AI helping doctors heal

In this opinion piece, Britta Verhey (Stanford University; CA, USA), explores the importance of the physician-patient relationship, how governmental involvement and inadequate technology disrupted the physician-patient relationship, and how AI can reestablish this fundamental and invaluable interaction. “I want to help. I want to connect with people. A doctor interacts with people at their most vulnerable. He offers treatment, be he also offers counsel and hope. That’s why I love the idea of being a doctor.” —Patch Adams Every year thousands of medical students across the country stand to recite the Hippocratic Oath, signifying the beginning of their training to become a physician. Most of these students choose to become a physician so they can help other people , matching the philosophy held by Patch Adams. Following many years of training, these newly minted physicians enter practice and most continue to hold onto their motivation to help others. This motivation carries through to what most physicians see as the most rewarding aspect of being a physician—their relationship with patients . Unfortunately, this unique human relationship and patient-focused model of healthcare was disrupted with the shift to a government-regulated model. Instead of focusing on reducing suffering and making a meaningful difference in the lives of patients, governmental involvement in health care relegated physicians to maintaining crushing regulatory and clerical responsibilities. Efforts to reduce these burdens through technology were attempted over the last two decades, but the technology applied only led to increased clerical burdens for the physician and further disruption of the physician-patient relationship, leading to physician burnout and patient dissatisfaction. As technology has advanced, however, a new tool has become available and is being investigated and developed as a potential solution to reduce physician clerical burdens and help improve physician-patient interactions—artificial intelligence . In contrast to prior technology solutions, artificial intelligence (AI) can unify disparate software platforms and harness the power of the cloud, natural language processing, and large databases. These advantages of artificial intelligence may be leveraged to reduce physician clerical burdens, improve physician efficiency, and increase the quality of care, potentially restoring the traditional physician-patient relationship. It is my goal that the following discussion will provide perspective on the importance of the physician-patient relationship, how governmental involvement and inadequate technology disrupted the physician-patient relationship, how AI can reestablish this fundamental and invaluable interaction, and the importance of physicians embracing AI and participating as leaders in its development and application. The traditional doctor-patient relationship was established in the early 1800s and incorporated the idea of a social contract between physician and patient, where physicians were held to ideals such as altruism, competence, and scientific rigor . The relationship is built on trust, respect, communication, and shared goals . Importantly, a good physician-patient relationship has long been understood as the foundation of patient satisfaction, leading to better adherence to treatment plans, and consequently, improved health outcomes . Conversely, poor, or weak physician-patient relationships can have negative implications on clinical care. What constitutes a good physician-patient relationship? A recent study from Stanford University examined this question and determined that effective communication was the most important component of a positive physician-patient relationship. Effective communication between the physician and patient includes both verbal and non-verbal cues, such as maintaining eye contact, listening intently and completely, responding to emotional signals, connecting to the patient’s story, and collaborative decision-making . Given the importance of the doctor-patient relationship, much time is spent in training medical students in the skills necessary to create successful relationships with their patients, with training focused on building effective communication skills, empathy, respect, and trust. Unfortunately, the traditional, altruistic nature of the physician-patient relationship was disrupted when government deliberately transitioned from its traditional healthcare oversight role to an active role in healthcare delivery, regulation, and reimbursement. In 1965, the United States Congress implemented two new programs, Medicare and Medicaid, with the goal of providing governmental-subsidized health care to anyone over 65 years of age, as well as for those of any age without means . While the altruistic intention of these programs was noble, the practical application created a cascade of perverse incentives, bureaucratic regulations, and economic pressures. Patients and providers were no longer cost-conscious, leading to year-after-year increasing health care GDP expenditures . In response, instead of allowing natural market force corrections, the government became more involved by imposing rigid controls on the programs, including through new price control methods such as DRG’s (Diagnosis Related Groups) and CPT (Current Procedural Terminology) codes. With the application of these price control methods, additional metrics and regulatory powers became available to the government bureaucrats, including forcing providers to follow care algorithms and respond to quality standards, Consequently, priorities in medical care shifted to incorporate such economic principles as cost containment and efficiency—health care became a health transaction. Physicians were made into “expense-gatekeepers” for the government and for-profit insurance companies, who