 Thanks, Renana. All right, hello, welcome to the McLean Center's noon lecture series. I think this is the third lecture for this year. And this year we have, or this week, we have the honor of our own Dr. Will Parker presenting. So Dr. Parker is an assistant professor of medicine and public health sciences, and assistant director of the McLean Center at the University of Chicago. He's a pulmonary and critical care physician, clinical medical ethicist and health services researcher who studies the allocation of scarce medical resources. In addition to co-directing the clinical medical ethics consult service, he runs an NIH and Grewall Foundation funded quantitative bioethics lab focused on absolute scarcity problems where demand greatly exceeds supply and healthcare systems triage patients for treatment using algorithms. The labs work on deceased donor organ allocation policy, life support triage under crisis standards of care and the allocation of novel scarce therapeutics has been published in top clinical journals such as JAMA, AJRCCM and Jack and Dr. Parker's normative analysis has been featured in the Hastings Center Report, Health Affairs, JAMA Health Forum, USA Today, The Washington Post and The New York Times. He's been recognized with the National Young Investigator Award from the American Society of Clinical Investigation and the American Thoracic Society. And as many of you know, he is a good friend of mine and it is an honor to have him here today as a colleague and as a friend. So thanks, Will. All right. You guys hear me? Well, thank you so much for that introduction, Micah and to Dr. Angelos and Dr. Siegler for the invitation to speak in the seminar series. As you know, I got started at the McLean Center as a medical student and I think my entire path was laid out in what wasn't called a doctor-patient relationship course at that time but that's what it is now. So I sort of owe everything and all my success to Mark Siegler and the McLean Center. So with that being said, my only, my disclosures are the grant support that Micah mentioned. I'm gonna talk specifically about work funded by this NIH R01 and the Greenwell Foundation today but I've no other financial conflicts of interest. Oh yeah. I need to do, make this small, right? And then, well, this just go away or? Well, we'll move that up there. Okay. So what my goals are is to really engage the bioethical controversies surrounding crisis standards of care. So we're gonna skip a lot of the more procedural and logistical issues focused specifically in on the absolute scarcity problem sort of at the core of crisis standards of care and by doing that, ignore incredibly important issues about load sharing between healthcare systems, other things that will certainly come across your mind when I'm talking. So with that limitation, I'm gonna dive in and actually talk about what the definition of crisis standards of care is. This is Memorial Hospital and for fellows, I know you've heard this before over the summer so forgive me for some of the redundancy but this is Memorial Hospital after Hurricane Katrina, after levees had broke. As you can notice, this is several days later and the hospital is completely flooded and temperature had risen to over a hundred degrees and what happened inside Memorial Hospital over that couple of day period while they waited rescue was one of the major prompts for the institution of medicine that's what it's called at the time now called the National Academy of Medicine to define a concept of how a clinician should behave in these disaster scenarios where you can't take care of all of the patients. There are some that will not be treated. Some will come to harm and some may even die because of lack of resources. Ironically, this isn't a new problem at the most abstract level because the deceased donor organ allocation system in the United States deals with this every day. There's a chronic scarcity of organs but it was sort of a new concept that routine healthcare, particularly life support for critically ill patients may be unavailable and patients have to actually be triaged. So that's the problem we're gonna be focusing on today. And before we talk about the controversies in crisis standards of care and life support triage, we're not just to make sure we're on the same page. I'm gonna lecture for a while on what the ethical values are that are sort of acceptable or possible candidate values for life support allocation. And as I mentioned before this, I think these values are broadly applicable to a number of problems, that problems where the resource is so scarce that we as a society have decided to control it or a healthcare system is controlling it and algorithmically allocating it. So people are getting rank ordered for treatment based on various different ethical values and ideas that we'll go into next. So the algorithmic allocation is kind of a central feature of the problems that my lab focuses on. So here's the big four categories. These are actually sort of sets of ethical values. These values can be turned into principles like rules that you would behave but I'll talk about them as values, things that it's good to have more of, right? And kind of go through the four categories one by one but what I hope is obvious from just looking at these pictures and thinking about these words that these ideas are in conflict with one another. It's going to be often impossible to construct an allocation system and that completely satisfies them all. So diving into the first concept or of treating people equally, there's the first idea is that respect for persons, everyone's a human being who's in need of a scarce medical resource who will has some benefit from that resource, right? You all have end-stage organ failure, without an organ you'll die eventually at some point, right? And so we should just assign the resource randomly, right? Treat everybody the same, use a lottery system. The problem with that is that in practice you have to define when the need for the resource starts at some point, right? When does somebody cross the threshold and sort of enter into the lottery? So even if you want to conceptualize this abstractly as random selection, there's still a timing element here, right? Like when did the patient present to the healthcare system with critical illness, with end-stage organ failure? You know, I guess in the case of vaccine allocation sort of truly random lottery allocation was possible, right? Because all of a sudden 280 million adults, U.S. adults would benefit from a COVID vaccine instantly but in most problems there's a first come first served element of treating people equally by constructing a queue. However, how you construct this queue is incredibly important from an ethical perspective. This was shown no more concretely and forcefully in the deceased donor kidney allocation system. The people involved in organ transplant in the room probably familiar with this large change to the kidney allocation system in 2014. Before 2014, you can't hear me? Okay, I'll try to be, how about now? Is that better? Oh yeah, I'm kind of a wanderer, so. All right, how about now? Is that better? No? Can you hear me? Yeah, okay. Yeah, this will help because I kind of move around. All right, we're talking about deceased donor kidney allocation. Before 2014, the queue, your starting time was based on your point of listing at the transplant set, right? So the date that you got to see a transplant professional is when your clock started, when you were determined to have the need for the resource. And this is a treating people equally idea, right? That, well, that's sort of a fair way to do it, treats