AI Could Raise U.S. Health Care Costs

Despite whether AI is viewed from a chaos-theorist or techno-libertarian perspective, big data arguably rivals the introduction of the encyclopedia. We’re in the honeymoon phase, day-dreaming about big data’s infinite problem solving capabilities and profit-potential. In the healthcare economy, AI will generate $6,662.2 billion in revenue in 2021, according to a Frost and Sullivan report.
But The Hasting’s Center research indicates health care costs are rising 7 percent per year. For employer-sponsored health insurance plans, if the national U.S. average employee contribution is $1,427, what will it be in the future?
Without a major intervention, Medicare will dry-out in nine years. Companies are shifting cost burdens to employees through reduced wage increases, less extensive coverage and higher deductibles. Compared to Organization for Economic Co-operation and Development (OECD) countries, health spending per person in the U.S. was $10,224 in 2017, surpassing Switzerland, the next highest per capita spender, by 28 percent. Yet we’re still behind in the quality of health care, which, ironically, is supposed to be our inviolable defense.
The hidden driver of soaring costs is medical innovation, which contributes 40–50 percent to annual cost health-care cost increases, according to the same The Hastings Center report. Despite big data’s prodigy, it can’t fix a collapsing health-care system and may even accelerate its insolvency.
This is a difficult white elephant to address, especially in medicine as AI applications are saving lives. Practitioners, for instance, can compare a patient’s medical profile, gene structure and other diagnostic data stored in the cloud to millions of other patients in seconds. Vice Presdident & Global Head of Healthcare Domain Practice at Tata Consultancy Services, Magna Hadley, said, “Doctors can now analyze the DNA of a patient or a tumor in a matter of hours, spending only a few hundred dollars. This is something that took 10 years and cost billions when the first genome was sequenced.”
IBM Watson Health is the most notable celebrity, being used at 16 of the most prestigious cancer institutes in the U.S.
Watson, however, is just one subset of AI being used in the medical field. Dr. Anthony Chang, chief intelligence and innovation officer at Children’s Hospital in Orange County, California, said, “Think of Watson as the string section of an orchestra. There are many sections that comprise an orchestra. Deep learning will be a big part of AI in medicine. But data visualization, data mining and data science are other areas as well.”
The logic is that this data will also ease medical costs. Workflows will be improved and redundancies will be eliminated; patients will receive more nuanced diagnoses saving them from unnecessary treatment and its sticker shock; and big data will eliminate costly and elaborate drug trials, the reasoning goes.
But patients are billed for labor, lab work, procedures that use medical devices and drugs — all of which will use AI more extensively.
Dr. Mark Wolff, Ph.D., chief health analytics strategist for the Health & Life Sciences Global Practice at SAS and author of The Quantum Patient: Analytics and the Future of Health Care, consults for medical device manufacturers. “Think of HAL in 2001: A Space Odyssey and having a CT scanner with the cognition to anticipate operational failure and safety compromises,” he said.
According to a 2015 ECRI Institute report, however, CT scanners cost from $300,000 to as much as $2 million, depending on their level of technological sophistication. An average CT scan costs a national average of $896 compared to $97 in Canada, according to a Journal of the American Medical Association study . And another report in the Journal of American Medical Association indicates the rate of diagnostic medical imaging use continues to increase.
Although physician practices, smaller hospitals and clinics will have equal access to big data services on the internet, the best technology, however, will continue to reside in larger hospitals, according to Wolff.
Robert Wachter, M.D., department chair of medicine at the University of California, San Francisco, in this book, Digital Doctor: Hope, Hype, and Harm at the Dawn of Medicine’s Computer Age, anticipates smaller hospitals will close out-performed by larger hospital networks more amenable to AI integration. And according to a New York Times article, hospital mergers set elevated prices in private markets because of their dominance.
In pharmacology, AI promises to reduce the costly efforts in bringing a new drug to market from drug discovery, to drug development, to commercialization. But no regulations exist that govern how manufacturers set prices. So a best-in-class drug created with AI could capture staggering rates. This doesn’t factor in individual insurance policies and other business-cycle variables.
The insurance industry will increasingly leverage AI in the form of behavioral analytics, adding another class of cost-variables. Wolff points to Adobe Systems, commonly known for its server, web and design platforms, with the company’s application of “Audience Intelligence” analytic software that collects data on human behavior. “The customer is replaced with ‘patient,’” said Wolff. “It makes no difference to Adobe.”
He envisions a complex Fitbit-like device, with curated health data periodically fed to an online-based customer profile. “Think of the car insurance industry offering a lower rate with an installed GPS tracker. Monthly premiums will rise or lower based on how good or bad of a driver the customer is,” said Wolff. Based on this prediction, healthy patients may be incentivized with lower premiums, co-pays and deductibles while sicker patients, by no fault of their own, may end up paying more.
A study addresses the issue from a macro-economic perspective. A nation’s GDP has direct influence on the prevalence of medical technology and per household healthcare affordibility.
Various policies have tried to stymie these trends. We introduced Medicare revisions to hospital payouts in the 80’s, instituted public and private healthcare organizations in the 90’s, and reduced expenses for patients with chronic illnesses in the 2000’s.
One analysis confronts the national health care debate by demonstrating that expanding state-sponsored coverage to only low income populations actually raises costs. Government agencies, such as ACA, Medicare and Medicaid are more liable if medical technology raises expenses. But spreading coverage to wealthier demographics as well lowers medical innovation costs without affecting volume. This is because higher income households can afford higher markups with private insurance, such as in single-payer European systems.
“So, Watson, when will our healthcare system become insolvent?” Watson: “I’m sorry. I do not have an answer for your problem.”