Dr. Christopher Landrigan - Resident Sleep, Safety, and the Design of Academic Health Systems

All right. All right. All right. All right. Good morning, everyone. And welcome to Grand Rounds. It's my pleasure to introduce Dr. Christopher Landrigan. He did medical school at Mount Sinai, followed by residency in pediatrics and fellowship in hospital medicine and health services research at Boston Children's as well as a MPH degree from the Harvard School of Public Health. He then stayed on his faculty and is the chief of the Division of General Pediatrics at Boston Children's, the William Barronburg Professor of Pediatrics and Associate Professor of Medicine at Harvard Medical School and the director of the Sleep and Patient Safety Program at Brigham and Women's Hospital. He'll be speaking on resident sleep, safety and the design of academic health systems. Dr. Landrigan, thank you for being here and for sharing your expertise with us today. Thanks very much, Esther. Let me see if I can make this work. Can you see the slides? Excellent. Okay. Thanks very much. I'll get started. So what I wanted to do today is to talk a little bit about really the work that I've been engaged in for the better part of the past 20 years, thinking about both sleep deprivation in residencies as well as I think the many interacting elements of the health care system and what happens as we begin to try to design safer system to care. As I'll get to, there are many complexities to this and what we know a lot about the effects of sleep deprivation on safety and performance, the impacts of policies and efforts to change works have been complex and have been influenced by many of the things. I'll try to touch on that quite a bit as I go on. So first of all, just a couple of disclosures. I'm going to speak a bit about I pass today and while I pass began as a purely academic endeavor in our efforts to try to spread it to other hospitals, we did form a spin off company here from Boston Children's and that doesn't work around. Just by way of very high level overarching introduction. I want to start with a discussion of the epidemiology of adverse events and medical errors, which was work that really began to come to the four in the US. Just as I entered my fellowship in the late 1990s and you know at that time it was estimated and really broadcasted a national level for the first time that there were large numbers of deaths each in the US due to medical errors. I think that because we don't have a national surveillance system to carefully address this issue, we don't really know what the precise numbers are, but I think it's fair to say that it's years have gone by. None of the estimates that have come out have been particularly reassuring. So we started back in the late 90s with an estimate of 44 to 98,000 deaths each year in the US due to medical errors. About 10 years after that, the office of the inspector general of the Army of Health and Human Services estimated that maybe the true number was about 180,000. I was involved in a study in North Carolina tracking over a six year period and about that same time trying to see what was happening to rates of medical errors, not adverse events. And we found that they were high and were remaining high as the years went by without a very encouraging looking trend lines. You can see there. And then perhaps most recently, British Medical Journal, a couple of investigators at Johns Hopkins estimated that perhaps the true number of deaths is 251,000 per year due to medical error. These are extrapolations. I think again, there's no precise understanding of what your numbers are here, but I think it's fair to say that they're common and have continued to become. And it's years have gone by due to, I think the complexity of our health systems and lots of things, frankly, that happen within this that are not terrible today necessarily. Now that being said, there have been while the overall rates of adverse events have been made high, there's no doubt that we have been very successful in tackling certain focused issues within within patient safety. One of those is cat related bloodstream infections and other types of hospital acquired infections, where from the work of Peter Pronebos and other folks, there has just been tremendous reduction in the frequency of these types of things over the past 20 years since I was training. In fact, if you look at the data from agency for healthcare research and quality, it looks as if the rates of cats related bloodstream infections and similar types of issues have dropped much to 80% over that time period. It's really a very different place from where it was. When I was a resident, and we just sort of presume that these types of things were unavoidable, those kind of the prices of business, if you will. I think the attitude about hospital acquired infections is not quite different. And on the surgical side of things, surgical safety checklists, while they have not always been perfectly implemented or perfectly successful, there's certainly great evidence that they can be extremely effective when done properly. One of the issues that has been pervasive, I think, for many, many years in healthcare, although it's certainly been a shifting target, has been the issue of sleep deprivation. And what the impact of sleep deprivation is on patient safety and we are our own sleep deprivation. I began to get interested in this many years ago and did my cable war really focusing on understanding this a little bit better and trying to meld an understanding of sleep insurcaating biology, what we know about patient safety. And if you look to the fields of sleep insurcaidian medicine, there are really four primary factors that are thought to drive alertness and performance that could potentially have an impact on patient safety. Those include biologic time of day or where you are in your circadian rhythm. The number of hours that you've been awake in a row, consecutive waiting hours, the amount of sleep that you're getting in a regular basis, and then lastly a phenomenon called sleep inertia that I'll describe in a moment. For decades now, it's been recognized that the human circadian pacemaker, which is located in the superciskmatic nucleus of the hypothalamus, the central pacemaker, drives a host of secondary pacemakers that are located in organ systems around the body. And that in turn drives a whole host of biologic processes. Some of these are going to be familiar to everybody, just who went through an obstetric stratation in medical school. You know that core body temperature, for example, has this 24 hour pattern up and down over the circadian cycle. Likewise, there are many hormonal outfits melatonin, for example, and growth hormone and others that follow this 24 hour circadian pattern, where there's an output that's higher at certain times of data than others. And it turns out that these rhythms are preserved, even in the absence of any type of an external stimuli. They're endogenous rhythms, where it much like the heart rhythm is endogenous. That one beats about once per second. These would beat about once or 24 hours. And just like the heart rhythm, although they are endogenous rhythms, they can be affected by external factors. In the case of the circadian system, the primary synchronizer of the circadian system is light. And some of our social behaviors can also potentially influence it. From a safety research standpoint, more interestable perhaps even than the hormonal outfits is what it does to neurobeaproperformance of alertness. So for example, just focusing on the fourth panel down on the left, the slow eye movements. Slow eye movements are the occurrence of doing this kind of thing, where your eyes are really running your socket and your head against the nod. And that happens much more often in the middle of the biological night than it does during the daytime. Likewise, reaction time, if you look at second motor