Speech Recognition and Digital Dictation

Digital wave‘More for less’ is a current motif of the healthcare headlines. The use of digital dictation and speech / voice recognition in hospitals around the UK is already demonstrating a rare combination - an increase in activity (productivity) and cash savings. The BCS Health Informatics Health (Northern) Specialist Group tapped into local North West expertise to find out what is happening. Phil Paterson reports.

The speakers at the Group’s meeting in February were Margaret Cosens, then at NHS Connecting for Health and previously Programme Manager at The Countess of Chester Hospital, and Keith Richardson, PACS, DD and VR Lead, Chief Information and Knowledge Office, NHS North West.

Margaret Cosens led a project at the Countess of Chester Hospital to implement speech recognition in radiology. The results far outstripped expectations and led to other hospital clinicians asking to be included in the implementation. She contended that NHS trusts now have a tool available to:

  • reduce turnaround times;
  • help deliver the 18-week and other waiting times;
  • reduce administrative costs - by a great deal;
  • free up skilled staff;
  • enhance the delivery of high quality patient care.

Keith Richardson is leading a project across the North West to promote the digital management of correspondence [clinical letters, workflow, digital dictation, voice recognition and electronic discharge management] in all 63 acute, mental health, and community trusts. Keith spoke about the current usage in the North West and the scale of the benefit that could be realised within the context of QIPP (quality, innovation, productivity and prevention).

Margaret Cosens demonstrated speech recognition by speaking to the audience and giving control commands to the computer alongside her, such as ‘Open’ and ‘Close’. The software included Microsoft Word and Outlook. Her words appeared on screen in a Word document immediately after she had spoken them. This demonstration was performed using Talking Point software. One of the control commands is ‘change language’ which allows the user to switch from plain English to special terminologies such as clinical terminology for radiology reports.

Margaret started on digital dictation and speech recognition at the Countess of Chester Hospital where speech recognition was deployed as part of the PACS (Picture Archiving and Communications System) project. The Speech Recognition System linked the PACS with the Radiology Information System (RIS) and became a treasured tool, providing the rare combination of clinical benefit and business benefit.

The Countess of Chester Hospital was the first in the North West and West Midlands cluster to implement the NHS National Programme for IT PACS, in July 2006. The speech recognition programme started in 2006/7 and even before implementation had been fully rolled-out there was a queue of radiologists wanting to use it. Radiology reporting was automated and linked directly with the RIS. Radiologists could produce their reports then sign them off straight away.

The hospital measured the number of days taken to get reports signed off. Over the one year 2006 - 2007, the use of digital dictation reduced average turnaround times from about seven days to three and a half days, and the addition of speech recognition reduced this further to between one and one and a half days. By September 2008, 91 per cent of all the radiologists were reporting by speech recognition.

Clinical and business benefits

Clinical benefits accrued because, after digital dictation and speech recognition were introduced on top of PACS, the radiologists suffered far fewer non-clinical interruptions. Reporting was quicker and questions were fewer so they could be clinically more productive.

Business benefits were achieved through a big reduction in staff costs. The annual cost of radiology department staff to type reports, etc was reduced from about £120,000 pa to £45,000 pa.

Secretaries left naturally, but were not replaced. The workload for the remaining staff went up in terms of numbers of reports each, but with digital dictation and speech recognition making the job so much easier there were no objections.

Take up

Digital dictation users have got a mix of two products, Winscribe and Talking Point.

Take-up was fast and there was a 30 per cent saving in costs. The speech recognition take-up was slower. The key to the success of speech recognition was PACS. Speech recognition suddenly became beneficial as radiologists could write their own reports alongside the scanned patient images in their own office.

The hospital moved on to hospital-wide digital correspondence, the principal drivers being:

  • to reduce the time for key clinical information to get to GPs or other requestors to within 24 hours;
  • to release senior secretarial staff to manage patient care activities and the overall business;
  • to reduce staff costs by providing support for typists.

The Hospital uses MedisecNET, which captures clinical letters from PAS (e.g. outpatient clinic letters) and sends them to over 100 GP surgeries in Cheshire. The implementation rollout needs handholding, personal support, and encouragement to clinicians.

Speech recognition results in reports being ready the same day, at the end of a clinic, produced and signed off by the clinician. Digital dictation results in letters being signed off a few days later, but it too speeds up the typing process.

There are challenges:

  • funding for the products and the people - needs a commitment to support the implementation before the benefit is delivered and to the ongoing, recurring costs, as well as finding the capital;
  • vision - letting others catch it, the heart of change management;
  • time - resisting the pressures to pull away from it.

The big positive is that the technology works and, with a robust approach, every trust could reduce their turnaround times, help deliver waiting time targets, reduce administration costs by a great deal, free up skilled staff and enhance patient care.

