Every health system leader has unprecedented executive management challenges facing their organization in the wake of the pandemic. The Baldrige Foundation and ABOUT Healthcare welcome you to LeaderDialogue Radio, where leaders glean valuable insights and practical takeaways to help navigate effectively through these challenging times. The show airs on the 1st and 3rd Tuesdays of every month at 1:00 pm (ET) on Business RadioX.
Apurv Gupta MD, MPH, an expert on the topic of empathetic automation and driving care transformation, will be joining Leader Dialogue co-hosts Dr. Charles (Chuck) Peck, Dr. Darin Vercillo, and Ben Sawyer on the April 5th, 2022 podcast. The podcast topic is Achieving Systemness: Empathetic Automation Opportunities and will be exploring how empathetic automation can help to address clinical workforce overload and shortages. Dr. Gupta shares his perspective and insights on this important topic, providing practical suggestions that can be applied immediately.
—
Listen to the podcast here
Achieving Systemness: Empathetic Automation Opportunities With Apurv Gupta MD, MPH
As always, it’s great to be here with some friends of mine. Ben Sawyer is here with me and we’ll be talking with a special guest, Dr. Apurv Gupta. Ben, as you probably know, is one of the executives at ABOUT healthcare and it’s great to be here with everybody. I’m particularly excited about our guest because Dr. Apurv Gupta is a colleague of mine at Guidehouse. We’ve worked together for a number of years and he’s here to continue to talk about some things that we’ve been talking about for the last several episodes or so. That topic is around achieving systemness.
In this episode, what we’re going to focus on is something that it’s a term that Apurv has coined. The term is empathetic automation. I’m going to introduce him first and give you a sample of what this is going to be about. Dr. Gupta completed his internal medicine training at Beth Israel Deaconess Medical Center in Boston and received an MD and a Bachelor of Science from Brown University and an MPH from Harvard University.
He leads care transformation at Guidehouse Consulting as an expert in driving transformational change, physician engagement, change management and leadership development. He’s led multiple projects in clinical operating model design, clinical variation, service line optimization, etc. He has a wide experience as a clinician manager, executive educator and thought leader.
He also co-hosts his own podcast called Speaking Of, Making Healthcare Work for You. I would encourage you to tune into that podcast. He’s got some great guests and talks about a variety of difficult issues that everybody’s facing in healthcare. Apurv, it’s great to have you here. Let me start with a couple of comments that revolve around technology, particularly the electronic medical record. Those of us who have done both sides of healthcare, one side being the provider of it in the trenches and the other side being the administrator of it in a variety of leadership roles. I think we would all agree that the EMR is a good, bad and ugly story.
Most clinicians would agree that it had a tremendous benefit for patients in terms of quality, safety and, in some cases, producing efficiencies that weren’t there prior to its inception. On the other hand, it got its bad or ugly side for clinicians. It means that it’s taken away the opportunity to have that face-to-face interaction that used to occur. I remember when I used to take care of patients. It was all about me looking at the patient and the patient looking at me.
I visited my PCP and it was all about me sitting on the exam table and him facing the computer during the visit, not to mention the additional burden of hours that it puts on nurses and physicians potentially at the end of the day in terms of completing a lot of their work that they didn’t have time to do when they were entering it into the medical record.
This idea that Dr. Gupta is going to speak about is the notion of empathetic automation. In other words, how to make automation work for clinical folks. Also, to help with the syndrome that’s occurred now in healthcare, which is serious with people leaving their jobs, burnout, etc. I think this is a timely topic. I’m going to be quiet and welcome Apurv. It’s great to have you. I’ll start with a simple question. Could you explain to the audience what you mean by empathetic automation and maybe give a couple of salient examples of how you think empathetic automation can help.
Thank you so much, Chuck. Honestly, it’s such an honor and a pleasure to be here with you. It’s doubly sweet because you’re my boss. It’s been an amazing opportunity for me to come here and have an opportunity to talk to you about this and to this amazing audience. Thanks to Ben for making this happen as well. I think you captured it so well, Chuck, that when we’re talking about systemness and what we’re struggling with in the post-pandemic era, try to figure out how to make our organizations work.
You would think that electronic medical records would be a part of the solution and unfortunately, they’ve only been a small part of it in that they’ve also created this distancing between clinicians and between clinicians and administrators. Also, as the clinicians and the patients. As you were describing, where we go in to see our physicians, unfortunately, we’re often sitting with their back to us as they’re entering in data.
