 I think you're on mute, Mark. Can you hear me? We wait till about 1203, 1204. Oh, good. I like your tie. Thanks, I was saying, you know, we gotta wear it at least one day of the year, right? So why not today? Good afternoon, everybody. It's Mark Sigler chatting. Welcome to the next to last talk in this year's annual series, lecture series on ethics and the COVID-19 pandemic, medical, social and political issues. But believe it or not, this is the 26th talk in that series. Our speaker today is William F. Parker, MD, MS, who's an assistant professor of pulmonary and critical care medicine and is also an assistant director of the McLean Center for Clinical and Medical Ethics here at the University of Chicago. Will received his bachelor's degree in physics from Williams College and is MD from the Pritzker School of Medicine. Here at the University of Chicago, Will completed his residency in internal medicine in fellowships, both in medical ethics and in pulmonary and critical care medicine, earned a master's degree in public health and currently is completing a doctoral degree, a PhD in health services research in the Department of Public Health Sciences. As a physician scientist who works clinically in the intensive care unit, Dr. Parker's research focuses on the allocation of scarce medical resources. He is specifically interested in applying advanced empirical methods to design allocation systems for multi-principled ethical frameworks. Will began his research career as a medical student working with us at the McLean Center under myself and Lainey Ross on issues such as deceased donor organ allocation and with their guidance, he secured an NIH K08 career development award from the National Heart, Blood and Lung Institute to develop a novel heart allocation system. Today, Will Parker will be discussing scarce healthcare resource allocation specifically in the context of the COVID-19 pandemic. During the COVID pandemic, Dr. Parker was selected by the University of Chicago Medical Center Hospital incident commanders to form and chair the UCMC COVID-19 Ethics Resource Group where he has developed critical care allocation and CPR guidance protocols that have been adopted by multiple other institutions in the city of Chicago. This group of the city of Chicago group has forged strong ties with the local community and has worked to incorporate community input into pandemic decision-making. Will Parker also served on the UCMC COVID-19 Vaccine Allocation Committee and is writing on the COVID-19 vaccine has been featured in health affairs, JAMA Health Forum, USA Today, The Washington Post and I Could Keep Going. Dr. Parker's talk today is entitled Empirical Assessment of Scarce Healthcare Resource Allocation Protocols. It's a delight to welcome Will Parker to our program. Dr. Parker. Thanks so much, Mark. It's a really, it's an honor to be here as somebody who's trained in the McLean Center for so many years to be giving a talk in the seminar series is a true privilege. So I'm gonna jump right into it. I got a lot of slides and a lot I wanna share with you guys and get to the discussion as fast as I can. My only disclosure is that career development award that Mark had mentioned earlier. So what I hope to do today is define the concept of absolute scarcity of healthcare resources. It's actually a little trickier than most people assume. Consider the ethical frameworks that currently exist for the allocation of absolutely scarce medical resources. We're not gonna spend a lot of time debating the relative merits of one framework or the other, but it's important to get that sort of background and context because what really the meat of the talk is about the central role I believe empiricism plays in developing one should play in developing these practical allocation protocols. And really point to a lot of examples where the empirical methodologies used were insufficient or inadequate to actually design protocols that were sufficient to meet the ethical frameworks they were created to fulfill. And we're gonna focus specifically on two examples from allocation of scarce medical resources in COVID-19. All right, so what is absolute scarcity? So it's important to consider two concepts or the difference between absolute and relative scarcity. I think helps define it. Relative scarcity is something we're all too familiar with in the broken US healthcare system. There's structural inequities in the access to healthcare based on racial and socioeconomic lines. This leads to preventable violence and excess deaths. The policy and politics of relative scarcity are very hard, but the ethics are uncontroversial. Any basic clinical medical ethics analysis would lead to obviously these relative scarcities of access to healthcare should be overcome. And here's an example of relative scarcity in the COVID-19 pandemic. Access to high quality ICU care. Many of the community hospitals throughout the city were getting overwhelmed with COVID-19 patients. To be honest, they would have trouble taking care of a COVID-19 patient with severe ARDS in the best of times just because they require such intensive levels of high quality ICU care. But in the height of the surges, both in the spring and the winter, they were unable to transfer their patients out to higher level tertiary care center. And why was this? Well, hospitals weren't accepting them. There are structural financial barriers that don't incentivize hospitals to accept transfers from patients, particularly those who don't have insurance or who have Medicaid. Instead, they would rather preserve their capacity to do lucrative procedures on the well-insured and wealthy. So this is an example that we wrote up and described in a health affairs blog with Harold Pollack who's also associated with McLean Center and Kelly and Kelly who's an incoming PhD student at the university of structural violence on across racial and socioeconomic lines mediated by relative scarcity, right? There were ICU beds available. We weren't out of them in the city of Chicago at any point. We never hit crisis standards of care or absolute scarcity, but certain people just didn't get access to it because of market failures. So that's very different than what I'm gonna spend the rest of the talk talking about, which is this concept of absolute scarcity where there's a fundamental hard supply limit. We can't fix this with some more egalitarian, clever market solution to properly align incentives and increase supply of whatever the healthcare resource we're concerned about. There just isn't enough to go around. And then therefore violence from lack of care is unavoidable. Some people just aren't gonna get it. And you have to make value judgments and this is where the ethics starts to get controversial. So what's the classic example of an absolutely scarce healthcare resource in the United States to see donor organs? There's over 100,000 people on the wait list that kind of went down a little bit because of the opioid epidemic. And there's something, there's like 33 people on this wait list who die every day. So it's kind of over one an hour die waiting for an organ daily. And so this is a chronic absolute scarcity of an important healthcare resource that I got my, started with in the McLean Center what got me interested in studying clinical medical ethics to begin with. But unfortunately we have absolute scarcity of oxygen right now in India. As I'm sure you've read the horrible news reports, there, this example up here in this top left for this picture option not available cannot admit patients, right? There's, they're getting oxygen supplies that are roughly half of what they need to support patients in their hypoxic respiratory failure. And although this absolute scarcity is of course downstream from a poor infrastructure and other oxygen could be made not absolutely scarce in this instant in India it is. So absolute scarcity can also be temporary, which is also another confusing, but that's I think the best way to think of what's going on over there, they have to ration and it appears they're doing it mostly by ability to pay in a combination of that and first come, first serve. All right, so how would we approach the rationing of absolutely scarce healthcare resources if we find ourselves in a situation, heart supply limit, we have to decide which patients are gonna get it, which patients won't, who's