Transformational projects

Keith Richardson explained that he was now working at the NHS Northwest Strategic Health Authority, but that he had also worked at the Countess of Chester Hospital and was a supporter and enthusiast for digital dictation and voice recognition. This area has been identified by NHS Northwest as one of the ‘Top 10 Transformational Projects’ under the QIPP Programme.

The NHS in the North West produces about 15 million letters per annum, about half of which are dictated by a clinician. Some 63 NHS trusts communicate with 1,315 GP practices which receive an average of 5,750 clinical letters a year each, or 110 per week.

The consequences at the GP practice are that practice staff currently spend about five hours a day processing and scanning paper clinical letters received from hospitals into their GP computer systems. In the North West that equates to 812 staff, costing about £16.25 million a year, simply processing hospital-generated paper.

The consequences for hospitals are that staff members are employed stuffing envelopes at a cost of about £60,000 a year per hospital. About 75 per cent of letters are sent with hospital transport such as the pathology tests pickup van but the rest are posted - costing about £1 million a year per trust for postage.

One option is to try to move the trusts and GP practices to improve on the traditional, paper-based process. Digital dictation offers more methods for dictation, at the office or home or on the move, with reports going onto a computer network, with an interface to PAS for demographics and reference data, which gives more options for typing - again at the office, home or remote.

After initial transcription what is required is an editor more than a typist, who can email a document to the author for online review and sign-off. The process becomes one of e-letters, e-processed.

A more advanced option is to replace the secretary with a background computer and the person who dictates the report edits it via online access - but the consultants want to keep their secretaries.

A further option, in radiology, has the consultants dictating directly onto the screen using voice recognition technology; they can verify their reports, sign them off and send them to the GP practice the same day. Both these options present cultural challenges. Responses vary with exposure to change, IT awareness etc.

Digitised reports and letters enable the concept of an integrated clinical communications hub between hospitals and GP practices, routing electronic letters to the right place. Cheshire trusts are using this approach to send e-discharge letters.

Consistency is all

Typing errors can arise due to the author of a letter typing and sending it off without a proof reader. You do see what you hear yourself say. Some radiologists will dictate, leave it for a while, and come back to their reports as a batch at the end of a session.

Systems are voice-profile based and very quickly pick up the profile for new users. Consistency is a factor. The systems are very accurate. Microsoft Windows 7 includes free voice recognition software. Basic professional voice recognition software is commercially available for about £300 to £650.

What is the difference between digital dictation and speech/ voice recognition?

The more established approach is to dictate to a tape and give it to a typist. Tapes are then stored until the typist has time to get round to them.

If the consultant wants a particular report to be found and typed urgently, it can take the typist hours to search all the tapes to find the right one. Work scheduling is disrupted. This technology uses analogue recording to tape. Most of the NHS is still on this old technology.

Going digital

Digital dictation, which really came in with PACS, produces a ‘digital tape’ instead of an analogue tape. Digital recordings can be stored on a computer network and made available wherever is most appropriate for typing.

The consultant talks, a digital recording is produced and the secretary types it. The secretary gets an indexed list of dictated letters with priority reports highlighted in red at the top of the list. Management can provide performance statistics on turnaround times etc.


Speech-to-text technology allows a secretary to open up some hard-to-understand reports to find keywords that are converted to text - the words are typed out for the secretary - allowing the secretary to edit this into a proper report.

Speech/voice recognition takes it further and provides a full process. The words come straight back on screen for the clinician to read and edit on the spot.

For further information please visit: www.hinorth.bcs.org.uk

This article also appears in the book Health Informatics.

Autumn 2010

Comments (2)

Leave Comment
  • 1
    Arabella wrote on 8th Apr 2015

    This article is utter rubbish. VRC does not give an accurante transcription and does not save money. The whole thing is a waste of money and only serves to make NHS managers and would-be 'tech heads' justify their existence. The fact that so many ignorant, middle-aged, computer illiterate men admire these systeems and are in love with them only goes to illustrate how useless they are.

    Report Comment

  • 2
    Harry's Mad wrote on 6th May 2015

    I couldn't agree more Arabella! I trained to be a medical transcription its and used to enjoy my job. I prided myself on a fast turnaround but since my hospital implemented VRC my work is over a month behind because the software doesn't understand what the doctor is saying. It takes me longer to sit and listen to the voice file while editing the document that it would have taken me to type it from scratch. As usual, the people responsible for spending exorbitant amounts of money on these systems know nothing about the people who actually do he job. They may only be able to type at a snails pace but the secretaries and transcriptionists type at the speed people speak. More importantly, you turn previously happy workforces into disillusioned staff with a huge backlog of work.

    Report Comment

Post a comment