What Can We Do About the Hardest Patients? (Replay)

To kick off the new year, we’re revisiting one of our favorite episodes from 2022. It’s about a program designed to help coordinate care for the most complicated patients — and a study that rigorously tested its efficacy, with surprising results.

Before we jump into the episode, I wanted to thank you for spending the last year with me. I’ve learned a lot from some really interesting people, including many of you who’ve emailed me about the show. I hope you have a wonderful year, connecting with friends and family, and enjoying the things you love.

And on that note, I hope one of the things you’ll enjoy is today’s episode, which originally came out last summer. It’s called: “What Can We Do About the Hardest Patients?”

*      *      *

As a young family physician living and working in Camden, N.J., in the early 2000s, Dr. Jeff Brenner started to notice something about his patients.

BRENNER: We were in the most dangerous city in the country. And I had people come into my office and I would take staples out. I’d take stitches out. And they had been beaten. They’d been stabbed. They’d been shot. And I’d ask them, “This is horrible. Did you report your crime to the police?” And they’d sort of laugh at me and say, ” I’d never call the police here.” Honestly at the time I couldn’t get my head around that and I realized despite the fact that statistically, we were the most dangerous city in the country that the true crime rate was probably much, much higher than we realized.

Jeff had an idea: perhaps emergency room and hospital data could paint a more accurate picture of crime in Camden, N.J. He asked for, and received, claims data from local hospitals and health centers.

BRENNER: We had no idea what we’re doing. We stuck it in Microsoft Access. We mapped it, graphed it, charted it, and the data was eye popping. 

Among other findings, the data suggested that between 2002 and 2007, 50 percent of the city’s residents used an emergency room or hospital in a single year, which was twice the national rate at the time. The leading utilizer averaged 113 visits annually. In his quest to better understand public safety using data, Jeff had discovered something else about hospital use.

BRENNER: It started to tell a story that was horrible, which is that a small sliver of patients were going back over and over for all kinds of reasons.

This small sliver of patients goes by different names. Sometimes they’re called complex patients and sometimes, “super-utilizers.” But they share things in common. They typically have multiple chronic medical conditions, often mental-health problems, as well as housing insecurity, addiction, and other challenges.

BRENNER: They felt so disrespected and they still felt so unwell and they were going back and forth all the time to the hospital. But they weren’t getting care. They were getting treatment, but they weren’t getting care.

Jeff’s data had found these complex patients. Could he also help them?

*      *      *

From the Freakonomics Radio Network, this is Freakonomics, M.D. I’m Bapu Jena. Today on the show, we’ll hear about Dr. Jeffrey Brenner’s groundbreaking approach to caring for complex patients:

BRENNER: We spent about five, six years building what I think is the preeminent community-based care coordination model. And every day, my staff would come back and be heartbroken by what they were seeing.

Jeff and his team were intensely dedicated to this project. They received attention, and acclaim. And then, they put it to the test.

DOYLE: They were getting a lot of praise for their program and they honestly wanted to rigorously determine what’s working and what’s not?

*      *      *

BRENNER: My name’s Jeff Brenner. I’m a family doctor by training. And I spent most of my professional career in Camden, New Jersey, a small city just across from Philadelphia. I spent about 10 years on the front line, seeing kids, adults, and delivering babies, and taking care of big, big families. I was in a small three-exam room office. I did home visits and slowly built up a citywide healthcare coalition. It started as a health provider breakfast group, and it’s still running today, 20 years later.

JENA: So, it serves breakfast, or no?

BRENNER: There still is a monthly care coordination care management meeting citywide and still some breakfast. Yep.

When Jeff was deciding where to practice medicine, Camden, N.J., felt like a natural choice. He grew up nearby, in South Jersey, and in medical school developed an affinity for a certain type of patient.

BRENNER: I always loved taking care of my most complex patients. They were incredibly interesting and I always felt like you could make a really profound difference when you leaned in as a provider. And I got really intrigued with what it is that makes them so complex and why our system seems to fall short so often. As you look at healthcare, we do miraculous things for people every day. We pull people back from the brink of death. We transplant hearts and lungs. We do incredible cancer treatment, but we really stumble when it comes to taking care of people living with complexity.

JENA: So, what is it that makes patients, like the ones you’re describing, so complex?

