Crisis Standards of Care: preparing for the next pandemic

October 31, 2023

Information

October 4, 2023

Crisis Standards of Care: preparing for the next pandemic

William F. Parker, MD, PhD

Assistant Professor of Medicine and Public Health Sciences

Assistant Director, MacLean Center for Clinical Medical Ethics

Section of Pulmonary and Critical Care Medicine, University of Chicago

ID10927

To CiteDCA Citation Guide

00:00All right. So we're happening.
00:01Yes, we're happening.
00:02Fantastic. Well, welcome
00:04everybody. Thanks so much for coming.
00:06I'll speak just for a
00:07minute and then introduce
00:08our our guest for tonight.
00:10My name is Mark Mercury.
00:11I'm director of the Program
00:12for Biomedical Ethics here.
00:14And I'll start with
00:16a very brief story,
00:18very brief. So in
00:21March of 2020, when the pandemic was very,
00:24it seemed very suddenly upon us and
00:26we saw what had happened in Italy. We
00:28saw what was happening in New York.
00:30I reached out to the chief medical
00:31officer of the hospital and said,
00:33do we have a plan if we run
00:35out of stuff like ventilators?
00:37And he said, well, no, we don't, But
00:39there's some people who are working on it
00:41and I'd like you to be part of that.
00:42And I said sure.
00:43So he assembled and the the
00:46Ethics Committee leadership
00:46were here with us tonight,
00:48assembled a small group of
00:49folks who were then reporting
00:51to a large group of folks.
00:52And we very quickly,
00:55it felt very quickly,
00:55we tried to work up a price of
00:57standard of care, a triage plan.
00:59What are we going to do when there's
01:00two people who need a ventilator and
01:01we don't have only one ventilator?
01:03What exactly is the plan?
01:05And of course it wasn't just Yale.
01:06New Haven Hospital didn't
01:07have a specific plan.
01:08People all over the country
01:09were caught off guard,
01:11some more than others.
01:11We had no guidance specifically
01:13from the state.
01:14The health system need to put
01:16something together.
01:16And it was a very remarkable time for
01:18a lot of reasons. We had
01:21terrific leadership here in particular
01:22by Ben Tolch and who really organized
01:24our efforts here to come up with the
01:26crisis standards of care and many of
01:28the people who worked on those are here.
01:30But there was a lot of cooperation
01:31between people who were working on these
01:33things. And I'll tell you I was
01:34leading a double life at the time.
01:36I was chief of neonatology and
01:40and running this ethics program
01:41And so I was doing both.
01:43And one of the things that
01:45fascinated me is thankfully there
01:46was a clinical director for the
01:48newborn ICU and and an acting chief
01:50during that time as well because.
01:52But it seemed like the CDC every
01:545 minutes was coming out with new
01:57recommendations for what we're supposed
01:58to do which babies we isolate how.
02:00And thankfully the NICU was
02:01largely spared trouble from COVID.
02:03Every time you turn around CDC had new
02:06recommendations making everybody crazy.
02:07But the flip side of that when it came to
02:09the allocation of the scarce resources,
02:11when it came to crisis standards
02:13of care or triage plan.
02:15The federal government was quiet as
02:17a mouse and we were an occupancy.
02:19Where's the CDC on this one?
02:21And so we were doing our best.
02:22But what happened was there were others,
02:25some very smart people from all over
02:26the country and all over the world who
02:28were working on these same questions.
02:30And so we found each other online on Zoom,
02:33and we got help from each other a lot.
02:34And so it was during that time that
02:37I had reconnected with Will Parker,
02:40who I've known since he was
02:41a young medical student
02:43back in the day. Chicago and Will was
02:46helpful to me and we've kept in touch.
02:48And so I'm delighted that he's agreed to
02:50come here today because as you'll hear
02:51when I read his his CV,
02:53he's got some serious expertise that's
02:55going to help us because importantly,
02:58we got caught. We worked very hard,
03:00very fast to come up with some
03:01crisis standard of cares.
03:02And we built a plan.
03:04But I mean, the Ben who who leads
03:06the show would be the first to admit
03:08that our plan ain't perfect.