vigorous performance tasks, the second down on the right. That is a, this is really a very simple reaction time test where I would put a red light up on the screen. And as soon as you see the red light, your job is a research subject just to hit a button saying I see it. It's the simple, super reaction time test really. And for most of us responding to that stimulus in the middle of the day is about 250 milliseconds or a quarter of a second. But as you can see, as you get into the middle of the biological night, let's say three o'clock in the morning when you respond to that test, on average, you will rise to about a half second or a second to respond to that stimulus. And if you look at the slowest 10% of responses in the group of all of us together, you see some people not responding until six seconds past. Right? So the red light goes off and that is one, two, three, four, five, six before you. You see it and respond. Which if you're sitting at a laboratory in a protected environment, it's really an interesting phenomenon, but not so concerning. If on the other hand, you happen to be a trucker, this is a disaster. So this is the temporal distribution of the teapulated single vehicle trucking accents from the natural transportation safety board. And what you're seeing is that there's this massive peak of these types of accidents in the middle of the night from three o'clock to seven o'clock in the morning as compared with the middle of the day. And the mechanism is thought to be precisely this that a trucker's eyes glaze over for a few seconds. Excuse me, they happen to be on a curve on the road or what have you to go over the rumble strip into the trees and and the disaster and so. Entirely separate from this or a biologic system is something called the sleep homeostat, which is really a sea sauce system in the brain that says the longer that you've been awake, the greater the drive to sleep and the longer that you've been asleep, the greater the drive to wake up. This follows this more or less linear pattern over time independent of the circadian system. And again, just sort of making it a little bit practical. Again, from the trucking literature. So this is a figure that's displayed not by time of day, but by number of consecutive hours of driving. What you're seeing is that the risk of a sleepiness related accident was roughly flat for about the first eight hours that trucker drives. But after the eighth hour of consecutive driving, you begin to see this exponential rise, such that by the 13th hour, there's a 1500% increased risk in a crash compared to what it was in baseline. In fact, the amount of impairment that's induced by sleepiness has been very consistently shown to mimic that, which is induced by alcohol. So that at about 17 to 19 hours of sustained weightfulness depending on the study, performance tends to degrade to about the level of induced by alcohol concentration of 0.5. And at the 24 hour mark on average, mean performance is equivalent to that of this by blood alcohol concentration of 0.1. About 30 years ago now, excuse me again, it was recognized that not only sleep home is tragic pressure driven by the amount of sleep you've gotten in the prior 24 hours, but it's also driven actually by the amount of sleep that you're getting on a regular basis. So in this experiment, what was done is there was a group of research subjects who were brought into a very protective environment, a kind of a cave like lab. And they were, you know, where you can very carefully control how much they're sleeping and waking and activities and so forth. And one group of these subjects was limited to eight hours time in bed, a second group, six hours time in bed, a third group, four hours time in bed, and they were compared with a final group who was forced to remain awake for 72 hours straight, zero hours time in bed over three days span. And what you see is that of course the zero hour time in bed group, the performance you grades very rapidly, where they start having more reaction time lapses and other types of neuro behavioral failures that occur as that period stretches on. But what you're also finding is that you focus on the six hour group, four hour group over time, let's say over 14 days of working on these types of schedules. Even the six hour group begins to look like a group that's gotten no sleep at all for 24 hours straight and the four hour group after 10 to 14 days begins to look like a group that's had no sleep whatsoever for 48 hours straight, which is a really pretty profound level impairment. This is of course without any chance for weekend cats of sleep or other recovery. Interestingly, chronic sleep deprivation unlike total acute sleep deprivation seems to sneak up on us in a way that can be a little bit difficult to get a hand more. So what you're seeing here in this graph is subjective awareness and reporting of sleepiness. And as you focus just on the first couple of days, you see the slope of this line is not so different from the slope of the objectively measured performance over here on the left. There is this relatively rapid degradation in perceived alertness. But after the first couple of days, there's this flattening so that people are reporting sure maybe a little bit worse than it was yesterday. But it's not that bad. I basically feel like I'm doing fine. And yet objectively, they're getting worse and worse and worse and worse and very rapid rate of not breaking this. Lastly, a phenomenon called sleep inertia, which is really a griggyness in the first few minutes after awakening. And it's been demonstrated that the degree of impairment in those first few minutes can actually be even worse than that, which is induced by 24 hours of total acute sleep deprivation, which is what you're seeing in the left hand side of the graph over here. This is performance decrements in those first few minutes just after awakening, especially from deeper stages of sleep, stage three, stage four sleep. And then you reach a higher level of performance, which dips in the afternoon and then again after the night of sleep deprivation. But never actually put that level of those first few minutes of just profound impairment on first waking up. On the right hand side of the slide, which you're seeing is some very cool, more recent studies that have looked at this using FMRI to essentially understand what brain metabolism looks like in those first few minutes compared to later in the day. And what you're seeing in red or hot spots in the brain or really cold spots in the brain, where there's glucose metabolism is poor, and there's a buildup of metabolites. So as healthcare workers, obviously all four of these factors are in play all the time. You know, we are often working in the middle of the night having been awake for many hours in a row, perhaps not getting great sleep on a regular basis. And then if you are so fortunate enough to fall asleep for a few minutes while you're on duty, if you're woken up for a crisis, sleep the nurse potentially plays in as well. There are studies now going back more than 40 years that have demonstrated that this is potentially problemat. So this is the first study in the medical literature of which I'm aware was published in the New England Journal way back in 1971, where medical interns, internal medicine interns were asked to read EKGs the night after they had been on call versus on nights where they were arrested. And what was found is that they made about twice as many mistakes, both a commission and omission, they're reading those EKGs when they were sleep deprived, as compared with one they were arrested. There have been very similar studies conducted in surgery programs. This one was done again quite a while ago now, 20 years ago, published in the British Medical Journal where surgical interns were being tracked while they did simulated laparoscopic pulse effectively. And what was found is that both the amount of