Voice-to-Text

Enabling clinicians to dictate in a computerised system so that we can provide our patients immediate access to clinic letters and other health information

Opportunity

Methods for the acquisition of data in health systems are improving rapidly - future vision includes the real-time capture of dictated information at the point of care with rapid transition to text - included are other forms of data acquisition using mobile devices and data collection without human intervention.

The latest speech recognition engines are capable of achieving near 100% accuracy for a wide range of authors.  This is achieved by employing Artificial Intelligence (AI) algorithms to recognise and learn the idiosyncrasies of an individual author’s speech patterns.  While speech recognition has been around for at least a decade it is only in the last 2 years that it has come of age, having a demonstrable impact in international healthcare organisations.

The average person types at 39 words per minute, 19 words per minute if editing is required. Yet, the average person speaks at 200+ words per minute. This fact when combined with New Zealand technology allows Winscribe’s automatic speech recognition system to accelerate processes, enable  a higher number of patients to be treated, decrease reporting backlogs, reduce administration costs, and improve overall patient care.

Aim

Improve patient outcomes through increased accuracy and earlier delivery of dictated clinician notes by employing speech recognition technology. We aim to see:

  • Faster turnaround of transcribed documents, such as clinic letters, operation reports and ward rounds
  • The ability to highlight and prioritise urgent documentation
  • Full visibility and audit trail of each document from dictation to authorisation
  • The ability to confirm all patients in a clinic have had appropriate dictation completed
  • Validation of up to date patient demographics at point of dictation and merging into subsequent correspondence
  • Consistent quality of DHB-wide templates for standardised documentation
  • Improved communication between author and transcriber through workflow
  • Enhanced workload management for transcription services

Intervention

This initiative was part of Phase One of the Leapfrog Programme. We upgraded our current New Zealand-developed products, Winscribe and Soprano Medical Documents, to Winscribe Text and Winscribe Speech Recognition, moving both typists and authors to the upgraded system.

The new solution was piloted in the Renal Service. Renal physicians have been the testers of both 'back end' (dictation as usual) and 'front end' (self-editing while dictating) services, and have contributed to the design of templates and training programmes for other services.

The upgrade entailed setting up a completely separate system and gradually moving departments over one at a time minimise the impact on daily operations.  To ensure the full benefits are realised from the system the initial stage was limited to 200 authors. At this point we want to fully examine the benefits, improve and refine processes, and plan for further rollout as required.

A production system was introduced in January 2016 and the system went fully live in June 2016.

Impact

Implementation of new speech recognition software for transcription has reduced the number of hours of transcription required to transcribe and edit dictation, increasing efficiencies in the transcription service.

We have reduced the number of hours required to transcribe one hour of transcription from five hours to three hours (from 5:1 to 3:1 hours of transcription:dictation).

Project Team

Sponsor
  • Robyn Whittaker, Clinical Director of Innovation

Team Members
  • Janak de Zoysa, Clinical Director Renal 
  • Jane Wright, Manager Transcription Service
  • Arnah Pearson, Integration
  • Lenore Robers, Project Manager
  • Boris Burges, Project Manager

Robyn Whittaker

Clinical Director of Innovation

Janak de Zoysa

Clinical Director Renal Services
Back to top