everyone the same. Except who gets access to a deceased donor kidney transplant center first, it's people with private insurance, it's people with connections, and patients who are from historically marginalized groups who don't get referred to a kidney transplant center when they're developing end-stage chronic kidney disease, have their kidneys fail and have to be referred by their dialysis center later, were sometimes being listed years after starting hemodialysis. In 2014, patients who were previously on dialysis got credit for sort of time served prior to listing for when they started hemodialysis for end-stage kidney disease. After that policy was implemented in 2014, almost overnight, the structural racism of the previous regime was eliminated. So this, I bring this point up, even though it has nothing to do with crisis standards of care, as an example that queuing in first come first serve has critical empirical dimensions about when you start the clock. And we'll come back to the concept of queuing at the very end of the lecture, when we talk about how in practice in a pandemic scenario, most of the established crisis standards of cares would actually just be a simple first come first serve queue. All right, so that's treating people equally. So next we'll move into maximizing total benefits. I think this is what most people intuitively think about when they think about triage, a battlefield situation where there's 100 wounded people and one medic and they have to decide who they treat bypassing patients with who are sort of too far gone to save, skipping people who don't need treatment at all and focusing most of their effort on the patients with the greatest benefit of treatment. But already there, that's not so simple to define benefit, right? And there's sort of two classical conceptions, saving lives and saving life years. If you're just interested in saving lives as defined by the total number of people at the end of their critical illness will be alive at the end of that specific month of the pandemic surge, you would, of course, treat the patient on the left here, right? But in reality, or but there might be a different ethical conception of benefits, which is that we wanna save the total number of life years that it doesn't, it matters not just whether or not someone's alive to the end of the hospital stay, but how long they're gonna live at after that. And in kidney allocation, saving life years plays a huge role in a so-called longevity matching system where the kidneys with the longest expected benefit are matched with the people with the greatest number of expected life years following transplant. But in crisis standards of care, this concept of saving life years was rejected kind of wholeheartedly. We'll get into that when we start talking about the bioethical controversies. But, and also missing from this slide completely is any idea of quality adjustment or morbidity or any other type of benefit we typically think of in a healthcare situation. So that's maximizing total benefits. The last thing I'll say here, and this will come up again, is that this scenario of a young person with a lower probability of survival than an older person for critical illness is sort of, is completely unrealistic and inconsistent with the empirical data. Almost always the young person would have a higher percentage of survival to discharge. And I'll show you some data to prove that later on. So in some ways, this is a false trade-off constructed that bioethics has spent a lot of energy in crisis standards of care debating that wasn't really end up being empirically relevant. All right, next is the concept that we should favor the worse off in allocation of life support during a crisis or in general, that there's some people who enter in to the scarce resource allocation problem at a disadvantage. The first idea here that the rule of rescue, Mike, this is of which we do normal sort of triage in the emergency room, right? When everyone is getting life support, it's just about who's going up to the ICU first or who needs to go to the OR first. Everyone's gonna go, everyone's gonna get treated. We treat the sickest first because they don't have time to wait, right? But the patients who are less sick have time to wait. But of course, in a true crisis scenario where not everyone's receiving life support, this leads to fewer benefits than a lottery. There, this means that you would be saving, you'd be prioritizing resources from the people least likely to survive. So, sickest first fails completely in crisis standards of care scenarios, but it actually works pretty well for clients because the people who are sickest, those with the like highest model for end stage liver disease scores have the highest benefit from the transplant. So, the empirical and clinical details of the problem are key here. And for crisis standards of care and life support allocating to the sickest patients would lead to sort of a catastrophic failure of maximizing benefits. So, what about the young, right? If somebody develops critical illness or end stage organ failure earlier in their life, the argument is that they are systemat, they are worse off than someone who develops it later in their life. Their disease threatens to cut their life short, prevent them from playing all nine innings of baseball, right? And I think this is very intuitive, particularly for children who are being denied access to healthcare resources all the time in our society and across the world as something that's morally wrong because they're not allowed to live a full lifespan, right? And an older adult by contrast has had the opportunity to live through the stages of life and therefore it's ethical to prioritize a younger person over an older person for an absolutely scarce healthcare resource. As you'll see, with that ends up being very controversial in the way it's actually protocolized in the crisis standards of care. And then the last idea in favoring the worst off is that there are people who are systematically disadvantaged for historical purposes prior to the current crisis. And some of the structural inequities are exacerbating the pandemic for the crisis for particular marginalized populations. And we need to look no further than Chicago to find ample evidence of this. The figure on the right is a map of deaths during the alpha wave. So this was after vaccines were available but had been distributed predominantly to the wealthy areas of the city. You can see that Hyde Park is no exception to that. Were that, I think my clicker thing isn't really working but were that bubble on the South side of lower COVID mortality and that was driven almost exclusively by vaccination rates, although there was disparities in COVID mortality before vaccines became widely available. But these lines on the map, they just didn't happen. By the way, humans tend to group together into similar groups. They were explicitly designed by the federal government by systematically promoting investment in certain neighborhoods and disinvestment in others, like explicitly racializing residential segregation and redlining. And that's what the Mapping Inequality website and materials will make abundantly clear if you live in this neighborhood and you see where the lines were drawn, the historical legacy of those racist laws and policies is quite obvious even today. So how do we account for this, right? And when we have to allocate life support when these patients from these, from people from these historically marginalized communities living in historically purposefully disadvantaged neighborhoods present with critical illness, how do we handle that? That's the bioethical