Speaking up for speech recognition

The National Programme for IT’s Picture Archiving and Communications System (PACS) enables images such as X-rays and scans to be stored electronically and viewed on screens, creating a near filmless process and improved diagnosis methods. As well as improving patient care, the introduction of PACS is saving money – with Trusts reporting an average saving of £250,000 in their first year of using the technology. However, research undertaken at the Countess of Chester Hospital NHS Foundation Trust suggests that further savings can be achieved with the introduction of speech recognition alongside PACS. The Countess of Chester Hospital NHS Foundation Trust was the first in the North West to implement and go live with PACS in 2006. In March 2008, the Trust undertook a PACS benefits realisation project and discovered that they were achieving significant benefits with the implementation of speech recognition – including major cost savings, faster reporting and diagnosis – as well as some surprising results such as fewer clinical and non-clinical interruptions for radiologists. The Trust already had a track record of being at the forefront of technology adoption and scored highly in assessments in terms of clinician’s usage of IT systems. Following the introduction of PACS and an electronic patient record system, the decision was made to implement speech recognition into the radiology department, under the direction of Margaret Cosens – the information management and technology programme manager at the time. Since 2000/2001, the radiology department had used a digital dictation system. However, the Trust wanted a system that could enable automated reporting direct from PACS into the radiology module of the patient record. "We wanted to use voice recognition to facilitate this, but many of the consultants were wary. Between 2003 and 2004, there had been three trials conducted using such systems – all of which failed dismally," Margaret Cosens commented. She explained that in all of these cases the recognition function was "not accurate enough" and the words took too long to appear on screen. In addition, the secretary had to edit and correct the dictation, which proved difficult to manage. Despite these initial failures, the team remained convinced that speech recognition could be made to work in radiology, as well as other clinical settings. "We knew that speech recognition had potential and had heard of successful projects elsewhere. Then the PACS manager from Milton Keynes General Hospital NHS Trust brought to my attention a system called TalkingPoint, which was based on the ‘Dragon’ speech recognition engine. We went to look at the system installed at Milton Keynes and found that it worked effectively," she commented.

Potential in radiology The Countess of Chester decided to introduce the system, which took five days to implement in total. Provided by GHG Software Developments, the solution uses English radiology vocabulary ensuring it is specially tailored to the reporting requirements of radiology departments. The radiologists are now able to complete examinations on the spot, viewing both images and text at the same time, while alterations and adjustments can be made and signed off quickly; making them instantly available to the clinician who asked for the diagnostic opinion. TalkingPoint’s managing director, Malcolm Grant, explained that in emergency admissions the increased speed in reporting is a significant clinical benefit – particularly in the case of a suspected thrombosis, for example, when a fast response is critical to patient outcome. "The radiologist may produce the report immediately, but delays or inaccuracies can occur when tapes are sent off or even lost, as well as during the process of typing up digital dictation files," he commented. Margaret Cosens also pointed out that the Trust has been able to take a number of steps out of the reporting process. "At many other Trusts, a radiologist speaks into a tape or digital dictation system, the secretary types this up; prints out the report; then puts it in a tray awaiting the radiologist’s corrections. The radiologist looks at the report, edits it, gives it back to the secretary, who makes any necessary changes; then this is returned to the radiologist to be signed off. The process can take anything up to a week to complete. "At the Countess of Chester, the radiologist dictates using voice recognition, the words appear immediately on screen in front of them while they still have the PACS image open, they are able to check that the words tie in with the image and can view the whole report at a glance. There is no need to play back the tape or digital recording. They can make any fine tuning to the report, then sign it off directly – saving a significant amount of time and effort." Before PACS was introduced, the turn-around time between patients receiving a scan to the report being signed off and available to the clinician, was approximately seven days. After PACS, with digital dictation, the turn-around was between three and four days. Now, with PACS and speech recognition, this has been further reduced to between one and one-and-a-half days at the Trust. "The radiologists were amazed at how good the recognition was. After initially rolling out the system to half of our radiologists, the rest started to approach our IT department with requests to adopt the system," Margaret Cosens commented.

Programme expansion All 12 radiologists are now using the voice recognition system. However, after seeing it being used by the radiologists, the Trust’s ultra sonographers also wanted to convert to voice recognition. All 10 are now using the system, while reporting radiographers will also soon have access to the technology. "I believe that there will always be some people for whom speech recognition isn’t suitable. However, even our visiting specialist registrars, who only stay for six months, feel comfortable using the system," said Margaret Cosens. She also pointed out that radiologists are given a choice of picking up the speech recognition headset or the digital dictation system, when they are reporting from their workstations. However, the new technology has quickly become the preferred option – between January and July 2008, 89% of cases were reported using speech recognition. Further research also revealed some surprising results: "Before PACS was introduced, we looked at the number of times that a radiologist was interrupted while reporting. We found that 57 minutes out of every four-hour reporting session, were lost due to interruptions. Thirty minutes were due to non-clinical interruptions including questions such as: ‘Have you got this film?’ Twenty-seven minutes were attributed to clinical interruptions, where a doctor needed additional information or an opinion on a patient," Margaret Cosens commented. "After the introduction of PACS and speech recognition, these interruptions were reduced to ten minutes in a fourhour reporting session. Furthermore, they are now only clinical in their content. Although we cannot be certain of the reason, it is likely that staff need to interrupt less because turnaround times are much quicker and they already have the information they need in the report." In addition to clinical benefits, she also reported significant cost savings. In the financial year of 2005/2006, the Trust spent £120,000 per year on typing radiology reports. In 2007/2008, it spent just £43,000. Secretarial time has been freed to provide extra clinical benefit, such as helping with the organisation of multi-disciplinary staff meetings. The speech recognition system is also now being trialled in the pathology department and there are plans to implement voice recognition in the Trust’s theatres enabling surgeons to make notes during an operation. The commands are very simple – the clinician simply says: "start reporting" to commence, then "confirm" or "reject". The consultant surgeon can enter the notes on the electronic patient record so that the nurse on the ward can establish exactly what post-operative care the patient is going to need. Reports can also be retrieved by the GP. A third trial is planned in the area of clinic letters, which shows the potential to increase efficiency and improve continuity of care following discharge. "A number of consultants remarked that they like to leave the clinic with all their letters completed," Margaret Cosens explained. "We are currently setting up a system where the consultant has the ability to use a voice recognition system to dictate a letter while the patient is fresh in their mind. They can put this into a ‘draft’ folder or send the letter straight away to the GP via the electronic system. By the time the patient has returned home, in the evening, the letter will already be with their GP."