AI’s ability to see and hear patients holds great promise
Artificial intelligence is rapidly spreading across healthcare, with applications large and small finding their way into workflows across the industry.
Whether it’s supporting clinicians during telemedicine visits, transcribing entire conversations between doctors and patients, writing notes for nurses in response to questions in the patient portal, helping patients assess their concerns via chatbots, or any of a host of other applications, AI is proving its worth for many healthcare stakeholders.
Narinder Singh has been working with AI for years. He is the CEO and co-founder of LookDeep Health, a virtual seating, nursing and care services company. Previous roles include working in Accenture’s Center for Strategy Technology, a corporate strategy function in the office of the CEO of SAP, co-founder of Appirio, president of Topcoder and vice president of engineering at webMethods.
Healthcare IT News spoke with Singh about how AI can help expand capacity in telemedicine, the risks of generative AI for hospitals and health systems, how healthcare providers can overcome these risks, and the role AI is playing in writing technologies.
Q. You note that telemedicine of course takes the burden of distance out of healthcare interactions. But you say it doesn’t increase its capacity. How do you think AI can help with that?
A. Let me start with some context for why this is an important question, and perhaps the question for the future of hospital care. Every week we speak to hospitals who are noticing either increased patient acuity or staff shortages, and most say both.
The US population over 65 grew five times faster than the total population from 2010 to 2020 – the fastest pace in more than a hundred years. This is part of a longer-term trend, highlighting the increasing age and associated acuity of patients that hospitals will treat in the future.
At the same time, we have repeatedly seen predictions of worrying to a disaster for nursing and other functions in the hospital – not to mention the financial pressures that make it almost impossible to expand the workforce.
We have now had a generation of telemedicine in the hospital – from eICU to teleconsultations and now virtual sitting and virtual nursing. On a project level, there have been many successes, but on a macro level, telemedicine in the hospital has collectively had a very limited impact on care, with one major exception – COVID.
During the pandemic, we learned that seamless access through telemedicine creates flexibility that allows a system to adapt. Yet, it did not expand our resource capacity. Telecapacity can span vast distances, but it does not change the underlying work units needed to deliver care.
Now, AI can mean many things, but let’s start with what relates to telemedicine: the ability to expand our observational capacity (rather than how it impacts decision-making). Today, a nurse treating six patients will be in a random patient’s room for one to two hours. Doctors will generally only be in an individual patient’s room for a few minutes each day.
Therefore, in the vast majority of cases, a patient is not in sight of a health care provider, despite the fact that so much of what happens to the patient can only be assessed and understood at the bedside.
Are they less active? Are they trying to get out of bed less? Is their breathing more labored? Did the alarm go off because the sensor slipped off their finger or the breathing tube slipped out of their neck?
One branch of AI, computer vision, could enable us to keep tabs on every patient at all times. This could help better allocate the hospital’s scarcest resource – the clinical attention of nurses and doctors.
We have decades of evidence that increasing clinical bandwidth has a positive impact on patients. Video alone – even in legitimately attractive areas like virtual nursing – will simply repeat the disappointments of the past. With AI, we can make better use of our most important limitation, time and expertise.
Imagine a world where AI acts as a guardian angel for patients and their caregivers, identifying potential problems and alerting healthcare professionals before a small problem becomes a big one. This isn’t just about efficiency; it’s about fundamentally changing the way we deliver care.
AI can provide that extra layer of support, so that no patient is left unattended, even for a moment. It’s not about replacing the human touch, but about augmenting it, making our healthcare system more responsive, more resilient, and ultimately more human.
Q. You warn that generative AI poses real risks to hospitals and healthcare systems. What are those?
A. Generative AI can streamline prior authorizations, patient coding, and the complex interactions between insurances and providers. But it can also spark an epic civil war between them.
This productivity could lead to a faster but more complex litigation landscape, ultimately requiring more human adjudicators to resolve disputes. Rather than reducing administrative work, it could actually increase it. Generative AI could infinitely scale the most cynical stereotypes of overuse and aggressive denial of claims.