dilemma. And the last set of values is that we need to reward socially useful people and prioritize them already sounds a little icky, right? But we do it. So in organ allocation, patients who are people who become living kidney donors receive the equivalent of four years of waiting time. So if they develop end-stage renal disease later in their life and need a transplant, a deceased or a kidney transplant. So this is reciprocity, it's payback for prior good behavior. But the other version of rewarding social usefulness is instrumental value, that there are some special people who do so much good for society that, God forbid they were critically ill, we would need to allocate them more life support, right? Or prioritize them for life support so they can go back into the world and spread the good. I think you see, I've chosen a particular candidate for instrumental value that may trigger some in the audience, right? And make you question this concept and how one would actually calculate the so-called multiplier effect. But while we all laugh when it came time to get our COVID vaccines, we were happily engaged this ethical value, right? Leaning heavily on our amazing multiplier effect as clinicians to take scarce vaccines before our vulnerable patients. I stay up at night thinking about how many people died because healthy ICU physicians like myself and nurses were vaccinated before sick vulnerable patients in a very categorical way, right? Ignoring all other ethically relevant principles. Now, that's not totally fair because patients in long-term care facilities, which of course are this combination of really high personal risk and really high risk of infection from the congregate living setting were prioritized equal to healthcare workers. But that's a different talk. I think that vaccine allocation represented a tremendous and disproportionate way on multiplier effects and instrumental value. All right, so hopefully going through those one by one was interesting and a refresher for those who are familiar with these ideas. This entire kind of landscape or this way of thinking about the problem of laying out all the ethical values and showing how there in conflict comes from the work of Govan Prasad and Zeke-Manuel and now heavily influenced, fortunately, by Monica Peek, our McLean Center Associate Director for Research. And one of the, I think the main points, which I made at the beginning and I'll reiterate here, is that you have to pull on multiple principles. If you're making an actual protocol, turning these values into action statements, if you're going to create a satisfactory framework that lots of stakeholders will buy into. If you put all your eggs in one ethical basket, there's gonna usually be a serious problem with your allocation system, a problem that makes it sort of unacceptable, both legally, politically and socially. So with that background, what I'm hoping to do is go through four bioethical controversies. I think it would be kind of fun to stop for questions and commentary after each one, rather than like stacking it all up. Since we have a, you know, the fellows are here and this is a seminar, right? So let's see how that works. So here are the four, the four things that I think a lot about as I'm working on this problem. And starting with one that's really not a question in my mind, but it's something I hope to convince you of, which is that SOFA is biased and inefficient and needs to go. It should be eliminated from all crisis standards of care protocols. So early in the pandemic, Gina Pisticello, McLean Center, a fellowship graduate, now at the University of Pittsburgh and a emerging national leader in serious illness conversation ethics, did a Sentinel review of all the state of CSC protocols that were on the books. This was in like April, the collection period is sort of April, 2020, to, you know, May, 2020. So very early, right? And what we found is that only about half the states had a plan, anything written down. A lot of states like Illinois had, had a crisis standards of care, but there was no meat in there, right? There was no actual life support triage protocol. So it was a lot of procedural stuff, which is important, not like Valerie's like kind of glaring me, like no, the procedural stuff in the legal framework is incredibly important. But the ethical meat of the life support triage was sort of unspecified, incredibly vague in the Illinois plan. But the states that did have a plan, they were relying heavily on the SOFA score. And here's an example of one of the CSC plans. This is a very famous, or famous, or, well, now sort of infamous, but a widely imitated plan from University or from Pennsylvania that was replicated in many different places. There was a couple of these plans that disseminated like the New York ventilator guidelines was imitated across a lot of states. Maryland also had their own crisis standards of care that was written down before, but what's common amongst all these protocols is that the sequential organ failure assessment score, SOFA score was the key tool for identifying who to save the most lives, right? To identify who is gonna survive the end of their critical illness. And before I move on and rant about how bad SOFA is for a while, I hope you, this example drives home how complex the process of taking those ethical values and actually turning them into a specific protocol is. So there's a lot of criticisms of this protocol, but they at least have something written down, right? Some concrete attempt, mainly driven by Doug White at Pittsburgh, who is a friend of the McLean Center, to formalize the ethical values into a real concrete protocol. Something that the C-Stoner organ allocation system is forced to do every day. So now to complain about SOFA, it's old. It was made up in the 90s at a critical care conference by a lot of very brilliant critical care physicians, love ICU physicians who use their clinical expertise to create a scoring system that could be calculated at the bedside, right? It's supposed to be basic labs, nothing too fancy. The worse your labs are in a particular area, the worse your hypoxic respiratory failure measured by your PO2 to FiO2 ratio, the more points you get. It's all very intuitive. The only thing, a lot of this stuff is still relatively up to date, to be honest, except for the cardiovascular component, which includes old pressers, do not use dopamine for shock to the trainees, especially cardiogenic shock as it increases mortality, but that's another lecture itself. So, but the one problem with this is that it is in no way based on a regression model. So these points are not mathematically related to saving lives, to who's gonna die. That each point is just, the cut-offs were arbitrarily decided. And with that caveat, it's pretty remarkable that it works at all, right? It speaks to the clinical expertise of the people developing the score and deciding these cut-offs, that if a patient is already in intensive care unit and you calculate their SOFA score, it predicts their mortality pretty well. If you've been there for 48 hours, you calculated their score across all the different domains, it gives you a good sense of the trajectory of their critical illness for a patient who's already receiving life support. Particularly if the score goes up, then you know, okay, things are not looking good, right? But that's not the triage problem. The triage problem starts with the patient arriving with critical illness. And you have very limited information. It's much more akin to the trauma bay, right? Where you have to make a triage decision based on a very limited set of information. You don't have the luxury