In Conversation With… Harlan Krumholz, MD, SM
Editor's note: Dr. Krumholz, Harold H. Hines, Jr. Professor of Medicine at the Yale School of Medicine and Director of the Yale-New Haven Hospital Center for Outcomes Research and Evaluation, is an international leader in outcomes research and health policy. We spoke with him about readmissions and post-hospital syndrome, a term he coined to describe the risk of adverse health events in recently hospitalized patients.
Dr. Robert M. Wachter: What got you interested in what happens to patients after hospitalization?
Harlan Krumholz: In the 1990s, I was trying to characterize the health care system and was drawn to consider neglected areas that patients care about. Much work has been done on mortality and at that time we were focusing a lot on processes of care—whether people were getting the right treatments or not. But this issue of readmissions nagged me. At the time, a landmark article had been published on readmission drawing attention to the issue—but almost a decade had passed with no action. I had noticed, even before I discovered that article, that many patients were coming back to the hospital and published a paper looking at patients with heart failure—and saw that about half the patients were admitted again within 6 months. I haven't stopped thinking about it since then.
You and I trained together. At that time, a lot of emphasis was on the venue in which you were practicing. In the hospital, getting people out of the hospital was our goal. When we were in clinic, we thought about people as outpatients. We never considered their journey across venues and how our actions might have affected that. A perfect example was how we treated discharge summaries. We never saw them as a tool for communication or as essential to communicating what happened in the hospital and to helping to pave the path for the patient subsequently. That led me to want to study their journey—and that led me to the postdischarge period and readmission.
RW: Talk about the work you did on measures for CMS.
HK: Many people argue that we have too many measures. What we found with process measures was that when you start honing down on specific populations and looking at particular actions, the denominators for these measures shrink, the ability to measure a process well is eroded, and the measures become less meaningful. You also simply cannot measure all the important processes, especially when it comes to communication, collaboration, and coordination. At some point, we decided to pivot to focus on outcomes, providing an incentive for hospitals and health systems to do their own analysis of their processes in order to improve their outcomes. Thus, we aimed to illuminate the actual outcomes achieved, like mortality. Our central challenge initially was to show that we could produce a measure with claims that was a good proxy for a measure with higher quality data. In fact, it turned out that the measure results from claims were highly correlated with measure results with medical record data—giving us confidence to proceed. CMS came back to us after we successfully launched publicly reported mortality measures in heart failure, pneumonia, and myocardial infarction and asked us to do a payment measure. We were reluctant because we felt that a payment measure needs to be surrounded by other measures that provide some sense of whether the amount of expenditure is producing anything of value. We felt that payment in isolation might not be sufficiently helpful—at least at that stage. So, we said, "what about readmission?" I thought that could be a win–win. If readmissions drop, then savings accrue, and patients—who are generally not eager to come back to the hospital—would benefit. So we were able to launch publicly reported readmissions measures for those same three conditions, and these measures were ultimately incorporated into the Hospital Readmissions Reduction Program.
RW: What led you to characterize what happens to patients after they leave the hospital as a syndrome?
HK: I've been working in this space for a long time trying to figure out how to reduce readmission. But it was bothering me that although the rates were declining, it still seemed that we should be doing even better. Moreover, there was a perplexing observation. Patients were coming back to the hospital for a range of problems often not directly related to why they were initially admitted. It occurred to me that maybe we didn't understand this phenomenon well enough. Maybe we were observing that people were going home in a transient state of generalized vulnerability whereby all of their physiologic systems were in a bit of disarray. We had deconditioned, malnourished, sleep deprived, and stressed them to the max. It seemed plausible that the result of all that toxicity in the hospital had perturbed their systems in a way that made them susceptible to a wide range of problems. The rates of infection, bleeding, blood clots, and falls were increasing—in addition to the risk of whatever they had come in with originally. This led me to think about what might be the mechanism.
Where the idea came from is interesting. One time I came back from an overseas trip. I was particularly jet lagged and I realized—I feel just like my patients do when they go home. People were talking to me, and I couldn't concentrate. I was a little clumsy. I could have fallen. We are not really taking into account what happens to people when they are hospitalized. That jet-lag experience led me to think about this toxicity.
