Genie Solutions releases integrated MedSpeech voice recognition
Genie Solutions has upgraded its flagship GP and Specialist practice software, Genie, giving users the option to interface their software with the MedSpeech voice recognition system. The integrated solution was officially launched in mid-October with the release of Genie version 7.7.1 following six months of development and beta testing.
MedSpeech utilises the SpeechMagic engine, which was originally developed by Philips but later purchased by Nuance, who also retail the consumer-oriented Dragon range of voice recognition products.
To commence dictating a letter, the doctor simply opens Genie’s standard letter-writing screen and clicks a ‘MedSpeech’ button. This launches the MedSpeech dictation window, allowing the doctor to construct the body of the letter using their voice alone. As is the case with typed letters, patient demographic details are already present in the Genie letter template, negating the need for this information to be dictated.
Doctors can dictate their letters starting from a blank page, or can create highly structured dictation ‘Autotext’ templates and recall these as required. Templates can be both opened and populated entirely using speech commands, making this a very efficient way to generate letters, particularly for specialists that routinely perform and report on a limited number of procedure types.
Corrections can be made by the doctor during the dictation process if required, or optionally, any errors can be left for a practice staff member to rectify. During the correction process, the MedSpeech software makes both the original speech recording and the interpreted text available to the reviewer. These visual and audio representations of the dictation are synchronised by the software, allowing the user to skip to the relevant part of the voice recording by simply placing their cursor next to any word that needs to be corrected.
Following review, the dictation text is imported into the appropriate part of the Genie letter template. At this stage the letter can be cued within Genie for a final review by the doctor, printed or sent electronically.
Unlike other voice recognition solutions requiring users to undertake a prolonged period of training, MedSpeech has the user dictate just five minutes worth of prescribed text. However, to help the system rapidly build a vocabulary relevant to the user, MedSpeech can also import and analyse a selection of documents, such as previously typed letters. As with other sophisticated voice recognition systems, the software is designed to learn the nuances of the user’s voice and vocabulary, becoming increasingly accurate with ongoing usage.
MedSpeech can be run within a Terminal Services environment across both local networks or the Internet. While the software can only be installed in a Microsoft Windows environment, Genie Solutions and Speech Solutions are working on a solution that will allow Mac OS X users to continue to run Genie in their preferred operating system, with the software capable of interacting with a MedSpeech installation running within a Windows virtual machine installed on the same system.
Distributed by Speech Solutions, the software — which comes with a Philips microphone — retails for a considerable $5500+GST for each user needing to dictate. An annual fee of $856+GST is also charged after the first year to cover ongoing software upgrades and support. Typists and other practice support staff are able to install a free version of the software, which allows them to review and edit dictations, but not dictate notes themselves.
Dr Joseph Macdessi, a product beta tester turned evangelist for the solution historically used a handheld digital voice recorder, with dictations transcribed by an in-house typist. A paediatrician at the Children’s Hospital at Westmead, Dr Macdessi typically generates around 150 letters per week, which he described as having a high level of variability when compared to the types of reports his proceduralist colleagues typically produce. He indicated that the system typically requires three to four corrections for each of his letters, which are generally four to five paragraphs in length. Dr Macdessi estimated that he self corrects around half of his letters, with the rest of his letters being assigned to his typist to review in advance of the letters being signed.
Despite the system not operating at a perfect level of accuracy for Dr Macdessi, he indicated the solution was delivering significant time savings for both himself and his practice staff.
"I am handling messages with extreme efficiency and not having to review letters several days after dictating them is an absolute blessing. Letters can be turned around in a day, meaning I don’t have to spend time recalling each patient encounter days after seeing them as part of the letter review process."
Posted in Australian eHealth