It appears our fears that machines would steal our jobs were grossly misplaced. Or, is that what machines want us to think?
Joking aside, the dollar amount flowing into the research and development of artificial intelligence (AI) is huge—exceeding $30 billion a year according to reports by McKinsey and PWC. With big players like Amazon, Google, Facebook, Microsoft and Apple investing the most, the race is on for the future of AI. So far though, the hype around AI hasn’t quite lived up to reality.
Both McKinsey Global Institute and MIT/Boston Consulting Group have reported that only about 20 percent of companies have put AI technology to use in a meaningful way. So while the devices in our hands can perform facial recognition, there are still many more developments to come.
With its ability to parse through raw data to make informed inferences, artificial intelligence seems tailor-made to impact healthcare for many years. So far, machines have been trained on huge datasets, where most of the potential insight has been buried. For instance, machine learning, a branch of AI specializing in data analysis, is particularly well-suited to image recognition.
In fact, diagnostic support is one of the more promising areas of development in the near-term since it could aid or augment physicians’ diagnoses. AI is already being used in early detection such as enabling review and translation of mammograms 30 times faster than physicians with 99 percent accuracy, according to a 2016 study published in the journal Cancer.
In the near future, it isn’t hard to imagine AI being a transformative force in alleviating the time it takes to read imaging results for thousands of practitioners. Although machines aren’t likely to take over for radiologists anytime soon, if they are equally as accurate as humans, it might not be long until they’re working hand-in-hand.
AI is also showing promise in tackling some of the bigger problems in healthcare, such as identifying similarities or connections between healthy and sick patients from massive biological data.
Despite the significant investments in AI, making machine learning pay off in more practical ways for physicians will take time. One reason is because the iteration process applied to healthcare has typically started with ideas or possibilities rather than how to meet the needs of clinicians. While the industry might want to implement innovative technology applications, change isn’t coming quite as fast as in other sectors.
Case in point, as recently as 2013, nearly half of all physicians in the U.S. still relied on paper records for patient care, according to a report by Kaiser Health News. Outdated fax machines have yet to be replaced in healthcare primarily because they are secure and reliable, whereas email isn’t.
Perhaps the most promising and hoped for innovation potential of AI in healthcare will be the ability to alleviate the increased time demands on physicians and combat burnout, which is growing disconcertingly high.
“The data from a year or so ago suggests that over 50 percent of U.S. physicians have some degree of burnout, so we do have an epidemic,” said Mark Linzer, MD, professor of medicine, University of Minnesota Medical School and director of the Division of General Internal Medicine and director of the Office of Professional Worklife, Hennepin County Medical Center, Minneapolis, Minnesota.
The quandary is that technology is partly to blame for the increased levels of physician burnout. According to a 2016 study by the American Medical Association and Dartmouth-Hitchcock Health Care System, close to half of a physician’s workday is spent on electronic health record data entry and other administrative desk work. (1)
Ultimately, AI should take some of the burden off of physicians’ shoulders by automating their most time-consuming administrative tasks to allow them to focus on what matters, caring for patients. But how soon can the free flow of data be achieved so that it can be aggregated and available in real time, when physicians need it most? That might be an answer only a machine would know.
However, the gap between what AI technology can do and what clinicians or physicians actually want is slowly beginning to close. One recent development that is meeting the practical needs of physicians is speech-enabled documentation platforms that integrate with EHRs to streamline data entry, freeing up more time to spend with patients. Another advancement called artificial emotional intelligence (AEI) uses machines to analyze verbal and non-verbal cues to determine a person’s emotional or psychological state in order to monitor and guide treatment.
An additional area where AI is taking hold in healthcare is using pattern recognition to identify patients at risk of developing specific conditions. More and more we are realizing that improving care will require aligning big data with the ability to make more timely decisions and the application of predictive analytics can support clinical decision-making and actions.
However, the proliferation of unstructured patient data, which resides outside of organized databases, will need to be accessed by AI in order to make the most of data analytics to manage populations better and make care more cost-effective. The first step might be finding ways to put this data into the workflow of clinicians for real-time access.
If AI could alleviate even a portion of the tedious data entry and uncompensated tasks of physicians, not only could it impact those who are practicing now, but future physicians as well. Just as in other industries, once AI-based systems begin to surpass human performance in certain tasks, they are likelier to take over quickly.
So even as AI gets increasingly sophisticated at doing what humans do, we’re still not at risk of taking the human element out of medicine altogether. Or maybe that’s what machines want us to think.
- Improving Working Conditions Reduces Physician Burnout. Agency for Healthcare Research and Quality, Rockville, MD. http://www.ahrq.gov/funding/grantee-profiles/grtprofile-linzer.html