Information is digital. It should be easy to find, but it isn't. Click To TweetThat’s where the genesis of empathetic automation came from. It was the notion that this is an era in 2022. We have amazing technology at our disposal, not only in healthcare but throughout many different industries. Some of our healthcare organizations are starting to leverage those technologies, but they’re doing so in the back office arenas like revenue cycle or supply chain. They have not yet thought about how to bring those technologies to bear on the frontend, which is where, as you were alluding to, our clinicians are struggling.
The electronic medical record, even though it has brought some value in terms of improving communication and enhancing patient quality and safety, has created a documentation burden. It has become complicated for people to manage electronic medical records. Unfortunately, they spend a lot of time hunting for information in the electronic medical record. You would think that information is digital. It should be easy to find, but unfortunately, it isn’t. You have to go to multiple places to find the information.
Clinicians are often still chasing after each other. The doctor asks looking for the nurse to find out what happened with the patient, the nurse walks into the patient and the patient is saying, “Nurse, what’s going on with me?” The nurse has to say, “I don’t know. Did your doctor tell you what’s going on with you?” These are some of the symptoms of a dysfunctional system that are largely engendered because our electronic medical record, which should be able to be a better hub of communication and information is unfortunately not able to function that way.
We came up with the idea of empathetic automation because we wanted to put the focus of healthcare executives on the front-line clinical activities that are going on to get them out of necessarily looking at the back room. Also, helping them to think, there are clinical workflows, complex as they may be, that can also be automated. If we do so, we can help improve the lives of these clinicians who are overwhelmed right now with the amount of paperwork, documentation and communication that they have to undertake.
How do you distinguish automation from artificial intelligence? There’s this immediate, “Wait a minute. Stop. We can’t trust a computer to make decisions and that computer’s not going to think better than a physician or a nurse.” How do you distinguish between those two?
That’s a critical question, Chuck, for us to address because so many of our team members and so many of our health system executives are also grappling with that. The risk is that all of these wind up becoming buzzwords. They’re cool, fancy and gimmicky and people throw these terms around without necessarily a lot of clarification. People may be tempted to think that automation is the same as artificial intelligence, natural language processing, or machine learning.
While all of these technologies are on somewhat of a continuum, the way we distinguish them is we’re referring to automation as simply automating the mundane, repetitive tasks that a clinician has to undertake. These don’t require clinical intelligence, which is why they’re overwhelming the clinicians.
They’re creating a sense of burnout because it’s taking them away from what they want to be doing, which is taking care of patients. When we speak with clinicians, as you do, and others that are on the show would surely recognize us. They estimate that anywhere from 25% to 33% of their workday is spent on tasks that are nonclinical value-add. It’s the same as chasing after information, hunting for the electronic medical record or waiting for information to flow through to them.
When they can be automated, those tasks are more of what we call the dumb bot level of the solution. Whereas, on the artificial intelligence side, you’re talking about using the data that’s coming in, analyzing the data, maybe using machine learning protocols and trying to predict what needs to happen. Trying to emulate the mind of the clinician and trying to make some sense of the trends of the data.
That is more on the intelligence side. What we’re trying to do is simply say, “Clinicians are busy doing very repetitive, rote, mundane tasks that they don’t need to be doing. In fact, no one needs to be doing it, a software program to carry out the same algorithm and do it minute after minute and keep repeating that algorithm without having to think.
It’s an if/then statement or go get this data and then go plug in the data here. Pull the data from multiple places and present it to the clinician in one view. Go take the data and match it up to a particular table that tells us what to do and gives us an answer. These are some of the things that we can do on more of the so-called non-intelligent side of the spectrum.
I described my visit to my PCP for the first time. As you described, he basically and this is not his fault. The way the rooms were set up, which I’m sure they are in most places, his back was to me about 90% of the time. He was asking me the typical sort of questions. I had filled out my forms in advance with my history and etc. Could you pick a diagnosis and maybe give a specific example so people understand what you’re talking about and how this might work.
Yes, absolutely. I’m happy to, and hopefully, that will make it very illustrative. One of the examples that we’ve been speaking with our clients about is congestive heart failure. This is a condition where the heart is not pumping forward as effectively as it could be. As a result, the fluid backs up into the lungs, maybe into the abdomen, as well as the legs. It can cause difficulties with breathing and it can cause difficulties with swelling.