03:10We need this.
03:11This plan still needs work.
03:12So there's one approach which
03:14could be let's just wait
03:16until and the next pandemic is
03:17upon us and we're drowning to say,
03:19well, we should really try and
03:20figure out what we're going to do.
03:22Or maybe now between crises
03:24we can try and figure out what
03:26exactly the plan should be.
03:27So I would like for us and
03:29that's why those of you who
03:30have worked so hard on this,
03:31I would like for us to keep
03:32the conversation going and
03:34and and Mike, I appreciate you
03:35being here. You were so supportive
03:36during so much of this stuff.
03:39I know. So a lot of important people who are
03:41in that effort were here,
03:42and a lot of people who had nothing
03:43to do with that effort but may in
03:45fact be leaders for the next one.
03:47So pay attention and when you
03:49have a good idea, share it.
03:50So with that, we're going to talk
03:52about crisis standards of care,
03:53preparing for the next pandemic. Dr.
03:56Will Parker is an assistant professor
03:58of medicine and public Health Sciences
04:00and assistant director of the
04:01McLean Center for Clinical Medical
04:03Ethics at the University of Chicago.
04:06And by the way,
04:06I just have to say because
04:07I'm sweating in this thing,
04:09the mask is because I've been exposed,
04:12just found out, not because I'm sick.
04:14And so this with the CDC site
04:15assures us is the
04:17is the adequate precaution indoors.
04:19So I'll try not to get close
04:20to you, but that's what's
04:21going on. And I look around here, there's
04:22very few of us wearing the mask today.
04:24I think I look good in it. But, you know,
04:28so, so Will really brings all the skills
04:30to the place. He's a, he's a pulmonary
04:32critical care physician.
04:33He's a clinical medical ethicist.
04:35He's a health service researcher
04:36who studies the allocation
04:37of scarce medical resources.
04:40He's specifically interested in
04:41absolute scarcity problems where
04:43demand greatly exceeds supplies and
04:45algorithms triage patients for treatment.
04:48He runs an NIH and Greenwald Foundation
04:51funded quantitative bioethics lab.
04:53That's not nothing that applies advanced
04:55empirical methods to evaluate and
04:57design allocation systems according
05:00to the underlying ethical principles.
05:02This is his academic work.
05:04That sentence again,
05:06OK advanced empirical methods to evaluate
05:09and design allocation systems according
05:12to the underlying ethical Princess.
05:14Current lab projects include
05:16deceased donor organ allocations,
05:18policy life support,
05:20triage under crisis standards of care,
05:22and the allocation of novel
05:24scarce therapeutics.
05:25Will is a graduate from Williams College,
05:28and from then he's been
05:29at University of Chicago,
05:30where he got his MD,
05:31where he did his medicine residency
05:33and critical care fellowship,
05:35where he got a master's degree in public
05:37health, where he got a PhD in
05:39public health and completed a
05:41fellowship in medical ethics.
05:44So Will is the perfect guy to
05:46help guide this conversation.
05:47I'm so grateful you came
05:48all the way from Chicago.
05:49And with that, I introduced Dr.
05:51Will Parker.
05:58All right, you guys hear me.
06:00Thank you so much,
06:01Mark and program for bioethics for the
06:04invitation to give the seminars series
06:07and that really kind of reduction.
06:08I hope you guys can help me
06:10think about this problem,
06:11which I think is incredibly challenging
06:14and I've been fortunate not to get
06:16some support to to try and take it on.
06:18So hold on, just look in here,
06:21let's see if this works.
06:24You know, my,
06:25it's difficult to start these talks with.
06:27This is where I've gone and been,
06:29but I've basically been at UFC
06:30the whole time this month.
06:31OK, so it'll be very boring
06:33with this one new C slide,
06:35but there was a transformational
06:37experience that I had in medical school.
06:39I participated in the fellowship at Outreach
06:41in the study of professional ethics.
06:43This is the memorial for the murder of
06:45Jersey used in front of Europe in Berlin,
06:48where we're learning where the
06:49the current processes are.
