Durcan, medical secretary,
  5. Owen Epstein, consultant gastroenterologist (epstein{at}
  1. Centre for Gastroenterology, Royal Free Hospital, Hampstead, London NW3 2QG

    Clinic letters and discharge summaries provide the fundamental link between hospitals and primary care. In 1997–8 there were 41.635 million outpatient visits to NHS hospitals in England,1 with each visit generating a typed letter. The widespread implementation of voice recognition systems should be considered as a response to the increasing typing load experienced by medical secretaries. However, currently available desktop systems require speech enrolment and considerable care to produce text with error rates of 3-5%.

    CyberTranscriber2 is a convergent technology, which exploits the internet, speech recognition, and email to generate letters and reports. Unlike conventional speech recognition software, it does not need speech enrolment, and the accuracy of the final report is equivalent to that achieved by an experienced medical typist. The user dictates down a telephone line, using the telephone keypad to control the record, pause, and “rewind” functions. The dictation is remotely processed into a rough draft by speech recognition software, and both the text and voice files are automatically routed via the internet to medical transcriptionists who correct the rough draft. Within 24 hours (range 4-24), all the text is returned by email as a Microsoft Word attachment.

    We have explored the possible role for CyberTranscriber in NHS hospital practice by comparing the efficiency of generation of clinic letters by an experienced medical typist with letters generated by CyberTranscriber. A total of 441 consecutive outpatient clinic letters of similar length were generated, of which 215 were produced by conventional dictation and manual typing and 226 by telephone dictation and CyberTranscriber. The average time taken by the medical secretary to generate and print completed letters was 7.7 minutes per letter with manual typing and 2.2 minutes a letter with CyberTranscriber. There was no difference in the error rate between the two methods.

    Thus, CyberTranscriber was three times more efficient in generating letters than an experienced typist. In practical terms CyberTranscriber allowed the medical secretary to complete the week's clinic letters in the equivalent of two days' work rather than the five to six days associated with manual typing. Further improvement in efficiency will be possible once NHSnet is operational, as the returned email could be forwarded to the relevant general practitioner without the need to print and post a letter.

    CyberTranscriber could make a considerable impact on the efficiency of communication between primary and secondary care. The system costs about 1p per word but represents an irresistible technological response to the need to improve efficiency in the NHS.

    For more information about the project see


    We thank SpeechMachines for developing the TalkForms and providing free access to CyberTranscriber for the duration of the trial. Competing interests: None declared.


    1. 1.
    2. 2.
    View Abstract