Matt Allen is a medical student at UC San Diego interested in clinical informatics and digital health. He is interested in improving the relationship between clinicians and computers.
Digital technologies have become critical to human health. As such, these technologies should not be considered only when other, “harder” science has already spoken, rather, they should be paramount from the onset. Digital technologies are shaping and changing the practice and progress of medicine, and technological advances are not passive participants. Technology often begins as a servant only to quickly become the master. Accordingly, the evidentiary standards and attention to development of digital health technologies needs to be as robust as that behind medicine itself.
Electronic Health Records (EHRs)
The EHR is perfect for the metaphor of a technology beginning as a passenger—giving helpful direction, passing snacks, and providing good conversation—only to quickly end up in the driver’s seat. EHRs were supposed to improve patient health outcomes while also making physicians’ lives easier. The results have been mixed. Many of the physicians I meet as a medical student feel chained to a computer that tells them what to do and keeps them from practicing medicine as they wish.
Typically, we do not need to encourage uptake of well-designed technology (picture people waiting in line for hours to pay hundreds of dollars for a new iPhone). It should tell us something that the US government had to pay billions of dollars to encourage physicians to adopt EHRs. I believe EHRs have not been welcomed, in part at least, because they do not merely facilitate the practice of medicine but actually influence how medicine is practiced. Additionally, physicians may feel more burnout as they lose control working in a less human and more technological environment.
Why do a group of highly educated, well-compensated professionals, many of whom are likely early adopters in other tech spaces, resist the EHR? It is probably because physicians rarely design this technology that changes their workflow and practice. Accordingly, the interests of insurers, administrators, and billing professionals are prioritized. When this happens, physicians often feel helpless and discouraged.
Many areas of medicine are currently set to be similarly disrupted by digital technologies in the coming years. However, if physicians pay attention, use their voices, and become directly involved in the development of those technologies, the results may become more health and physician-centered.
Randomized Controlled Trials (RCTs)
Let’s consider something near and dear to practitioners of evidence-based medicine: the randomized-controlled trial. Technological advancement is currently set to revolutionize this cornerstone of modern medicine in numerous ways. I briefly highlight two here:
In Decentralized Clinical Trials (DCTs), patient enrollment, instruction, monitoring, and even treatment can be carried out virtually. Such an approach lowers the barrier of entry and increases the number of potential trial participants.
Synthetic Data are artificially generated datasets designed to mimic real-world scenarios and patient data. This approach can improve data accessibility while maintaining patient privacy. Large quantities of synthesized data also enable machine learning or in silico clinical trials.
Such methods have the potential to drastically lower costs and shorten timelines associated with RCTs. Getting RCT-level insight faster and cheaper could revolutionize and increase the practice of evidence-based medicine. However, relevant experts must ensure that technological modifications to RCTs preserve or improve the scientific integrity of the approach. We must identify and correct for new biases or common pitfalls introduced by these new research methods. Technology will make RCTs easier to run and therefore more available to explore a wider range of questions, but we must be careful not to let technology corrupt the RCT.
Clinical Decision Support Systems (CDSSs)
The way in which physicians interact with evidence is increasingly being influenced by clinical decision support systems. Ideally, these programs provide physicians with timely, targeted clinical knowledge to improve their decision making. Accordingly, designers of such systems must determine what clinical knowledge will be disseminated as well as when and how this knowledge will be updated. In a world of exponential growth in medical knowledge, relying entirely on practicing clinicians to keep up is unrealistic. More and more, we are likely to rely on computers to examine data and decide what merits the clinician’s attention. We will need to educate Epic Employees and Machine Learning Engineers on how to evaluate evidence just as urgently as we do practicing clinicians.
In addition to ensuring the robust quality of evidence provided by clinical decision support systems, the use of these systems and other digital health interventions themselves needs to be evidence based. Medicine has well established ways of assessing new medications or surgical techniques. Doing the same with digital health interventions, however, is a new ballgame. The stakes are just as high if inappropriate or out-of-date guidelines get baked into CDSSs and lead to patient harm. Vetting and monitoring these systems will become even more complicated as they become dynamic, generated by artificial intelligence.
Thankfully, the digital nature of these tools often lends itself to quick and low-cost trials. Rapid Randomized Controlled Trials employ principles of A/B testing, distributed over multiple medical centers, to rigorously evaluate the effectiveness of clinical decision support systems. These creative, digital applications of the principles behind RCTs need to be applied to all areas of digital health to ensure that our technological advances are evidence based.
Physician Perspectives Are Needed
Developing technology that is helpful and user-friendly requires engaging end-users in the design process. When it comes to health technologies, that end-user is the physician. Physicians will be the users of clinical decision support systems and the prescribers of digital therapeutics. Physicians will be the interpreters and appliers of the findings of new age clinical trials.
Greater distance between designers and users means greater distance between what systems offer and what users want. In the health technology space, the voices of payors, regulators, or industry are often louder than the voices of physicians. This is apparent in the design of modern EHRs. If physicians do not contribute, this will also be the case with the next generation of digital health technologies currently being developed.
To illustrate, would you rather have a clinical decision support tool that uses Natural Language Processing to analyze your chart and ensure compliance with requirements set by insurers, or would you prefer a tool that identifies a recent rise in your next patient’s blood pressure and automatically adds 10 minutes onto their next appointment to discuss management?
Would you want DCTs to be used to bring profitable “me-too” drugs to market faster, or would you want DCTs to enable increased clinical trial participation from historically marginalized communities?
Clinical informatics is the medical subspecialty that deals with these issues and there are already several medical journals dedicated to the field. However, health technology cannot merely be a subspecialty. It needs to be like cell theory—foundational for the practice of all physicians. This is because advances in health technology are unique in that they often change physician workflow. While a new medication option for diabetes most likely won’t drastically change physicians’ day-to-day practice, advances in health technology likely will.
Unfortunately, there seems to be a trend that the more time one spends taking care of patients, the less likely they are to be invested in the progress of digital health. Often, the physicians that do take interest are the doctorpreneurs. Having some physicians take care of patients while others focus on technology is the wrong solution. Individuals who understand clinical downstream effects are essential. Only practicing physicians understand the context in which CDSSs and other digital health interventions will be applied. Another pop up might seem like a great idea to an engineer but not to the doctor clicking through 20 of them to order Tylenol. Only practicing physicians can answer questions like “will the information provided by this algorithm be clinically relevant or useful?” or “will clinicians actually use this system in the way it is being proposed?”
Technological progress cannot be resisted or ignored. Further, technological progress can lead to better outcomes for patients as well as more productive, sustainable careers for physicians. This outcome, however, is not assured. Physicians need to determine their own technological fate. Doing so will halt the trend toward computer dominance over physician lives and instead allow physicians to partner with computers that augment their abilities while improving the physician experience.
As a long-time devotee of substantive documentation in the medical record, written and electronic, I am interested in Matt's perspective on the single greatest failing of the EHR in today's modern medicine - the copy and paste sloth. Until insurers agree to omit payment for progress notes that include >90% copy and paste from previous note, the immense vacuity of today's medical record will persist. We are doing our learners a terrible disservice by accepting the current state of practice.
Glad to see a real look. The EHR has been designed for others. In order to develop a better system, voice must dominate design and use but various text tools can add to that. As you note software developers must integrate with the primary user, behind the scenes automation can extract for others. Spiral development should be the approach, We can see EPIC is more than tedious for both practitioners and users.
No experience with diagnostic aid tools, but it is true that knowledge grows exponentially so that keeping up is quite hard for busy practitioners. As before such tools need to embed designers with users.