Within in five years primary care providers will begin being replaced by sociable humanoid robots, avatars, and computer programs. Within ten years you will no longer hear any complaints about medical students choosing specialty residencies over family practice because the role of the physician will be completely redefined to complement a rules based approach to the diagnosis and treatment of many diseases. This transformation is inevitable because of demographics, economics, and progress in artificial intelligence, but the academic leaders of medical education and health policy are largely ignorant and unprepared for this massive disruption.
It is hard to understand this change because it requires us to challenge the conventional wisdom that we need more doctors and that a trusting relationship can only be had between two human beings. It will also require us to understand the implications of the evolving relationship between human beings and technology, and how disruptive innovations change industries.
Many do not see this future because they fear it. Samuel Stenes, MD spoke for many physicians when he posted the following comment on a blog I wrote about this subject: “You are truly out there. People do not want that cow waste, but the profession has been sufficiently depreciated that anarchy care will rule.”
Philosopher Roger Scruton makes a similar eloquent argument about trusting relationships requiring two human beings:
In real life, friendship involves risk. The reward is great: help in times of need, joy in times of celebration. But the cost is also great: self-sacrifice, accountability, the risk of embarrassment and anger, the effort of winning another’s trust. Hence I can become friends with you only by seeking your company. I must attend to your business … When I relate to you through the screen there is a marked shift in emphasis. Now I have my finger on the button. At any moment I can turn you off … Of course I may stay glued to the screen. Nevertheless, it is a screen that I am glued to, not the person behind it.
Professor Andy Clark of Edinburgh University works in the “embodied cognition” and “extended mind” fields of philosophy, and he has an entirely different take on this issue. Citing studies that show that research subjects who are prevented from using hand gestures perform poorly on tests of mental ability, he concludes that some of the activities that enable us to think and create trusting relationships occur outside of our body and brain: “Evolution and learning don’t give a jot what resources are used to solve a problem. There is no more reason, from the perspective of evolution or learning, to favor the use of brain-only cognitive strategy than there is to favor the use of canny (but messy, complex, hard-to-understand) combinations of brain, body, and world.”
The health care problem that the United States needs to solve is how to increase the quality and decrease the per-capita cost in an era of aging Baby Boomers, inadequate numbers of primary care providers, and a struggling economy that cannot support spending 18% of its GDP on health care. The messy, complex, hard to understand solution is that sociable humanoid robots and computer programs will replace human providers in diagnosing and treating some diseases.
Clayton M. Christensen, Jerome H. Grossman, MD, and Jason Hwang, MD in The Innovator’s Prescription provide a framework for understanding how disruptive technologies are making the current business models for diagnosing and treating chronic disease obsolete. They divide chronic diseases into two types: intuitive chronic diseases and rules-based chronic diseases. Intuitive chronics diseases such as lupus, back pain, and schizophrenia lack clarity in diagnosis and treatment and so at the present time still require a human physician working in a multidisciplinary solution shop such as a tertiary care hospital. Rules-based chronic diseases such as Type II diabetes, GERD, and heart failure are well enough understood scientifically that a computer following an evidence-based medicine protocol can diagnose and treat the patient.
But computers are not that smart and patients will never accept such an arrangement. The first objection does not stand up to scrutiny when one thinks about the implications of the four year Watson/Jeopardy! Project at IBM. Columbia University Professor Herbert Chase, MD believes computers will replace physicians as expert diagnosticians by utilizing the same methods that defeated human Jeopardy champions: taking a question about a patient’s symptoms, analyzing the question, generating a differential diagnosis, collecting and evaluating the entire medical literature on the subject, and coming up with a diagnosis with a measurable level of confidence.
David Ferrucci the leader of the IBM Watson project explains how Watson can do this in a 20 minute video, and Stephen Baker’s book Final Jeopardy: Man Vs. Machine and the Quest to Know Everything is a fascinating description of the development of Watson by a twenty person team at IBM.
United Hospital in Saint Paul, Minnesota is already taking a similar approach by using Isabel to search the medical literature in real time to generate a differential diagnosis for the hospitalized patient based on the patient’s symptoms and medical history. “Isabel is like a medical Google. It doesn’t think but it searches and sorts and shows you what it’s found,” says Scott Tongen, medical director for quality at United. Tongen notes that such a system can help with the number one reason for mistakes in diagnosis: early closure on a diagnosis with confirmation bias preventing the physician from considering alternative causes for the patient’s illness.
Studies are starting to appear that document improvements in diagnosis with the use of computer technology. Computers outperformed radiologists in diagnosing Alzheimer’s Disease; face identification software did better at diagnosing acromegaly than endocrinologists; a 2010 British Medical Journal study of a pediatric computer-aided diagnosis tool revealed a higher rate of accurate diagnosis of urinary tract infections, pneumonia, and bacteremia compared to the performance of emergency room physicians.
