Sponsored by an educational grant from Agfa HealthCare and EMC Corporation.
Southern Ohio Medical Center is a voice recognition success story. The center’s interest in voice recognition technology dates back to 1995 as part of a project to improve efficiency among radiologists and reduce final report turnaround time. The center made some gains, dropping turnaround time to 12 hours, by monitoring radiologists and transcriptionist, but, ultimately it realized it could not make further improvements without investing in technology. In addition, a transcriptionist shortage forced the hospital to send some transcription offsite, an expensive and inefficient proposition.
Enter Agfa HealthCare and Agfa’s TalkStation. The hospital went live with the technology in 2002, and its radiologists use 100 percent voice recognition. The results are impressive. For starters, SOMC reports true live turnaround for emergency room and critical cases with a sub-30 minute turnaround time. Overall turnaround time sits at two hours. Costs are reduced, too, as the hospital no longer requires four to five transcriptionists to complete radiology reports. In addition, SOMC no longer prints and distributes preliminary reports, which translates into improved workflow.
Project manager, RIS/PACS Administrator Howard Stewart, offers some pointers for his colleagues:
Speech recognition promises significant benefits, but it is also a major undertaking. How should sites lay the groundwork for a successful implementation?
Stewart: Speech recognition is an easy sell for administrators because of the cost savings, but it does represent significant changes to radiologists’ workflow. To boost the odds of acceptance and a successful implementation, radiologists should lead the project.
SOMC also built support mechanisms into our speech project. We installed three transcriptionist workstations, so radiologists could send reports to transcription for editing if necessary. Within a few weeks, all of our radiologists bypassed the workstations and self-edited all reports.
There are a fair number of speech systems on the market today. How can hospitals make the best decision for their needs?
Stewart: Not all systems are created equally; some products do work better than others. What’s more, transitioning from transcription to speech recognition is complex. It’s important to talk to customers using the products and complete site visits to see the systems in action. Radiologists should be active participants in this process and see what speech entails including training, typing and proofing. Information systems also should participate in the evaluation and decision-making because speech entails some database maintenance and user profile challenges.
Are there other IT issues to consider with speech?
Stewart: Find out if the vendor can integrate the speech recognition solution to the hospital RIS and PACS. In an ideal environment, the correct patient report window automatically opens as the radiologist views a patient’s images on the PACS workstation. The radiologist should not have to select a patient from multiple systems.
Review the mechanism for addendums, too. Addendums should not constitute separate reports.
Speech does represent such a drastic change for radiologists. What type of training is needed?
Stewart: Ideally, training includes systems applications and keyboarding. Our radiologists jumped in with both feet, so we found we needed more upfront training than we planned on originally. Agfa provided additional training after the implementation to make sure radiologists’ skills matched their enthusiasm. In addition, speech recognition combines the keyboard, mouse and microphone. Many radiologists, however, don’t type efficiently and don’t know how to proof their own reports efficiently. It’s important to focus on these needs particularly if the system requires a lot of typing. (Agfa Talk does not require doctors to type.)
What about the role of the vendor?
Stewart: Find a vendor that will work well with the hospital, and detail specific needs including benchmarks and dates in the RFP.