going to live it or die. This is a paper, a table that I've summarized from a very nice Lancet article 2009 by Gobind Persaud principles of allocation of scarce medical interventions. And I think many people in this audience will have problems with particularly the way Gobind framed principle X or Y, but I think it's a nice article because it sort of lays out a big range of ethically salient values and principles that one could use to construct an ethical framework. The first category is this idea that everyone's a human being and should be treated equally. That sort of has two mechanisms why either randomly assigning the resource or first comes first service fair in theory, right? But in practice, usually the rich and well connected will cut to the front of the line. Prioritarianism is this idea that we should favor the worse off. It's sort of the rule of rescue that motivates us allocating more resources towards the sickest first. That's actually what's used in liberal allocation with the meld score. So the sickest patients are at the top of the list. It turns out that those are the same patients who have the greatest benefit from liver transplantation but that's not necessarily true. And so a lot of this article, Govind talks about how overemphasis on the sickest first could lead to very inefficient situation where saving far fewer lives than would be optimal. If you sort of ignore the relative benefit each patient's getting from the resource, which of course is the next tier here, right? With utilitarianism or maximizing total benefits. And then finally there's this category of promoting or rewarding social usefulness, pretty sticky concept there, right? What does that really mean? But played a heavy role in vaccine allocation in particular. So it's the idea of physicians in other frontline healthcare workers who are on the, taking care of COVID-19 patients deserve reciprocity or payback for their time serve. Or the other concept was we should vaccinate all the healthcare workers and people who are central to pandemic response to keep society functional, right? That's instrumental value sort of paying it forward. I would argue that's sort of like a maximizing benefits arguments maybe should be in the category above. But what I think this article does a nice job of is sort of laying out a wide range of options and pointing out that there's gonna be conflict between ethical principles and values that I think a lot of stakeholders in our society would point to is important. So maximizing benefits would mean you prioritize people based on improvement in survival with treatment. That's in direct conflict with the concept of treating people equally. Young, fair innings is this idea that everyone deserves to play their nine meanings of baseball live to be 65 or whatever we define a complete life. And therefore we should allocate more resources to the younger people compared to the older people. That's in direct conflict with treating people equally. And then often the COVID-19 pandemic with vaccines would be in conflict with maximizing benefits because obviously the older people have such a higher risk of COVID-19 death that vaccinating them first preferentially would maximize benefits to society. So you're not gonna escape conflict in almost all scenarios where you're gonna have to allocate a scarce resource. And therefore, we need to develop a multi-principle framework to somehow balance these principles. You can't just put all your eggs in one basket. There's gonna create serious ethical violations and some other important principle. You have to deal with nuance and complicated balancing of ideas that needs to be formalized somehow in a protocol. And I think one good example where there's sort of an immediate problem with the concept of maximizing benefits is you have to talk about are you trying to save lives or life years? So here's an example of three patients of varying ages with different expected life expectancies multiplied by their probability of survival of a hospital discharge. And if you were trying to maximize the number of life years saved you would allocate to the 28 year old. If you were trying to maximize the number of lives saved defined as survival to hospital discharge you'd allocate to the 80 year old with advanced dimension in the center. And so therefore, you can't and both of these ideas have some purchase with most people, right? So to totally ignore other ethically important principles will ultimately get you into trouble. And you're gonna have to construct frameworks, multi-principle frameworks. And this is the one that Govan came up with. You know, I definitely don't agree with all of this, right? This, you know, there's this idea that this I don't really like that I'm younger we shouldn't allocate so many resources to very, very young children because we haven't like invested that much as society in them. So that's why like the maximum number of resources peaks out in between 15 and 20. But, and I think Govan would say he would, he says that he's lots of things he'd like to change about the system. I don't know if that's one of them, but this is an idea this what I wanted to show this because this gets at the necessity complexity that's required when you actually work out your protocol in order to make one that won't create major ethical objections to one of these key values that people dear. And here's an example of another multi-principle framework the first one I worked on with, with Dr. Ross. It's one for deceased owner kidney allocation. It's a lexically ordered or a framework. So there's one principle that's more important that's equal opportunity that everybody has an equal chance of receiving a deceased owner kidney transplant but it's supplemented by fair inning. So within that constraint that equal opportunity we give the youngest kidneys which is sort of a proxy for the best kidneys to the younger patients. And therefore we maximize the probability that these people can live play their nighting as a baseball for younger people who are there who are worse off for going into end stage we know it sees as a younger age. So multi-principle frameworks are necessary important and also necessarily complicated, right? So Govind and Zeke Emanuel revamped their paper for the COVID-19 pandemic. You guys probably all read this one back in the spring in New England Journal as you'll notice it's very similar to the Lancet paper. The one thing that I thought was kind of funny is that the concept of prioritizing younger people fell way down used only when it's, you know aligns with maximizing benefits basically has no weight in their recommendations. And I wonder if it has anything to do with the fact that they all both got 10 years older between the two papers or not. But I'm not sure, but you know I think this is again a good way to think about it a good starting point to start to think about this problem and the necessary complexities involved. So how do we go from these principles to a protocol? And so, you know there's this idea that you could do this without data, right? You could just, you know you've done all this really careful ethical analysis you've developed a normative ethical framework you've balanced your ethical principles beautifully. You know, you set up the correct electrical constraints and created a rigorous philosophical argument for that framework, right? And then the protocol should just fall out, right? But this only works if there's consequences are sort of irrelevant, right? If you know, if you're if you don't really care about maximizing benefits and you know you've made an argument that a lottery a pure lottery is the only way to go then I guess you don't really need to look at data but is if you're going to invoke any other principle except one, usually you're gonna have to you're gonna have to look at data and empirical work is gonna have to be directly involved in developing the protocol and the sort of iterative process. And so I would argue with our equal opportunity supplemented by Fair Innings the data of that situation was intricately involved with the framework we've developed. The fact that the age distribution of deceased donor kidneys and age distribution of candidates on the wait list was critical to the actual protocol we developed. But I think I would actually go further and say that the data, the empirical outcomes may actually once