required physicians to submit extensive documentation and obtain prior authorizations for treatments . Government interference and for-profit business involvement drove a wedge between physicians and patients, disrupting the sacred physician-patient relationship. As doctors became overwhelmed with paperwork and coding requirements, the tech-boom of the 1990s offered the electronic health record (EHR) as a solution. Although the new technology was initially billed and embraced as beneficial to both doctors and patients, with promised improvements in reimbursement, efficiency, accuracy, and quality, these purported benefits never materialized. Instead, there was an enormous increase in administrative tasks, as the computer software was inadequate and had one-size-fits-all requirements. Additionally, it was now essential to have a computer in the exam room, with the data entry requirements for the EHR and insurance company fulfilled by the physician. Physicians paid more attention to the computer screen (or what Abraham Vergehese calls the “iPatient”), instead of the real patient . Billing and documentation requirements that used to take doctors a few minutes to complete, now occupied a large part of each patient visit. One study found that family medicine doctors spent 6 of 11 hours each workday looking at the EHR . The combination of government involvement in health care and the subsequent implementation of obtrusive technology into the exam room, intrinsically changed the traditional physician-patient interaction from a face-to-face, conversational, personalized, therapeutic alliance to a depersonalized, fragmented interaction, focused on technologic interface and data entry. Physicians lamented that the focus on costs and business-oriented technology reduced them to cogs in the health care business machine, with productivity expectations and requirements to fulfill regulatory and clerical burdens . The result of the shift away from the caregiver orientation to the business orientation included increasing dissatisfaction with work and ultimately physician burnout, resulting in an additional $4.6 billion in costs to the system . The Physicians Foundation, in its 2018 Survey of America’s Physicians, reported that less than one-half of doctors reported they were satisfied with their jobs, that three out of four felt some level of burnout, and 62 percent were pessimistic about the future of medicine . Patients are also dissatisfied with the shift away from the traditional physician-patient relationship, with one recent study indicating that 71% of patients preferred the physician who didn’t use the computer and another showing that 75% of patients wishing their healthcare experiences were more personalized . With the EHR the source of so many negative effects on physicians, patients, and health care in general, including failing on the original intentions of increasing efficiency and reducing costs, there have been many efforts directed toward finding a better solution. Fortunately, technology has advanced, specifically with the development of artificial intelligence, giving a new and completely different tool genuinely capable of helping physicians. In contrast to prior technology and the EHR, AI doesn’t require keyboard input, up-front processing, or dependence on a monitor. AI runs in the background and can be trained based on processing of database information . AI utilizes tools such as the cloud, natural language processing, and database mining to complete tasks and develop solutions. AI’s greater specificity for tasks, unlimited potential, customization, and diverse applications will likely help restore the patient-physician relationship. AI is already being used to remove redundant, repetitive tasks from a physician, including EHR entries, ordering labs and imaging, writing prescriptions, and communicating next steps to the patient. Removing these tasks alone will give physicians a significant amount of time back to the physician and enable he or she to focus on face-to-face interaction with the patient. As AI technology and applications develop, additional background tasks may provide even additional time for the physician-patient interaction, as well as enhance the interaction, through integrating health information to generate a more accurate diagnosis and treatment plan. Despite being in its infancy, AI applications are exponentially growing, with already an estimated 50 health care tasks being performed or assisted by AI . For example, the AI technology of voice recognition eliminates the task of manually entering chart information into the electronic system. AI-powered digital scribes can automatically generate a clinical note and unshackle physicians from the EHR, restoring one of the fundamental components of a good physician-patient interaction— eye contact. Other tasks that AI is already able to perform in the background while a physician is communicating with a patient include, determining insurance eligibility, making insurance claims, obtaining prior authorization, generating appointment reminders/scheduling, and completing the required coding for billing purposes. Beyond these repetitive clerical tasks, AI is also already performing physician practice management tasks. Some of these include examining patient visit numbers for the scheduled day based on the complexity of the patient needs and visit requirements, ensuring adequate time and staffing support for each patient. Future applications of AI are currently being developed that will focus on more advanced clinical and patient management duties, including medication management, mining medical records, designing treatment plans, drug creation, and patient triage . One of the largest areas of research and development for future AI applications include processing the vast amount and rapidly changing scientific research