06:50Learning about the role that
06:52the medical professionals,
06:53the medical profession at
06:54large played in the Holocaust,
06:56learning that they weren't just
06:58bystanders but in fact active perpetrators
07:00of key elements of the genocide.
07:02And this experience,
07:04as you might imagine,
07:05is not something that leaves you lightly,
07:08especially when you get to spend
07:10the week hanging out with this guy.
07:12This is what you look like then.
07:13So I think this is,
07:14this is how I remember you in my
07:16mind with a full beard and and
07:19of of full week of seminars and
07:22dedicated tutorial style ethics
07:24teaching which was really shaped
07:27the way I think about clinical
07:29medical ethics and bioethics overall.
07:31And so naturally,
07:32when I was asked to help draft
07:34a crisis standard here,
07:36a triad protocol just like Mark
07:38was saying for my hospital,
07:40I I emailed him and I was like,
07:42hey, I'm, I'm sure you Remember Me,
07:44but I've been looking at your the
07:47Your Yell protocol that you put online,
07:49and it's been a very helpful guide.
07:52And so I think this story just tells
07:56a little bit about where I where I
07:57come from and my perspective on all of this.
08:00And another amazing thing is that now
08:02one of my medical students went on
08:05Vasby this year, Mark Kevin Lazenby,
08:07who's worked in my lab,
08:09let's all come full circle.
08:12So with that hopefully Mike's anecdote aside,
08:16I just want to talk about my support
08:18and funding for this talk I have.
08:21I'm unfortunate that I have a KOA
08:23from NHLBI that is focused on
08:24the heart allocation problem.
08:26I'm not going to talk about directly today.
08:28And then also from the National
08:29Library of Medicine,
08:30the Green Wall Foundation that
08:32directly supports this work,
08:33but no other conflicts of interest.
08:36So what I hope to get through today
08:39and and open the questions and
08:40interruptions at any time is defined
08:42prices as the standards of care.
08:43What are we talking about?
08:44Right?
08:45Then go through didactically
08:46the ethical values for life
08:49support allocation in the crisis.
08:51Just make sure we're all on the same
08:53page from a normative perspective.
08:54And then finally there's four
08:56active bioethical controversies
08:57and crisis standards of care.
08:59I hope that we can pause sort of after
09:01each one and have a little discussion,
09:03because they especially need
09:04help with like the third one.
09:06And so I'm looking to get as much out
09:09of this for as seminars as I can.
09:12All right, So what do we What
09:14is crisis standards of care?
09:15Bernie Lowe,
09:16who's one of the leaders in bioethics,
09:18is probably known the most in this room.
09:21I presented it this way at a talk to those
09:25very apartment and right to the point.
09:27This is Memorial Hospital in New Orleans.
09:30Several days after the levees
09:32break and Hurricane Katrina.
09:34You can see that the hospital
09:36is completely flooded and they
09:38were losing power completely,
09:39running out of most of the resources
09:41to provide life support and hospital.
09:43And what happened in Memorial Hospital
09:45is still contentious and debated.
09:47It's been made into a Netflix series,
09:49but it clearly is not in accordance
09:52with the principles of bioethics
09:54released at several different levels.
09:56And this event and the 1st H1A1 influenza
10:00pandemic spurred the Institute of
10:02Medicine that is now called the National
10:04Academy of Medicine at the time,
10:06to form a ad hoc committee and
10:08define crisis standard of care,
10:10which is a recognition that a disaster
10:12is making it so we can't give
10:14everyone the treatment they need.
10:16In particular,
10:17we can't give them life support
10:19even if they're in acute respiratory
10:22cardiac failure and needed to
10:24prevent them from dying.
10:26So truly a tragic and horrible circumstance.
10:30So how does one approach
10:32such a terrible problem?
10:34Either when you have an acute crisis
10:36names of care like Hurricane Katrina
10:38or a perhaps subacute one with a COVID
10:41pandemic surge where the patients,
10:43as those of us who worked in
10:44the ICU that time,
10:45seemed to keep coming faster and faster
10:48each day and the panic that we were
10:50going to run out of life support rose.
10:53You know,
10:54how do we approach the stereo problem?