you run the protocol a little bit make you think twice about your ethical framework, right? You find some results based on a simulation model for example of your norm of ethical framework and it wasn't actually that the protocol you derived wasn't representing that framework. It's just that you didn't really realize what that framework would really mean. And perhaps the weight on one principle or the other was too high or too low. And so that's the more controversial point. I think everybody here would agree with this point but whether or not these empirical outcomes should actually influence the norm of ethical reasoning directly is a point I would like to an assertion I'd like to make and hopefully some of the examples we'll bring that home over the course of the talk. All right, so crisis standards of care, what is that? This is a picture of Memorial Hospital on August 31st, 2005. Several days after Hurricane Katrina hit the city of New Orleans causing levees to fail and massive flooding. This point that hospital lost power they just had a backup generator and they could take care of I think roughly it was like a third of their ventilated patients could continue to receive mechanical ventilation. So they literally had to decide who was going to live and who was going to die. And this tragedy as well as the influenza pandemic and what influenza pandemic led the Institute of Medicine to develop the concept of crisis standards of care. And just to make this more specific to go back to our example of these three patients it's this horrible tragic choice literally having to decide which one will receive care and the other two will not receive life support and die instantly which unfortunately happened in that situation and was always a threat to happen in the U.S. during the COVID-19 pandemic. So what's this concept? How do you formalize this situation? The Institute of Medicine called it a crisis standards of care which is when a disaster either an acute one like Hurricane Katrina or a pervasive one like a viral pandemic causing respiratory failure leads to a substantial change in the ability of healthcare operations. Basically disaster causes absolute scarcity necessitating rationing, right? You're past the point where with different contingency measures you could really theoretically take care of anyone. We're not talking about relative scarcity issues anymore we're in the zone of absolute scarcity of life support therapies. So what practical protocols have been proposed throughout the United States? So Gina Pisticello, former McLean Ethics Fellow wrote a really nice article along with Mark and other co-authors from the University of a systematic review of all the ventilator protocols. This was back in the summer of 2020. At the time only 26 states even had a plan about what to do despite the pandemic still being still raging and this always being a threat and getting very close to crisis standards of care in various states throughout the entire pandemic. And the documents varied widely. Here's an example. I think the first time Gina made this map basically every state was a different color because they're so unique in their own way. Everyone put their little wrinkle on their protocol. Again, without necessarily reference to a specific ethical framework that they were derived from, right? Just various manipulations to the actual algorithm that patients would be sorted by. But predominantly one thing that carries through this is this idea of saving the most lives by rank ordering patients according to the sequential organ failure assessment score you might have seen that in a couple of my slides earlier. So I only explanation of what that is. It's a bedside and laboratory scoring system where you incorporate the patients. It's kind of intuitive. Each organ system, you get a certain number of points from zero to four. You add them all up. The more points you have the sicker you are in the higher probability of death. It's pretty well validated in non-COVID patients. Its performance in COVID patients is somewhat worse but reasonably well calibrated where because of its intuitive nature as you pick up more organ failures the probability of survival to hospital discharge declines relatively linearly in the middle of the scale. And so most of these state protocols were built around this score and this concept. Here's the New York protocol that Cuomo decided under no circumstances would he activate which is another interesting political dimension to this but I think we'll stay focused on the ethics here. The people get color codes according to their level of SOFA score. Red is the top priority group with the SOFA score of less than seven and they would go first fall by yellow and then blue and then green are people who don't need ventilation. But this is again sort of derived from military triage is where these color codes come from. And there's no, if this is kind of a pure save the most live system very little consideration of other principles. There are some exclusion criteria but not many and the tiebreaker is a random lottery. So it's sort of save the most lives then treat people equally lexically in order, right? Pennsylvania and the model policy put forth by Doug White early in the pandemic and Scott Halpern is a multi-principle framework which includes saving the most lives and saving the most life years. Now, interestingly, they don't conceptualize saving the most life years by using the patient's age despite that being probably the most important factor for determining how many life years someone has left is how old they are. But rather they use their chronic conditions whether they had major chronic conditions which means death is likely within five years or severe chronic conditions. So meaning death is likely within one year. So it's saved the most life years in this very constrained sense, right? Over this like five year timeframe and then save the most lives was operationalized by various solvateurs. And you in adding these points together from these two different rows essentially you could sort of back out the relative weight of both of these principles in the ethical framework. But it doesn't really seem like there was the classic normative ethical reasoning, here's my ethical framework and now I'm gonna pull out this protocol from that, right? It sort of seemed like let's put this out there as an idea. And I think when Doug was here giving his talk fully admitted that the actual weights in scores of each of these categories certainly not set in stone. But I think they were trying to lay down a marker and explain operationalize how you would actually construct an algorithm that would incorporate multiple ethically relevant principles. And for that, which they should be tremendously lauded. So this they actually wrote after the H1N1 pandemic and then re-upped for the COVID pandemic. So one problematic issue immediately with this policy and why it got a lot of heat from various groups is that they had some examples of what would be a major and severely life-winning condition to help guide people and actually assigning the points for both of those disease processes. And you see like one example is a major comorbidity is end stage renal disease in patients less than 75. Well, operationally that would lead to some very problematic examples. Here's one where a 40 year old patient on dialysis who has SOFA score six, so pretty likely to survive to hospital discharge from their COVID-19 respiratory failure ends up with a higher score, so lower priority than an 80 year old patient with a much lower probability of survival to discharge who gets three points, right? So this is a situation where the protocol has actually putting much more weight on, I mean, here it's sort of a protocol failure, I would say to operationalize either of the principles that they hold dear, right? But it also shows that there are in embedded in the way that you're writing this down and turning it into a mathematical system, you're constructing a relative weight between these two principles, sort of after the fact and instead of beforehand, right? And so Dwight Miller led a pulmonary care fellow and Monica Peek and I wrote an article about how that these types of scarce resource allocation scores particularly with the major chronic conditions which are so much harder to define what is death likely within five years? How