Best of: A more thoughtful approach to technology can improve medical care
Russ Altman: Hey, everyone. It's your host, Russ Altman from The Future of Everything here. Today, we're rerunning a fascinating conversation I had with Sara Singer, a Stanford professor of medicine and an expert on integrated health care.
Anyone who's had to navigate the health care system knows it's extremely complex and care can often feel disjointed and inefficient. In this episode, Sara highlights new technologies that could improve integration within the healthcare system, ultimately enhancing a practitioner's ability to care for their patients. I hope you'll take another listen and enjoy. Before we get started, please remember to follow the show to ensure that you get alerted to all of our new episodes and never miss an episode on the future of anything. Healthcare is a complicated business.
It involves many different professionals including physicians, nurses, pharmacists, physical therapists, social workers, care managers, hospital business units, and many others. The complexity of healthcare also varies widely across patients. Some patients may only have a single medical problem that is managed in a straightforward way. Others may have several interacting problems, heart disease, hypertension, diabetes and a bad knee.
Leading to relationships with several clinics, a long list of medications, a full calendar of appointments, and lots of opportunities for inefficiencies and problems. Now imagine that we want to change health care by introducing a new technology to help any of those different professionals or situations. The technology may be powerful, but how can the technology developer figure out who is the best user of this new capability and how should it be integrated into their job so that it minimizes disruption and provides the best value to improve their productivity and outcomes, and the outcomes for all the patients.
Professor Sara Singer is a professor of medicine and organizational behavior at Stanford University. She studies health care management and policy. In particular, how organizational leadership and culture impacts the implementation of health innovation. She has a special interest in understanding both provider and patient perspectives on their experience personally and within the organization in order to understand how to improve health care more rapidly.
One focus of her work is integrated health care. Sara, can you tell me, at least in the context of health care, what is integrated care and why is it so important to understand and study? Sara Singer: Uh, so the idea of integrated care I think is best, um, understood in the context of people's everyday experience of health care, especially when you're somebody who's sick. We developed the concept when thinking specifically about patients who have complex situations.
Those patients have to go see, often have to go see lots of different doctors in lots of different kinds of organizations. And very often, those doctors and organizations, they don't talk to each other. And so the experience of care is highly fragmented. And that's really the opposite of what we're trying to achieve.
And so it's that opposite that is integrated care. And the way we've defined it is to suggest that what should be integrated is not only the information about the patient and the treatments that they receive and across the different providers and settings where they're getting it. But also what should be integrated is what the patients value and what they believe and what they're able to do themselves. All of that should be integrated as part of the patient's experience.
Russ Altman: So that's really interesting. And so it sounds like there's integration from the side of the system to make sure that the system is integrated. But you also are adding this very important component of integrated in the patient's experience and the patient's perspective. So how would, how are we doing in America or in the United States or globally, if you prefer to answer it that way, at creating integrated healthcare? Or are there places that are doing it better and other places that are perhaps still working on it? Sara Singer: It's a pretty universal challenge and I think that there aren't a lot of places that claim to do it well. I think that there are some systems of care that do it better.
My own research suggests that the extent to which people are receiving their care within an integrated kind of a system. You know, within the VA, within Kaiser Permanente or other systems that are vertically integrated so that they have their primary care doctor is connected contractually with the specialists and with the hospitals. In those settings, care is better integrated, but I think it's a pretty universal challenge and nobody claims to do it perfectly well.
Russ Altman: I think I have to ask, because I know you have an appointment at the business school at Stanford as well. Does integrated care automatically mean more expensive because of all this coordination or not? Sara Singer: Oh, you know, you know, ironically, we think that it probably means the opposite, right? So if people are getting the right care at the right time from the right providers, then we are hoping that they stay healthier and they aren't going to need the kind of, um, additional care that you get when information isn't communicated. Or, you know, result tests don't come back to the people who need to see them in order to act upon them.
