Conventional wisdom has it that recent expansions in coverage of telehealth will never be fully reversed. The addition of artificial intelligence (AI) into telehealth could solve several issues faced by doctors and hospitals. There is some concern, however, that the blending of AI and telehealth will industrialize the practice of medicine, dissuading patients from seeking critically needed care.
The American Telemedicine Association, of Arlington, Va., held its annual meeting virtually June 22 to 26, and one of the sessions took up the question of the pitfalls of a convergence between AI and telehealth. Terrence Lewis, senior associate counsel for the University of Pittsburgh Medical Center (UPMC), said the introduction of AI into clinical care has prompted “a spectrum of different reactions.” A few clinicians may adopt the technologies more quickly than can be appropriately institutionally supported. However, Lewis said, “the vast majority of our providers have begun to embrace the technology” of AI and telehealth, in large part due to the COVID-19 pandemic.
This is the case for both primary and specialty care, but Lewis said clinicians are wary of allowing the practice of medicine to become industrialized to the point at which the patient is converted into a mere logical input. Ultimately, the decision-making by the provider is still the crux of medical practice, and Lewis said the architects of care systems must ensure that the role of the health care professional (HCP) doesn’t become secondary to an algorithm that is seen as having morphed into a virtual doctor.
Hans Arora, a pediatric urologist at the Lurie Children’s Hospital of Chicago, said physicians initially were concerned about loss of jobs to AI. This concern has more or less dissipated and been replaced by concerns over safety. More experienced clinicians may be less enthusiastic about adopting novel technologies than their younger counterparts. However, clinicians of all inclinations are wary of possible compromises of privacy and data security with the clinical use of AI.
Doctors must retain ability to fly solo
Arora said there is also some concern that AI might erode clinician skill and boost the educational requirements for aspiring physicians, who will have to possess more than a glancing familiarity with AI in the years to come. “We have to feel comfortable managing patient care in the absence of some tools we have today,” he said, adding, “we have to be comfortable enough that we can take care of the patient without AI.” This latter consideration may prod clinicians to conclude that they must lead AI into clinical care rather than allow it to be inserted into medical practice in the absence clinician governance.
Terri Casterton, senior director for innovation at SCL Health Inc., of Broomfield, Colo., reminded attendees that the amount of time clinicians spend to get through the patient’s self-reporting of medical history and current symptoms is enormous. This becomes especially problematic as the demand for clinical care continues to outstrip capacity, something she said has been a big point of emphasis at SCL. To deal with this problem, SCL is using telehealth blended with AI to give the patient more time to think things through prior to the appointment, something that should help ensure the individual exhaustively characterizes his or her health problems in advance.
“When they’re in the physician’s office, they enjoy that one-to-one engagement, but some patients come in with some anxiety and some pressure,” Casterton said. Consequently, patients may lose track of some of the symptoms they’ve experienced. However, the combination of telehealth and AI can be used to help patients provide more exhaustive information about their health history than they might remember to divulge when face-to-face with an HCP. This approach might not only help the clinician avoid missing a diagnosis altogether, but also yield more accurate diagnoses.
Patients wary of, but not averse to technology
Casterton said patients have become more selective about the HCPs they engage with, often with an eye toward those who are familiar with the latest drugs, devices and approaches to clinical care. She said a number of patients have responded with some enthusiasm to the e-visit program at SCL Health, which gives the organization at least some competitive advantage. Most patients have little problem with the use of technology if they are assured that the doctors and nurses “are always doing their best to care for me,” she said. She noted that the flip side of this equation is that the patient will be put off if it appears that the technology has taken control of the interaction.
Casterton said health care systems might do their marketing efforts a favor by offering would-be patients “a portfolio of options to pick from for the one that fits them.” The key going forward is in part to make it so personable “that they don’t even know AI is part of it,” she said. However, she advised that this is an achievement that requires a fair amount of time and effort applied toward training the algorithm.
Lewis said the COVID-19 pandemic “is a real opportunity for AI to have a big impact on clinical care,” adding that UPMC is working on a module for a smart clinical visit summary that captures a lot of key data from the visit. This summary, which would be forwarded to the patient, will be more efficient for clinicians and gives the patient more assurance that the visit was worth their time. The summary would provide more than the standard advice about losing weight and is designed to give the patient a take-home that can be reviewed repeatedly – an aspect that is especially helpful for patients whose conditions require multiple interactions with HCPs.
Arora noted that radiology is still one of the primary areas of early application of AI, such as when an algorithm is applied to images prior to a radiologist’s examination. This early analysis could make the radiologist more efficient, especially when the current imaging study’s results are paired with any prior images of that patient. This kind of efficiency is becoming increasingly relevant to radiologists, given that the sheer volume of patient image files continues to grow over time.
Physicians are wary of possible losses of privacy and data security, but there are also concerns about whether an algorithm is unintentionally biased toward or against a particular diagnosis. Still, Arora urged developers to not be put off. “While we’re going to keep poking holes in new technologies and new ideas,” he said, adding that clinical skepticism is not Luddite hostility, but rather an interest in ensuring no harm comes to the patient.