can we accurately predict that? Could exacerbate health disparities because of structural inequity and the burden of disease of chronic diseases in disadvantaged communities? And so, lots of uproar across the country about these points, Massachusetts' plan was very similar to the Penn plan originally and then because of disability groups, other groups advocating for racial equity got rid of the major chronic condition points. And in fact, they've sort of fallen out of favor altogether, I think later on, I'll show you Doug White's new framework which again removes these major chronic condition points. And the idea here is that it creates a violation of treating people equally or, you know, giving priority to the Warsaw who have because of structural disadvantage from racial inequity that pre-existed COVID in a very problematic way that wasn't acknowledged in the initial framework. Maryland has a slightly different system, right? Again, dropping the chronic conditions. And but I just show you this example too about all of the details and complexity that are embedded in this. As you notice, all the SOFA points are different, right? Then both in the New York and Pennsylvania system. And why does this one have SOFA 9-11 for two points compared to Pennsylvania, which has SOFA 6-9 for two points? What ethical, what's the ethical foundation for that decision? None's really specified. And so the arbitrariness of this is I think really problematic as well, despite these people doing their best, obviously. And the tie breakers, which is how somebody has the same primary score, very way more than the scoring systems themselves. Again, this is another map that had like almost a different color for every state, right? And the green here is not mentioned. No clear way to break the tie. So hopefully at least they would have defaulted to a lottery, but they didn't write anything down. But ideas for lotteries that were popular were age, using it as a secondary consideration, only as a lottery or first come, first serve or first responders or healthcare workers. So what did take away from all these protocols that came out? I mean, basically every components of all the protocols varied, different SOFA cutoffs, different number of possible scores. And I feel like these would have very an ethically important differences and underscore this idea that a lack of formal empirical thinking about these issues can lead to a lot of variation in arbitrary decision making that doesn't necessarily have a solid foundation ethical reasoning. And I think one really important tool to help illuminate the ethical consequences of these practical choices is simulation modeling. And so what I'm gonna do now is show you some of the work we've been up to. And over the past year, literally I've been working on this for a year to actually use real data from COVID-19 patients to simulate the allocation of scarce resources and some of the established state protocols that exist. And ultimately this is the framework that I think should be used to actually develop a protocol de novo and help refine, start with an ethical framework come with a protocol simulated, refine both the protocol and perhaps the ethical framework if the unintended consequences were quite severe. And I'll show you some data from our results. So what do we do? Well, we got data from all the U of C and Northwestern system, we're up to 2400 patients now through both surges of the pandemic. These are critically ill people with COVID-19 who are admitted to the ICU. So not all were on mechanical ventilators from high flow, these canula in this setting but we're in hypoxic respiratory failure pretty much gonna die without life support of some kind not necessarily delivered in a critical care setting but need life support to survive. And here's an example, two randomly selected people, right? And so what the algorithm does is you score for each of these two people and you allocate the resource the algorithm allocates in the simulation model the resource to the person with the lower score or however the algorithm says to do it, right? So in this example, I think this is the Pennsylvania scores that this 28 year old sort of wins the pair wise comparison with the 80 year old and they would be allocated the resource and the way you calculate the total number of lives saved of the system overall is you observe the actual survival outcome of the recipient because thankfully we know what happened to all 2400 people in our data set, right? Because we didn't hit crisis standards of care in Chicago so we gave everybody critical care and we can observe their actual outcomes and then we assume that the person in the simulation who didn't get critical care died. And so you repeat this process for the entire sample and you can see, here's a 50% allocation shortage rule. So half of the patients are assigned to palliative care, no critical care resources assumed to die and then of the patients who are assigned to critical care or allocated the resource you observe their survival. And so obviously, and here's the data that we had. I think if you look at this table closely you can figure out which hospital was USC in which is Northwestern. If you look at the insurance status and the racial demographics, but just to show you the reason I'm showing you all these numbers is to show you that we had a very diverse data set which I think was a pretty good sample, representative sample of COVID-19 patients sort of across Chicago and across various so-so economic groups. And they were sick though. They all had, these are median SOFA scores within the first 24 hours of presentation. Sickest at tertiary hospital A, all requiring life support of some kind with high mortality rates. The lowest was ranging from 16 to 25% by center. So consistent with people who likely would have died without critical care resources. And what protocols did we simulate? Here's the six, what were the first is sickest first which nobody, not some protocol that anybody would actually recommend. It's just there to show you that sort of the default rule of rescue was somehow operationalized in a pandemic in a crisis standard of care where you're only treating the patients with the highest SOFA scores, how inefficient that would be. A lottery is random allocation, youngest first is again, pretty simple. And then three of the state inspired protocols. New York one, which is sort of lowest SOFA first, right? Based on SOFA tiers with random assignment within tier. So it's kind of, it's both save lives and equal opportunity within that SOFA tier, right? And then two different multi-principled frameworks. So the Maryland one and the Penn one, which have different weights on saving lives and saving life years. And then secondary considerations like age, which they call life cycle, but I think it's close equivalent to fair innings, right? All right, so when we apply these scores to our dataset, one thing that kind of jumped out right away to us is how many people got a score of one? It's like, it's like vaccine allocation. Everybody was phase one, one A, one B, one C, right? The same thing here, pretty much the predominant score, particularly in Maryland, most of the patients, even though they had organ failure, we had not progressed up to the point where their SOFA scores were nine or 10, the things, the type of points that you would require multiple, you get multiple points in the system, right? So the majority of patients end up getting the same score or the plurality in Pennsylvania of one, meaning that the tie-breakers are gonna dominate this process, right? And you wouldn't know this unless you applied, you apply this protocol to data, right? And so that's why that step is so critical. And so how efficient are these systems? Well, they range from the sickest first, if you take the sickest patients, you only treat them and you don't treat the healthy patients. Well, obviously the survival of that cohort is the lowest. So again, that's just there not as a serious system anyone's proposing, but as a counter example. And then lottery improves survival significantly up to 80%. That's about the average survival of the dataset, which makes sense. Youngest first, low SOFA score first in the multi-principle frameworks all improve survival above and beyond what a lottery can't because they're selecting for people who are more likely to survive