time it was taken, number of unnecessary movements as well as the frequency of frank errors according to the people who were assessing the study were greatly increased in the group that have not slept versus those that have. In fact, if you look at the literature as a whole, there have now been well over a hundred studies that have looked at the performance impairments induced by sleep deprivation in one way or another, sometimes using very simple reaction time tests, sometimes using more than 40 years. Sometimes doing more sophisticated medical tasks like simulated surgery or history and physical quality or reading EKGs, what have you. A meta analysis of 60 of these studies was done in the early 2000s by the AC chimney in Green Phillip Earth. And what she found, this is sort of a typical meta analytic slide, is that really across the board, these studies showed that sleep deprivation impair performance. And you can see that a few of the individual studies that cross that zero line in the middle of the graph would have concluded the sleep deprivation had no performance. But if you look at them in the aggregate, I think the pattern is pretty obvious. And in fact, the magnitude that the the decrements of performance is profound. This is a standard deviation and an effort performance on average across all tasks. And if you focus in on a more clinical subset of these studies, it was about a two standard deviation drop of performance down to the seventh percentile of mean rested performance, which is. I think not right. I just want to. But a very reasonable criticism about these studies up until the time they were conducted in the early 2000s was that for the most part they had been done in very protected laboratory types of environments. And so, you know, you know, you've got an intern, it's just come off duty and I've been sitting down in a choir room and having read EKGs. Well, you know, you may be able to demonstrate some impairment there, but how applicable is that to the real world? You know, if you were actually on duty, and a patient is coming into the emergency department and crashing and the adrenaline is popping and you're taking immersion action, perhaps that's a very different thing. And that's like a simulated environment. And my group wanted to get at this through real world studies in a couple of different ways. And the first thing that we did was we conducted a national cohort study, which began way back in about 2002, where we got about 3000 interns around the country to report for us on a regular basis. Exactly how much they were sleeping and working and then we validated those things in various ways. And as the year went on, anytime they had a motor vehicle crash or stuck themselves with a needle or a scalpel or thought they made a matglare or various of their outcomes. We had them send us reports of those things and we validated them in various ways. And what we found actually pretty consistent with the tracking literature and the literature from across other fields that have looked at this. The rates of motor vehicle crashes were roughly doubled on that post called day as compared with the rate for the same person on how the day is to be. Excuse me, so called within subject's type of design where you serve as your own control on days when you're not close call versus days when you were close call. Likewise, the risk of percutaneous injuries, needle injuries and scalpel injuries was increased by about 60% on that post called day as compared with yourself on other days of the week. At the same time that we were conducting a cohort study, we also wanted to really delve into what the impact was on patient safety, not just resident safety. And to do that, we thought that the best approach was really to conduct a randomized control trial. And so in two intensive carry-in, it's over brigham and limits. The medical ICU and the cardiac unit back in about 2003-2004. We conducted a randomized control trial where we had the residents work every other month. We had the unit schedule under traditional schedule, which at that time was a Q3 schedule every third night they were on call overnight. And then we compared that with the schedule where essentially what we did is we took the traditional log shift that would happen on an overnight call and split in the nap between two interns. So just to explain that if you panel A up here is the traditional schedule and B is the intervention schedule. Really all we did on this intervention schedule is we took interns C's call on this Wednesday to Thursday shift, that first log shift that you can see on the graph. And we gave the first half bits of intern four down here in the second half goes to intern three just to split in half with about an hour for them to kind of overlap and do a little bit of a handoff. And that's it. I mean very simple type of intervention. But one that is consistent with principles of sleep and circadian biology and you're not keeping people awake for too many hours to row and you're sort of graphically cycling in and out of days and nights, which is not ideal. But maybe as good a solution as there is to dealing with nights. And importantly, I think in looking at this and this will be in contrast to some of the more recent work that we've done that I'll get to in a little bit. You'll see here that we were able to keep the staffing on the unit. It's really quite constant by this really simple design where we just took this log shift and split between two people. What we ended up with is a situation where the number of interns who were present on the unit at any given hour of the day was the same in schedule A versus schedule B. It was just that there were more people rotating through schedule B and we're going to make that work. As we did this, we had research nurses who were monitoring the charts every day. We actually had continuous direct observation of the interns as well by position of servers who were following them throughout their shifts and taking down anything that they thought might have been an error that occurred in the course of their care. We also had access to a couple of different hospital systems. In other words, we did really comprehensive and detailed surveillance and look for adverse semantic layers. And then, anything that was picked up by any of those methods were subsequently reviewed by a couple of investigators who were blinded as to which schedule this came up on. And we're able to make pretty reliable classifications about whether there was an error and it's varied so far. We also had the residents hooked up to EEG while they worked so that we could get very direct measures about sleeping. They were using electro-ocular gram leads, which would be leads that are relatively near the eyes. And although you can't see it on this research subject because she's got a nice long hair that covers it. She actually had a full panel of EEG leads under her hair as well. As we rolled this out, we measured what happened to sleep in work duration. And I'm very surprised when we scheduled people to work about 20 hours a year for week. In fact, they did work 20 hours a year for week. We also found that they slept significantly more about an hour or more per night. And as they did that, the attentional failures measured by the EEG and the EOG fell by about 50%. So we know that from a neuro-behadriel standpoint, having them work fewer hours and getting them to sleep more actually did make an impact on physiologically how they were performing from an attentional standpoint. But the bottom line, I think, from a patient's safety perspective, what matters the most, I think, to all of us is that it did appear to work. So we found that the total rate of serious medical errors decreased by about 36% on the intervention schedule. And that included both the significant reduction in medication errors as well as a really dramatic drop in diagnostic errors that these residents were making when they