This futuristic scenario is an example of how artificial intelligence can become part of healthcare. In fact, artificial intelligence systems are already being developed to read medical scans and tissue samples to determine whether a patient has a disease. Future software could analyze patterns in thousands of health records to identify the most effective treatment for a particular patient—for example, which cancer treatment might work best given their genetic makeup.
In a recent paper, David Dranov and Craig Garthwaiteprofessors of strategy at Kellogg, explored the implications of integrating artificial intelligence into healthcare – specifically, how such software will affect the central role of the physician.
For now, the need for human interaction in healthcare is likely to keep AI on the sidelines as a supplement, rather than a replacement, for doctors, Dranove says. But perhaps in a few decades, patients will be comfortable interacting with computers and even trusting them as the primary source of medical guidance. “Maybe in the long run, that will change,” he says.
Mixed evidence
Proponents of this new technology believe that artificial intelligence could help in two main ways.
The first area where AI could penetrate is processing plans informed by data mining. The software could extract patterns from electronic records of past patient characteristics, genetic variants, symptoms, treatment and health outcomes. Based on a new patient’s similarities to previous cases, the AI program may then be able to predict the most effective drugs to prescribe or perform surgery.
The second area is in diagnosis, particularly in the fields of radiology and pathology. A computer could be given a large set of images from previous patients with known diagnoses. The software program could then be trained on these images to identify features that indicate a positive or negative result.
Some studies suggest that AI can perform such tasks quite well—and sometimes spot signs of disease that doctors miss. For example, one group reported that an AI program detected breast cancer on mammogram—especially invasive cancers in the early stages—more accurately than radiologists ever did.
Other studies have explored whether it is better for AI to supplement or replace the expertise of doctors when it comes to diagnoses. But that research has come to conflicting conclusions, says Dranove. In some cases, such as the breast cancer study, AI-guided doctors made less accurate decisions than AI alone.
But in other cases, the combination of the doctor’s expertise and artificial intelligence was the best option. For example, one team tested artificial intelligence software trained on it detection of hip fracture on radiographs. Two experienced radiologists who incorporated the results of the AI program into their assessments performed better than the software alone.
“The evidence is mixed,” says Dranove.
Need for compassion
But, even if the evidence ends up showing that AI can do as well or better than doctors in some cases, will AI really replace doctors? The answer depends in part on how critical human interaction is, says Dranove. For example, doctors elicit information from patients, explain why a procedure is necessary, and provide instructions for aftercare. Dranove believes most seniors today, and perhaps younger adults too, still want to hear from a human about their health.
“There is a need for compassion in communication that artificial intelligence cannot contribute,” he says.
Healthcare organizations may decide that a lower-paid medical professional, such as a nurse practitioner or physician assistant, can fill this role, with their decisions guided by AI. But that, too, will depend on whether doctors’ duties can be boiled down to standard questions and answers, or whether more nuance and expertise is required, Dranove says. For example, a doctor may be more adept at helping a patient feel comfortable discussing their health status and determining how much an illness is really affecting a person’s quality of life.
Even in radiology, one of the specialties that seems most threatened by artificial intelligence, the job still involves substantial human interaction. Dranove and Garthwaite reviewed a list of tasks that radiologists charge for. These included services such as CT scans, CT scans, ultrasounds, mammograms and so on. At first glance, radiologists seemed to largely spend their days using technology.
But a more comprehensive list of jobs, from the Professional Information Network, showed that the job also involved a lot of interpersonal exchanges. For example, radiologists must discuss results with other medical staff and explain risks, benefits, and treatment options to patients.
“It’s not just reading a movie and writing a report,” says Dranove.
Who gets the profits?
The researchers also looked at what would happen to the healthcare value chain if AI became a complement to doctors rather than a replacement. The value chain includes all the parties that contribute to and benefit from it: the patient, the doctor, the nurse, the health care system, the pharmaceutical company, the insurance company, etc. As with the production of any good or service, health care can create value—including better health for patients, wages for providers, and profits for companies—and incur costs.
Because physicians play such a central role, they often capture a large portion of the value in the form of very high salaries. If artificial intelligence took over diagnosis and treatment decisions, one would expect doctors to become less valued and their salaries to decrease accordingly. On the other hand, could doctors end up receiving even higher salaries if they can make faster or more accurate medical decisions with the help of artificial intelligence?
While doctors may become more productive, they won’t necessarily reap financial benefits, Dranove says. Instead, the health care system is more likely to extract added value through higher profits. For example, the organization can improve health care quality metrics and thereby argue to an insurance company that they should be paid more.
“Doctors won’t be replaced by AI, but they may not directly benefit from it either,” says Dranove.
And it is unclear whether even the health care organization will receive monetary rewards. Medical care in the United States is often based on a fee-for-service model. If AI reduces overtreatment and leads to fewer procedures, “you’re losing money,” he says.
Therefore, organizations may not have strong financial incentives to develop and use AI, even if it improves patient outcomes. The exception would be stand-alone systems like the Veterans Health Administration. if they save money, they reap all the benefits.
A patchwork quilt
Embedding AI in healthcare faces many other hurdles. One of the biggest is the lack of access to data. “You can’t beat a doctor based on reams of data unless you have lots and lots of patients to train the computer on,” says Dranove.
In the United States, medical records are scattered across healthcare systems, and HIPAA limits the ability to share information. As a result, most AI developments so far have taken place in medical organizations that only use their own patient records. This means that large healthcare systems have an advantage over smaller ones, which may not have enough data to effectively train the software.
“I think we’re going to see a patchwork quilt where AI is applied,” he says.
While it is possible that these large organizations could share their trained software with others, they may be reluctant to do so. “From a societal perspective, I should share this information” because it could improve health outcomes for patients elsewhere, Dranove says. But the organization’s perspective might be, “why give away something for free that makes my system so much more valuable?” Without a federal law requiring data or software sharing, he says, “I think this is going to be an extremely fragmented process for a long time.”
That doesn’t mean large healthcare organizations should be slow to deploy AI, he says. But a coordinated approach will spread the benefits of AI more evenly.
“If the data can be shared, then everyone will have that opportunity,” he says.