of 48 hours of observation on life support. And when you actually look at SOFA score in that context, it does very badly, much worse than just using someone's age to rank order them. So this is an area under the receiver operating curve. Many of you have seen these, these are a way to measure a predictive score's performance. The vertical line or the horizontal, excuse me, the diagonal line here is a coin flip. A perfect predictor would look like a box and have a hundred percent sensitivity and a hundred percent specificity at a particular cutoff threshold. And SOFA is just a little bit better than a coin flip. And this is again, because we don't have the time, it doesn't have the time to get this level of accuracy, not that anything is different about COVID. COVID is just one of the most common cause of critical illness. COVID critical illness is not too much different than other causes of critical illness. And even worse and related to its poor performance, the SOFA score is statistically biased against particularly historically marginalized groups, particularly patients who identify as black and why is that it uses the absolute creatinine value, which we know on average is higher in people who self-identify as black. And that's, you know, that difference between black and white people in terms of their measured serum creatinine, not their EGFR that's different or estimated glomerular flotation rate is why there's been a lot of controversy about algorithmic fairness and using race in the calculation of EGFR in the first place, right? But SOFA score by ignoring that mathematical difference between groups across a protected subgroup creates a disadvantage for black patients. They would have higher SOFA scores than they should. SOFA predicts black people are more likely to die than they actually are in practice. And this has been shown not just by our lab but by a bunch of different groups across the country. It was one of those things that just sort of jumped off the page in a variety of data sets, both pre-COVID and post-COVID. And this is two other quantifications of what this particular statistical bias is called a calibration bias, right? So SOFA is miscalibrated for black patients. And that leads to, would lead to under allocation of life support in a crisis. And this is part of why SOFA is a bad inefficient predictor. When you have a biased score in this sense, it's worse than identifying ICU survivors overall and leads to violations of maximizing benefits and problems for considering historical disadvantage because you're exacerbating a previously existed health and equity. So again, more empirical data showing the same thing. And this is in a combined data of University of Chicago and Northwestern from our group, that if you run a simple month in Carlos Simulacia that the published crisis stands of care protocols that rely heavily on the SOFA score create racial and ethnic inequality without any significant gains in efficiency or saving lives. It penalizes patients who, they would penalize patients who are black and not be very effective at saving lives basically marginally better than a lottery. So that's the status quo in the US right now. Obviously it's one that I think is deeply problematic as we'll get to in a bit other considerations like those life year considerations have been stripped out of most of these plans. And so out of the 31 now we're up from 26 to 31 as of 2022 states that have a plan, 86% have the SOFA and most 25 out of 31 were like heavily on the SOFA score, many exclusively. So this is the central gap of the R1 and the Greenwall Foundation that we're trying to take on that we need to make a new triage tool to replace SOFA. And I argued now is a good time to do it when everything's not burning down around us like it was during the pandemic. So to conclude, to reiterate some of the points I've made SOFA is outdated severity of illness tool for patients already receiving life support. It's not a triage score, it never was. There's this thing called modified SOFA that people claim is a triage score. It's not. It's the same thing, they always calculate after 40 hours. It's a nice rhetorical trick, but there's no easy way to take SOFA and manipulate a little bit and turn it into a triage score. It'd be much better to just throw it out. And that's my plan and come up with a new one because it's less accurate than youngest first and it's statistically bias against black patients which makes it even more inaccurate. So yeah, I mean, we could stop for discussion at that point. I think there's less, in my mind, there's less controversy about SOFA score continuing to exist, doing okay on time in crisis standards of care. I don't know if anyone has any burning comments or questions at this point. Otherwise I can just keep rolling and talk about using age. Yeah, that's a great question. It's when initially in the pandemic we were very afraid of ventilators the actual machine becoming shorter. There was also a clinical practice that turned out to be wrong at the time of intubating patients after six liters of nasal cannula oxygen. A practice we abandoned at University of Chicago after about two patients. Thanks to some of my colleagues who just thought this is bananas but some very elite East Coast institutions in Boston kept doing it really egregiously late in the pandemic in my opinion. And this has fueled a lot of misinformation about COVID pneumonia and its severity saying it's always only just because they were intubating patients who didn't need it that mortality was so high. That's of course ridiculous, right? But the way I think about life support now for this grant and for future work is a package. Life support is not one machine. Life support is something you get when you have a one-to-one or one-to-two nurse who's titrating vasoactive drips. Or if you're not on a ventilator but you're on high-flown nasal cannula on the precipice of requiring mechanical ventilation, you need pretty much constant attention to survive. So life support is more about the people that deliver it. And the number of people who can do it is limited. And that's the scarce resource rather than the actual machines. Valerie? Yeah, I mean, it's actually not a hard predictive modeling task at all. I mean, we are hard at work doing it but it's not, it's something that's been done in using observational healthcare data a million times over. All it takes is a little bit of attention to the specifics of the problem. Understanding what the training dataset has to be. It has to be patients who have just started life support. We're doing like a tale of four hours, like the idea you'd have some temporary ventilators or something you could stabilize patients with to get a little more information after life support has started. And constructing a new score that we're designing to be parsimonious that's actually based on the relationship of these clinical variables with someone's actual survival. They're sort of like this mafia, Sophia mafia. I don't say that just cause it was started in Italy in the critical care world where it's reported in all the clinical trials. It's just something everyone's used to thinking about. And it's like the only, it's the idea that that could be the only bedside metric of severity of illness one could calculate easily in a triage scenario is kind of ridiculous. I think it's only because of a group thing that we've been stuck on so far. Yeah, I wasn't planning on writing this grant but let's get so mad about it when like many of you in the room were helping me with this we had to write our own crisis standards of care and started to think more about it and Monica and I were