were working on intervention schedule. Subsequent to the completion of this study, and really the completion of some similar studies by other investigators around the country, the Institute of Medicine was asked to formally review what resident work hours looked like. And to make some determinations about whether we should be doing anything different than we were doing at that time. And so the island does what it always does, which is to say that it convened a panel of extra children across fields that reviewed the world literature on this and held hearings and tried to put together some recommendations based on the best sciences available. And after a year long study, essentially what they concluded was that it was unsafe for residents to be working for more than 16 hours or really that sleep. At that time they proposed a couple of solutions, either building a truly protected period for sleep into the middle of a traditional 28 or 30 hour shift, or potentially just limiting work hours to a maximum of 16 consecutive hours. And then they also called for improved handoff processes, which will come back to in a moment, supervision and ongoing implementation research to understand what were the best ways to approach this problem and how could we do this in a way that we could really robust. And then the ACGME responding to these recommendations as well as its own internal review process, as I think most people here are aware in 2011 decided they were going to limit the work hours for interns, interns only to 16 consecutive hours, but they largely left the workstables for PGY2 as an above unchanged. They also called for improved handoffs and teamwork and standards for supervision and workload, although all of these later, what I consider to be infrastructural changes, were sort of big to be honest. There was not a whole lot of meat to them or a specification about what they meant by that. After these changes went live, I think there's a lot of concern in many parts of the medical community that maybe this was not such a great idea, and further study was needed to figure out if in fact this was a beneficial policy or not. And a couple of large studies, something that came out trying to take a look at this. The first was called the first study, which was a randomized controlled trial of a large number of surgical programs around the United States, I think there were over 100 programs that were involved with this. Where programs were essentially randomized to either be on a schedule that was adherent with the ACGME's 2011 rules, or to be on a schedule that was called flexible, which basically meant they were allowed to go back to, you know, the shift length could be as long as it was allowed in 2003 rules, in other words, up to about 30 hours or so. And what the investigators of this study found was that there was really no difference in either deaths or serious surgical complications if resident positions were on the flexible schedule versus the 2011 schedule. But a limitation of the study despite its massive size was really a little bit unclear what kind of power they had to detect the results that they were looking for. And in particular, there was I think concern that the measures of safety that they were using here were pretty distal from what interns activities are. They were looking at death rates and major surgical complications rates. I think you could make the reasonable argument that interns activities and operating room probably have relatively little bearing on either of those things. And there was no direct measure of intern errors or direct measures of the safety of the residents themselves, but that being said, you know, the still of the large and very influential study. Along a very similar line in internal medicine, the eye-compared study was carried out very similar type of its sign program for randomized to either flexible schedules or adherent with the 2011 duty hour schedules. And here they did do some measures of sleep and found that it was not significantly different actually other two schedules. And they found that like the first study, there was no difference in mortality on one schedule versus the other. Again, there were some limitations in I think both power of this particular particular analysis as well as the fact that there were no measures of resident safety. There was not a lot of detail about how schedules were actually carried out across programs and so forth, which is a difficult thing to do when you're doing a randomized control trial that's disbarged to control all those variables. But nevertheless in the wake of the eye compare and the first trial, as I'm sure everybody here is aware in 2017, the ACG and the essentially said, look, we don't have compelling data that the policy we put in place back in 2011 with beneficial or something to do away with the ship. Let's go ahead and do a way with the shift limit that we added place for interns and then we will go on from there. So what were the concerns? I think if you take a look at the broader literature that I've tried to present at the beginning of the talk here, I think it'd be fair to say that there is compelling data that sleep deprivation adversely affects the performance of resident physicians just like it affects the performance of everybody else from across other specialties in this. Specialties and disciplines. And so, you know, so what's the concern with I think putting a program in place that potentially limits work hours as a means of potentially improving patient safety? And I think one of the biggest concerns that is out there still, but was certainly present 20 years ago, was this notion of the handoff and the idea that if you shorten the shifts of resident physicians, one of the major potential consequences of that is you've got more transitions of care happening between people who are working shorter shifts. And we believe that those transitions of care are of themselves, has this fact there's pretty good data to support that notion. These are reports of sentinel events that are collected by the joint commission from hospitals all across the United States. And what you can see here is that miscommunications are at the top of the list of root causes of the worst of the worst adverse events that happen in hospital responsible for a fact about 65% of all their joint commission sentinel events. Have communications as one of their root causes. And again, there's short shifts. You're going to have increased frequency of handoffs if you think that handoffs and communication are real problem. Perhaps that's a very good reason not to shorten the work shifts of resident physicians. I would say that as I thought about this and our group really began working on this issue in earnest about a dozen years ago now, to me it didn't make a whole lot of sense that if we're worried about handoffs, we should continue to perpetuate long work shifts and we also think our has to this. You know, shouldn't it be possible for us to fix both of these things simultaneously? So in other words, put a safer work schedule in place, but also to address the issue of handoffs so that it's not quite the hazard that it was perhaps a baseline. And so we began to put together what eventually became the eye pass program where initially on in the medical residency program here in the PDF presses for here, we put together a bundle of interventions with the intent of trying to improve the quality of transitions. And the bundle itself was really pretty simple. We had to we had to understand in that none of us, including the trainees of that time, but none of us is faculty either never really been formally trained and how to do a good and high quality handoff and so we introduced training program. We also introduced an amonic, which was basically just a way of structuring the information at change your shifts. We thought a little bit about the verbal handoff process, you know, was it happening and acquired