10:56I think it's actually one of a
10:58set of problems as Mark moved
11:00to in his introduction,
11:01a set of problems where we've
11:03we've recognized the scarcity,
11:05we've recognized that the the treatments
11:07are incredibly important and valuable
11:09and life saving and a central authority.
11:11Maybe it's a health system like Yale.
11:14Maybe it's the entire United States
11:17government in deceased or organs has
11:20taken control of the resource and
11:23is algorithmically allocating it
11:25according to an explicit protocol.
11:27So there's something written down on
11:29paper which takes patients and puts
11:31them in a list and triages the treatment.
11:34So that's the central focus of my
11:36lab and I hope the parallel between
11:39the different clinical domains,
11:40what I think is the same bioethical
11:43challenge fundamentally is clear.
11:46So how do we, how do we,
11:49starting from the ethics,
11:50how do we approach this problem?
11:52How do we construct A protocol
11:55based on what ethical principles?
11:56Where, Where to begin?
11:59I think about this this way that
12:02several several of my mentors
12:04had written and described.
12:06Govind Prasad is sort of chief probably
12:08among them and I think laying out the
12:11space of reasonable ethical principles
12:13that should be considered when you're
12:15allocating scarce healthcare resources.
12:17I think this framework has also been
12:20adapted substantially by my mentor at
12:22the University of Chicago, Monica Peek.
12:24And of course Zeke Emanuel has been
12:26involved with this from the beginning.
12:28So what I'm going to do now is just go
12:31through these four sets of of values and
12:34and describe them in greater detail.
12:37So the first is that we should treat people
12:39equally coming from respects with persons,
12:41right?
12:42We don't have enough treatment to go around.
12:45Everybody's a human being.
12:45They all need it, right?
12:47They're all in the in the
12:49case of crisis and care,
12:51they need life support and they'll
12:53die of respiratory failure.
12:55So we should treat them equally.
12:56So a lottery would do that,
12:57right?
12:58You would just randomly assign the treatment
13:01and that sort of respects this principle.
13:05So that in here lotteries and
13:08actually in a protocol too.
13:10That's in contrast with the
13:11idea of first come first serve,
13:13which is that patients queue up
13:15for treatment and then they sort
13:17of survive for as long as they can
13:19on the wait list before they get
13:21they get treated and in practice
13:23first come first served.
13:25You know,
13:25while it might be a good way to
13:27allocate dinner reservations,
13:28so we can talk about that,
13:30I think it's a pretty bad way
13:32to allocate scarce health care
13:34resources specifically because the
13:35people who end up at the front of
13:38the line usually use their socio
13:40economic advantage to get there.
13:41And one of the I think greatest reversal
13:44of the structurally racist healthcare
13:46policy in recent history was the
13:502014 change the kidney allocation system,
13:52which which Romenka,
13:53who's here at Yale was very involved
13:56with this where pre dialysis waiting
13:57time started to be counted as points
14:00for patients in the king transplant list.
14:02So let's say you'd been listed
14:04at a transplant center and you'd
14:05waited for five years,
14:07then you would when you
14:08finally got on the list,
14:08you'd get five years of credit.
14:10Before that you'd start with 0.
14:12So it's a cue,
14:14but inherently unfair and skewed
14:16towards people who can list
14:18preemptively before their kidneys fail,
14:20who are predominantly privately
14:22insured and white.
14:24And so once they fix this,
14:26this huge racial disparity in kidney
14:29transplantation rates went away overnight.
14:32So this is an example of where and the
14:35idea of treating people equally but with
14:37a with a first come first served cue
14:40doesn't actually work out in practice.
14:43So that's the first set
14:45treating people equally.
14:46The next set of principles is
14:49maximizing total benefits, right?
14:51We have a,
14:51we have a security healthcare resource.
14:53We want to use it not just sort of
14:55randomly across the population.
14:57We want to use it to maximize the benefit,
14:59which can be formalized in a bunch
15:01of different ways and just listed to
15:03save lives and save life years. Here
15:07what you can imagine what interaction,
15:12excuse me, quality just these years.
15:18So in this example you would
15:20if you wanted to save lives,
15:22you clearly would allocate to the
15:24gentleman on the on the left here who
15:26has an 80% survivor of the discharge.