like, we need to work on this. But all right, so next, now we're getting into more I think what are more controversies and I'm really curious to hear people's opinion is how can age be used to allocate life support in crisis standards of care? Here was what Utah was gonna do. So Utah was one of those ones that wasn't all in on Sophia. They were gonna make a scoring system that was based on the person's age you're older, you get three points. All these scores are like golf, right? More points is worse, you don't want more points even though the older, more likely to play golf it doesn't still get more points. But then the next one is their ASA score which is about their level of impairment in fact, or perhaps ability or disability, right? And so what's interesting and problematic about this tool is that estimated survival is now it's own row with the bottom. So what this does is essentially double penalizes patients with chronic diseases. It double penalizes older people. It says their lives are less valuable than those who are younger and able body, right? Because estimated survival is a third of the score and then you get more points tacked on if you're disabled or older. So you see why this raised objections and concerns, right? There's an explicit valuation of human lives treat different people differently based on how old they are in their age in the sense that their life is actually worth less. And that's what another example of how when you write down these protocols sometimes your bioethics becomes more explicit even if you don't really intend to deprioritize the elderly this much this score amounts to a severe empirical deprioritization of the elderly. And so because of protocols like that one and throughout the country, the Department of Health and Human Service has an office of civil rights, which is trying to make sure that states are following federal laws like the Anti-Age Discrimination and Disability Act. And they went through and engaged all these state governments and said you have to get rid of age and disability to lump them together, right? So they said any language about deprioritizing or excluding patients based on their disability status or their age has to be removed. It's inconsistent with federal law. And this is sort of sums what they said. So you got to get rid of the life year's ideas. You got to get rid of categorical exclusions based on someone's disability status. And the resource intensity and duration of need is a criteria for allocation. You have to take that out too, which is, if you think about it as a real handcuffing of your ability to maximize benefits and save the total number of lives, it seems to suggest you can reallocate ventilators. And the last one is important, right? Evaluate people with disabilities based on their actual mortality risk, not discipline related characteristics. They said the same thing about age. Evaluate older people based on their actual mortality risk, not based on, you know, age related characteristics unrelated to their likelihood of survival. So a lot of states said they O complied with the HSS order, which usually means they just took down their old crisis standards of care and there's nothing on their website anymore, that the replacement isn't there. But in theory, what happened or what these states would do if forced to come up with something is they just got rid of those considerations, right? Turn it into a pure estimated survival system. Most of the time with SOFA, as we talked about, it's a terrible way to do that. And this gets at, but this brings up that there's actually two ethical justifications for using age to allocate scared healthcare resources. One is the one that's rejected by HHS, but for crisis standards of care, but not rejected by the deceased owner or kidney allocation system or organ allocation systems overall, which give categorical priority to children. It would be interesting to see how this would be tested if we had a pandemic that actually affected children and children or adults were competing for the same life support. But right now, they said, according to the federal law, you can't value younger lives higher. So okay, but even taking that as given, age is a strong independent predictor of short-term survival. In any parsimonious triage score, if you include age, you're gonna much better identify ICU survivors than if you don't include age. And back to SOFA, my, this is by SOFA. SOFA ignores age, right? It's just these biological and physiological markers of critical illness arbitrarily decided and cut off. And so this is data that were, well, should be published soon, hopefully, fingers crossed, but of the SOFA predicted mortality across age groups in combined University of Chicago Northwestern data set. And that's the red bar, right? We define this cohort as patients who are on life support on a ventilator, severe hypoxic respiratory failure on pressers. And so they all kind of had high predicted mortality by SOFA, right? But if you look at their actual mortality, that's in the black bars across age groups, unsurprising to any doctor in the room is taking care of a critically ill adult, the younger people conditional upon this high SOFA barrier to entry to this cohort were much more likely to survive. And that's because there are, you know, their organs are younger, their bodies are more resilient, younger people faced with the same insult of critical illness are more likely to survive with the same treatment. And so we have a new score, this score is just like SOFA plus age, but we're gonna throw out SOFA completely in our next iteration. And you can show that you can sort of recover this calibration bias against young people and fix that by using someone's age. So using, I think age as one variable among many, I'm not saying as the only one, not appear youngest first situation has a pretty robust justification. And to my understanding would be consistent with federal law and what's consistent with the statements that the office of civil rights has put out and removing age from, I'm gonna go a little further, saying, if you take age out of life support allocation, it's actually anti-young ages, right? There are younger people are more likely to survive. If their lives have equal value and the only way we can identify them as being more likely to survive is by using their age, we should use that if we just wanna save lives, right? That the status quo is actually biased against young people. One way to see that the converse of that in organ allocation is that the model for end stage liver disease, right? Got them, it has all these biological factors to predict who's gonna die in the next six months from liver disease, but it doesn't include age. I'm sure age would be an independent predictor of who's gonna die in the liver transplant waiting list, conditional upon all the other components to meld, but we don't put it in meld and why don't we put it in meld? It's cause there's actually some fair innings baked in there. I would argue that liver allocation has an implicit anti-elderly ages bias built in or fair innings concept built in to liver allocation, but the converse is true in crisis standards of care. By not using age, you're actually by making the score biased against younger people. And then finally, there are people find fair innings, potential lifespan, equity, saving life years, appealing, and they have brought appeal and they're actually implemented in organ allocation policy nationally right now. So if they were so, if these ideas are so repulsive and inconsistent with federal law, why are the organ allocation policies getting away with them, which rely heavily on them to give