So this, uh, utopian vision of cost and quality going hand in hand is very much connected to, um, the aims of integrating care. Russ Altman: We have a lot to talk about, but I have to ask, especially since you mentioned complex healthcare with lots of interacting parts. We're in the middle of a pandemic still. How has the pandemic been for integrated care or for your general area of expertise? What have you seen and has the system been reassuring and good or has it exposed cracks that maybe you knew about but other people didn't know about? Or maybe even that you didn't know about? Sara Singer: Both, right? So we did a survey, uh, because we had done a previous survey. We went back to the providers who we, um, had surveyed previously to ask them how they responded in the context of COVID.
We asked them about the safety precautions they took. We asked them if they, for their patients with multiple complex conditions, if they reached out to them. To make sure that they were okay with regard to taking precautions around COVID, but also were they okay with respect to getting the care that they needed while they were afraid of coming into the healthcare setting. And what we found was not so many organizations had the infrastructure that enabled them to do it.
And so that exposed some of the challenges around integrated care. At the same time, I think the big news around what happened when we faced the COVID pandemic is that everybody shifted to tele medicine and tele health. And that opened up additional possibilities for more integrated care and doing integrated care in better ways. It also exposed opportunities for doing it worse. Um, but the idea that you could dial in and see somebody's home and get a sense of what some of the challenges might be for that patient so that you could better assess what kinds of support that patient needed. That was a benefit.
The ability to connect with someone in a way that was easy for them to access so they weren't having to come in, enabled for more continuity of service. So those were a variety of benefits around integrated care and to the extent that we continue to use those advantages and fix some of the disadvantages, I think we might come out on the other end with some benefits around integrated care. Russ Altman: That's good to hear. Speaking of COVID and not only the patient's experience changed, but healthcare workers were in many cases having to turn their job upside down. And there are other forces leading to healthcare jobs being turned upside down these days, like the introduction of AI, different payment plans.
And I know that you have new projects in what you're calling the future of work, specifically for healthcare. So can you tell me about those projects and what are they? What are the questions that you're asking and what are you trying to learn from them? Sara Singer: There are a couple of different projects. Maybe we should talk about the one that's been published first. You know, there, uh, together with colleagues from, uh, from MIT and the institutions where we did the work.
We were interested because it was an opportunity to understand how the technology developers were working with the end users, the people who were actually using the tools that the technologists were producing, to understand how they could develop a product together, you know, that would be a better product. We were interested in that, uh, in doing that research because it was, it's still pretty rare to have the opportunity to do research where the technology isn't completely separate from the people who are using it. Russ Altman: Right.
Sara Singer: So we wanted to understand what that dynamic was like. And you're right, part of the vision for machine learning is that we're going to be able to contribute tools that enhance the ability of workers to do their work and to offload some of the work that they're doing. But you know, so far that doesn't seem to be the case. In research that we do, we hear over and over again from the clinicians that, you know, technology tends to add work without necessarily offloading the work. So they're very eager to work with technology developers to find ways, uh, that the technology could really add value for this.
Russ Altman: It's actually quite amazing that what I heard you say is that it's not always the case that new technologies are developed hand in hand with those who will use them. And that's kind of shocking because anybody who's been in a health care environment, it's almost like a ballet, especially when it's working well. The workflows are very well defined and people know their alleys, so to speak. And so the disruption that can come, no matter how great a tool is, if it hasn't been specifically designed to accommodate the workflow patterns, uh, or to retrain the healthcare workers, it's going to be a disaster. So tell me a little bit more about what are the types of applications where people are actually being foresightful enough to say, hey, we better make sure that what we build is actually like, even close to useful and compatible with the clinical environment.
Sara Singer: Yeah. And I should say, you know, that's sort of 101 at the business school too, right? You got, uh, implementation matters. But in the world of, you know, AI and machine learning, there's so much technological challenge to that work that a lot of it has gone on in a, kind of retrospective or a simulation modeling kind of an environment. And outside, you know, they're working to advance the technology. And I don't do that work, so I don't know, I know enough about it to be dangerous, you know, through, through this work. But I value and appreciate what they do, I just know that it would be so much more enhanced if they did it hand in glove with the end users.