And I started thinking that transient cognitive dysfunction was part of this syndrome. Leora Horwitz (now at NYU) led a study that we did together in which we documented that people, soon after the hospitalization, often forget a lot of what they were told in the hospital. Our study showed that they cannot recount the basic facts of their condition or what happened to them.
That also led me to think that our strategy of doubling down on the information we were giving patients may not be matching what they are prepared to receive. Christiana Care Health System had this wonderful video demonstration where they were teaching about discharge summaries. They went into a patient's room and said, "10 people are going to introduce themselves at discharge: the dietitian, the nurse, the doctor, and the work clerk, etc." They each went down the line. The patient was an actor. This was a simulation, and they asked, "Do you have any other questions?" The patient looked up puzzled and said, "When am I getting out of here?" I remember thinking that this actor is channeling what is on the mind of our patients. Many patients are not ready receivers of the information overload.
RW: I got the sense that you transitioned a bit from considering post-hospital syndrome as a characteristic of the health system—in that we're not very good at everything from discharge summaries to having any sense of our patient's condition after they leave—to it becoming a pathophysiologic entity that the health system was unaware of. Is that right? Calling it syndrome seemed to gather more attention for this predicament than the more typical academic approach.
HK: Here is how I conceptualize it. People come in the hospital with a condition. We immediately try to jump in to mitigate and to cure if possible. In the course of that care, we are causing a lot of collateral damage, which we've tended to discount as "they may be a little uncomfortable." They may have roomed with someone who was up all night. We maybe poked them at 4 AM. But the big thing is that we are saving their lives—so we just push forward. We have a mentality that focuses on rushing in and addressing the primary illness, without regard to any harm that accrues. So, my conceptual model is that all of the stress, discomfort, inactivity, poor diet, missed meals, and inadequate rest and sleep has created a toxicity that incurs substantial collateral damage.
At discharge, people are weakened and susceptible. Still, with all these things we do in terms of communication, collaboration, coordination… we can either ease their path, catch them gently, and help them make a successful transition. Or we can say you are punch-drunk from everything we have done to you, and now we're going to set an obstacle course in front of you, wish you luck, and see how this stress test plays out for you in the next couple of weeks. But even better, we could make the hospitalization more soothing, more healing, more supportive, more restful—and, maybe, better position people for the post-hospital period.
Both sides merit working on. In the end, it's not about readmission; it's about improving recovery. The path to recovery is making people as strong as possible throughout the course of the hospitalization while attending to the acute illness. And then making the transition as smooth, seamless, and easy as possible. I see it as a continuum. Are we trying to get them as well nourished, as well rested, as well conditioned, as strong and oriented as possible by the time they leave, to make it easy for them to take the next step? Or are we keeping them up all night and generally disrupting their routine and interfering with their recovery? There is immense opportunity for us to take sleep as a vital sign, for example.
One important discovery for me was learning about the allostatic load model. Bruce McEwen came up with the concept about how stress can perturb physiologic systems and created animal models that would be able to reflect that. Deena Goldwater, a former fellow in Bruce's lab who is also trained as a geriatrician and cardiologist, came up to me after one of my talks and told me what I was describing about patients reminded her of experiments she had done in the lab. She told me that they would take healthy rats, put them in a box, expose them to noise at odd times, and poke them in order to give them pain. They would sleep deprive them, force them to be inactive, and malnourish them. And she said, "Your description of what happens to patients sounds very much like what we did to the rats. When we evaluated them, we could see that we had screwed up their immune system, their hematologic system, their cardiovascular system, and their endocrine system." That made me think that elevated levels of cortisol, catechols, inflammatory markers—all of these come together to influence the hypothalamic–pituitary–adrenal axis, and may be the mechanism by which the toxicity of the hospitalization generates pathophysiologic consequences to multiple organ systems. And the result is this heightened period of vulnerability to a wide range of health conditions in the period after hospital discharge, often for conditions unrelated to the one that led to the initial admission.
