AI supported medical processes can help to save human health care professionals

With reports streaming in of health care workers falling ill from COVID-19 as they battle heroically against the coronavirus, we must take a cold, hard look at the math.

As the number of people infected with the virus goes up and the number of available health care professionals goes down, we soon won’t have enough doctors and nurses to treat us, no matter how many ventilators we can produce. This problem, like the disease itself, will be solved once there’s a workable vaccine, but we don’t have a year to wait. And we can’t just throw our invaluable health care professionals at this problem, like soldiers rushing from World War I trenches. There’s a better way, but only if we act aggressively and fast.

For years, futurists like me have foretold the coming intersection of telemedicine, artificial intelligence (AI) and health care, as the reality on the ground has only inched forward. The misaligned financial interests of big medicine, and the licensing and regulatory inertia of big government, have guaranteed that the health care sector moved at a fraction of the speed of the tech sector. To help us reduce suffering, save lives and get through this crisis, that’s got to change. 

Here’s the emergency plan: 

First, we agree on a basic set of inexpensive health diagnostic tools everyone should have at home. These could include a smartphone-enabled mobile electrocardiogram and a digital thermometer, a fingertip pulse oximeter, a blood pressure cuff, a scale and a digital stethoscope. The items could all be packaged together in a government-approved bundle, sold as a single unit, and delivered through online retailers or provided for free to those who can’t afford it.

Second, the U.S. government should oversee the establishment of a single, AI-driven national health care triage platform that would be the first point of contact for anyone experiencing new symptoms. These people would first do app-guided home tests on themselves using their home diagnostic tools. They would then input these readings on the platform and fill out an electronic questionnaire that would start broadly and then narrow with each answer. If the program determines the potential patient can be treated with AI-generated advice, that recommendation would be made and an automated text message follow-up procedure would immediately kick in. If the text-message exchanges indicate a worsening problem needing additional attention, the AI program would be designed to pick that up.

Third, people whose symptoms are determined by the AI program to need additional attention, or those referred through the text-message process, would then be referred to telemedicine doctors or nurses to discuss the situation in a video call. These health care providers could be in the U.S., but if the COVID-19 crisis continues to worsen here and alleviates elsewhere, we could also hire foreign doctors or nurses to do this work. These telemedicine providers could handle some situations through video consultations alone. In other cases, they could ask patients to send stool, urine or even pinprick blood samples to local testing centers through kits that could be mailed both to and from patients. The telemedicine providers could refer the patients for further testing at a designated facility, or to specialists who would see the patients first by video and then, if necessary, in person.

All of this could be done in conjunction with and as an overlay to our current mishmash of insurance companies and other providers using technology that mostly already exists. By overseeing the quality and consistency of the AI backbone and taking responsibility for errors that may be made in this AI-generated triage process, the government will free the entire U.S. health care infrastructure to focus more on both the COVID-19 crisis and the neediest patients not experiencing this disease but still suffering from diabetes, heart disease, cancer and other acute, chronic diseases. The federal government would need to work with states to ease telemedicine licensing barriers that currently prevent doctors licensed in one state from seeing patients in another state, even if only virtually. 

Even though the United States was spending more than 18 percent of our GDP on health care, more than twice the percentage of GDP of any other country, we had some of the worst collective outcomes of any country even before this crisis began. With our need for health care rising and our supply of critical health care professionals likely to decline precipitously over the coming months, we must go significantly virtual now.

Having the government step in to set up a human-designed AI screening mechanism will be uncomfortable for many people — but that is where we already are heading. Medicine already has become far too complex for humans to perform alone. Virtualizing some of the first points of patient contact, however imperfect, could also provide an invaluable service to the many millions of Americans who have no insurance and extremely limited access to health care.

Ceding what may be life-or-death decisions to our AI agents is a less than ideal choice. But if we want to have enough humans to provide the essential human care we need, we must help our essential health care providers let go of everything else.

AI plus human telemedicine has always been our future. To save our health care professionals and ourselves, we must make that future happen now. 

Jamie Metzl is a member of the World Health Organization’s international advisory committee on human genome editing, a Singularity University Exponential Medicine faculty member, and a senior fellow at the Atlantic Council. He previously served in the National Security Council and State Department during the Clinton Administration and is the author of “Hacking Darwin: Genetic Engineering and the Future of Humanity,” to be released in paperback on April 7. Follow him on Twitter @jamiemetzl.