children categorical priority for transplantation and explicitly try to save the most life years with the longevity matching to C-stoner kidney allocation. So with that, I'll stop, except I don't think I made the point that of course we use age to allocate vaccines quite explicitly as like the only factor because older people had higher expected benefit. And so it seems crazy to not do the opposite with life support allocation or in fact, it's anti-young ageism to not do the opposite with life support allocation. And yeah, see what thoughts people have. We can go to the next controversy, which is, it may take a while. Does anybody want to call me Agist or Jabot, it does. I wonder if a bandage, is that a bandage, do you look at the bag, what is this? And then leaving with others to choose it. If it's a protocol, it's not extended Monday, but still life came to me. Say bye, I'm a doctor in an emergency. I wanted to know what's the right thing to do at that time. So, I would only do environmental math equality and everything, but then also the people who are in the hospital are going to come and get brought in. There are so many people. Yeah, no, that's exactly, there's so many. No, yeah, and I think that's exactly right. So these are government policies, right? These are the state. The society and the government, I think it's the right answer that these policies are supposed to give tools for the bedside position. And of course, ethical values of, you know, and these is sort of a, I guess it's definitely a political question at some point. What things as a society are we going to value when we're allocating life support during crisis standards of care? When we're giving that bedside position explicit instructions about who to prioritize. So, yes, ultimately this is a health policy question, but I think hopefully I've convinced you that the principles of clinical medical ethics are incredibly relevant to that health policy. All right, I don't know this goes to 130, but you know, don't keep you guys here for too long. So I'll keep moving along because this is some, this is a topic that I think we're struggling with in this work and a lot of people are struggling with. As I mentioned, and as everyone knows, Chicago, across the country, every urban area has a similar map. The areas that are historically marginal and the people who have been historically marginalized by structural inequity, we're more likely to acquire COVID and die from it. We're likely to acquire it because they were living in congregate housing, having to actually work and be an essential worker instead of zooming the whole time and being outside and working with other people. And so that's another form of structural disadvantage and then more likely to die or get seriously ill from COVID adjusted for age because of existing chronic conditions that made one at higher risk of severe illness from COVID like diabetes and hypertension. And the fact that those occur structurally in certain areas are because of inequal access to healthcare. So how do we think about the status quo of the US healthcare system and the relative scarcity, disadvantaged populations have all the time, right? There are people who need healthcare, can't get it every day in the US but now we're in this extreme situation where it's explicit that we're triaging. How do we deal with that? And there's actually at least four different ways to address it. And I'm gonna kind of go through these one by one and explain them. The first is demographic parity, which is to make allocation and perhaps even the outcome of allocation here which is survival equal across groups of the protected characteristics. So you're gonna just do assign ventilators randomly if you have to choose between two patients, you're gonna do first come first serve and hope that that means that patients who are not gonna be treated any differently according to their race or ethnicity, you're gonna design an allocation score that the access to the resource is the same across the groups. The problem with this is that the only way to achieve it is to completely ignore benefits, right? There's no way to maximize total or even get near the maximum number of lives saved if you have demographic parity. You have to balance the two of those, right? And there are attempts, which I took the slide out to using deep learning and machine learning techniques to balance demographic parity with saving the most lives. So that's one concept, right? That the allocation has to be the same across the protected characteristic. The next is non-discrimination and the non-discrimination is don't make it worse, right? Your score should not exacerbate underlying structural healthcare disparities. And this is the main goal of the algorithmic fairness literature in machine learning, right? When they're designing a new prediction tool for who's gonna be testing for a new disease or doing an intervention in EHR based early warning score, you wanna make sure that your warning score is not going to exacerbate underlying structural inequity and that the patients are experiencing. And hopefully I've convinced you that SOFA violates this statistical concept of non-discrimination. The technical term for it is calibration parity, right? That SOFA is not calibrated across racial and ethnic groups. And this would lead to a misallocation of resources at the advantage of patients who are white and at the disadvantage of patients who are black, violating non-discrimination and exacerbating underlying health inequities. So that's one concept. But one issue about fixing non-discrimination if you're actually trying to use someone's race explicitly in the algorithm to achieve non-discrimination to let's say de-biased the SOFA score by minusing one point if the person's black, like that was proposed. That's challenging for multiple reasons. Legally, the threshold for using someone's race explicitly in a law or in a decision or a policy is much higher, the strict scrutiny standard. I hope I didn't botch that, Valerie, but higher than the standard for using age. And of course, there's the political dimensions of demagogue to use attempts to prioritize particular vulnerable groups, misconstrue them, misrepresent them and for political gain, right? To rally up an uninformed mob to their cause. So I think using race identity, or ethnicity explicitly in allocation algorithm isn't gonna work. The last practical issue is if having a particular, membership of a particular race ethnic group gets you higher priority for something, how do you stop people just saying, well, I'm part of the, I'm BIPOC, right? Me, I don't imagine I said that, like how would you validate that, right? Would you racialize me into a different group based on my physical features, name, the way I speak? So there's a lot of practical concerns too about the use of race ethnicity explicitly in allocation algorithms. So, but there's other ways to fix this problem, right? And one is just to like fix the parts of SOFA that are broken, that are causing the bias to actually do good multivariable prediction modeling actually resolves a lot of these issues. The only reason we have this problem with SOFA is because people, it's sloppy. It's sloppy thinking, it's not following the best standards of data science and that's what causes the bias. So, stay tuned for this, but I actually think non-discrimination as a standard in life support allocation will be relatively easily achieved. My hypothesis is our new score, which hopefully I'll be talking to you guys later on about will satisfy non-discrimination and not have the same bias SOFA score. Because we're gonna use better statistical approaches or