BRENNER: If you have one medical problem, we do a decent job for you. And if you have two medical problems, things start to break down. If you have three or more, you begin to get conflicting advice, conflicting specialists, conflicting medications. If you add in other kinds of complexity so, mental health complexity, addiction complexity, social service complexity, activities in daily living, or even, issues around longevity — the whole system starts to break down and you see it happen in tragic ways for people of all income strata.

JENA: One of the challenges with complex patients is that they tend to use a lot of healthcare, but are all complex patients what you’d consider to be super-utilizers?

BRENNER: No, actually what’s interesting is there are a lot of complex patients who are under-utilizers. They actually need more care than they’re getting. They’re as interesting as over utilizers. I think one of the fundamental drivers of all of the over-utilization and under-utilization for people living with complexity is trauma. It’s adverse childhood experiences and later-life trauma. They have a hard time navigating systems of care. And, as a result, they either underutilize the systems or overutilize them.

Jeff’s interest in complex patients married his interest in data and the Camden Coalition was born. As he told us before, he used claims data initially to identify trends in public safety, and then later, to find patients who were seeking care often, and repeatedly. He put together a team and devised an intensive program to intervene with these patients on multiple levels. The goal was to improve their health and the quality of their care, but also to drive down spending. These so called “super-utilizers” are high-need, and they’re also high-cost. You may have even heard of the shocking statistic that in the U.S., just five percent of the population accounts for 50 percent of annual healthcare spending. Super-utilizers like the ones Jeff described are an important part of this group.

JENA: What made you think that you could fix these problems using data?

BRENNER: How else would you fix it? Every hard problem in medical history has always been solved by testing — by rigor, by data, by science. The scientific method is why in a really good way, we’re curing cancer now. Improving the care of people living with complexity is as hard as curing cancer.

JENA: You were finding out who was high cost. You were noticing that the total healthcare spending in that community was concentrated among smaller and smaller amounts of people. How did you take that data-driven approach to actually try to redesign care?

BRENNER: The dominant hypothesis at the time — both for us and nationally — was that healthcare is disorganized, which it is. And that the fix to that would be to coordinate care. So, we spent about five, six years building what I think is the preeminent community-based care coordination model. We built a health information exchange. We had real-time data feeding in every day every minute. We knew exactly who was admitted and discharged. We had permission from each of the hospitals and badges. We could walk in and out of the hospital, go right to the bedside. We enrolled people if they’d been admitted two or more times in six months. They stayed in our model for, two to three months. And we went with them to every one of their appointments, helped them apply for benefits, we coordinated their meds. We did all of the things that you read about in the literature, all the things that people were talking about in conferences. This was an incredibly sophisticated, very detailed care coordination model. And every day, my staff would come back from going out with people to their appointments and visiting them at home and be heartbroken by what they were seeing. We had done an amazing job of finding everyone, but the care that they were getting was atrocious. So I think, you know, it’s sort of obvious in retrospect, but the story of our model was that it’s the care, stupid!

JENA: Can you tell me about the results that you started to see, as the program started to develop?

BRENNER: From the earliest days, I was going out seeing people, we were a tiny little team, we were talking about every case, and you could see patients that we’d reach out to, work with very closely, and their utilization would just go away. It was so beautiful to watch and they would reconnect with friends and family. They would sort of turn their life around. We saw enough examples of that to think that we were onto something. And then, as we got deeper and deeper into this and enrolled more clients, I think I started to realize like what a gigantic messy problem we were taking on. There was no way care coordination is going to make any impact on those folks.  Let me give an example. We matched all the jail data in Camden to all the healthcare data in Camden — which is an amazing data set And if you were jail involved, your outcomes in the program just fell off a cliff. We were not making any impact on people who were going in and out of jail. it’s like, enrolling people into a cancer trial when you absolutely know the medication doesn’t work.

By the early 2010s, Jeff and his team at the Camden Coalition wanted to rigorously test their care coordination model, which had come to be known as “healthcare hot-spotting.” The term borrows from a policing strategy, where data is used to identify geographic locations that are hotspots for crime. In this case, the coalition was using data to identify patients who were heavy users of the healthcare system, and to then develop a specially-tailored program to meet their needs and prevent them from going back to the hospital over and over again.

The Camden Coalition’s raw, internal data suggested that, overall, their approach was working to keep super-utilizers out of the hospital. But it was based on a “pre-post” analysis, meaning they only looked at whether there was a difference in outcomes before and after an intervention. A lot of factors could have influenced how patients did beyond just their involvement with the Camden Coalition. The only way to know for sure would be to conduct a randomized controlled trial.

