In medicine, we have routinely accepted the growth of our knowledge base, diagnostic capabilities, and therapeutic alternatives as measures of progress. But there is growing evidence that we need more help managing this “progress” to make sure that each patient receives the best care for their unique setting—the idealized, but as yet unavailable, “optimal personalized medicine.”
Few would argue that we have a growing gap between the knowledge and information that could be brought to bear in a patient’s care. What fraction of this is actually considered at the moment of truth and then what is actually done? We need help solving this “know-do gap,” as the enormity and complexity of healthcare data and healthcare delivery processes have far exceeded our aggregate ability to keep up:
- We have more than 141,000 unique diagnostic codes in ICD-10.
- Each year, more than 500,000 new publications are added to our medical knowledge base in PubMed. And while we all admit that we cannot keep up with that, even if we just focused on approved/published guidelines, we would still need to keep up with 6,400 separate guidelines, with most being updated every few years.
- We have more than 21,000 semantic branded drugs in RxNorm, each with as many as a several dozen potential adverse side-effects, and, just to add to the complexity, they are most often given in largely untested combinations.
- And it is estimated that there are upwards of 50,000 unique medical devices approved for use in various medical and surgical conditions.
It is beyond optimistic to imagine that this year’s (admittedly gifted) medical students will be able to store, update, analyze, and retrieve all of this information (including the nearly 27 million records currently stored in PubMed) exactly when needed, and then tailor the composite understanding to the unique clinical and socioeconomic setting for each patient. We simply have not evolved quickly enough as a species to fully exploit the progress in our accumulation of relevant medical knowledge.
The promise of assistants
But there is hope in the form of technology which we have not been shy about using to help manage this deluge of data. “Personal digital assistants” or PDAs, have been around for almost two decades (remember the PalmPilot?). In their humble beginnings, they served to help busy clinicians by handy storage and fact reminders, providing ready access to lists, protocols, and some useful calculations.
More recently, we have watched a more connected version of ‘virtual assistants’ come of age and explode on the consumer market: Microsoft’s Cortana, Apple’s Siri, Alphabet’s Google Assistant, Samsung’s Bixby—all featuring voice recognition integrated with internet searching and variable capability to monitor and control other connected devices in the growing Internet of Things (IoT). The value of these virtual assistants in healthcare is being rapidly explored.
Healthcare is quickly entering the era of “cognitive assistants” connected to vast collections of multi-modal information, powered by artificial intelligence and machine learning to provide real value in healthcare delivery. These modern electronic agents can provide ready access to relevant literature for any medical condition or circumstance as they provide well-referenced and assisted interpretation of available lab results, medical images, and historical details to guide next steps in a patient’s care. IBM Watson has garnered substantial mind sharing in this area, even if the hopeful announcement, “Welcome to the Cognitive Era of Health,” leaves some wondering if all prior efforts were somehow free of cognition.
Going beyond cognitive
And some assistants are already moving beyond the “thinking” to the “doing.” At Reflexion Health, we have developed a Virtual Exercise Rehabilitation Assistant (VERA™) that employs an engaging avatar to help patients understand their prescribed exercise regimen. It, then, goes to the next step of carefully measuring their adherence to the instructed movement while comparing the specifics of their movement to learned constraints, which then provides personalized, real-time form-related feedback to help ensure patients get the most from their at-home rehabilitation experience. VERA helps to extend the wisdom and judgment of skilled clinicians to patients—independent of location—enabling patients to recover in the safety and comfort of their own home, while saving time, steps, and money for all involved.
Surgical robots represent another more familiar example of healthcare assistants that go beyond the cognitive to the physical. At present, surgical robots are adjunctive aids to skilled surgeons, but smart, autonomous robotic assistants are being developed and tested, with results that, in some settings, appear superior to gifted surgeons.
And, a word on the singularity
Much has been written (often with some sense of dread) about a coming “singularity,” when our artificially intelligent, massively connected and relentlessly learning systems outstrip human ability to reason and create. In healthcare, there is an enormous upside to an integrated, coordinated, and more automatic healthcare infrastructure that helps deliver on the unfulfilled promise of the investments already made and the learnings already had; not to mention the accelerated pace of learning that such an infrastructure will provide. We owe it to ourselves, our profession, and our patients to fill the “know-do gap,” learn as much as we can from every patient’s experience, and then systematically apply those learnings so that we can all do an incrementally better job with the very next patient seen.
It’s abundantly clear that we need help to get this done, and while not rushing to “welcome our new computer overlords” as Ken Jennings famously said when losing Jeopardy!® to IBM Watson, we should not shrink from developing and exploiting the full suite of smart and learning tools necessary to move us forward.