It started seeming to me like a syndrome. That's why I called it post-hospital syndrome, a transient period of generalized vulnerability—and consider it likely a response to a high level of acute stress that was accompanied by malnutrition, deconditioning, sleep deprivation, and so forth. I don't think the term has caught on enough because hospitals are not yet understanding this as a strategy to improve recovery (and reduce readmission). We certainly need more studies—and it ought to be tested around readmission. But then again, what is the harm in making the hospital experience more hospitable, more healing, more patient friendly. We have so far to go to make the hospital a truly healing environment rather than one in which we (implicitly) say, "Tough it out. We're taking care of your acute problem. Be grateful that you're getting the attention you're getting." Obviously, that is an overstatement and many places now have Patient Experience Officers—but I just don't believe we are doing enough fast enough.
RW: Imagine you have been given an unlimited amount of money to build and design a hospital of the future, both the physical plant and the things we do that will minimize the chances of post-hospital syndrome.
HK: Now that would be a great opportunity. We need to start thinking about how to support patients—to make the experience one that, at every opportunity, strengthens them. Say a patient comes in with pneumonia. It starts from making sure we are tracking parameters (sleep, nutrition, activity, etc.); it's about how we are creating the context and experience for a strong recovery. Rest and the environment that promotes it is one thing that we need to ensure. We need to make sure that the patient gets adequate sleep, in an environment conducive to that. I would have patients in their own room. I would be sure people are well nourished. We would encourage, as preferred by the patient, social support, and visits. We would surround them by bright colors and sounds and odors designed to lift their mood (I have written that adult hospitals should follow the lead of pediatric hospitals in this way). We would avoid blood draws, Foley catheters, tests, and procedures except what is absolutely necessary. We would give people a schedule every day so they know what to expect and when, enabling them to have a sense of control and understanding. We would avoid a lot of the uncertainty. For example, on the consult service, we don't tell people when we're going to visit them. So, they are stuck in the room most of the day because they are afraid to miss us. It is as if we implement systems that give people anxiety, and we're making it hard for them to be active in any way. This hospital of the future not only delivers outstanding care, efficiently and effectively, but delights the patient by every action, taking into account whether it is better positioning them for leaving the hospital. Even in the midst of acute illness we can think of rest, nutrition, strength, support, cognition, and a healing environment.
When I really got into this, I started remembering all these situations where I had a patient, maybe their roommate died the night before or the roommate had diarrhea or vomiting. Never once did we go to the patient when someone had died next to them and asked, "What was that like? Are you okay?" Because these people may never have witnessed someone dying before. I never remember considering the toll of that stress.
RW: We've had readmission penalties now for a few years. Where have you come down on their effects? It sounds like you were part of the driver for the penalties and there have been benefits and some criticisms. What is your scorecard on it now?
HK: It turned out not to be a positive experience for the provider community—because it was about penalties rather than rewards. But for the patients, we improved readmission rates. They have gotten better without any evidence of harm. And hospitals that reduce readmission rates have tended to reduce their mortality rates. Although the policy could be improved a lot, particularly the way that payments are distributed. But the penalties have led places to make investments in improving the transition, improving safety. And that is a good thing.
RW: My sense is that, you're right, the penalties have led to considerable investment on the transitional part. But maybe less attention on what happens in the hospital that may set patients up to be at risk for post-hospital syndrome.
HK: We are at the point where we need to find some institutions willing to test this—I firmly believe that it will pay dividends for the patients and the institutions. The corollary here was the work in delirium by Sharon Inouye and others, which demonstrated that when you addressed these issues, delirium risk declined substantially. They couldn't eliminate it, but they decreased the incidence by getting people sleep, not disrupting them, these same themes. If we could get some institutions to experiment on a ward and ask every day: How much rest did patients get? Did they get enough nutrition? Are they being active enough? Are we doing everything we can to decrease their stress level and make things predictable and help them through, recognizing the end game is to make them as strong as possible at the point that they leave. And in addition, continue our focus on improving the transition. I'm convinced it would work, but we should demonstrate it.