And that's what our first study enabled us to explore and understand. Russ Altman: Do you drag the technologists into this kicking and screaming, or are they happy to kind of have the upfront added cost, which will hopefully lead to downstream reduced cost of introducing whatever the technology is? Sara Singer: The technology developers that I've worked with have been enthusiastic, which is not to say that they have the capacity for it. You know, you have people with very different mindsets and very different, um, very different objectives. Uh, you know, the clinicians are delivering great patient care and the technologists are developing really cool technology.
That's not necessarily the same thing. Um, having said that, the technologists in the organization in this particular study were interested. Not to say that they didn't encounter frustration along the way, they certainly did. Uh, but in the end, I think they saw that, uh, they produced a much better product as a result.
Russ Altman: So I'm interested because I know from reviewing your previous work that one of the areas you focused on is getting consumer kind of suggestions about how healthcare could be redesigned, consumer ideas about how their experience was perhaps not as optimal. And so I'm wondering as you look at AI and machine learning. We've talked about the technologists and the providers. Do you imagine that this work could also extend to getting the patients or the consumers perspective on where AI is welcome? And where maybe they're not so anxious to see AI? Or at least having a conversation about the pros and cons there? Sara Singer: Certainly.
I mean, in the same way that I think integrated care, um, should include the patient perspective because the patient's the one that's experiencing all of the care that they're receiving to the extent that, um, AI or machine learning is creating technology that is going to, that is being built on patient data. Or that is being used in the context of patient care. I think that patient's perspective could certainly add value. Russ Altman: I noticed one particular study that you did perhaps a while ago in this area was consumer suggestions for improving mental health care.
And that's an area that during the COVID pandemic, mental health has by all measures, and for very understandable reasons, as people have been locked up in houses, not able to interact and socialize. And I'm wondering, can you tell me a little bit about that work? Because that's one of the most sensitive areas. And so I think it's a great early look at how people might think about AI and machine learning.
So I'm wondering, what did people learn from asking? And did it change the direction of mental health care in some cases because of the consumer or patient ideas about what they wanted that they were not getting? I just found that to be really fascinating. Sara Singer: So, I'm trying to recall the particular paper, uh, honestly that you're talking about. A lot of my research has gone to the patients and asked them about their experience.
Russ Altman: That's the issue. Yes. Sara Singer: And it has completely changed what we have been studying and the approach that we're taking. The most recent one that I'm thinking of, and it's not, I apologize, it's not in the context of mental health care.
And maybe you can remind me a little bit more about it. But it is in the context of very complex kids who needed spinal fusion surgery. And we worked with a patient and family on the team that was working to kind of think about how that process ought to be designed to optimize the patient's perioperative period. Russ Altman: Right, right.
Sara Singer: And that includes kind of both before the patient goes in for surgery, during surgery, and then the follow up to surgery. And having her present, um, as we had these conversations, changed almost everything that we did in the context of that work. And in fact, it, you know, ultimately shifted the, our attention to being almost entirely, um, focused on the way that the clinicians and the patients should be interacting over the course of that entire period in order to make the care, Russ Altman: Yes. Sara Singer: Um, optimized. Russ Altman: Thats exactly what I was wondering.
And so what I would suggest just thinking about is that that's gonna be a very important dynamic also when they're introducing AI type technologies, is that you could imagine that this might change the whole pattern of interaction between the clinicians and the patients. Just out of curiosity, were those patients like very small kids who mainly you were interacting with parents? Or were the kids also able to pipe up with their preferences? Sara Singer: These were super sick kids who, you know, had other challenges who certainly could communicate their interests. We didn't have them on the team, but they were definitely, , their needs and preferences were definitely represented through their parents. Russ Altman: This is The Future of Everything with Russ Altman. More with Sara Singer, next.
Welcome back to The Future of Everything. I'm Russ Altman and I continue my conversation with Professor Sara Singer of Stanford University about how to best introduce new technologies into healthcare. It turns out that sometimes, technologists may have a good technology, but they are focusing on the wrong user. And sometimes even the right user needs the technology to change in unexpected ways, in order for it to really add value.