the hospital discharge. So again, the ethical principle of maximizing benefits, that's how much pain for your buck getting there between eight and 11% improvement in survival with each ICU resource allocated. And as the degree of shortage increases, the multi-principle frameworks in particular perform even better. So if you're only allocating ventilators to one out of four people, for example, and you're using a system to identify that healthy is 25%, then the efficiency of that system improves relative to lower degrees of shortage where you're treating almost most of the patients, the efficiency gain in terms of life saved for multi-principle frameworks for principles or frameworks based on SOFA score is lower. So that's what that's supposed to show you here. And the comorbidity penalties, this is kind of by construction based on the way we define them because we use the Alex Houser score for which is a measure of chronic disease and sort of defined it by percentiles. But in both the Maryland and Pennsylvania system, if you had a chronic condition, you got those extra points for having a chronic condition, nobody in our simulation received critical care. So the designers of the system didn't mean for chronic conditions for the life year saved principle to override saving most lives. But that's what happened here, right? That's what happens empirically when you apply the score to actual data, the actual distribution of the data. And so that's why the point values here are critically important. And the design of the algorithm is so important to actually execute the ethical framework that you sought out to fulfill. And then as I alluded to earlier because everybody's got a score of one, the tiebreakers are huge, right? And what are the tiebreakers? Well, they're age-based in both the Maryland and Pennsylvania system. And I should say the Pennsylvania system has a tiebreaker, a minus one priority for healthcare workers and frontline essential workers, which we weren't able to simulate because of data limitation. So whether you can't really call any of these protocols the actual state policy, I think it's better to refer to them in general. But the idea here is if there's an age tiebreaker and everybody ends up in the first category, that age tiebreaker is actually gonna be the dominant principle ethically. And this system in practice would be a youngest first operation for what's actually usually occurs amongst the patients. And so further refinement of the point system and different sulfur cutoffs may tilt things back towards saving the most lives, right? But without simulation and applying this to data, you can't know that. And then disturbingly, but not unexpectedly, black patients, particularly the ones arriving at UFC, had higher sulfur scores at presentation because it delayed access to care in structural racism, causing chronic diseases. And so when you, especially if you look at the cohort where mechanically ventilated, black patients would systematically be allocated less critical care. And then if you looked at just the survival by racial and ethnic group across the whole cohort, this is exacerbated even further. So this is the way that these scarce healthcare resource allocation systems could exacerbate healthcare disparities by adding sort of one more structural inequity at the end of the long line of disparities all the way to the critical care stage. And why was there this disparity? I think I just said this earlier, sulfur scores amongst non-Hispanic black patients were significantly higher, as well as the prevalence of major and severe chronic conditions. So nothing, unfortunately, nothing unexpected there. And so what are the conclusions of this modeling exercise? Treating the sickest first, would say the fewest lives are in crisis standards of care. I think that's pretty obvious. We didn't really, nothing too insightful there, but by prioritizing younger and healthier patients, multi-principled allocation protocols could save more lives than a lot or in more life years with the HTI breakers. But this benefit comes with the detriment of black patients, patients preexisting, medical conditions or patients. So this truly is sort of an equity efficiency trade-off to use that terminology. And that policymakers need to just strike a difficult balance and an explicit balance. And I think these types of tools are critical for defining that point between efficiency and equity. And I would argue it's actually impossible to get there without this type of simulation modeling. All right. So, and here's an example. So Doug White and Bernie Lowe have proposed a revised algorithm, dropping the major chronic condition points that were so problematic, and including negative points for people who live in areas with high area deprivation index. This idea of using geography to counteract the forces of structural racism, structural inequity in the US. And that is awesome, awesome idea. But how many points does you subtract from the score, right? That doesn't fall out of ethical theory. That's got to be derived from data. So I think this is again, what I think Doug has done such a great job at and Bernie Lowe has done such a great job at is laying down a marker for this field and putting something on paper, recognizing that a lot more works needs to be done to get all the points right and all the mechanics of this system. And that process may also reveal that the underlying ethical framework has one or two issues that we need to deal with. All right, so what I actually think, one last point about ICU before I do a couple of examples of vaccine to wrap up here is that withdrawal of critical care resources would be the critical process. So when the pandemic hits, you fill up your ICU and you take care of as many people as you can, right? You don't start rationing before you run out because you don't know, I mean, it's a pain in the process. So the ICU would become completely full and new people would be showing up in the ER and the question would be whether or not to withdraw. And so here's an example, ICU's full, you're sick as patient currently in intensive care unit, 54 year old guy who's been on the bed for eight days. His sofa score has gone up from six to 11 because now he has pressor requirement and he's on dialysis. Let's say, but we're optimistic ICU doctors. We think we can get him through this. COVID takes a really long time to get better, sometimes months. And but in the ER, there's a new patient who's 75 years old and his sofa score is only five. There's literally no room in the end, no space left. Critical care is absolutely scarce. Who, what do you do? Do you withdraw critical care support from patient B to give it to patient A? Well, depending on the protocol, you get dramatically different answers. So I guess a lottery system would just randomly assign or withdraw care from not just patient B, but anybody in the ICU, right? That seems like that would have many serious flaws from that ethical, ethical perspective. Youngest first would always prioritize the youngest patient. New York has this very strict sofa evolution-based system when you apply it to this case, means that you would allocate to the new patient who's arrived in the ER. And Maryland constructs a very high barrier to withdraw, like the family can appeal. I think in practice, there would be very few withdraws under the Maryland system, although there is a mechanism for withdrawal for people who are getting incredibly ill. And so in that case, patient B remains on the vent. So in order to actually simulate these extubation rules and figure out what would be happened in sort of a real-time situation, we built a dynamic microsimulation model developed and coded by Burhan Siddiqui from the Booth School where, and what happens here is patients are added dynamically to the wait list. They have a, you apply the allocation rule and the extubation rule at every time step. This model is amazing. It fits great retrospectively on non-COVID data. And now we finally have enough COVID data to make it work in real time. So what we hope to get at this is really how many more lives could you save by executing some of these extubation rules, as well as deal with some of the other limitations of our simple Monte Carlo model I showed you earlier. All right, so I'm gonna spend about, five, 10 more minutes, Max, on talking about protocol failures