15:29But if you wanted to save life years,
15:33you also have to know how old the patient is.
15:35So here we have an 80 year old with an
15:3880% survival discharge and a 40 year
15:41old with a 40% survival discharge.
15:43In this situation,
15:45if your goal had to save life years,
15:47the total number of lives gained from
15:49the resource you would allocate to the
15:51second patient because their expected
15:52life years gained from treatment,
15:54in this case with mechanical
15:56ventilator for COVID-19 pneumonia,
15:57is 20 compared to 8 to the other patient.
16:01So already the utilitarian derived idea
16:04of maximizing total benefits has some
16:06problems here because we have to specify
16:09exactly what benefits we're after.
16:11Next is this concept that there's
16:13certain people who enter the
16:15allocation being worse off, right?
16:16They've been sort of screwed over by
16:19society or by their disease process,
16:21and we should account for that in
16:24the allocation protocol we developed.
16:26Now, one idea is the rule of rescue, right?
16:29You're going to treat the person
16:31who's the sickest 1st.
16:33And of course I think we can all
16:35imagine in a crisis standards
16:37and care scenario where basically
16:38everyone will die without treatment.
16:40If you treated the sickest people with the
16:42highest predicted probability of death,
16:44that would lead to enormously low benefits,
16:47right?
16:47So while sickest first is actually used
16:50in liver allocation like the melt score,
16:52that's only because those patients actually
16:54have high benefits from transplant.
16:56In a crisis standards and care scenario,
16:59sickest first would lead to the
17:01least optimal solution in terms of
17:02with respect to maximizing benefits.
17:04So that's in general is out.
17:07So what other classes of people
17:08are worse off?
17:09Well, the Youngs,
17:10if you develop end stage organ
17:12failure or achieve respiratory
17:13failure when you're young,
17:15a life threatening medical
17:16condition and you die young,
17:18then you haven't got to
17:19live your whole life right.
17:20You haven't got to play your
17:229 innings of baseball.
17:23This is the concept of Fair innings
17:25that every person is deserve some of
17:27the full life and we should allocate
17:30resources in order to ensure that it happens.
17:32So this is a more General Healthcare
17:35allocation argument than just
17:37the absolute scarcity problem we
17:39might articulated,
17:40probably perhaps passed by Norm,
17:42Norm Daniels,
17:43but applied here,
17:45this would end up with ideas
17:47like pediatric candidates for
17:49organ transplantation should be
17:51categorically prioritized over adults,
17:53which is actually the way we do.
17:54We do things right.
17:56But there's another group of
17:58patients who are worse off,
17:59and those are people who have been
18:01structurally disadvantaged by society
18:04and in structural laws and rules.
18:08I think.
18:08I don't know if many people are having
18:10any familiar area in Chicago here,
18:12but the map's pretty clear.
18:16All of the areas that are dark here,
18:20the highest are the highest
18:23percentage of African Americans
18:24or people identify who are black.
18:27And there are also areas that have been
18:30structurally disadvantaged by du jour,
18:33structural racist policies like,
18:35and I'm going to go into this more later
18:38in the talk, like detailed
18:41well in the color of the law,
18:43color of law or redlining specifically,
18:47and we'll talk about this more.
18:49But you can imagine if you're if you're
18:51living in one of these neighbourhoods
18:52and the pandemic is hitting you unequally
18:54because the city has been designed
18:56to make your neighbourhood worse off,
18:58should we account for that?
19:00So this is the concept of favouring
19:03the disadvantage somehow in in
19:05in your allocation protocol.
19:07And finally, the last category
19:09is rewarding social usefulness,
19:11which already kind of seems a little
19:13icky when you just say it right.
19:15But we actually,
19:16in order an allocation,
19:18use this principle pretty in
19:21a very concrete and big way.
19:24So if you are a living Kitty donor
19:26and your Kitty goes on to fail,
19:27you get 4 years of waiting time points.
19:30And the idea there is that you're
19:33getting paid back for being good
19:36in the past right Reciprocity
19:38for your previous good deeds.
19:39But the the other idea
19:44here is that there's some people who are
19:46like very valuable to society, right?
