This site is intended for health professionals only


How I improved the accuracy of our digital dictation software

The problem

As a former PCT medical director, I have been extensively involved in GP performance cases, dealings with the GMC and complaints, and one common thread that runs through this a high percentage of this type of case is the poor quality of clinical records. Unfortunately, producing quality type-written records takes time which is obviously increasingly at a premium in modern general practice.

Getting started

Over the last five years or so I have been experimenting with voice recognition software as an alternative to learning to type and whilst I have had a modicum of success, there have always been limitations with regard to accuracy, responsiveness and particularly medical vocabulary.

At a conference I attended recently, I came across a professional company promoting a comprehensive package and resolved to explore this further but the costs turned out to be considerable (approximately £2,000 per user, which represented a substantial outlay in a nine-doctor practice).

What I did

I now have an effective method which works within SystmOne, our practice clinical system (even though it is not officially supported by the system software). It has improved the quality of my clinical records, decreased the turnaround time for referrals and letters and saves me a substantial amount of time during an average working day.

At the core of the system is the software, Dragon Naturally Speaking version 12. I purchased the premium edition which has a full retail price of £149.99 and includes a rudimentary headset, but is available for less than this by shopping around.

Training the software to recognise your voice and speech patterns is surprisingly quick and easy and following installation of the software you can have it working with a high degree of accuracy within 20 minutes.

One potential reservation around using a generic speech recognition system is that we use a lot of medical jargon which is not recognised by the software and there is no doubt that, initially, it produces some amusing alternatives to everyday medical words. A more comprehensive and specific medical version of Dragon is available but, at a cost of around £1,000, represents a substantial investment.

By experimenting, I have realised that the amount of medical jargon included in most of my correspondence and records is relatively small and I do not think justifies the substantial outlay for the expensive medical version. On eBay, I came across a vendor from New Zealand selling a substantial medical vocabulary for around £40 which can be imported into the basic software. This has been very successful for me and I have noticed a substantial increase in accuracy in transcription since importing this vocabulary.

The most critical success factor, however, has been the purchase of a professional quality dictating microphone. This was the largest expense at around £250 but it enables the user to have the software running constantly in the background and just pick up the microphone to dictate swiftly into the clinical record with a high level of accuracy. I use the Olympus DR1200 Directrec USB mic.

Results

I would strongly recommend that anyone who is unable to touchtype and feels under time pressure during surgeries to investigate speech recognition as an option. Even the basic installation can substantially improve your time efficiency for an outlay of less than £200, and if you then feel inclined to refine things further, I believe that a professional microphone is a worthwhile investment, especially as an allowable business expense.

I believe however, that the best justification for making this move is that my clinical record keeping is already substantially improved. Any clinician who has ever reflected on the quality of their notes produced when under time pressure will understand the sense of security and satisfaction this can provide.

Dr Paul Cook is a GP in Chesterfield, Derbyshire