In many ways, we’ve all grown accustomed to lightning-fast exponential leaps in capability. Since the turn of the century, we’ve been fed a daily diet of technological wonders that “disrupt” existing systems and promise a new and better world in previously unimaginable ways.
There are apps for everything and smarter smart phones and virtual assistants and bots and drones. It’s a brave new world everywhere. Industry from A to Z is evolving rapidly — and that includes the world of medicine and healthcare.
It seems like one minute we’re using the web to map the human genome, and the next we’re modifying patient DNA to treat devastating diseases, like sickle-cell disease, beta thalassemia, multiple myeloma, and sarcoma.
Awe-inspiring new capabilities are truly evident everywhere in modern medicine. But it is equally evident that the real-world day-to-day impact of innovation is often grossly hyperbolized. For example, amazing CRISPR technology isn’t much use to the average physician. And even workaday EMR/EHR technology hasn’t simplified medical practice for anyone I know, and there’s plenty of evidence such “helpful” tools actually drain physician time and energy.
In the healthcare realm, there’s often an enormous disconnect between those creating the new technology (inventors, vendors, and the developer community) and the intended end users of that technology (physicians, healthcare professionals, and patients). There’s a desperate need for more thoughtful consideration of user design best practices that’s rarely being met.
There exists incredible opportunity to leverage user-centric design in healthcare innovation for broader impact. We need to better utilize some of our new awe-inspiring technological tools among the enormous numbers of physicians and patients who aren’t being well-served by the digital revolution.
But there is light on the horizon.
Consider chronic disease, which currently accounts “for nearly 75 percent of aggregate healthcare spending” in the United States. While our technology has been advancing, so has incidence of diabetes and prediabetes, which now impact a third of the country’s adult population. Refocusing advances in healthcare technology to better address this particular challenge presents opportunity to:
1. Improve diagnosis
2. Prevent converting prediabetes to diabetes
3. Improve population health
There’s a lot of promise in AI-enhanced diagnostics tools for radiology and imaging and as they enter the practice of medicine. With regard to diabetes, the first autonomous AI-enabled FDA authorized diagnostic system is a retinal screening tool called IDx-DR, which is “able to correctly identify the presence of more than mild diabetic retinopathy 87.4 percent of the time and was able to correctly identify those patients who did not have more than mild diabetic retinopathy 89.5 percent of the time.” Those are great numbers, but the example showcases valuable “democratized” diagnostic models on other fronts as well.
The IDx-DR device carries its own malpractice insurance, which is a legally prudent and optimal approach for bringing such new technologies to market since liability and risk for malfunction of the tool are absorbed by the developer. This helps reduce practitioner apprehension and adoption friction.
Further, the technology basically allows for embedding highly specialized diagnostic capabilities into every primary care clinic in the country. It simply is not feasible to put a human specialty physician into every practice, so this capability touches on the issue of scale. To turn the tide on population-level challenges such as diabetes, scalable systems are important. The opportunity is not in focusing only on introducing clinically validated high-tech diagnostics to large medical centers across the country, but in myriad small settings as well. The aim is to broadly increase skill and capacity to get to scale.
A note of caution is required. I’m excited about systems for amplifying human intelligence, not in schemes for replacing it. No system, no matter how powerful, can substitute for the human specialist. These diagnostic tools should simply aid in using specialty resources more effectively. As we forge ahead integrating AI-enabled diagnostic tools into medical practice, we must do so with tightly focused purpose and great caution. There needs to be explicit discussion about risks in AI systems on many levels, and certifying that the companies building such tools adhere to prescribed ethical principles is not enough for qualification. Each and every product, regardless of who developed it, must be stringently reviewed and must adhere to governing principles.
High-tech diagnostics are powerful tools that will augment human intelligence, but they are tools wielded by and for humans.
Over 33% of U.S. adults are estimated to have prediabetes. Using technology for better chronic care management and improving patient engagement may prevent those cases from ever escalating to diabetes.
The way we currently manage prediabetes is reactive, not proactive. It is not patient centered, but that can be remedied with new and more thoughtful approaches gleaned from user-centered design. For example, the customer experience technology that in now familiar in consumer interactions with brands like Amazon or Netflix could be mirrored for working toward more successful patient involvement and better health outcomes.
Moreover, a quarter of U.S. adults have more than one chronic condition. The tools we use to treat them should reflect that reality. Technology that solves a third of the problem in isolation will miss the mark. Additive, interoperable, and interchangeable should all be words to describe our developing healthcare tools. Say you had a business serving five clients. If you had to have a different email system for each of their accounts, it would drive you to distraction. But there are myriad tools like Gmail or Outlook that you can use to pull in multiple email sources and aggregate them for efficient and manageable communication. You rarely see that approach in healthcare, and it’s holding us back.
Existing capabilities enable us to better provide the right care at the right time. The opportunity in healthcare is to apply that technology to support people outside the hospital, practice, or treatment facility. User-centered patient engagement tools and systems will be incredibly powerful in seizing that opportunity.
3. Population health
Modern data science and machine learning technology presents an enormous opportunity for optimized utilization of resources at the population level. But there’s also a huge discrepancy in that plenty of people think an Excel spreadsheet is an AI tool! The core challenge is using incredibly sophisticated and evolving systems to solve foundational operational problems to, for example, better predict who is at risk for developing prediabetes further up-stream in disease management efforts epidemiologically.
There are questions today’s technology can help us answer: How to manage limited resources. How to develop and use population health management systems to create timely clinical intervention. How to empower healthcare teams to make better in-the-moment decisions. How to more effectively sequence work or workflows in population identification, predictive modeling, and precision treatment approaches.
The possibilities here are rightly very interesting at the moment, and moving forward, we must promote development of clinically validated high-quality designs and systems to enable implementation and adoption.
There are several barriers to fulfilling all this promise, and they should not be overlooked.
For example, emerging AI tools often present real challenges concerning transparency and explication. It is difficult to earn consideration for technology that is not well explained to the physician community; this is also a barrier to adoption.
Physicians and practices are not going to adopt technology they do not trust nor understand. Commercial and government payors will not adopt technology without evidence of clinical effectiveness, safety, and value. Evidence-based documentation is required to support payment and coverage. Technology developers do not tend to think about that. New tools and technology are not reimbursable if they haven’t gone through these processes. There are also considerations that must be made for common data models and foundational liability concerning the deployment of new tools.
But in a world where we’ve grown accustomed to all those lightning-fast exponential leaps in capability, we’re more than ready to meet these challenges — and seize upon opportunity for new healthcare tools of the trade.