Who are you imagining are the end users? Is it the providers or the patients or the technicians or the nurses? I just want to make sure we understand that as we think about these interactions. Sara Singer: Yeah. So it's a great question and I'm glad you asked. Um, one answer is it can be, I think it can be any of the above. The other way I will answer that question is that it needs to be someone specific. You know, as you're developing a technology, it's really helpful to know who you're developing it for.
And then, as I said before, to be developing it with that, uh, with that end user, the clinician. If in the context of the research that, that I had done previously, we had been working with care managers primarily as our end user. But ironically, one of the, uh, one of the things that we learned is that, um, that the right end user for the technology that was developing was actually a different care manager. So it's part, so I'd love to move to talking about the, um, the kind of big takeaway from that study. Russ Altman: Yes.
Sara Singer: I'll just keep going. Russ Altman: Please do. Yes. We love big takeaways. Sara Singer: It was that the technology comes out best if the technology is built in concert with the group that you intend to have to use it. And the way that it gets built is through an iterative back and forth.
It's not just a, you sit down with them and you have a conversation. It needs to be an iterative back and forth because there's a process that the technology goes through in its development where you're learning. Russ Altman: So it's not just one focus group. Sara Singer: Right. It's not one focus group.
It's a, it's an iterative process. So in that first instance, what we learned is, you know, the problem came out of the emergency room. And the emergency care manager said, you know, we could really use a tool that would allow us to understand when the surges are coming. Technologists tried to start developing that and then we learned that the, um, that surges actually happen for pretty random reasons like exogenous shocks that you can't predict.
Russ Altman: Right. Sara Singer: And so it wasn't super, the technology couldn't be good enough to do what the ED care managers needed. Russ Altman: Right. Sara Singer: So we asked the question like, we've got this interesting beginning of a technology, is there a place where it could be used? And we actually found that it could be very helpful for predicting in the ICU. So we created in that instance something called a low bed tool. And the, um, and it was the, um, ICU manager, who became the end user that we worked with to continue to iterate over time.
Russ Altman: And the tool was a good match to the needs of the ICU manager. Whereas it wasn't a good match to the elect, the emergency room manager. Sara Singer: Well, so that's it. So it was a better match. But it still wasn't a good match.
It was only a good match over time through this continued iteration that went on. Russ Altman: Yes. Sara Singer: So there are two reasons why you need to continue iterating these people. One is because the end users understand the data that the technology people are using.
And, you know, with machine learning, you've got to take all, like lots of different sources of data, you need lots of it. And then the technologists try to, you know, try to apply algorithms to make sense of it. But you need the end user to help you understand the differences in the technology. In the case of, um, the bed, you know, predicting the beds in the ICU, some of the data had used different identifiers. Um, they, uh, had different measures of when beds were available, you know, beds will be available or beds are, you know, beds are, uh, you know, the bed is empty and it needs to be cleaned versus it's empty and you can show up. Russ Altman: And those differences all matter.
Sara Singer: They matter to make the, to improve the accuracy of the technology. And so, you know, you need to continue that, that iterative process. Um, the second tool that we developed was one that, uh, focused on predicting readmissions and particularly who needed additional help, uh, after discharge to make sure that they weren't going to show up again in, uh, in the hospital. Russ Altman: I know that readmissions are considered a very bad thing because it means that the hospital didn't really get the patient in a good condition and then they're coming back having decompensated. And so it's not good for anybody.
Sara Singer: It's not good for anybody and finally, policy makers have made it not good for the hospitals either. And you could imagine a situation where a patient would come back and a hospital would say, okay, we get more money because the patient is back. Well, the policy makers have, uh, added legislation to penalize hospitals when people come back. So now everybody is aligned to trying to make sure that people don't come back.
And so this readmission risk tool, you know, is further along in its technology development and we're working there with care managers who are, you know, working with patients, uh, who are, uh, who are going to be discharged. And one of the things at that point we learned is that, you know, the technology was doing its best to predict. But it, the people who they were predicting would come back, weren't marrying up with what the care managers and the other clinicians kind of knew to be true about the patients that they would expect to come back. And so what they realized is that the technologists didn't have access to certain data that was only available in medical records in like the notes section. Russ Altman: Which is harder to analyze than the kind of structured stuff. Sara Singer: Yeah.