AI tools are making progress in reducing the time doctors spend on paperwork, especially outside of the hospital. But in a hospital setting, the complexity of care and the lack of defined “visits” mean these tools are not yet as effective.
We’ve had years to learn how difficult and specific the development and application of machine learning algorithms in hospitals is. The allure of a magic bullet to remove that tedious hard work and integrate it into clinical workflows is tempting, but naive.
“Generative” patterns are relevant to many areas of health care, but they are not a golden ticket. They do not yet meet the need to synthesize defined sets of information and repeatedly draw the same conclusions. Predictability of inputs and outputs is crucial for evaluation and certainty in clinical decision making.
Q. How can hospitals and healthcare systems overcome these risks of generative AI?
A. On the first point, related to the battles between insurers and providers, I don’t see an immediate solution. You simply can’t afford to have humans trying to handle the volume of AI-generated requests or responses, so participation in this arms race is inevitable.
However, acting in a way that provides a foundation for evaluating and incorporating generative models into workflows provides leverage for the future. Key steps include securing PHI, ensuring checks and balances on outputs, evaluating models both in and out of scope, and not alienating your staff with premature claims to replace their roles for a few dollars an hour.
This is just the beginning.
We are already seeing insiders like Sequoia and Goldman questioning the hype and benefit of generative AI. We will go through a valley of despair; but by focusing on the pragmatic and not falling in love with the broad proclamation, many innovation teams will stay off the cutting board. Hospitals need two antagonistic mindsets.
First, experimentation is essential. Generating non-clinical content (emails, communications), evaluating EHR context summaries, improving language translation and transcription – these are all areas where generative AI can be safely tuned and targeted for improvement. These applications can free up valuable time for healthcare professionals to focus on more critical tasks.
Second, hospitals should enforce rigorous evaluations and demand repeatability. For clinical scenarios, expect evidence of claims of capability. Better yet, have an approach for continuous evaluation of AI capabilities within the solution. Concrete claims should ensure that the same set of inputs produces the same results, maintaining consistency and reliability in clinical decision-making.
In other industries, technologists, as Norman Vincent Peale once said, “aim for the moon and settle for a landing in the stars.” In health care, we’ve seen the disastrous consequences of such strategies, setting industries back a decade or more (Theranos for blood testing, Watson for AI for cancer).
You can be pragmatic without being slow – the right leaders provide that balance.
Q. You’ve seen more than half a dozen transcription companies raise over $30 million in recent years. Why is that? And what role does AI play in these scribe technologies?
A. There are over a million physicians in the United States. Their time is incredibly valuable, and a generation where they were treated as both experts and junior data analysts has led to massive burnout.
The math is simple and the technology is now more accessible than ever. The story that “the time is now” is not new, but it may finally become a reality. It is a beautiful use of technological advancement.
AI plays a crucial role in these scribe technologies by drastically improving the accuracy and efficiency of transcriptions. With AI, transcription can be done in real-time, with higher accuracy and at a fraction of the cost.
The challenge is that AI developments have continued their breakneck pace of progress in recent months – and are redefining the starting point for building such solutions. It is clear that transcription solutions themselves are not fundamental AI models; rather, they are solutions built on fundamental AI models.
The cost of developing competitive solutions has likely dropped by 95%. Better integration with clinical workflows, exceptional go-to-market models, and innovative offshoot solutions will continue to be huge differentiators. However, the quality of the difference between top solutions in the AI aspects of transcription itself will essentially become zero.
As a result, in this future, it’s only inertia that’s preventing prices from dropping dramatically, which should be great for healthcare providers. Lower costs make these advanced transcription solutions accessible to more practices, further reducing the administrative burden on physicians and allowing them to focus more on patient care.
The surge in investment in transcription companies is a testament to the transformative potential of AI in healthcare. The risk, however, is that the commoditization of the category results in desperate overpromises to meet investor expectations.
Follow Bill’s HIT reporting on LinkedIn: Bill Siwicki
Send him an email: bsiwicki@himss.org
Healthcare IT News is a publication of HIMSS Media.