BRENNER: I knew enough to know this was not going to be a one and done problem that this is going to be a lifetime of work to figure out what care models make a difference in people’s lives. It’s about time we start testing what actually makes care better for sick people.

In 2014, that time came. After the break, we’ll hear from one of the leaders of that research.

DOYLE: Finding out what types of programs or what types of patients these programs work for is at the frontier of trying to learn how to improve the lives of these patients and potentially save money at the same time.

I’m Bapu Jena, and this is Freakonomics, M.D.

*      *      *

DOYLE: A lot of my research is in trying to find where there is value and where there is waste in healthcare

M.I.T economist Joe Doyle studies how, where, and why we spend money on healthcare in the United States. He also sits on the board of the Jameel Poverty Action Lab at M.I.T., a global research center that uses scientific evidence to inform policy on poverty. Joe and I also have some shared history.

JENA: So you and I, we both went to the University of Chicago, right?

DOYLE: That’s right.

JENA: And we both shared an illustrious advisor. Who was that?

DOYLE: Steve Levitt of Freakonomics fame.

In 2014, Joe and another MIT economist, Amy Finkelstein, partnered with Dr. Jeff Brenner and his team at the Camden Coalition to determine if their approach to caring for complex patients was working—and, if it was saving money. They used the gold-standard of medical research—a randomized controlled trial, or R.C.T.—for the study, and began enrolling patients.

BRENNER: We just simply didn’t have the infrastructure to be able to pull off an R.C.T. by ourselves. So, they were fantastic partners in doing this.

Prior internal research had suggested that when people entered the Camden Coalition program, their healthcare utilization dropped dramatically. Lots of things could explain that, including the coalition having its intended effect on keeping people out of the hospital. But also, when you find people who start at a very high level of healthcare use, sometimes they just naturally get better over time, an idea economists call “regression to the mean.” So Joe and Jeff and a lot of other people began the hard work of conducting an R.C.T. to see if this program was, in fact, working to reduce healthcare use and healthcare costs. They looked to hospital readmissions to tell the tale.

DOYLE: When we measure readmissions, it’s typically did someone show up in the hospital again within 30 days? That’s a very coarse definition because some people need to go back to the hospital it’s good for their health to go back to the hospital. Whereas other people, it’s a reflection of a lack of coordination of care that could have prevented hospitalization. We want to prevent hospitalization because they’re just so expensive. And if we had low-cost ways of coordinating care in ways that would prevent re-escalation, we should implement those.

JENA: Could you tell me about the Camden Coalition hot spotting approach, and why you wanted to study it?

DOYLE: So the Camden coalition is extremely inspiring. They were known to hotspot their city. So which census tracts were driving the most healthcare spending. And then at the individual level, using data to find out well, who are the patients with the most complex needs, who are the ones who are driving a lot of our spending, identifying them in the data. And then they had this program to address their clinical and social needs. To be eligible for the program, they needed to be in the hospital and they had to have been in the hospital at least one other time in the six months prior to that hospitalization.

JENA: What was known about the program in terms of this efficacy prior to the study that you and your colleagues did?

DOYLE: There was a lot of discussion of this program because it was so inspiring and data driven. And there were a small pre-post analysis showing that when people went into this program, their healthcare utilization fell dramatically after they entered it.

JENA: Tell me about the randomized control trial. How long did it take to do? What was the process behind it?

DOYLE: So first I give the Camden Coalition and Dr. Brenner, a lot of credit. They were getting a lot of praise for their program and they honestly wanted to rigorously determine whether it was saving money on its own. If we found that it did, we expect that payers around the country would just — it would sort of takeoff even more than it has. And if we find that it didn’t, then we would learn that, well, we need to go back to the drawing board and think about other ways we can help these patients, not to stop trying. The Camden coalition would recruit patients into their program at bedside in the hospital. Some patients would get the program and some wouldn’t in this flip of a coin way, so that we could then compare the outcomes of people who were randomized into the program to those who weren’t to get, very credible, very easy to understand differences in their utilization, so we could test whether the program, was keeping people out of the hospital.

JENA: And how many patients ended up being randomized in the trial?

DOYLE: Prior to the study, we had done some analysis to determine how large the study would need to be in order to have meaningful results. And we calculated that 800 was the number that we needed to recruit.

JENA: Was it national or only in New Jersey?

DOYLE: It was only in the hospitals in, Camden, New Jersey.