19:48They have a multiplier effect,
19:51like for example a famous CEO who is
19:53a job creator or something, right?
19:56And if we should give them the
19:58resource because then they'll keep
19:59them alive and help other people.
20:01That seems, I think, why I picked a CEO.
20:05That may not be the most popular
20:06on on purpose, but that reasoning
20:11actually overwhelms.
20:13The COVID-19 vaccine allocation
20:15aside from elderly patients and long
20:18term care facilities who went first?
20:21Us. I remember getting a second
20:24dose in mid January,
20:26well before weeks before any of my
20:29vulnerable patients and I realized that
20:31the weight on instrumental value and
20:34reciprocity was severely miscalibrated.
20:35But that's a different talk I see.
20:38So hopefully what's become obvious is
20:41I've laid these values and criticisms
20:43out is that they're inherently in
20:45conflict with each other, right?
20:47There's there's certain times
20:48where they go hand in hand,
20:50but if you're trying to
20:51maximize total benefits,
20:52you're by definition not
20:54treating people equally.
20:56There's no way around that.
20:57So how do you, how do you move forward?
21:00This is terrible.
21:02Well fortunately Gobin has thought about
21:04this a lot and he's a lawyer bioethicist,
21:07not at the University of Denver but has
21:09spent a lot of time at on the East Coast.
21:11And so some of you may have come across
21:14in different times is amazing thinker
21:16and you know his point is that you
21:19have to you can't some some may be
21:22better than others and there could be
21:24arguments based on more fundamental
21:26principles that make may help you
21:27rank order the four big categories
21:29but you can't ignore them all.
21:31You can't ignore ones and you have
21:33to do the hard bioethical work
21:35to combine them with the multi
21:37principle allocation systems.
21:38And I think that's very much true.
21:40And we'll see as we look at some
21:42examples of crisis standards of
21:43care and attempts to do just this,
21:45invoke multiple ethically relevant
21:48principles into a protocol.
21:51OK, I got through that quicker than I hoped,
21:53which is good because now now we
21:56get to the hard part, which is OK,
21:58what are the key bioethical controversies?
22:00And you know,
22:01where this is the bioethic seminar.
22:03So we're gonna focus on the
22:06life support triage protocols,
22:08a lot of hypothetical situations,
22:10and engage these deep,
22:12deep bioethical issues.
22:13I want to say there's entirely
22:15another set of equally important,
22:17maybe even more important,
22:19practical considerations during crisis,
22:21tangent care and procedural
22:22considerations about load sharing and
22:25how how would the triage team work,
22:27for example?
22:27I'm going to set those all aside so
22:29we can just kind of do more thought
22:31experiment stuff because that's
22:32where we're doing the bio or do it.
22:34We're bioethics tonight, right?
22:38So these are the four big problems
22:40and I'm hoping maybe we can just
22:42pause after each one for a brief
22:44round of discussion.
22:45We never end up getting whatever.
22:46I'm worrying when I get to the last one,
22:48and that's perhaps the most
22:49important that I think we need to
22:51resolve before the next pandemic.
22:53We need to improve crisis standards of
22:55care and deal with these four questions.
22:57So first let's take down sofa together, huh?
23:01I think we have a lot of friendly
23:04people in this room for this particular
23:07point so early in the pandemic.
23:09Gina Pistacello is now in rush.
23:10She's emerging leader in the serious
23:13illness conversation space or so.
23:16She's now at Pittsburgh, excuse me,
23:17since Pittsburgh, Pittsburgh.
23:18So watch out for what she's going to do
23:21next in terms of clinical medical ethics.
23:24She read every single state crisis
23:27standard of care protocol in like a week,
23:31accurately categorized them,
23:32convinced like three other people
23:34to check everything she did,
23:36and and published the My Eyes Cited
23:39paper ever the landmark survey of
23:43US ventilator allocation guidelines.
23:46And what we found is that everybody,
23:48for the most part, was using SOFA.
23:50And I'm a pulmonary critical care doctor,
23:52so I knew what SOFA was.
23:54And this is what, of course,
23:55we started to write into our algorithm too.
23:58And here's an example of the
24:01way SOFA was going to be used.