RW: One of the themes of the last 20 years of improvement work has been unanticipated consequences, and we've now learned that they're more common than not. Pain as a vital sign turned out to have unanticipated consequences (in that it helped fuel the opioid epidemic). If sleep becomes a vital sign, what would you worry about or do you feel like that only has an upside?
HK: This is a good point. It should be a primary focus on nonpharmacologically-induced sleep. In the sense that the interventions are all about not disturbing people, trying to keep things quiet, creating a restful environment—maybe an opportunity for people to hear music or other similar enhancements. Every patient is different and things happen in the hospitalization, so to get the acute situation under control takes some work. But if, from the moment patients came in to the hospital, we were intently focused on their experience, then it's possible they would be stronger and better prepared for a successful recovery at the time of discharge. Whether that would actually make a difference is worthy of testing.
My hunch based on our reading of this entire literature and our research is that this could be a game changer. Now we are discussing disposition from day one. But if we were discussing these other factors critical to recovery, how we are going to make sure patients are as strong as possible and at the same time addressing their acute illness, we would be far ahead of where we are today.
RW: What do you think about hospital-at-home programs and whether they contribute to a restful, less stressful environment?
HK: The digital revolution is going to be transformative in health care. If we can use it to monitor, understand, and help make the experiences people have with the health care system more positive and focused not only on acute management but also on recovery and prevention, then yes, we're better off keeping people out of the hospital. But we have to monitor carefully anything we implement to be sure that with those good intentions we don't somehow make things worse.
RW: Obviously one effect of digital is we have the ability to monitor a patient's status at home; whereas in the old days, it would have been in the hospital. But let's say the patient still needs to come in to the hospital and will be discharged either to home or some other setting, how does the digital environment differ from the experience of a patient today?
HK: Let's say people come in with a patch. The patch will tell us something about where they are. So if I'm rounding, they may also have a handheld notification device—just like at restaurants—or something that says, "The cardiology team is going to be in your room in 15 minutes." If I'm trying to round on a patient, I can also look and see that patient is in radiology. We will be able to communicate more seamlessly. We will know how much rest they got, how much activity, how much nutrition—even how much noise they have been exposed to. I want us to have respect for the importance of those parameters. The digital revolution is not going to do anything if all it does is produce data; it has to produce information that we pay attention to.
The question then is the degree to which the data can help us understand how people are recovering. To what extent are we weakening them, and where are they in their rest, nutrition, activity, and stress. Getting better at monitoring that can help us meet our targets for people to be strong. We need to be able to use digital to make sure that people don't fall off during the time in between leaving the hospital and seeing their own doctor.
RW: As a learning health system, what do you think the lessons are from readmission penalties? Here you have a safety or quality measure and an efficiency measure that now influences every hospital in the country, and you have many consequences, some anticipated, some unanticipated. What have we learned from that measure that is generalizable to our efforts to change the payment and the transparency system around other things in the world of improvement?
HK: Our general approach to measurement is that we see it as a form of social engineering in medicine: making visible what has not been visible. We now know that people end up back in the hospital quite frequently, and some of the ways the system is configured may be contributing to that. Illuminating this can help us focus on areas of improvement. But, we have been awash in measures, and there can be some measure fatigue. We have to separate measurement for quality improvement from reasonable measurement for accountability and incentivizing change. For the future, we need to create the means by which consortia of institutions can come together and be insulated from penalties if they are willing to innovate and can achieve certain targets collectively. If 10 institutions are willing to work together and could collectively achieve a gain, then they could disseminate what they learned to others—and receive some incentives for that type of collaboration.
Another theme is that our current knowledge generation process within medicine is moving too slowly for the information needs of everyone involved. We need to find ways for more rapid-cycle learning that can be tried in a group of institutions and then, if found to be favorable, scaled rapidly. We need to have policies that encourage and reward them. Then we need pathways to share so that we can quickly address the problems. When you look at our hospitals and health systems, a wide range of areas needs attention. I'd like to get people working on them in ways that reward them to come up with innovations that can scale, so that we find cooperative and technology-enabled means by which we're able to transform medical practice such that it really does achieve the best for each individual.