The Watson approach to improving diagnosis by searching the entire medical literature is not the only way computers will change clinical practice. Dr. Joseph C. Kvedar of Partners HealthCare’s Center for Connected Health has developed the concept of emotional automation, and he believes that humans can and will develop trusting relationships with sociable humanoid robots. Kvedar notes that humans have a tendency to anthropomorphize objects such as pet rocks and tomagotchis, and Sherry Turkle of MIT has famously observed that smartphones and tablet computers have transformed our relationship with technology: “It is different now that we carry our second self with us. We think with the objects we love and we love the objects we think with.”
Kvedar believes that Karen the virtual wellness coach/avatar who motivates human patients to exercise more than a control group and hospital patients who preferred a robot discharge planner to a human one are examples of humans learning to trust technology in a healthcare setting. The robot was favored because it was not in a hurry and did not talk down to the patients being processed for release from the hospital.
Technology is also beginning to change clinical practice when therapists use digital worlds with autonomous, virtual human avatars to help patients work through problems with social anxiety, drinking, gambling, post-traumatic stress, and agoraphobia. Therapists can discuss the patient’s feelings when the virtual bartender asks the patient’s avatar if he wants to order another drink. Different coping mechanisms can be practiced in real time in virtual situations that are experienced as real. In one study, people with social anxiety disorder confessed more of their personal flaws, fears, and fantasies to virtual avatars programmed to be socially sensitive than to live real therapists. B.J. Fogg of Stanford has written extensively about what he calls “captology,” computers as persuasive technology, and his thinking needs to be integrated into clinical practice.
Nick Yee has described the Proteus Effect where video game or digital world avatars change how we behave in virtual environments and in real life. Subjects given more attractive or taller avatars disclosed more personal information and bargained more aggressively than unattractive, shorter avatars. Yee has shown that the subject’s perception of his or her own attractiveness persists outside of the game or virtual world to affect participation in real life online dating. Yee writes that providing users with “fit, athletic avatars in exergames may encourage longer and more engaged exercise sessions than if they were provided with normal-looking avatars or avatars modeled from their own bodies.”
Will patients embrace technology or refuse to form trusting relationships with sociable humanoid robots? I don’t know, but the rapid adoption of smartphones and initial studies indicate some may welcome technological solutions to common problems. A qualitative Beth Israel Deaconess Medical Center study utilized focus groups in Boston, Portland, Tampa, and Denver, and the subjects fully expect that computers will be increasingly used in health care and substitute for face-to-face physician encounters: “Patients know how busy their doctors are and they want to reserve us for what they really need us for – treating serious illness … They may be more than happy to rely on computer protocols and ‘faceless doctors’ to help them manage garden-variety medical problems,” says senior author Tom Delbanco, MD. Stephen Downs of the Robert Wood Johnson Foundation commented on the study, “Year after year, people have seen information technology transform one industry after another and, more to the point, transform their everyday experiences. This is the age of the iPhone, Facebook and Google Maps, yet health care feels very much the same. This study suggests that people are ready for change – they want a modern health care experience.”
During Man-Made Minds: Living With Thinking Machines at the World Science Festival in New York City, Eric Horvitz of Microsoft showed a tape of a medical kiosk avatar interviewing a mother and her child who has diarrhea. After asking several questions, the avatar decides the child is not sick enough for immediate attention and makes a clinic appointment for the child for later in the week. The moderator of the panel of artificial intelligence experts got a big laugh from the audience when she noted that the avatar was much more compassionate in relating to the child and his mother than human triage nurses she has encountered as a mother taking her child to New York City Hospital emergency rooms. Anyone who is not familiar with recent advances in artificial intelligence needs to acquaint himself or herself with Horvitz’s avatar assistant on a computer screen outside his Microsoft office. Because this computer program has been learning about Horvitz for over 10 years, she can predict that he will not attend a meeting on his schedule for 4:00 PM and she can offer that time slot to a colleague who she recognizes by name and who needs to meet with Horvitz. It is perhaps relevant that all four artificial intelligence expert panelists predict that robots will be built that will be hard to tell apart from real human beings.
It would be wise for health policy and academic health care experts to educate themselves about recent developments in artificial intelligence and computer technology; the implications for health care delivery are enormous and may change projections about cost, workforce requirements, and the training of health care professionals.
Kent Bottles provides health care leadership consulting and blogs at Kent Bottles Private Views.
Submit a guest post and be heard on social media’s leading physician voice.