and protected space where people really prioritizing this transition as a time when you shouldn't be interrupted. And we wanted to make sure we had all the right people in the room. And then lastly, we worked with the IT folks here to build something into electronic health records that there was a tool to help in writing what we're trying to get the residents to do verbally. And as we put this in place, low and behold, we found really in a very short period of time a pretty massive reduction in both medical errors and injuries to medical errors, so called preventable adverse events. You know, there were a lot of limitations to study. It was pretty small, but it was a pre-plosed study. It was conducted just on a couple of units at Boston Children's. But it was sufficient. They can balance it became their preliminary data for the IFA study, which was a nine center study in children's hospitals across the US and Canada, where we sought to get the input of medical educators and hospitalists and health service researchers and others to really take this basic idea that we developed and make it more robust. And then drill it out across institutions. And so we, you know, we refined our mnemonic, which became the IFAAS mnemonic, just an organizing framework, grab information on it, be passed off, which was based on the literature as well as feedback from our initial efforts. And then that was framed within really a change management framework, where we thought to ourselves, you know, it's one thing to kind of introduce into training program or in mnemonic, but if you really wanted to stick and want this to be something that's real, there's a lot of other stuff that has to go into that. You have to have hospital leadership. You have to think about how I'm going to redesign these handoff processes across different corners of the institution, and how am I going to do some faculty development so that I get people to provide observation and feedback to the residents that are doing it and reinforcing, you know, how are we going to, could we have some kind of campaign to push this after the hospital? All these different pieces that really go into any type of possible change process. And we put all that stuff together and called that the IFAAS handoff bottle. And then as we rolled this out and sought to measure what his impact was, we really went back to the same type of surveillance for serious medical errors that I talked about in our prior randomized control trial of regular women. And then a daily chart review looked at hospital incident reports. We debriefed the residents after they were coming off call and really sought to be as comprehensive as possible in capturing medical errors. And then anything that was captured those processes, again, we did the same type of a two-step process where those were then rated by individuals who were blinded as to when and where they were captured that made judgments about their severity, preventability, and so forth. And as you was paired with some direct observation to look in this case at what the processes looked like to really directly measure the quality of verbal handoff processes, the quality of written handoff processes. And then lastly, we had research assistance following the rest of the around real time doing so called time motion analysis where they were looking at, you know, sort of how are they spending their time. With the particular focus on was, you know, the worry here was that as we introduced this process for handoff, perhaps we just make care a little less efficient for them. So they're spending a whole lot of time talking about patients, but they're spending less time at the bedside, more time at the computer, not doing the kinds of things that we wanted. And so we tracked those things very carefully by having somebody follow them around and document moment by moment exactly how they were spending their time with the course of 24-hour cycle. And what we found as we did this was that, you know, first of all, the quality of information transmission improved quite significantly across all five categories that we were most interested in. You know, I think for those who were a little bit more critical, I think it'd be fair to say that we certainly didn't get to the levels of adherence with the process that we would have liked. And how many of these elements you'll see were just happening 50% of the time, let's say, even post-intervention, but it was certainly an improvement across all the categories. And from a patient's 80 standpoint, this was associated with just as in our pilot study, very significant reductions, both in the overall rate of medical errors, as well as most importantly, the reduction in preventable adverse events, which is injuries to the medical errors by about 30%. From a time-motion standpoint, we found that really there was no impact on resident workflow after this got up and running. You know, certainly there's some time and effort to do the training initially and get people to understand how to do this. But after it's up and running, the duration of the verbal handoff itself was unchanged, the amount of time spent at the computer, either updating the written handoff side can in order to general was unchanged, and the amount of time at the patient and family bedside was unchanged. Since that study came out in the around 2014, we've been working a lot on just how you adapt this program for other groups as well. This is a study that involved nurses share of hospital children as well as nurses at a couple of other medical centers where we found that as we interviewed side pass, the rates of nursing reported handoff, that care failure is felt very significantly, both in change of shifts and change of location of patients. And then, excuse me, we received funding from agency for health service, and quality to roll the program out in 32 more hospitals, which included adult hospitals as well as pediatric hospitals, looking at internal medicine services, a couple of surgical services, a lot of pediatric services trying to understand how this played out at community settings versus academic settings. And again, much in line with both the pilot study and the nine center study, we saw that as the course of the program went on, rates of adherence with I-FAS rose quite significantly, and that was associated with market decreases in patient safety problems. So we saw a reduction of about 47% both in minor and major harms related to handoff failures, according to resident monthly reports. And then lastly, on this topic, we have, as we've expanded into the I-FAS Institute and continue to collect data, and I think, as we've gotten a little bit better at our processes for rolling the program out, the improvements look even better, where we're now seeing about a 76% reduction, the same types of medicine we're seeing before. And then on one final nod, as we began to develop the communication program, and think through how to improve this infrastructural process within healthcare, it occurred to us that really a lot of the processes that we developed to improve communication amongst ourselves were also relevant for our communication with patients and families. And we do a better job of being clear with them, making sure that we were all mutually engaged in patient care and have a shared understanding of what the plans for today are going to be, and what the goals of care can be, and so forth. And so we began to work with a number of patient and family collaborators, as well as health literacy experts and others, to adapt I-FAS, for patient and family, since we did that, as published a couple of years ago now in the British Medical Journal, we found that we were able to achieve significant additional reductions in progress. And additional reductions in preventable adverse events, harmful errors, even in agricultural hospitals, it was already using I-FAS for a resident hand-dose. Okay, so with all of that, really