and medical information. Unlike the limitations that humans have for remembering and implementing knowledge, AI can easily store, access, and integrate the data intelligently, giving doctors the ability to give their patients more effective and personalized treatment plans in a shorter time frame . AI is also being trained to look at patients holistically, integrating their social determinants of health into the record, reducing stress for physicians, who are increasingly responsible for the total outcome of the patient. With current AI applications and those to yet be developed, doctors will be able to diagnose and treat their patients more efficiently and effectively, helping to restore the traditional physician-patient interaction, improving physician and patient satisfaction, increasing accuracy, reducing errors, and decreasing physician burnout and improving mental health. Implemented wisely, AI has great potential to restore the patient-physician connection by shifting the physicians’ responsibilities away from transactional tasks toward personalized care that is the foundation of human healing. As AI technology advances and provides more intelligent support to the physician, the traditional role a physician plays in the physician-patient therapeutic alliance will evolve. Although AI is already performing many health care tasks and has a seemingly limitless potential for additional duties, there are critics that feel AI in medicine will fall far short of expectations. Those with a less enthusiastic perspective of AI point toward several hurdles that have yet to be addressed, including displacing entrenched workflows and proprietary EHR systems. Additionally, there are likely several legal, ethical, and privacy issues that will be faced, such as liability for medical errors and security of private health information . Finally, in addition to the potential limitations for AI to assist physicians by reducing administrative burdens, there are also questions about the impact for AI tools to provide clinical benefits and improve patient outcomes. A recent meta-analysis of all articles related to AI and clinical benefits through July 2021, found only a small percentage were quality, randomized controlled trials , indicating the need for additional high quality research to determine actual clinical utility and benefits, before being implemented into practice. Beyond the intrinsic potential limitations of AI, the successful development, implementation, and usage of AI is limited without the proper training. AI relies on large data sets and if there isn’t enough data, then AI less accurate. Additionally, since AI trains using real-world data, the economic and social biases contributing to health disparities may be included and perpetuated , although this risk can be mitigated by increasing the diversity of databases and customizing algorithms. Finally, critics point to patient acceptance of AI as a potential major hindrance to AI success, as patients have already expressed their dissatisfaction with technology replacing the traditional physician role. In one recent study looking at this phenomenon, people were less receptive to advice given by an AI doctor that knows their name and medical history. Conversely, patients do want to be on a first-name basis with their human doctors . In another study looking at patients’ receptivity to medical AI for a variety of patient conditions, patients preferred to have a human provider perform the service even if there was a greater risk of inaccurate diagnosis or a surgical complication. These results are thought to be rooted in the patient perspective of their uniqueness and that AI providers were inadequate to account for their individual circumstances . In a recent Pew Research poll, patients also expressed concern that their healthcare provider will rely too much on AI to diagnose disease and recommend treatments (60%) . These patient preferences for a human doctor and no intrusive technology can be assuaged through communication and education about AI and how it is a background assistant only being used to help the human doctor provide better, more personalized care. Possibly instead of focusing on the traditional terminology of “artificial intelligence,” patients may be more receptive to the technology described in terms such as “augmented intelligence” or “human-centered AI” . Additionally, physicians could convey to patients that AI is able to tailor recommendations on treatment to each patient’s unique characteristics and medical history, while also considering the most recent literature and guidelines. Finally, patients feel more receptive if a physician remains in charge of the ultimate decision. Physicians could utilize phrasing such as, “Based on your unique profile and the most current evidence-based guidelines, I recommend following the recommended treatment plan.” While there are limitations to AI and multiple challenges to overcome, AI’s capabilities can eventually make it an integral part of health care. However, if AI is to be fully accepted and become a part of standard practice, it must take a human-centered approach, where patients and physicians are the primary focus of the technology, including seamless integration with patient workflow and ease of use. Physicians must understand the technology, be leaders in its development, and successfully apply it to enhance their practice and their patient’s health . To this end, how do we ensure that physicians and patients accept AI as an essential part of the physician-patient relationship? Acceptance of AI will be dependent on overcoming fear—in physicians and patients alike. One of the biggest problems of past efforts to introduce technology into a physician’s role, is that it was applied without physician input and without consideration of the patient’s needs. As a result, the technology