AI Talk: AI and future of health care, part 3

Feb. 11, 2022 / By V. “Juggy” Jagannathan, PhD

The first blog post in this series explored the trendlines impacting the interoperability of health systems and what is happening with the pharma and drug industry. The second blog post focused on the trends impacting hospitalization and acute care provisioning. In this third and final blog post in the series examining artificial intelligence (AI) and future of health care, the focus is on affordable care, examining the role of virtual care and home care. This blog post also investigates what is being done to address health inequities.

Part 3. AI and the future of health care: Virtual care, home care and health inequities

Virtual care. Telehealth and virtual care have been around for a very long time – since the early 1960s. You can peek at telehealth’s timeline in this article. There was a slow and steady adoption of technology that allowed for televisits using audio and video conferencing. The pandemic, as we all know, upended this situation and the Centers for Medicare & Medicaid Services (CMS) jumped into the fray to provide emergency authorization to pay for televisits as normal outpatient visits in the spring of 2020. That overnight change altered existing telehealth dynamics and there was a spike in the adoption of telehealth.

This McKinsey and Company report records that the use of telehealth spiked in April 2020, but by February 2021 had stabilized to 38 times the pre-pandemic level. The McKinsey report also noted that telehealth use varied by specialty and one area that saw the most adoption was mental health care. Telehealth is here to stay and is going to be one of the core aspects of how care is going to be provided moving forward. Take a look at this article from the Healthcare Information Management Systems Society (HIMSS) proclaiming “Telehealth isn’t a fad.”

Home care. Taking care of a patient at home is not a new phenomenon. For millennia that was the norm. Take a look at this detailed history of home care dating back to 1500. Now, with new technology, home care is back in full swing. To get a glimpse of what technology is at play in a smart home, take a look at this article in The Hospitalist.

Essentially, they include monitoring sleep, detecting falls, robots to assist in all manner of ways, sensors to monitor body and environment. Of course, we are all aware of what is happening in the personal gadget space, with smart watches are popular, tracking our steps, heart rate, sleep patterns, etc., but every year around this time, it is instructive to look at what is happening with the annual Consumer Electronics Show (CES).

For the first time in its history, a health care company provided a keynote at CES. Robert Ford, chairman and CEO of Abbott Labs, gave an instructive keynote outlining what is happening in home care. The talk highlighted at-home diagnostic tests, fall impact detection and rapid COVID-19 testing at home in partnership with eMed. Ford also announced a new category of bio-wearables dubbed Lingo. Lingo will monitor glucose, ketone, lactate and alcohol levels– representing a significant push into the consumer health space by Abbott. Other CES presentations featured announcements ranging from nutrition monitoring and light therapy for pain management, to a weighing scale that doubles as a body scan.

Outside the world of CES, I found this interesting posting by Dr. Bertalan Mesko, who has a video podcast called “The Medical Futurist.” Among his top predictions for 2022: vocal biomarkers (for instance for detecting COVID-19 from voice) and chatbots assisting clinicians and patients with all manner of tasks. The wearable tech industry is exploding and the startup ecosystem is filled with companies targeting the virtual care and home health monitoring space. AI is at the core of intelligent remote monitoring, chatbots and sensor analysis efforts.

Health inequities. This World Economic Forum article features an impressive interactive graphic listing around 100 factors that relate to inequality in the world scene. There are many, many reasons for inequality for sure. Let’s look at a few factors that exacerbate inequities in health care and delivery here in the U.S.

• Wired broadband. Johns Hopkins did a report last year mapping the availability of broadband in and around Baltimore. That mapping clearly showed pockets with a lack of broadband coverage. If this is the case for an urban city such as Baltimore, imagine what the situation is like in rural parts of this country. Broadband connectivity is a critical infrastructure piece that is essential for health care delivery. Hopefully, the infrastructure bill recently passed with bipartisan support will eliminate this inequity.
• Social determinants of health (SDoH). SDoH refers to factors such as financial strain, food insecurity, housing instability, transportation barriers and health literacy. The pandemic highlighted health inequities that afflict certain segments of the population in a very disproportionate manner. It has been shown repeatedly that patients who have to contend with one or more SDoH factors have worse health outcomes. CMS has now made a concerted effort to capture SDoH factors as Z codes when patients interact with health systems to better manage this population. Several analytic companies now provide various solutions to help manage this population alongside state Medicaid agencies.
• Algorithmic bias. The last area that I am going to touch in this blog post series is health inequities exacerbated by bias in AI technology. We have looked at the problem of bias introduced in algorithmic models and how it leads to predictions that are inaccurate. Here is a study from Stanford that looked at geographic distribution of patient cohorts that was the basis for developing deep learning models. They analyzed 56 studies with patient populations concentrated in California and northeastern states. The vast majority of Americans were not represented at all. If data from these excluded regions are not part of a machine learning algorithm, how does one expect the algorithms to perform well for that population? Not very well.

In an interesting opinion piece, the Harvard School of Public Health suggests that developers of such models get their act together by ensuring proper data collection and using multi-disciplinary teams to oversee model development. They also suggest big consequences for those who don’t get their act together: make it easy for malpractice lawsuits to be filed if it can be shown the models are flawed. Another approach is to have formal regulation to ensure the models are properly evaluated before being put into practice.

We have explored a range of major trends impacting health care delivery in U.S. in this blog series. The entire industry is transforming before our eyes. AI technology is continually getting better and the adoption of AI in every aspect of health is accelerating. There are a few trouble spots related to algorithmic bias and health inequities – but they are showing signs of improving as we improve our infrastructure and technology adoption takes hold. I am quite optimistic about a consumer-driven health care model taking hold this year.

I am always looking for feedback and if you would like me to cover a story, please let me know! Leave me a comment below or ask a question on my blogger profile page.

V. “Juggy” Jagannathan, PhD, is Director of Research for 3M M*Modal and is an AI Evangelist with four decades of experience in AI and Computer Science research.