has kind of a long background up to the present study, and sort of where we are now. You know, the roster study was an effort to do a multi-center version of what we had done 20 years ago, a pregnant women's, in other words, to roll out a reduced work hour schedule for resident physicians. But this time, in a context where we were controlling, you know, we had implemented a hand-off program across sites and sort of dealt with that as a potential issue. And then, um, it's a bit of a multi-center way, so that we could try to understand whether, in rolling out what ideally was intended to be a tightly controlled work schedule intervention, we could achieve reductions in adverse events and improvements in patient safety, similar to what we achieved the first time around in this program. And so the basic design here, which was funded by NHLBI, a five-year multi-center randomized cross-over trial, and the basic design, which is a little bit confusing to look at, but it's displayed at the bottom of the graph here, was just as, you know, just describe it verbally, I think it's easiest. It was just a cross-over design where for a year, each individual schedule worked on the traditional schedule, and then the following year after Washington period, they flipped over to the intervention schedule, or vice versa. The program for randomize that happened, well, with intervention schedule first, half minimum with traditional schedule first. And then, as we did that, in very much the same vein as we had with all the other studies, we did this very intensive surveillance, including direct observation, our time that they're working to understand what happened to rate them at a glare, and that sort of put all those events through a rigorous process to adjudicate them. A couple of nuances of this project that were different from the prior one that we'd done, the first is that this was staffed by units, you were staffed by PGY2 and higher residents, each with pediatric ICUs, who were studying in this case, which traditionally had just been staffed by junior residents rather than by interns. And another nuance that's important, and I'll show you some data on it a couple of minutes, is that while we were able to hold the handoff process, these pretty constantly across these sites, a second element that we attempted not very successfully to keep unchanged was resident workload on these two different scaffolds. So unlike the effort at the Brigham years ago, where with the help of the residency program there, we were able to assign an additional intern to the mix and therefore keep workload in any given time hour of the day is same. In this case, because the baseline infrastructure of each of these programs were sufficiently different, we asked each program to try to do its best to keep workload constant. But as you can see from the demographic data here that was not totally successful. So these are the results of the study just looking at patient unit characteristics on the two different schedules, the EDWR as the traditional schedule extend duration and the rapid cycles are intervention scale, the second column there. And what you're seeing is that the patient characteristics, kind of all the standard stuff that you look at was was really quite similar across the two schedules, but the very last line that it's bolded. The average number of patients that resists for care and for actually increased by 25% on the intervention schedule unfortunately. As we looked at the impact of the intervention schedule on work hours and sleep and so forth, for the most part, at least by our by our intermediate outcomes, things really looked quite positive. Work hours were lower with the intervention as we intended weekly hours and patient care were lower with the intervention. The amount of sleep by a couple of different measures, both wristwatch, worn devices, the activity as well as by diaries was significantly improved on the intervention schedule. And then sleep in our behavior performance, behavior performance was was improved as well. So we had in this case, we didn't have the residents hooked up the EG's, but we did have them doing the reaction time test, the PBT periodically traps the shifts and we also have them filling out standardized validated surveys, looking at subjective sleepiness. And all of those things looks significantly improved. We also found interestingly that as their number of neurobehavioral lapses increase, they were making more objectively documented errors in the real world as well. We were able to assess the validated measuring and demonstrate that it meant something. But despite that, despite the fact that sleep improved, work hours improved, neurobehavioral performance improved, the overall results of the study were disappointing. So the main results of the study, the thing to look at is sort of lowest blue dot, the overall measure of resident position related serious medical errors, which was our primary outcome measure for the study. And what you're seeing is that it actually increased. We carried out less safe as we rolled this program out. The odds of relative risk of a serious medical error occurring increased by about 50% 1.5% versus baseline rate. And then down below in red is the sort of overall unit wide measure patient safety, which looks very similar fundamentally. I think that, you know, sort of while this overall result was disappointing, it's also important to point out that there was just this massive variability here across sites. There's really not sort of a consistent message where you can see that site A, things actually got significantly better under intervention schedule. Sites B and C was largely the same. And then D E and F got work significantly worse, which ended up driving worse thing overall. So, you know, if we're able to implement a schedule that improves sleep and prefers improved air behavior performance, then why did things get worse? And we were asked by our data safety monitoring board to delve into that to figure out kind of what had gone wrong, in a sense. And we began to notice a couple of things. The first was this, which was what you're seeing here is on the left, the traditional schedule and then on the right, the intervention schedule. What was the number of ICU patients being cared for during the day by each resident division? You can see that in all six of the sites increased at least a little bit. The other thing to point out here is that the line that's green is the site where things significantly improved. The two lines in blue are the sites where there was no significant change in patient safety as we implemented our intervention. And the three sites in red are the sites where patients safety significantly worsened as our intervention schedule went live. And so we began in looking at these kinds of preliminary data to develop a hypothesis that maybe there was an interaction between workload and the impact of our intervention schedule, such that it seemed like it was working better for workloads low, but it's workload increase. Then it became a problematic thing. And sure enough, started doing some secondary analysis to look at that. And we found that there was a very strong relationship independent of work schedule in the number of serious medical areas being made by residents and the workload. Which, you know, if you think about it, take a half step back. It's really not that shocking at all. There's certainly is a ton of literature on the nursing side of things that their workload has a big impact on the frequency with which they make care. But there's really shocking them work that's been done looking at either resident conditions or for that matter of attending conditions and how patient safety is related to workload. And then when we conducted secondary analysis controlling for workload, we actually found that it looked as if very much like our prior studies. The safety of our intervention schedule was beneficial. Things