in vaccine allocation. There's a lot to critique about US vaccine allocation from a logistical failure. Perspective, but what I'm gonna try to focus on are problems in taking these ethical principles, which are the ones that the CDC, the Advisory Committee on Immunization Practices hung their hat on. They're mostly derived from the National Academy of Medicine framework, which was inspired by an article written by Govind Prasad, Monica Peek and Zeke-Manuel. Three principles here, maximize benefits, promote justice, which is this idea that we want to protect and advance equal opportunity for maximum health. So that might, that lets you treat people differently, right? Based on their health needs. And then the idea of mitigating health inequities as an explicit goal is an ethical good in and of itself as listed on this principle framework. So some good solid principles, not gonna spend time debating if these were the right ones or maybe there should have been more structure here, like they should have been ordered, rank ordered in some way. But I'm gonna hopefully convince you that if through a series of decisions, the actual protocols that were developed and the decisions that were made violate these principles in various ways. And so here's the phases that the Advisory Committee on Immunization Practices came out with that, you know, disseminated across the states, every state did it a little bit differently, but were incredibly influential. And a lot of problems with these phases in various ways. But I think focusing specifically on problems where the product deriving the protocol from the framework was flawed. I'm gonna talk about age-based allocation and the lack of geographic prioritization. So age-based allocation was this concept of COVID risk, we know increases exponentially with age, right? So let's just use simple age cutoffs and ignore all of their relevant factors. Well, the problem with that is that the first cutoff they proposed was 75 years, the immediate life expectancy for an African-American in the United States is 74 years. So they create enormous disparities in access just by construction, by using age as the only factor that you're gonna allocate vaccines on. Not to mention that it's not the most efficient system, right? The most efficient system would be a multi, a score that leads on many different factors to identify those at highest risk for death from COVID-19. Yeah, age may be the most important input into that prediction model, but there's everyone else knows that diabetes, other chronic conditions, and place-based risk, which we'll get to in a second, are huge factors as well for determining risk. So this kind of simplistic, let's just use age, throw our hands up in the air, approach very problematic and let made states kind of scramble to fix what they were given from the federal government. And I think Illinois did the right call, Dr. Z. J. from the Public Health Department, by lowering the age of eligibility to 65-year-olds for phase one B, explicitly for these equity considerations, and I would argue probably for maximizing benefits. You don't know the latter until you actually see the data simulated based on different types of approaches. And so Govan had a nice Washington Post article where he formalized this to the argument, in Maine, they got even crazier with the ages and they went like in 10-year intervals all the way down to 18-year-olds. And so their system would prioritize the healthy, work from home, 50-year-old, over a 45-year-old, 44-year-old who's working in the high-risk community, exposed to COVID all the time with diabetes. That sort of policy is not only inequitable, it's also inefficient. It violates the core ethical principles that the CDC has laid out and represents a protocol failure, right? And as I mentioned before, place-based risk is an enormous risk factor for COVID. If you look at the burden of COVID mortality across the city of Chicago on the left, there are some areas that literally have 10 times the rate of COVID mortality compared to others. That's, if you think about it on age, yeah, age is a huge risk factor, but that's bigger than the difference between a 75 and a 65-year-old. It's on the same order of magnitude. A more educated or more enlightened, empirically-based system of vaccine allocation would have explicitly incorporated geography into the allocation mechanism. But as you can see, unfortunately, in the city of Chicago, we actually did the opposite. We gave much more vaccine to the least hit areas of the city, neglecting the basic needs of thousands across the city and probably saving fewer lives. And one of my summer med students here in Zang is gonna be working on quantifying just how many were lost because of this inequity. And why? Well, it's structural, of course. It's not necessarily, I mean, there were some conscious policy choices, perhaps, but it's also like, well, let's just, we gotta send some vaccine to each of the pharmacies, right? Well, we're all the pharmacies. They're on the North side. They're in the well-off neighborhoods. And so if you just kind of let the status quo roll and you're not thoughtful about your allocation protocol, you're gonna violate your ethical principles. And that's what happened in the city and across the US. It wasn't just the Chicago problem. These are deep-seated structural inequities that without any conscious focus on equity, get exacerbated. And relating to that, we didn't pour the water where the fire was burning, right? There were hotspots in Michigan, Texas and other areas. And there was just this sticking with per-person vaccine allocation at all costs, right? I think mainly by political reasons, but there was not any, even people were trying to construct empirical arguments that it was too late to surge vaccines to Michigan. Well, if that's true, then where's the next Michigan? Why did Michigan happen? Why didn't we send more vaccines to Michigan a month ago? And the lack of empiricism, I think definitely cost lives. And so I'll just end with a comment about dosing strategies. So I think this represents a protocol failure, a failure of a protocol, in this case, how to distribute the first and second dose of the mRNA vaccines that violates the core CDC ethical principles. The first dose advocacy was obvious from the first RCTs. That you can see that after 12 days, after the first dose, the clearly the first dose of the mRNA vaccine kicks in. And the three and four-week intervals were totally arbitrarily designed. They were that short because it was like the minimum, a feasible interval that they could get past immunologists and get the trials done really quickly. And so the UK saw this and they decided to postpone the second dose to 12 weeks. We moved the share of their vaccinated population, which dramatically higher than the United States early on. And about two weeks after this point here, about two weeks after that those vaccination curves separate, go figure the case curve separate too. People will talk about the differences in lockdowns and non-pharmaceutical interventions. But the timing is pretty eerie. And if you actually look at the number of deaths potentially averted, if the US had followed the UK strategy, it's a staggering number and working on more formal quantifications of this going forward. And of course that exacerbated the same vaccine equities I was alluding to earlier. If you're sticking to short dosing intervals and you start out by giving all the first doses to the well-off and well-connected, then now you're given second doses, which have marginal benefit at best to those same people before you're providing access to the most disadvantaged in society. And so that's an article we wrote up in the USA today. You can sort of want to check that out. You can read this, read the argument more formally. And then the J&J pause is another thing I'll just say one minute about that I think was a deadly mistake. Here's an article Gov and I wrote up in the Washington Post the ASAP committee is composed of all these vaccine experts, which of course you need vaccine experts. But there was nobody on there to do risk-benefit analysis. There were no ethicists, there were no health services researchers who really were weighing the downsides of this policy decision. And this may have happened anyway without the pause. But I would have felt a lot better if this had not