JENA: Okay. And what did you find?

DOYLE: We found that the people who were in the program who were assigned a treatment, there was a ramp up of health needs prior to enrollment, which we expected, and then a dramatic decline after they were in the program. This is exactly what had been found previously. What was a bit surprising though, was that when we look at the control group, there’s also this ramp up in healthcare utilization. That’s how you become eligible for the program, but a decline that mirrored what was happening in the treatment group. There was a decline in utilization for people who were not in the program. And we reconciled these facts with this idea of a regression to the mean. Automatically, most people are going to become less sick if you’re starting at a very high level of illness. So this regression to the mean could reconcile this fact that both the treatment and control group appeared to get better they both reduced their hospitalization following entering the study.

JENA: So then if you were just looking at the control group and the treatment group and then you look at let’s say their admission rate, following that hospitalization, was it any different between the treatment group and the control group?

DOYLE: So we found that the treatment group and the control group had nearly identical readmission rates. We found that in the six months after enrollment in the study, about 62 percent of both the treatment and the control group reentered the hospital. That’s a very high rate of readmission, reflecting that these are true super utilizers and remarkably, there was no difference between the treatment and the controls.

JENA: Did you find any differences in mortality or anything else?

DOYLE: We didn’t. Although mortality’s a more rare outcome. So we wouldn’t have the statistical precision to determine if there were effects on mortality.

JENA: So the findings, were they surprising to you or not?

DOYLE: The findings were surprising I expected when I looked at the pre-post analysis that had been done before showing very large reductions in hospitalization if you entered the program were probably overstated for these reasons that people tend naturally to get better after an escalation of health problems. But I did expect to find that there was some difference because this program is highly intensive. They spend about three months with the program. It’s a particularly intense program relative to other campaigns to try to improve the health and wellbeing of patients like this.

JENA: Why do you think the trial showed the program was inefficient on these measures compared to earlier research that seemed to indicate that it worked?

DOYLE: The Camden Coalition project was really targeting people who are particularly sick. They have many different healthcare problems. They have this readmission rate over six months of about 60 percent. So if you’re thinking about hot spotting or super utilizers, these are the super duper utilizers that we’re looking at. We’re learning that the program may have more trouble at that far extent of the distribution. In addition, Camden is different than other places. So one issue when you were looking at studies is if you find the program works or doesn’t work in one location, it doesn’t automatically mean it won’t work in other locations. It means that for locations similar to that, we expect the best evidence suggests that it’s not going to reduce hospitalization, but it could work in other environments.

JENA: The study got a ton of attention and I’m curious what people who were passionate about this idea of intensive case management of a hot spotting approach, how did they react to the study?

DOYLE: When I heard the response in the media, there was pushback. Hey, this program has got to work. Maybe it didn’t work in Camden, but it can work other places. But then after our study, there was a bit of a pause and says, well, we can’t assume that it works. We have to go out and test whether it works, or we need to find out who it works for and who it doesn’t work for. And that’s part of my drive to find where there’s waste and value in healthcare is addressing the clinical and social needs of super utilizers is an inspiring idea. And I don’t think our study says you should stop trying. It says that it may not work for everyone and we need to find out how best to treat these patients And so finding out who it works for is one way to add value to healthcare.

BRENNER: Without significantly improved care models that address people’s medical, mental health, addiction, and social services challenges — really sophisticated care models — we’re not going to improve the cost and utilization for people living with complexity.

Jeff Brenner again.

BRENNER: For God’s sake, we can, you know, open people’s bodies up and transplant their organs. Like, this is doable. We’re not talking crazy talk here.

JENA: If you could design a perfect program, what would it look like?

BRENNER: A PACE program. It’s just perfect.

JENA: Tell me about it.

BRENNER: PACE is a 30-year-old model — spread all over the country now — it started in San Francisco. And when you walk into a PACE program, it’s for nursing home eligible seniors. They often have primary care on site, physical therapy on site. They send vans out to pick everyone up every day, bring them to the program. They are on the hook. It’s a fully risk-capitated model. And they have to pay — if they go to a nursing home, if they go to the I.C.U., they pay for all the meds. They pay for all the durable medical goods — for their wheelchair. And that team knows the family, knows the patient. They’ve got an ethical obligation. They have a relationship with that patient. And they have to decide, does it make sense to get a new wheelchair? Does it make sense to approve a certain medication? Does it make sense to approve a certain treatment? It’s a comprehensive model. It has all the attributes that I’ve been talking about. So, we can do this.