And so at that point, what they decided to do was to say, we're going to do the best we have with all the data we can put into our models and use the algorithms to predict that. And alongside of it, we're going to have a, uh, a free text note where you see the additional information. And on the basis of that, our managers are going to be able to predict.
So what's interesting there is that it wasn't necessary for the machine learning algorithm to be perfect. But it had to be created in a way that the managers understood. It couldn't be just this black box thing. Russ Altman: Right. Sara Singer: It had to be done hand in hand with the people who are using it in a way that would enable them to use it to come to the right answers.
They just wanted the right answer. Russ Altman: Right, right. No, that's really a great hybrid solution. 'Cause I thought you were going to tell me a story about very complicated text analysis, blah, blah, blah. But instead, simply making those notes available to the user with the little AI output, the two things together kind of solved or went a long way towards solving the problem. Sara Singer: The algorithms are going to get better and better.
But they're still limited. You know, in healthcare, we're limited by the, you know, data we collect and we don't collect everything that we need. And sometimes a lot of the important stuff is written in free text. You know, the patient lives alone. Well you worry more about those patients.
The patient doesn't have a car. You worry more about those patients. And when that's the case, you know, just a simple flag, um, would tell the manager, oh, I got to worry about this person anyway.
And, uh, it gives the manager a lot more confidence in the machine learning driven, uh, output, because they just know what's in there and what's not in there. And when they do, you know, they're happy to use it. Russ Altman: I love this description that you've given about this iterative approach, because even in this story, you took us from the ED, then you took us to the ICU, then you took us to a tool that wasn't quite right. And obviously there needed to be a million discussions or many discussions in order to finally, kind of settle on a tool that was truly useful.
And this sounds like a very general and useful lesson for technologists everywhere who want to have their technology used. I myself work in a field where we've been thinking about, it's genetics and I won't go into the details. But we were always thinking that doctors would be the recipient of our, of the knowledge we were creating.
And very late, it became very obvious that pharmacists were a much better group to target the information at. And so it was a similar story. But it sounds like you have general purpose kind of advice for technologists in healthcare for sure, but maybe more generally.
Sara Singer: Yeah. So I would love to land with that. I mean, one bit of advice is tech's going to be better if it's built in cooperation with the people who are going to use the technology. And then the second one, that I think is really important, is to keep in mind that technology should be aimed at adding value for the people who, um, who are going to be using it. Particularly in the context of healthcare and particularly in the context of, uh, healthcare environments where the clinicians are doing their best to keep us healthy.
We've got COVID going on now. We have a crisis of healthcare personnel. And when we introduce a new technology and we, and it results in, they're having to do new work, additional work, it's so painful to see these people experience, um, experience that. They love new technology and they want to use new technology, but they need it to add value. They need it to enhance their ability to care for patients.
So we've seen technology where, um, uh, where it can be really useful in that case. Allowing patients uh, allowing clinicians to kind of to do their work. Uh while, you know for example video monitoring is watching their patients and calling out loud if a patient is trying to get out of bed and at risk of a fall. Uh, and we've seen it where they're doing stuff anyway and if you're just trying to layer something on top that is calling their attention to something that they're already doing in order to, you know, document it better, something like that, they really, take offense. And at this time, we need the technology developers to be aware that, you know, it's really not okay.
And if they can take the lesson about finding better strategies for finding those opportunities to do something that is more useful for patients through the cooperation, I think we achieve kind of both of those goals. Russ Altman: Thanks to Sara Singer. That was the future of integrated health care.
Thanks for tuning into this episode. With over 250 episodes in our archive, you have instant access to a huge array of discussions on the future of pretty much everything. If you're enjoying the show, a reminder to please consider sharing it with your friends and colleagues. Personal recommendations are the best way for us to grow the show.
You can connect with me on X or Twitter @RBAltman, and you can connect with Stanford engineering @StanfordENG.
2024-08-16 16:19