Unfortunately, it’s a chronic progressive illness that, if it’s not well-taken care of, can lead to frequent hospitalizations and emergency room visits. Multiple medications that a patient needs to be on and unfortunately, every medication has complications. This tends to be one of the more common clinical conditions we see in outpatient and emergency room settings, as well as in hospital settings. This is problematic for our patients because it can be very progressive.
In this kind of setting, you need a lot of coordination of care and that’s exactly what automation can be helpful with. You can imagine your physician, the primary care physician, sitting with you in the office if you had congestive heart failure, which we’re postulating here. You don’t. It would have to then go into the medical record, “Why is he sitting with his back to you?” It’s because he needs to pull out information from multiple different places to understand what’s going on.
He needs to understand what medications you were on. When were you last in the hospital or in the emergency room? What were your recent labs? What was your weight? We call it the dry weight, which is what weight you’re at when you’re not fluid overloaded compared to your current weight, so he can understand whether you’re well-managed. All of these elements reside in different places in the electronic medical record.
That’s a simple function of pulling all of these elements together into one view. We call it a clinical dashboard. This can happen in the outpatient setting while he’s trying to help you understand whether you’ve been watching your weights, your diet and taking your fluid pills, so you don’t get into overload. The same thing happens in a hospital setting when you have an exacerbation of congestive heart failure and you have difficulty breathing. You get admitted to the hospital. Now, the clinicians have to manage the same issues, but in a more acute way.
That’s the first use case. It’s compiling information from different parts of the medical record and making it more readily available to the doctor and to the nurse so that they don’t have to go hunting for that information. They don’t have to make a decision based on incomplete information. Unfortunately, what winds up happening when doctors and nurses are busy is they will often spend so much time searching for what they need.
Clinicians are really busy doing very repetitive, rote, mundane tasks that they don't actually need to be doing. Click To TweetThey then give up the search at a certain point because it’s an opportunity cost of having to move on and take on the next task. That’s where we think automation can step in. Every time that doctor is seeing someone with CHF, they would automatically get a compilation view of the congestive heart failure to begin with. That bot can then go and do a simple lookup of a New York Heart Association classification for heart failure or American College of Cardiology classification and determine the patient’s classification because then that determines what medications the patient could be on that maybe they’re not on.
Again, sometimes doctors don’t have time to even do that simple lookup. First, to assess what classification this patient is and then to think, “Based on this classification, is there a medication they should be on that I don’t have them on?” It’s another use case for a bot. They can go look up the classification table and figure out that this patient should be on a lipid-lowering agent, but they’re not. Go and look up the patient’s medications. See if that medication is missing from the list and give a simple prompt back to the clinician, “Dr. Smith, did you know that this patient is a CHF Class II and they may not be on a lipid-lowering agent.
No intelligence is involved here. It’s a fairly complex activity that we build-out. First, compiling all of the necessary information and presenting it in one view and then going and doing a lookup, comparing it to what the patient already owns and then presenting it back to the doctor. They’re not predicting anything. They’re not substituting the judgment of the clinician.
All they’re doing is simply making the task easier for the clinician in synthesizing the information so that the clinician can do what they need to do, which is make the clinical judgment of whether the patient indeed should be on that right medication. There’s a lot more we can say about that, Chuck, because there are probably three or four additional aspects of that CHF use case but let me pause here, check in with you and make sure that this is answering the question you had in mind.
Ben, I know you probably have a couple of questions, so please go ahead.
I do and thanks for that, Apurv. That was insightful. For the readers, what I’m wondering is if you could then explain when there is this application of artificial intelligence to help support the provider and patient interface and what you described, how does that then translate into the physician? In Chuck’s case, facing him and having a discourse instead of being locked to the screen where they have to fill out fields. That’s one question.
The second question is related to an interaction I had with a good friend of mine who is an internist nephrologist. He was down visiting for a few days and he’s the chairman of the board of a health system. He said that one of the challenges that he’s seen with this as a provider is that the insight that used to happen in medical records with the consulting physician wasn’t there because it’s so rote. Everybody is filling in the same fields.
Even though there’s a lot of information there, the pearls, the insight that he’s looking for sometimes is not there. The third question is, as it relates to the next steps for patients, particularly in care transitions, if they’re going from a physician’s office to an acute care facility or into another level of care, how from an empathetic technology standpoint can those logistical triggers essentially be best managed? Is that too much stuff to address or do you have the three questions?