implemented was geared primarily toward the financial and business aspects of medical care at the expense of the physician-patient interaction. Although AI is more intrinsically amenable to helping the physician and patient, it will still be important for the physician to take an active role in development and application of AI to ensure focus on patient care and not the administrative priorities of finance. To this end, medical students and physicians must have a working knowledge of AI, so they can see beyond the hype, evaluate AI-based studies, develop, and assess AI applications, and identify AI limitations and potential negative implications. AI education should be a part of medical school curriculum, including basics in computer science. Physicians must become active collaborative partners with administrative leadership to identify the critical problems and then design and test solutions to best affect patient care . Although administrators may have difficulty with a change in focus from finances to the patient, physicians must convey the principle that optimizing the physician-patient relationship and refocusing attention to the patient, will have both short-term and long-term positive effects on reducing health care costs. For example, only implementing basic AI tools to reduce redundancy and unnecessary work attributable to fragmentation could save an estimated $25 billion to $45 billion per year . Employing more advanced AI tools to assist in managing preventable or chronic diseases could decrease hospitalizations by the millions, leading to $100 billion a year in savings . Finally, a recent analysis by the National Bureau of Economic Research indicates that $200-360 billion dollars a year could be saved in healthcare with implementation of AI, primarily through efficiency improvements . While the business aspects of medical care are a necessary component in the efficient delivery of healthcare, preserving the quintessentials of caregiving that speak to what is most deeply human in medicine and the physician-patient relationship should be a priority. In my opinion, establishing an appropriate balance between these differing perspectives with the aid of AI, is not only possible, but essential for achieving the quality and sustainability of healthcare that patients and providers deserve. In addition to ensuring the mistakes of prior technology implementation are not repeated by developing AI with a patient-centric focus, it is also important to recognize that AI will never replace the human role of a physician. Although AI has significant potential to assist the physician with many tasks, there are aspects of being a physician and components of the physician-patient relationship that AI will never be able to perform. For example, AI does not have the capability to replicate human emotions, judgement, intuition, and touch, all of which are essential skills of an excellent physician and critical to a good physician-patient relationship. Consider a scenario where a child is visiting a pediatrician due to an “illness.” The pediatrician can see through the feigned illness of a seven-year-old playing sick to avoid school. Pediatricians may also be able to discern that the act is instead a symptom of something going on at home or school. This intuition and nuanced understanding are something that an AI algorithm will not be able to replicate. Probably the most important trait of a physician that AI can’t reproduce is empathy. Physicians can listen, understand, and emotionally-connect to their patients, allowing their patients to trust and confide in their physician—something that AI will never replace. Although AI won’t replace the human aspects of the physician and the physician-patient relationship, it will shift medical care from one of intuitive guesswork and pattern recognition to one of precise, targeted medicine. Instead of trying to memorize and apply an insurmountable amount of information, physicians can utilize AI to access what they need to know and apply the information into a broader human context. In his book, The Most Human Human, Brian Christian suggests that instead of replacing humans, machines will give us the understanding of what it means to be human. Christian believes AI will remove only the non-human parts of a physician’s work, restoring the physician to health and consequently enhancing the physician-patient relationship . Given the enormous challenges facing the healthcare system of a physician shortage, rapidly aging population, a decline in the working population, and persistent high costs of care, AI can offer a transformative set of tools to meet these challenges. Although the idea of AI serving a role in the physician-patient interaction may be concerning to some, AI has the potential to restore the invaluable connections of the traditional-patient relationship, as well as embrace new ways to coordinate care, make diagnoses, develop treatment plans, and foster improved outcomes. Physicians won’t and shouldn’t be replaced—physician empathy, intuition, insight, and patient desire for human relationships will preserve the necessity for physicians and the physician-patient interaction. On the contrary, a physician’s role will be enhanced, and the physician-patient relationship will benefit. “AI will not replace physicians, but physicians who use AI and present social skills will replace those who don’t” . Therefore, it is paramount that physicians embrace the technology, participating in its development and determine when it strengthens and when it undermines patient care. AI will enable physicians to return to what brought them to medicine—the care of a fellow human. AI will help doctors to heal. Disclaimers:The opinions expressed in this feature are those of the interviewee and do not necessarily reflect the views of Future Medicine AI Hub or Future Science Group.