occurred if I was on ASAP. The day after the pause is enacted, first dose vaccination rates plummet. And when they finally did do the risk-benefit analysis, it was obviously strongly in favor of continuing J&J vaccines on an individual level and giving patients the option for a one-dose shot because assuming the risk of COVID in their community was reasonably high. All right, and so I just don't wonder with all this vaccine stuff, how much of this area under the curve of these cases could have been averted with better protocols and better adhered to the ethical principles that the CDC laid out. So I think I said all of this already and failed to properly hire its own ethical framework. And in contrast to critical care allocation, there weren't many equity-efficient trade-offs. These were all protocol failures that sort of violated all the principles that they laid out. And I really do think that you got to do the math if you're in this allocation of scarce medical, absolutely scarce medical resources game. You can't shy away from complexity. You have to engage these issues empirically and quantitatively. And I want to thank all my co-authors, starting with Gina, who's really the expert that I always talk to about all the details, all the protocols, and Govind and Monique for all our work on vaccine stuff. And of course, my perennial mentor is Dr. Siegeler and Dr. Ross at the McLean Center for all your work mentoring me over the years and also for the Vaccine Allocation Committee for your thoughts. All right, one minute, one minute under the hour, so. Great, thank you. Thank you very much, Will, that was a whirlwind. Got a couple of questions. I want to start with Pat Narikis, who says, would you consider lack of beds physically in an ICU or pushing nursing ratios beyond the safe level to constitute an absolute scarcity or more of a relative one? Feels like we are always stretching thinner boarding in the emergency department, but at least at UCMC have never had the number of physical ventilators to be the issue. Yeah, I think those are examples, good examples of relative scarcity. Why is the nursing staff ratio low? Well, it's just we're not paying nurses enough to get them in, right? Or why do we have too many ICU patients per house staff or per attending? It's because we're just not, people are at home not working. It's not like every critical care attending is working in the ICU all the time. So I think all of those things are self-imposed. I mean, this came up at one point, our McEu senses was in the high 40s. That's a lot of patients, but we got like 90 ICU beds to stay. You know, we still had a lot of capacity. I pointed that out. It was like, that's all federally available. So I feel like those all were in the category of relative scarcity. I do wanna say though, like my black and white, it's absolute a relative scarcity dichotomy is obviously not true. It's there are gray areas there. Great, thanks. We have anonymous attendee who wrote is scarcity in many parts of the developed world and national security and intelligence failure, especially during the early months of the COVID. Yeah, like how, why did we end up in situations of absolute scarcity? Yeah, that's sort of above my expertise and you know, exactly how these things occur, why there isn't enough oxygen in India, for example, and how that can be rectified. And of course, you know, I think for every hour you spend thinking about what to do when you're in absolute scarcity, you should spend a hundred hours trying to prevent absolute scarcity from occurring in the first place. Great, Dr. Bob Chung, right. Super interesting talk, thank you. Is it possible that you try to tune your model to find the parameters, sofa, cut off, what size bonuses to give certain subgroups, et cetera, that would maximize live save or whatever outcome you're looking at. That's the idea of, you know, that's why we wanna build the simulation model. And then also I think provide kind of rigorous justification, the nuts and bolts of ideas like, let's save the most lives we can subject to the constraint of equitable allocation across race and gender, for example. And so, you know, to prove your protocols actually meeting those that ethical framework, you need simulation model. But you also, you know, I think sometimes doing this empirical work makes you really think about the ethical framework that you're trying to implement and whether that makes sense and you've done the proper ethical analysis. Rawls called that to be in reflective equilibrium to go back to your models and modify them to go along with your empirical data. Kevin Dirksen said, this is a great presentation. Thank you. How has your modeling work accounted for the use or absence of ECMO? Is there any data on COVID-19 health inequities with ECMO allocation modeled or actual? As a critical care physician and health services researcher, I'm interested in your thoughts about how ethicists, public health experts and policymakers can better account for the complexities of ECMO deployment and its particularities, limited staff and stuff as the pandemic continues and may represent something of a chronically scarce resource in some communities at baseline. Yeah, Kevin, it's an awesome question. I feel like I've probably missed many opportunities to use ECMO as an example of an absolute, that was the only resource in the United States was like demonstratively absolutely scarce. It's sort of for political, medical political reasons sort of dominated by these teams and these arbitrary notions of candidacy for ECMO that are actually kind of value judgments based on the person's perceived quality of life and the sort of a clinical medical ethics issue about whether initiate someone on ECMO, right? If a 90 year old with a severe end-stage dementia is in COVID-19 RDS about to have an hypoxic respiratory arrest, putting that patient on ECMO would physiologically keep them alive. But often the ECMO team won't do that because they're making value judgments based on their quality of life. And that's status quo, even without the pandemic. And then so what I think sort of happened is then we added this added layer of actual absolute scarcity of ECMO on top of that and things got really messy. But yeah, it's an area we should be studying more. Gina has epistelos designed an amazing survey that like seven institutions have filled out about their ECMO practices and especially about informed consent for ECMO. But I think it's something that hopefully, the data, there were more, there's bigger numbers for my type of skill set would be really rich area. I don't know if I answered the question, but... Oh no, I think you answered the question very well. You know, it did make me think though, as you were talking, I think on part four of your six part talk, you talked about extubation rules. And I just wanna push you on that because while extubation might be a way to say take care of everyone at first. And then when you start hitting scarcity, switch to crisis standards of care, the difficulty of looking at a patient who might get better and removing them from the ventilator, how feasible do you really think that is? I don't think it's super feasible. I think, and there'd be a tremendous cost to that, both emotionally and then literally in terms of physical like time in capacity, right? You gotta take the ventilator off, clean the room, right? You know, there's logistical issues that make withdrawing different from withholding in this case, in a really practical, real way. So I think that's why the dynamic model is so important. If we're gonna withdraw care in anybody, we better be experiencing like huge improvements in expected survival in the model to justify it, right? There has to be, you know, the principle of saving the most lives, the way the improvement in that outcome has to be so large to justify because of the other concerns. Even if the improvements were really, really large, do you think doctors and nurses are going to be comfortable doing that? No, I mean- You take somebody off a machine where they will benefit, but it might take them 10 days where somebody else might be able to get on it off in four days, he can save two instead of one. Yeah, I mean, I think whether, how this would actually happen, any of this stuff would happen in practice, you know, I mean, the only thing we can sort of see in the Western world