got better. We intervened from a work hour standpoint. But this is the secondary analysis. I think it has to be taken with a grain of salt. And our current work is really think a little bit harder about this workload issue and how do we properly control this and control for this as we move forward with these types of interventions. So I'll just end with one final couple of slides on a final study, which was just to say that just as we had extended the work that we did way back in the early 2000s on randomized control trials trying to understand the patient's safety impacts of different schedule intervention. We've also continued to do our natural cohort study where we've had cohorts of residents from from really each year from initially from 2002 to 2007. And then we had sort of a period of time where we're collecting these data due to loss of funding. But then again from 2014 to 2007, they have continued to collect these data on resident work hours, sleep, and resident occupational safety as well as their reported patient safety outcome metrics. And then we've done some analyses recently comparing just how things have changed from that initial cohort from 2002 to 2007 to the more recent cohort after the 2011 policy changes one by. And what we found is that in that analysis actually things look like they really have been benefits across the board. What you're seeing here is the relative risk of various adverse events occurring post the intervention of these policies and medical errors, preventable adverse events, preventable adverse events that led to a patient death, as well as all of the measures that we have at resident safety, go to vehicle crashes, near crashes, and percutaneous injuries have really significantly improved over time. Of course, it's a cohort study, you can establish causality with any of these things. But it's interesting that in general those things have improved quite significantly as years have gone by, specifically preversus post the policy. Even though it's also pretty clear that the implementation of these policies has been problematic and challenging, and I think confounded in many cases as in our trial with issues around workload and just other systemic infrastructure elements. So with that I'll stop and I'm happy to take any questions or thoughts that you have. I think it's fair to say that from a roster's trial resident positions randomized to an intervention scheme, but the limited extent of the work shifts made more errors in the care of their patients, contrary to our hypothesis, but there is wide psychocycability. We found that sleep and nerve behavior performance improved on this schedule, so sleep deficiency certainly can't equine our results. We also found that on the intervention schedule, resident workload increased by about 25%, and secondary analysis suggested that this could be the reason that the intervention did not have its effect of the fact. And it certainly makes me wonder that in those larger trials that it looks at the impact of the policy across the country, what impact might workload and some of these other elements have in explaining the reason that we've not been able to effectively tackle this issue as a country. And I think the implications of future work needs to address this stuff and really think about it very comprehensively. It probably doesn't make sense to think about work hours as an isolated issue. It has to be thought about in the context of the design of the units and design of our rewards in general. You have to think about workload, you have to think about handouts, you have to sort of put all these things together into a package to save. And that's obviously a challenging thing to do. So if that will stop and thank my many collaborators and funders. Well, Chris, let me begin by thanking you. And I think for those who haven't seen your work or seen you speak, I think everybody now is aware that you really are a pioneer of international authority in this world. And it wasn't an easy area to delve into, not necessarily received well by colleagues, I would say particularly surgeons in the early time of your work. And as you can see, this is multiple papers in the New England Journal of Medicine. So the well peer reviewed. So your international recognition is just a surge. What you really have done is taken a data from truckers, which doctors, I'd say surgeons in particular say, well, we're not truckers. And you brought this into analysis scientific analysis in our world, not in the world of trucker accidents in the world of medical errors being performed by physicians. And it's really when you started the Southernist thought you were like doing to fairly to be able to even get in and cooperate with these with these studies, but it really did require a scientific approach. And we're going to do that. You had to get collaboration from multiple departments, multiple institutions, that actually changed the way they work, the change the way they care patients, changed the way that their staff are assigned. Pretty massive undertaking. And you did that. And honestly, that was required to convince so the old guard and looking at the screen, there's only a few of us left to walk uphill and forth. You just know both ways to school with a school bus and we're proud of it. But seriously, the training that we felt honored to have, but in fact, we endured being on every other night call, every third night call. No concept of going home the night after call was we didn't view it at times inhumane. We were privileged. And when it was taken away when those of us who had done it, you know, were forced to train people in a more modern fashion, there was, there was reluctance. There was almost some sadness about they're not going to get to experience what we got to experience. But you were work along with others has prevailed and convinced. We're all in the world. That we, although we're doctors, and we may not be truckers, we still are humans. And we still need to sleep. And it could, it could have shown that. I must say, although a few people question your data anymore rigorously, then you yourself have questioned and pointed out the challenges and deficiencies. There are still conversations in the hallways of our national meetings, mostly by the more senior folks. Well, yeah, there may be less errors in the way they measure things, but patient care isn't the same. And the word that most frequently gets tossed around is ownership. When we watch them, they don't seem to feel responsibility or ownership for the patients. And I don't even measure that. But it might have implications beyond the training period. So we're going through our exchange and training and learning how to be a true physician. The experiences and values that we engender become part of us. And we have questioned whether the current generation of practicing physicians in the outside world in in primary care practices, in specialty practice, whether we have become partly because of our social specialization and American medicine, partly because of the training experience, feeling less ownership of our patients. And I have seen from personal experience through family members and even myself when I've had some health challenges. You do have to advocate yourself quite a lot to get noticed, a little more than it might have been in the historic medicine. And I don't know if it's possible to study the connection between the safety that we've engendered in the training and the way that we sort of are as a profession in healthcare in these days. How do you respond to that and how would you even approach studying it? I think it's an important point, Steve. None of us want our doctors to be disconnected from us as patients. And I think I think issues of ownership and feeling that type of responsibility and devotion and passion about providing I quality care is really essential to medicine. I personally don't think that that, I think that there's