LAS VEGAS – At HIMSS21 earlier this month, Dr. Robert Bart, who has served as the chief medical information officer at Pittsburgh-based UPMC since 2017, spoke with Healthcare IT News about his to-do list during this busy past year and a half, and what he’s eyeing in terms of clinical priorities as he looks toward the next year and beyond.
Among the topics we discussed: the value of voice recognition, the path toward effective decision support, building a bigger comfort level for machine learning and the future of value-based contracting.
Q. We just wrapped up a series at HITN focused on CIOs and CMIOs nationwide, and the lessons they’ve learned over the past 18 months of the pandemic. What have been some of yours?
A. I don’t know if there’s something called post-pandemic. I’m not personally convinced yet. But I think probably the thing that I’ve learned the most in the past 18 months is just the adaptability of people. People still needed to receive care, and clinicians felt it was their duty to deliver care. We were one of those organizations that flipped over to telemedicine pretty seamlessly.
Q. How much of an increase was that over what you had in place before? There were some pretty eye-popping percentages in terms of scale-up.
A. We probably had about 3,000 people who were authenticated to deliver and use our telemedicine platform. During the last two weeks of March in 2020 … and the first two weeks of April, we went up to about 16,000. We were doing about 250 daily consumer-facing telemedicine visits pre-pandemic.
“One way healthcare can become much more comfortable with machine learning and AI is if we use more tools in the clinical operations area: driving efficiency in scheduling [and] throughput of patients and using smart learning algorithms that help us manage the operations of healthcare. As we get more comfortable with artificial intelligence in that space, I think the comfort within the clinical realm will follow.”
Dr. Robert Bart, UPMC
By about April 1, we’re doing about 10,000 a day. And across our system, we peaked at somewhere in the 15,000 range, plus or minus a little bit. We did ended up doing over 1.25 million telemedicine visit for ambulatory care last year.
Q. That’s a pretty substantial change.
A. We’re not doing the same volume anymore, not knowing when the end of the pandemic is going to come, or whether there truly will be a post-pandemic. What I’m seeing is there’s a fair number of our patient population that really prefer, ‘If things are calm enough right now that I can see my clinician face to face, I’m going to go do that.’
And so we’re still tracking at about, I would say, 15% our peak of telemedicine visits, which is still significantly higher than our pre-pandemic based baseline. But it certainly is not near what we were doing twelve months ago, 15 months ago.
Q. But it’s still going to be a piece of the pie, going forward?
A. Oh, yeah, I think so. The pandemic was just a forcing function, right? Some people need care. You’ve got to do it. This is how we’re going to do it.
But what I don’t want to become is: This clinician’s preference is to do telemedicine; this patient’s preference is to do visits via telemedicine. But we do need to factor in a little bit, I think, on the patient-preference side.
Certainly the high-level, consistent use of it we’re seeing is in the behavioral health space. And that’s just because in rural Pennsylvania, there’s not a lot of behavioral health specialists out in those communities.
The fact that they can get therapy and interaction with a behavioral health specialist via telemedicine – especially if part of your challenge is you have anxiety – being able to deliver it to you while you’re sitting in your home on your couch, and didn’t have to fret about driving and parking, it’s a really good thing.
What we’re trying to do to give sort of telemedicine, some staying-power, or legs, coming out of this is trying to work with each of the medical specialties and surgical specialties on how do we best place telemedicine within the care-delivery models that work well for the diseases they manage. It becomes part of the integrated care-delivery process.
So if a Type 1 diabetic typically is seen by their endocrinologist once every three months, can every other visit, at a minimum, maybe be done via telemedicine? So you’re seen face-to-face twice a year instead of four times a year. Or maybe even three of the visits are telemedicine, and you’re seen face-to-face once a year if you’re otherwise doing well.
That’s, I think, what we want to do is bake it into the cadence of the type of clinical specialty you are, the type of patients you manage over a long term period, and figure out what that opportune mix of face-to-face and virtual care is.
There’s a little bit of a higher bar to go over, because no one sets standards on, like, what’s the appropriate cadence of face-to-face? You can get 10 endocrinologists or 10 cardiologists who have the same patient in front of them, and some of them will say, ‘I just need to see this patient once a year.’ Some will say twice a year. Some will say three or four times. It varies quite a bit by practice.