where it's gone down is age cut-offs for the ICU in Italy, where they just, it's effectively kind of a youngest first approach and without much withdrawal of care. So yeah. So withholding rather than a withdrawing. All withholding. I'm not allowed to monopolize the time, so I'm gonna go to the next comment by Robert Sebesta. How would simulation model analysis handle mid-stream modifications to a framework and percly based measures for the sake of fairness, adapting GCS and SOFA for medically sedated patients halfway through the pandemic? And then he said, great work. Thanks, thanks, Robert. Yeah, no, I think the equity, like the equity corrections are a great example of this. So you wanna correct for the fact that black patients in your community are gonna have higher SOFA scores of presentation and therefore receive less access to care. And you're willing to save fewer lives to ensure equitable allocation across race, which I think there's a great ethic argument for that. The exact weight of the equity correction is gonna depend on the distribution of SOFA scores in the African-American white population in your dataset. Like, so it's gonna have to be dynamic. It might be one point in Chicago and three points in LA, I don't know, right? So what, you're totally correct. I mean, these systems have to be very smart if they're gonna work. And that's, I think what I'm pessimistic about this whole field is like, how realistic is that actually politically? I mean, I think the math exists and the ethics exist, but from a practical standpoint, you worry about it ever getting there. Awesome, Padela writes, wonderful work. I want to ask about the upstream ethical failures, meaning why do we focus on high-tech solutions ICU vaccines and then have ethical failures as opposed to public health measures, which may be more efficacious? Is there a meta ethical principle we have missed which prioritizes low-tech? Is that like a meta criticism of my entire talk which focused just on ICU vaccines? But, yeah, I think, right, like, so the- I need, I need to talk. Good. Yeah, so allocation, right. I mean, the allocation of absolutely scarce healthcare resources, ideally in a pandemic doesn't have to happen at all because you do all the low-tech stuff, the non-pharmaceutical interventions to prevent the ICU from being overwhelmed. And then, you know, then vaccine allocation becomes less salient because a pandemic is not raging in your area, right? So it's not quite as important, but I think absolutely scarce healthcare resources, you know, that problem is not going away. It's, I mean, to see donor organ allocation, fortunately, I think it's gonna keep me employed, hopefully. And, but yeah, I think you're right. Like, I would totally say for every hour, talk, you sit through like this, go to, you know, 100 times more effort should be placed towards preventing absolute scarcity. So Eugene Barreza, a former McLean fellow, writes the default normative assumption in these protocols is often saving the most lives. Did you consider how to consider communities that explicitly reject this principle for religious or cultural reasons? That's a great question. And, you know, I think I sort of, my talk, I jumped past the normative ethical debate, you know, about what principle's the right one. This is for whatever reason socially, that's a dominant principle in both, you know, the vaccine allocation frameworks and ICU allocation frameworks in the U.S. And, but yeah, and so I think that the actual product protocol you need to derive is built upon the foundation of the principles that are culturally and socially relevant in that community. So yeah, I think it's like, you know, it's what the added flexibility of using empirical methodologies that you can, you can sort of simulate whatever protocol derived from whatever framework you want. So it's somewhat agnostic. My approach is hopefully somewhat agnostic to the actual ethical values. Connie Xiao asked, what are your thoughts on vaccine passports? Namely, how to do them and letting vaccinated people be completely unmasked as incentive to the unvaccinated, assuming low rates of breakthrough infection of the vaccinated large uptake of vaccination with greater incentivization? Yeah, a little, a little, yeah, I think that's a little off topic from like the main thrust of my, That's why I saved the class. Yeah, no, but it's, I think it's an interesting question. I really worry about building a vaccine passport on an incredibly inequitable vaccine allocation. You know, like it's only been easy to get a vaccine in the city of Chicago for like a week and a half, two weeks. And now you're like, okay, if you haven't got it yet, you need a passport to, you know, participate in society. That doesn't seem fair to me. But as on the flip side, I think using paid time off cash incentives, you know, like if you can't go into a concert without getting a vaccine, that sort of stuff seems fine, you know, from a ethical perspective. Another project that you really need to model is the distance people traveled to get vaccines. I just have to say that for many people who, well, we talk about the inequities of the vaccines. There was also some hesitancy that led to some of the disparities and there was some overzealous. And I think it would be really interesting to see the distance that people were willing to travel to get vaccines. Yeah, that would be really cool. I don't think we can get that. You know, you know where, it's funny, you know where the vaccines were allocated based on freedom of information requests from reporters. The thing that actually gets reported on the website continuously is where the people actually live who got the vaccine. So that map I showed you was the home addresses of the vaccine recipients, but they might have gone, who knows where they went to get it. And, you know, that's why I was so worried about ZockDoc and other forms of online only registration for vaccination because the wealth or well-connected can use those to take vaccines that were intended for disadvantaged communities. Exactly. I think all the rest of the comments, there was only one other comment that is interesting, again, slightly unrelated, just like the passports is, in the original military triage is rank seniority considered? I asked that, I'm fascinated because we did include docs as the 1A group. So in a sense, if it were equivalent to military, does rank and seniority considered? Do you have any idea? Because I don't think so. I'm not like a super in military history expert on the triages, but no, I think it's pure saving the most lives. But it does bring up the issue of instrumental value, right? And so an instrumental value based system like we use for vaccine allocation would prioritize the general and stuff, right? Because, you know, they can do greater good for the military. And, you know, that might be problematic according to other principles. Well, on behalf of all of the people who asked questions, and on behalf of the whole audience, I just wanna say thank you. That was a real whirlwind. You covered lots of different issues about equity and allocation and showed the real point of the importance of empirical data to help us as we make or try to make ethical policies. So thank you very much. Mark, do you have any last words you wanna say? I'll be very quick. Thank you so much, Will. I agree with what Lainey said. The talk was extraordinary and your work on empirical data research, which we helped to stimulate in the early 1980s as a model for ethics, who was to be congratulated. So I wanna thank you enormously. I do wanna tell the crowd that's remaining that next week, as I said at the beginning, will be the last talk in the series of 27 talks on COVID-19. And the speaker will be Brian Callender and he'll be talking on the topic. It's called the COVID-19 pandemic, past, present and future. So we're looking forward to that final talk and Will, thank you so much. You'll be meeting with the fellows in 10 or 15 minutes. That's great. Thank you so much for the invitation, Mark. As I said, it's great to be a part of the McLean Center, continue to be part of the McLean Center and take forward such an amazing tradition of a great institution. I love the idea that you've been here since your medical school days. Over 10 years. It's wonderful, Chris. Great. Thanks, Will. Thank you, Lainey. Thank you.