Let’s go at it one at a time and if I forget something, then I’ll ask you to remind me. I love the questions because, at least for the first two, unfortunately, it comes back to a function of time. How does this help impact the physician? Ultimately, they’re taking a twelve-hour workday and unfortunately, they are that long or 10 to 12-hour work days and they’re spending 3 to 4 of those hours looking for information, chasing after the information, trying to document the information or waiting for information.
What do they need to do in order to do that? That means multiple times going into the chart trying to figure out if something has resulted. Maybe having to call radiology or the lab to say, “What’s going on? Why are my tests not back yet,” especially in the inpatient setting. I’ve got a patient who I’m trying to discharge. “Where’s my echocardiogram? When is that going to be done? Why did my patient get bumped?”
All of this expediting is going on behind the scenes and it’s necessary to work to some extent, but it’s not real clinical work because you can imagine it’s not intelligence function that’s being invoked here. It’s a lot of grunt work that’s involved in calling up people, following up with them and making sure that something is getting done. We think we can automate those things that will help relieve the clinician, the physician and the nurse from having to do a lot of that chasing themselves.
Again, we’re not talking about a panacea here. We know that we’re never going to replace completely all of the follow-ups that do need to be undertaken. We’ll still be going on, but we’re hoping that we can make a dent in it to an extent. It’s like if you can think of that congestive heart failure patient that I was using the example of and a particular doctor seeing a patient with congestive heart failure, maybe with chronic obstructive pulmonary disease, another with diabetes and then with stroke. A lot of very common chronic conditions, half of my patient panel might be made up of patients like this.
If I can have a bot helping me organize the materials, present them to me in a succinct fashion, helping me do some simple lookups or helping to follow up, maybe with radiology to say, “Is the echo back yet or send me a ping when the echo is back, so I know when to go back into the record to look for it. Better yet, pull that report over to my note so that I know that it’s what the documentation is and it’s available for me to review.”
Now, you can imagine that this is going to save 5 minutes per patient or maybe 10 minutes per patient. As you add it up over the course of the day, we’re hoping that this can save anywhere from 1 to 2 hours per clinician and that’s where how this stuff works. An individual bot may only be able to save you seconds to minutes, but as those activities pile up, it can add up to a significant impact for a busy doctor and a busy nurse.
When you free them up, that’s the real answer to the questions that you were posing. That gives them more time to be patient-facing, literally and figuratively. That’s what they want to be doing. They want to be interacting with patients. Doctors and nurses, God bless them, they love patient care. That’s what brought them into medicine and nursing.
Unfortunately, we have made their lives so administratively burdensome that they’re not able to spend enough time doing it. The secret for us with automation is no secret. It’s like, “How do we buy back the time and have someone else, a dumb bot program doing that work and serving it back up to them so they can have maybe that extra five minutes per patient to focus on having a more meaningful encounter?
Ben, I hope that answers at least one of your questions. Do you have another thought on that?
Before we go to that second one, let’s hang on with the first one for a minute. Is there a habit change that needs to take place with providers? I had the same experience as Chuck when I had my physical with my primary care physician. It was the exact description. It feels like they’ve gotten used to it. You face the computer, filling in the fields where you’re talking to someone.
Automation opportunities are returns on investment because when you think about clinician burnout and overload, it makes a significant impact. Click To TweetYour point is well taken, Apurv, that if the time is given back, it gives them the opportunity to move away from the computer because the dumb bots in the background are compiling the information and it should give them the opportunity then to have a much more interpersonal interaction with their patients.
I guess the logical question for me is, “Is that going to require a bit of a habit change given the ingrained approach now that we see across care settings where they’re so fixated on making sure they capture everything that they’re not moving their eyes off the screen much to be able to interact with the patients.
It’s another fantastic follow-up question. Thank you. I’m smiling. Although our readers can’t see us, it’s such a crux of the question because I have two answers to that. On the one hand, we got ingrained in this in about ten years. It’s sad. I think it’s a sign of the times. It’s what we’ve done to ourselves and what we’ve allowed the system to do to us because we thought we were fixing the system.
It’s how we thought we needed to go and as a result, we’ve created this very burdensome cumbersome system that’s now in between us and our patients. Chuck elaborated on that in the preamble. One response to your question is that they will probably be some habit change, but we’ve done it to ourselves and in a fairly short period of time, I’m hoping that that’s something we can unlearn as well, just as easily when the systems evolved. This is where systems need to go over the next decade. That’s one thought process.