What we’re being asked to do with telemedicine – and this is, I think, a lot by the payers, who are trying to also understand how that might change access to care – is to make sure that we are using it in a manner where we’re providing value to the patient, but we’re not also doing it just to drive some of the cost of care.
So I think there’s a way to do it where you can actually derive improved health for the patient, while not driving significant costs in the system. And that’s really what you want to do. I’d like to get beyond, like, managing the disease – but really to the point of using virtual care to help push someone towards a healthier existence, a healthier life, pushing them towards something where we’re managing wellness.
Q. Is UPMC also getting more deeply involved in remote patient monitoring and hospital at home?
A. We’ve been a long-term utilizer of remote patient monitoring. We’ve got a pretty broad practice. We focus a lot on patients with congestive heart failure, patients with inflammatory bowel disease.
Because, especially with IBD, there’s a medical component, but there’s also, because it’s a fairly heavy lifestyle burden, there’s a psychological component. And so that remote patient monitoring and that frequent contact, I think, really helps the behavioral health aspect of it.
And then you mentioned hospital at home. We are moving in that direction. Hospital at home, I think, you’re probably going to see people moving into it depending on how the payers in each region of the country support it, quite frankly.
We’ve had significant discussions. We’ve developed a prototype program that starts doing some of the hospital at home types of care, so we can start moving into that space and really learn about it as we move closer and closer to a true hospital at home, at least as CMS outlines it.
The other area that I think we’re quite interested in, because we do own some post-acute facilities at UPMC, is the SNF at home – sort of the post-acute care type of experience that can be done at home, different from just the traditional home health nurse.
It does the check-ins, but is much more involved, much more supportive of all the things that occur. And so that’s one of the things we’re interested in really exploring.
Q. Obviously, the pandemic has also really put a spotlight on social determinants of health.
A. UPMC is very interested in social determinants of health. I personally have a strong interest in it: My role immediately prior to joining UPMC four years ago, I was the CMIO at the Los Angeles County Department of Health.
So as the second-largest urban government health care system in the country, you can imagine that social determinants of health is a big component of the care-delivery process.
In Pennsylvania, where we are at UPMC, there’s still a fair amount of fee-for-service care delivery. And social determinants of health are important and impactful, even in the fee-for-service models of care delivery, but not nearly as important when you start moving into the value-based contracting.
In value-based contracting, where you’re getting per member, per month fee for some cohort of individuals, you’re very much incentivized to understand and play an active or interventional role within the social determinants of health for those individuals.
We’ve examined quite a few different companies in that space, as well as try to really improve the types of information that we’re gathering on our patients during ambulatory visits, so that we can understand how to make care decisions more impactful to who they are and what their living circumstance and experience is.
If you’re not mindful of those things when you’re trying to help them make the best decisions on care, I think that we can make poor recommendations for that individual. There is a right recommendation, maybe from a clinical/medical perspective, but they’re not necessarily that executable or easy for that individual to follow through, given the social circumstances. And we need to actually factor that piece in. And we’re trying to do that better with the information that we’re harvesting from the electronic health records.
Q. Let’s talk about EHRs a little bit. How can they do better?
A. We want to get to the point where we’re starting to use more AI-driven workflow. The two major vendors in the EHR space I’m a client of, and so we’re actually going through projects with both of them in modernizing and optimizing their footprints.
Both of those platforms have been at UPMC for more than 20 years, and, you know, we’re not just freshening it up like a new coat of paint. This is removing walls, remodeling, all very involved on both ambulatory and acute care platforms.
And doing it, I think, with one of the keys in mind: As we’re doing this, can we also improve or decrease physician stress and improve physician wellness?
One of the things that we did when we started on this program about a year before the pandemic started, in the spring of 2019, I asked the physician who oversees our physician wellness at UPMC. I asked her to sit on our committees for both of our projects, so we would be thinking about what we need to do from an informatics perspective – but making sure that we’re getting feedback from her about how that might impact the physician workflow.
I think the other thing that we recognize and realize is voice is a big saver here. It’s sort of this huge equalizer. It’s a time saver. And because of natural language processing, the leverage that you can get from the discrete data within the dictated or voice recorded note may allow you to fill out or complete many other work-related required things for billing or regulatory requirements that can be harvested from the note, as opposed to relying on the physician doing a bunch of different tasks.