we're gonna have a statistical approach first of SOFA then I don't think it's not even fair to say it has a statistical approach. All right, but that's just one of the four ways you can deal with present day structural inequity, right? The next idea is that you need to go above and beyond non-discrimination. You need to actually try to even out that map, right? That map that I showed you that where all of the death and morbidity and mortality from COVID was being concentrated in the most disadvantaged neighborhood. That you need to allocate life support in such a way so not all the deaths are gonna be in the most disadvantaged neighborhood. And that's by explicitly prioritizing patients who live in those neighborhoods over patients who live in advantaged neighborhoods. Who as you can see, it's funny when statistically we're trying to do this or some later waves in the pandemic. And sometimes like during a particular wave like nobody died in Lincoln Park. Like then nobody died in the whole and so that gives you give us like statistical problems, right? Because the burden of COVID was so low it was essentially zero in certain areas. So maybe we should deal with that in the way we allocate life support. Area Deprivation Index is one of many place-based disadvantaged industries that could be used to allocate more resources to a structurally disadvantaged neighborhoods. And this has been proposed. So this is Doug White's and Bernie Lowe's plan. If you are from an Area Deprivation Index neighborhood, eight, nine or 10 on the scale, you get minus one, right? And so that's roughly embedded in that. It's this idea that mitigating structural inequity at least in the tail of the ADI distribution is about a fourth of as important in the saving lives. Because a one point move on the saving lives scale is like 25% risk of death. So that's how I'm getting that out there, right? So this is not an attempt to normalize the map and spread make all the color the same in terms of mortality, but at least make it not so severe. And this is like, and so they actually put this into practice with how they allocated monoclonal antibodies where patients from high area deprivation index neighborhoods got twice the chance of being allocated a treatment compared to patients who lived in an advantaged neighborhood. And, you know, the state of Oregon where our lab's currently working with the Oregon Health Authority right now who's revising their crisis standards of care and Harold Schmidt from Penn has come up with this idea of equitable chances that would take the PIT model even further where the number of chances would be proportional to the burden of the crisis in that particular neighborhood working on, you know, trying to put some actual numbers on that, which is where my lab comes in right now. But this idea is not without its critics. So as I'm sure some of you guys will find some of these arguments compelling First, they criticized the paper on sort of technical grounds for including narrative descriptions of the patients which they thought really sort of violated the idea of an objective triage team. But setting that aside, they thought, you know, ADI using ADI to allocate vaccines is incredibly like perfect, right? And other large scale public health interventions before that prevent the development of critical illness but not at the point of care where you're actually triaging patients because then you're involving non-medical factors in the treatment of patients, right? Unclear, like is age a medical factor but setting that aside. And they say ADI is not granular enough. So even though ADI goes down to the census block level if you click on the map around our local neighborhood you'll really see the differences. You know, what if you're a super rich guy who just bought a condo in Woodlawn or something? So you're in an ADI neighborhood that's nine but you're not at the personal level subject to the same forces of structural disadvantages as everyone else who lives there. So gentrification is a real concern for ADI. And then finally, you know, you're not sure how this is gonna play out in general whenever you introduce social factors in triage. So that's where the debate's at. The last idea is kind of going above and beyond correcting the present day structural inequity and allocating even additional resources beyond that point because of prior wrongs that existed before the crisis. This is definitely the most controversial particularly when it comes to allocating life support during triage. So with that, I'll just do some conclusions on structural inequity. We can have some conversation and then I could tell you a little bit about exactly what our lab's doing right now after we do some questions. But, you know, I think it's quite obvious to me that we should get rid of discriminatory and inefficient triage scores like SOFA. True demographic parity, equal allocation can only be achieved during a lottery. And I do not think that that should be the ethical objective. I think sufficiency is the name of the statistical concept or calibration parity, non-discrimination should be not making it worse, should be the objective of a triage score. And I do think that there's some role, I'm not sure exactly how much I'm curious to what your thoughts about using play space, disadvantage and disease to mitigate the present day structural inequity in a crisis, to do something about that atrocious map of the burden of the pandemic that just falling so much more disproportionately heavily on the worst off. But I don't know exactly what the ideal weight is and I don't have a well formed argument for where, how much weight should be put on that idea. So with that, yeah, I'd love to hear your thoughts up before I, and then, you know, then we can go, I have another controversy, but I definitely want to hear everyone's opinions and reactions to this particular problem. It is justified that each person continues to get good people, like everybody that already thinks about it. But if you were in a certain situation like it's a lot of people are hungry and you are not really eating a lot of food and someone was not hungry and didn't need it, but you denied people who didn't really need it and feel this is a crisis or something like that, something only more here, something big. So you just reply a lot of people in circumstances such as illness. Yeah, no, I think that's a great criticism of lottery and treating people equally that when you have absolutely scarce healthcare resources like this scenario and you're just blindly allocating without any idea of what the person's benefit from the treatment would incorporate it into your protocol that so many more people are going to die because of that ignoring those relevant morally, I think morally relevant differences between people in terms of their expected benefit is deeply problematic. So yeah, I don't think a lottery is the solution although I do think if to Bob Trug who was here made this point if you don't include an active withdrawal mechanism from your crisis standard of care protocol essentially you're doing first come first serve just whoever comes in the door and you need to critical because people arrive at one at a time in the hospital. And so this thought experiment of 10 people in a room and you're running the triage thing wouldn't happen in practice. So that's part of what we're trying to actually rigorously empirically simulate with a big collaborative network of all of my pulmonary critical care data science nerd friends across the country and getting like a large clinical informatics approach to be able to run a lot of analysis and all our data without sharing it centrally. So