The other thought process is why I keep emphasizing the dumbness of the bot, not to antagonize the bots unless they turned on us. What I’m emphasizing there specifically is that we are trying to focus on automating those aspects of the workflow that should not require any major process change for the positions because that’s another reason why they become overwhelmed is, “Okay, doc. We think we’ve got a better way for you to do something. Instead of doing steps A, B and C, you’re not going to do steps 1, 2 and 3.
This is going to be great. It’s going to save you so much time, but you’re going to have to learn something completely new. What we’re trying to focus on and emphasize with automation is to take out work activities for them. Whey may not even realize it, 5 minutes or 2 minutes at a time. We’re hoping that this is invisible to the physician and to the nurse and it’s a bit of a lighter day for them. It’s not asking them to necessarily do something different. It’s helping them again and that’s why empathy comes in.
As system leaders, we are empathetic to their plight in making their workday a little bit lighter, enough so that they can then take a deeper breath. Maybe, grab lunch, go for a walk and spend that extra few minutes with the patient, hopefully, in a way that they want to, rather than us thinking, “We’re going to have to train them differently to do something different.”
For the readers, on the other two questions, I doubt we’ll have the time to address those now, but one of them is, as the time comes back to providers, perhaps they can then use that time and the interaction with each other as consulting positions to provide that keen insight that we may have missed.
That last question, which is probably the topic of another episode, is about the triggers for transitions of care and how those can be automated so that the physician or provider isn’t having to do a lot of work to figure that out. Instead, those logistical triggers are handled in the background to support the provider and the patient or consumer to get to the next best level of care and in an efficient way for them. I appreciate your response on that. I think that was super insightful.
Apurv, just quickly because I know people are thinking this in the audience. I assume that you’re also trying to develop an ROI because there’s probably an ROI associated with this work. Finally, I assume that because these bots are dumb and agnostic, they can be placed into any system that somebody might already have in place in terms of the electronic medical record. I know those are very basic questions, but I think that’s important for people to know very quickly.
We talk about it as automation, but a lot of this has to happen within the electronic medical record. Some of the most advanced electronic medical records we work with our clients have a lot of capabilities to do this optimization or automation. That’s the place we always start is within the electronic medical record, how can we make that workflow work more smoothly. That’s one quick answer.
To your first point about return on investment, hopefully, our readers will be convinced by the end of this that the return on investment is significant when you think about clinician burnout, nursing burnout and overload. It’s the experience of care, reducing turnover and better engaging the workforce because when the doctor or nurse is better engaged, then the patients will be better engaged and you’ll have better outcomes.
This leads to better patient safety and quality. Because you’re not spending time chasing after information, the right information is coming to you. You can make decisions in a better-informed way. Those are all very valuable aspects, but then we would be remiss if we didn’t count that there are actual hard dollars attached here. When you save the physicians and nurses time, they’ll be able to take a deeper breath, but they may also be able to see more patients doing more of the things they want to be doing and there’ll be less turnover. Those are the ways in which we’re thinking about ROI. We don’t want to be financially literate. We also want to be sensitive to the broader view of healthcare that we’re trying to engender.
Thank you, Apurv. This was a great conversation. I think we’re going to want to follow up at some point and Ben, I don’t know if you have any last words for folks in the audience before we sign off and talk to them again.
I would thank Apurv again for coming on. He provided some real insight and things that providers, in particular, can relate to. Your comment, “In some ways, we’ve done this to ourselves.” There’s this opportunity to figure out a path. Now that we have a baseline of electronic medical records, how can we optimize that for a better experience for all patients and consumers? That’s a great way to end it and look at next-step discussions that can start to tackle that even further.
Everybody, thanks so much for your time and thanks to our audience. We’ll see you next time.
Important Links
About Apurv Gupta MD, MPH
Dr. Gupta completed his Internal Medicine training at Beth Israel Deaconess Medical Center in Boston, MA, and received an M.D. and Sc.B. from Brown University, as well as a M.P.H. fromHarvard University. He leads care transformation at Guidehouse Consulting, and is an expert in driving transformational change, physician engagement, change management, and leadership development. He has led multiple projects in clinical operating model redesign, length of stay and throughput, clinical variation, and service line optimization. He has experience as a clinician, manager, executive, educator, and thought leader.