JENA: What’s the barrier to expanding that to other populations of patients?

BRENNER: I don’t think insurers want it expanded because it pulls people out of the risk pool. Secondly, it’s a very intimate model. Care is so local that the attributes of good care are going to look slightly different from place to place, the relationships that you need to build and I think that programs like PACE take time to scale. The interesting thing is that we have scaled open heart surgery. We have scaled rare and complex neurosurgery. We have scaled oncology treatment. I know we can scale hard stuff if we’ve got the political will to do it.

It’s been about 20 years since the Camden Coalition began as a monthly breakfast meeting for primary care providers in Camden, New Jersey. Back then, they would discuss how to improve care for their patients. Dr. Jeffrey Brenner has moved on from the Coalition, but he hasn’t stopped having that conversation.

He’s still trying to find the best care model for complex patients. One approach he and others have tried is focusing on a single feature of our lives that many of us take for granted: housing

BRENNER: Oh, it’s — it’s staggering. You see people who, one day they’re falling apart diabetic foot ulcer that’s open, sugars in the four and five hundreds. You move them into a unit. Suddenly they’re sleeping at night. They’re taking their meds. They’re beginning to re-engage with friends and family. So, it’s really — the most powerful work I’ve ever done in my career is housing, housing, housing. It’s far more powerful than any medication I’ve ever prescribed.

The results of the New England Journal of Medicine study that Joe Doyle co-authored don’t suggest that the Camden Coalition’s hot spotting approach was a failure. Costs are obviously important but it’s also probably true that the program has changed lives. It also put a spotlight on complexity, an issue in medicine that, for its tremendous costs, doesn’t get enough attention.

I remember reading the study when it first came out. I was also surprised by the findings but what really struck me was the willingness of Jeff and the Camden Coalition to rigorously evaluate their own program. What this research shows us is that it’s important to look at the data, and to listen to it. To build on it. And then, to pivot. Here’s Joe Doyle again.

DOYLE: There are a lot of promising ideas. One is for patients with these many complex needs to have even closer monitoring. So a lot of monitoring of patients at their home, could give us enough data to find out when they might be starting an escalation and when we could, short circuit that escalation. The other approach is to think about individual social needs and try to address them. There’s good evidence that housing homeless people improves their health. But there’s some good clinical trial research showing that it keeps them out of the hospital and actually out of prison in ways that can save money. So finding out what programs target those patients and improve their wellbeing and potentially save money or not, that’s where I think, there are a lot of promising ideas and where there’s a lot of work to be done.

The New England Journal of Medicine “hot spotting” study, as it’s come to be called, was one of the most important randomized trials to date that didn’t deal with drugs, devices, or surgical procedures. It dealt with care. And as Jeff put it, how to treat people isn’t the same as how to care for people. There’s a lot more to figure out and I’m sure data will continue to help people like Jeff Brenner, Joe Doyle, and maybe some of you figure it out. And while we wait for that, I’d like to thank Jeff and Joe for joining me today and all of your listening.

Coming up next week:

OBERMEYER: A lot of the things that we worry about the most in the emergency department are things that can actually look pretty subtle at first.

In the emergency department, the stakes are high. Doctors have to make critical decisions, and fast, about who they should admit, and who they should send home. Sometimes, they get it wrong.

OBERMEYER: It didn’t look like the doctors really had appreciated that this was a person who had a high risk of dying in the next, week.

So, who — or what — might help them get it right?

OBERMEYER: We were able to use the artificial intelligence algorithm to try to get inside the head of the physician.

That’s next week on Freakonomics, M.D. Thanks again for listening.

*      *      *

Freakonomics, M.D. is part of the Freakonomics Radio Network, which also includes Freakonomics Radio, No Stupid Questions, and People I (Mostly) Admire. All our shows are produced by Stitcher and Renbud Radio. You can find us on Twitter at @drbapupod. This episode was produced by Julie Kanfer and mixed by Eleanor Osborne, with help from Jeremy Johnston. Lyric Bowditch is our production associate. Our executive team is Neal Carruth, Gabriel Roth, and Stephen Dubner. Original music composed by Luis Guerra. If you like this show, or any other show in the Freakonomics Radio Network, please recommend it to your family and friends. That’s the best way to support the podcasts you love. As always, thanks for listening.

JENA: This sounds a lot like eHarmony, but for program evaluation.

DOYLE: It was a good match.