a little bit of kind of conflating sometimes of that issue with the work hours issue. I don't think you need to work 90 hours per week in order to be really committed and devoted and connected to your patient and to feel that measure of ownership. And while on the one hand, I understand the argument that we trained in a mirror that where that kind of became part of parcel of what we do and perhaps we see that fading a little bit today. Now I would agree that medicine has become a little bit less personal now than perhaps it was even when I was in training 25 years ago or so. And sometimes with electronic health records and just the pace of care where the length of standard hospitals is short and so much patients are in an outbreak rapidly. And the complexity of care will be typically have, you know, half dozen specialists who are involved in care rather than being sort of the one doctor that's sort of doing everything. You know, all that can lead to a lack of ownership lack of connectedness with families. And I do think we need to be very tuned to that and make sure that that does not degrade either Karen, general or the patient doctor relationship, which is a tricky thing. But I personally don't think that that's a work hours issue so much as a, you know, coming together many forces and medicine that have made that more challenge where they used to be. Now, I, I tend to agree with you. I think that a lot of changes have happened much for the benefit. So the, the training needs on the screen and many young faculty member may have never drawn a blood culture or obtained a blood guess or put an IV at three in the morning. So I'm going to put IVs in and that's actually what we did most of the time of the night when we were basically indentured servants, the hospitals used trainees. We've got the same salary per week, regardless of how many hours we worked and therefore we were free. And so the concept of a phlebotomy service, I mean, when I trained, it was one blood droid day at six a.m. you know, just to eight a.m. that went around the hospital. And everything after that, that, you know, the residents and the students drew blood. We put in all the IVs. There wasn't an IV team. You know, so that is not necessarily productive work in terms of sort of doctrine or knowledge base. But it may have brought us closer to our patients. So it's really, it's really hard to know. I'm sure others have questions and comments. Of course, I just like to echo Steve's sentiments. Thank you for being here and sharing this with us. Yeah, you were a bit of a pry in the surgical world. When you started this surgical work hours and I remember myself kind of responding to the 80 hour work week. What are you talking about? I happened to have that conversation in front of people who weren't involved in medicine and they kind of looked at me perplexed and said, you know, gee, the rest of the world, it's a 40 hour work week. You got 80 hours. You still can't train them and don't really turn things around on me. I am encouraged that that that some of your measures safety measures have improved over time and and the frustration of some of those you can't connect to to work our schedules. Well, I think think what you're underscoring is that it's a complex problem and a lot of different with a lot of different things impacting on that. Yeah, I worry like Steve has mentioned about what values we instill in our trainees, how we can promote efficiency, how we maintain that connection to our patients. But the fact of the matter is what's your data shown is that we're moving forward. It just means that we have to approach it on multiple different fronts. And sure, sleep was one of it and got got a lot of attention and handoffs are important things, but certainly there's plenty of work yet to be done. We don't live in the same world today that we lived in five years, 10 years, 20 years ago. Okay. In this last world, we haven't lived in the same same world for 12 months. So in any case, applaud your progress and thank you for continuing to push us forward. Thank you very much. Chris, this is Carrie Butler. I have an observation that was looking for some comments. I think also another thing that's happened is many services have gone to an attending presence in house 24, seven. Certainly our department has done that. Radiology is now doing that. Can you comment to see how that may potentially impact the patient safety issues if you have that attending presence and then also have you studied attenings with the same rigor that you have the residents as far as sleep deprivation because I think we're in a unique situation. Yeah, I think those are those are great questions. You know, the quick answer is that there's a little bit of data about both of those things you're asking, but it's much more limited than what we know about residents. I think that residents have been a focus of a lot of the research in large parts because historically they've had the longest hours of anyone. But that being said, there certainly are, you know, a subset of surgeons work very, very long hours as well. Still is, you know, well. And there's been a little bit of work that's looked there. So so we did a retrospective study at Brameman women's looking at about 10 years of surgeries a few years back now where we tried to understand whether surgeons who were doing operative elective cases, elective cases rather the morning after having been on call the night before had a higher rate of complications than, you know, if they hadn't been. And what we found in that study was that being post call all by itself didn't appear to convey any increased risk. But if you then drilled a little bit deeper and look at those situations based on, you know, the data that we could get on operative times overnight. If you look at the cases where surgeons were actually operating on the nights before when they were on call on those mornings, the mornings after those nights where they presumably gotten less sleep, they did appear to have a higher significantly higher complication rate. So the data fundamentally I think similar to what we saw with the residents. You know, and certainly looking at the broader literature on sleep and performance, there's no real reason to think that experience or age, particularly protects you from from those types of effects. Unfortunately, not something that you really get better at with practice. In terms of whether attending presence in the hospital matters, matters more, the literature there has been pretty thin as well. And, you know, the couple of studies that have been done have not really been able to establish a whole lot of obvious benefit of having the 24-7 presence. I would say is a quick answer. But there's a lot of indirect data that it may be helpful. I mean, certainly there's plenty of anecdotal cases you could come up with with with attending catches something that gets missed. And then there's also data on ICUs, for example, that are staffs 24-7 that appear in some cases to have better than those that don't, even though that's a little bit less tightly tied at the attendance presence per se. So I think it's an area where there's more work that's needed, but at least there's some suggestion there that it's beneficial. Well, we're at the offer. And I really want to thank you. I think the people on the screen who weren't familiar with her work are people who owe you and people like you that have gratitude for the widespread adoption of IV teams and phobomy teams and things that have removed the indentured, serviture portion of training. And I really focused on their health and well-being as well as the health and well-being of our patients. So thanks so much for spending your time with this. And although we've tried to limit the hours in which we in our department have activities outside of work hours, unfortunately 7 am is still surgical work hours. So thank you for getting that. Thank you. Thank you.