So we’re very heavily pushing on voice is one of the things as we modernize our platforms. We’ve used it consistently over the past few years, but we’re now taking it to a whole other level, leveraging some of the AI that our voice partner has embedded in their solution.
Q. I’ve written a lot over the years about various precision medicine initiatives at UPMC. What’s new on that front?
A. One of the reasons I came to HIMSS this year was to do a presentation. I work very closely with a PharmD by the name of Phil Empey, and we did a presentation on genomics, and really drilled in on pharmacogenomics. I’m really excited about pharmacogenomics and the ability to improve care by getting the drug-gene pairs right.
So we ingest the data in a discrete manner from the result, so we can utilize clinical decision support. Because one of the challenges is it’s such a new field that most of our physicians don’t really understand the interpretation of the results.
On top of that, it’s such a new field that what was the appropriate interpretation of a result, last year, with new knowledge might be different this year.
We really think that, when you’re moving into the world of pharmacogenomics or genomic medicine, that you really need to embed decision support into your electronic health record. And that you have to really insist on taking the results only in digital format.
So if we get external results from reference labs, we don’t want PDFs. We want to actually discrete data, so we can trigger the decision support, as well as provide supporting content for interpretation by our clinicians – and the content so the patient can understand what that result means for them.
One of the reasons I’m so high on pharmacogenomics is that there can be a big benefit on medication adherence. There can be a big benefit in the cost of medications, making sure that someone’s on the appropriate medication for whatever is being treated. And the payers are very interested in pharmacogenomics, and have been very supportive of the reimbursement of this testing.
So there’s a nice opportunity where the payers are aligned, because they feel there’s financial benefit in healthcare. The clinical side is aligned, because they feel that they can get better therapy and therapeutic treatment for the patient. And the patients are aligned because they want the right medication at the right dose for them personally. So I think it’s a really nice opportunity.
Q. It’s exciting. Do you think it’s going to continue to kind of be the province of big academic medical centers like UPMC, or do you think eventually it will filter down to smaller hospitals – and maybe eventually become the standard of care?
A. Definitely pharmacogenomics will filter down and become the standard of care. With more of the whole-genome genomics, I’m not sure how much that will get down to the standard care as we think about it.
There’s two types of healthcare systems: There will be those that are consumers of the information, which might be the bulk of the healthcare system, standalone hospitals, those types of things, and then there’ll be healthcare systems that are involved in producing genomic results.
UPMC right now is currently a hybrid. We do consume some external results, but we also have our own genome center and produce our own genomic results. We want to get more and more to being a producer, and less and less of being a consumer. But, particularly in the pharmacogenomics space, there’s opportunity for all levels of healthcare systems to be involved in that.
Even for certain aspects of genomic sequencing, there’ll be some of that – particularly, say, if any hospital has an oncology service. That’s certainly going to have tumor genomics as something that’s being offered, and studied quite aggressively.
I think we did see a sort of stagnancy with genomics through the first 18 months of the pandemic. But I think people are rekindling their interest, and we’re seeing it sort of get more and more focused, even over the course of the summer here of 2021.
Q. Any closing thoughts?
A. There’s two areas that I’m particularly interested in. We talk a lot about machine learning and AI in healthcare. But, as you probably know, the acceptance of use of it in the truly clinical diagnostic space, is very narrow.
One way healthcare can become much more comfortable with machine learning and AI is if we use more tools in the clinical operations area: driving efficiency in scheduling [and] throughput of patients and using smart learning algorithms that help us manage the operations of healthcare. As we get more comfortable with artificial intelligence in that space, I think the comfort within the clinical realm will follow.
The other area that I’m quite interested in is robotic process automation. A lot of other industries have been using it for quite a few years and have gotten a lot of benefit in areas where there’s heavy, redundant human labor.
By automating it, they can get rid of so much of the redundant tasks. Healthcare hasn’t done that as much. And I do think robotic-process automation is an area where I’m really excited to see much more of a stronghold, or grow some legs and desire to use it – not just in the care delivery process, but maybe a lot of the back office processes within healthcare systems.