that's what we're working on right now is exactly that question avoiding a de facto first come first serve triage scenario by ignoring this critical step of who can you withdraw now to be fair some of the protocols don't ignore it like New York ventilator guidelines had a very explicit withdrawal rule based on changes in sofa but some new empirical work which is coming out now from our lab has shown that would lead to tons of reallocation and inefficiency. So I think that's another huge open question in this space most people are just focusing on that weight that allocation rule like how you get an ICU the first place but I think the ethics of when withdrawing life support in a crisis scenario is permissible and for what reasons needs to work out a lot more. Good job. Well, this is going to be a question outside of your slides but I'm interested. So you started with Katrina you talked a lot about COVID at what point and who makes the decision and how is the decision to made to switch from sickest first first come first serve which are these models that we feel very comfortable with to crisis standards of care in a new model and is it the last ventilator? Is it the day before the storm hits? I'm wondering and then how do we practically do that? And then second and related question is at what point do we say, well, you know what actually the problem here is a resource problem we need to increase resources. So I hear a lot about blood scarcity and I think to myself like that's a solvable resource problem, right? That's not one in which we need to start allocating. We need to increase the resources and it's not doable obviously in a crisis but. Those are both great questions and sort of speaks to the artificial construction of this problem for the talk. In reality, a lot of us were during pandemic surges were in what so-called contingency care, right? Medically necessary time sensitive surgeries and procedures were being delayed. Patients were being harmed. We're prioritizing the care of critically ill adults mostly who had COVID right at their expense. We had, I think we topped out in the sixties at UFC and we probably could have done like a hundred. So, you know, it was quite bad. The normal mickey census is 20. So it was, we were under strain but it was clearly in this contingency standards of care. So I took that slide out. So the line of when to activate crisis, when to activate triage partially is an empirical one because you can sort of say like if you turn on your triage protocol too early and the crisis didn't end up being turned out so bad you could end up killing more people than just the default sickest first first come first serve. In the U.S. a lot of states activated crisis standards of care but almost no hospitals admitted to triaging. They activated crisis standards of care for the other parts of the document which are important but so, and for things like enforcing hospitals to load share which was something that you wouldn't think would need to be necessary but nevertheless the public health department saying you must accept those transfers. This isn't just about maximizing your nonprofits. This is about the entire state's COVID load. So it's funny the states that had like one health system that kind of was the only game in town like Oregon for example and Minnesota is sort of like this did a lot better job at load sharing and preventing crisis scenarios than in Chicago it is incredibly fractured system. There were community hospitals that were well above their sort of capacity to deliver care and any real quality were desperate to transfer patients but there was no coordination. And that's partly because there's all these academic centers fighting with one another all the time competing for business is their sort of status quo and that frame shift to contingent optimal load sharing contingency never really happened. Was there a second question? Well, I think that's obviously an orthogonal problem, unrelated problem to this one that's critically important. We should constantly be thinking about the ways to increase deceased owner or occasion utilization. A lot of my collaborators all they think about is like how can we get transplant programs to accept more organs? Cause we know they'll do well cause they do it in Europe and they do fine and those policies are critically important. Anything we can do to reduce scarcity when possible is very important. I'm glad really smart people are working on it. I weirdly like this problem where you can't, there's an absolute scarcity. You can't, you know, you can't dissolve the issue but yeah, critically important. In fact, more important than this perhaps. Other comments, questions? What? To either be appreciated what we do today in the hospital if you have critical scenario plan that is getting rid of sort of the city instead of what we do today in the hospital. And to either be made more boldly, sharp as the infractions we've heard that is then impressively used by the people that will result in the work to be carried out to be able to actually lessen the pain. Right. Yeah, so is your argument like that we should just do first come first served because that's what we do and that's what providers are used to and allow, it's better just allow some people to die sort of on the waiting list because if you actually move to a triage protocol in practice, that would be such that have so many unintended consequences perhaps moral injury and the whole system may fall apart. And yeah, I mean, that's certainly a huge concern, right? There's a lot in these Christ stands of care documents about triage teams, support systems, objective ways that these decisions will be made. So it's not the person at the bedside making the call. It's like this disembodied voice telling you whether the patient can be continue on life support or they need to come off. So that's what some people have thought might help with that problem, but no, I mean, it's a huge, huge concern. I mean, how in practices would go down this obviously my entire grant and research is at the very theoretical abstract level, even though we try to keep it informed by actual clinical experience still really far away. There's still just people like moving around in a simulation. So, you know, but that speaks to the more work that needs to happen in this area. And then hopefully it's applicability to other problems, but all right, see if I said anything else I wanted to talk about. Oh, thank you. I didn't say thank you. I was gonna get to the slide. To all of my bioethical mentors, I mentioned at the top, started with Mark Siegler's class. I think he might have had to step out, but and Laney Ross who really taught me how to think, I think properly about empirical bioethics work and how to write a paper for the clinical medical journals that's actually an ethics paper in disguise to the fellows. You can do that. In fact, you can get NIH funding for what and to make them because clinical medical ethics is clinical medicine, right? That's like a whole whole thing. So, Laney, help me do that. And then, you know, Govan, he's one of these guys who's I've been reading these papers for years, teaching me how to think through his papers and now teaching me how to think in real life, which is amazing. And then of course, Monica, as the mentor helped me make the jump sort of the next level in my career. So thank you to these four in particular as well, to the McLean community overall, just having the collaborative experience here at University of Chicago is so important to getting any of this done. So thanks. The drawing versus the folder. Guys, I think you didn't have some interesting things to say.