Speech recognition software can be leveraged to tell if a radiologist is stressed, prompting individualized interventions to improve quality and satisfaction.
What if your speech recognition software could tell that you’re stressed or tired, picking up on the subtleties of your voice as you dictated your reports?
It’s technology that exists today and could be used to help improve quality and job satisfaction., said Bruce Reiner, MD, a teleradiologist and director of research at the VA Maryland Health System, who details the approach in an article in the Journal of Digital Imaging.
Occupational stress and fatigue is common among radiologists as they face declining reimbursement and increased exam volume. That stress can lead to corner-cutting and introduction of errors. So Reiner asked, why isn’t any one addressing this reality?
“Why aren’t we measuring it, tracking it, and analyzing it for performance outcomes, and intervening at the point of care instead of after the fact,” Reiner asked, speaking in a recent interview. This technology, he added, could be implemented into existing speech recognition software.
Voice stress analysis can detect stress in the acoustic properties of speech, Reiner explained. Although this analysis is fairly limited in the range of speech it focuses on, another technology – layered voice analysis – uses a wider spectrum and can detect other emotions like excitement and confusion. Through the creation of personalized speech profiles, each individual’s characteristics can be taken into account.
Once stress or fatigue is detected, the user would receive the appropriate prompts and customized interventions. For some this could mean exercising or listening to soft music, while others may need to have their work flow altered, redirecting studies that are less challenging to the radiologist. It’s an adaptive, customizable system, he said.
“It’s not designed to shut you down,” he said. “It’s designed to allow you to learn more about yourself and your practice and how you can improve the outcomes.”
In a companion article, Reiner explores the use of speech recognition technology to also track uncertainty in radiology reports. There would also be personalized profiles and customizable interventions, with the goal of improving report quality.
Speech characterization is used in other industries, such as aviation, to detect fatigue and stress, but medicine is slow to implement this tool into existing speech recognition software, Reiner said. However, vendors he has approached with the idea argue there isn’t a market for it, so they aren’t interested in taking that first step, he said.
There’s also likely to be pushback from radiologists, Reiner concedes. Some may see it as a tool that could impact their livelihood or open them up to medical malpractice suits if they continue reading despite a stress warning from the system.
“Some people will look at this as a tool of empowerment and quality improvement,” he said, “and the glass-half-empty people will look at this as a nefarious tool that will be used against them.”
Widespread adoption of such technology might require a mandate from a hospital administration, teleradiology provider or large health system – similar to how speech recognition technology was adopted, Reiner said.
The first step, he said, is to recognize that the radiology industry has a problem with occupational stress and fatigue, and that customized solutions can be developed that address each end user and the tasks being performed.
“We need to stop saying the current model of how we operate is good enough,” he said, “It’s not.”