24:04This is from Pennsylvania.
24:05It's still on their website.
24:07A lot of these are still on the website,
24:09even though they've been.
24:10We, as we as all show,
24:11we've moved on in a big way
24:13for some of these ideas.
24:15But in order to save the most lives,
24:18remember that's the ethical principle.
24:20We're going to divide people up into
24:23categories based on their sofa score.
24:25And I'll explain what the Sofa score is.
24:26The next slide where if the sofa scores
24:29higher, then we're likely to die,
24:31right?
24:31So they get more points and it's like,
24:33oh, you want less points,
24:35lower score is better and people will
24:37be rank ordered by their scores.
24:39And one interesting thing that Mark and
24:41I were talking about on the way over
24:43here is by bidding sofa scores together,
24:45what you're doing is allowing tie
24:46Breakers to kind of kick in more, right?
24:48So all right, if you have the same points,
24:51two points, and and this primary calculation,
24:54then we start to do other considerations,
24:57life cycle considerations or
24:59fair endings considerations.
25:00But hopefully you guys can all
25:02appreciate how this is an attempt
25:04to take those ethical values and
25:06principles I discussed and force
25:08it into an actual protocol that
25:10could be used in in real life.
25:12So
25:15what I'm going to focus on is the sofa score.
25:18And the problem with the sofa score,
25:21the sequential organ failure
25:23assessment score is old.
25:24It's almost 30 years old now and
25:26it's based on expert opinion.
25:28So this table, which I see a lot
25:30of people squinting their eyes
25:31glazing over and I don't blame you,
25:33was made-up in the 90s at a
25:35critical care conference.
25:38It's not based on a regression model that
25:41this is to predict like the Apache 2 score,
25:44LEPS 2 score or LPS score.
25:47Both of those are predictive
25:49models designed to predict the
25:50outcome Survival ties to discharge,
25:52not so far, they just made it up.
25:54So it's actually kind of remarkable
25:55it predicts anything, right,
25:56because that means that means we must
25:58know what we're doing in stockers.
25:59So the the this first column is
26:02the degree of respiratory failure
26:04and the more the lower your PA,
26:06O2, FI, O2 ratio is the the work
26:10of hypoxic respiratory failure.
26:11So that's the first column In the
26:14in the third column here or the 4th
26:15column you'll see this cardiovascular
26:17column which is supposed to
26:18measure the severity of shock.
26:20And for those again in the critical
26:22care space or anybody who's really
26:23worked in a in a in a hospital,
26:25well, we don't use that much,
26:27don't need anymore for very good reasons
26:29and we have a lot of other vasoactive
26:31medicines that are not listed on that row.
26:33Speaking to that in practice people
26:35do not calculate this according to
26:37their original formula in any way.
26:40And but that being said about
26:42all those potential problems,
26:44it actually works pretty well
26:45for patients already in the ICU.
26:47If you make a couple corrections
26:49in that cardiovascular component,
26:51you calculate it.
26:52And if someone's been in the
26:53ICU for 48 hours and you have
26:56time to calculate all those,
26:57get all those laboratory measurements,
26:59calculate the score and take the
27:01maximum and worst value in all of them,
27:03it works pretty well.
27:05So this is the SOFA scores on the X axis.
27:07This is a large population of the
27:09patients with susceptive infection
27:10in Australia and New Zealand.
27:12ICU and the locality should have
27:14like logistic function right then.
27:16The higher sofa score,
27:18the more likelier to die,
27:19each one of these points turning into
27:22like a 5% or so increase in mortality.
27:26However, that's not the triage situation.
27:29That's the triage situation is
27:31that the patients in front of you,
27:33you have much you don't have
27:3548 hours of information of them
27:37already receiving life support.
27:38You have to decide whether or not to put
27:40them on life support in the 1st place.
27:41So
27:46when you actually evaluate it as a triage
27:48score, SOFA performs quite poorly.
27:50So this is the area of the receiver
27:53under the receiver operating curve
27:55or measure of discrimination.
27:56A coin flip is, you know,
27:58this has this dotted line here and
28:01as you can see sofa's not doing
28:03much better than flipping a coin.
28:04It's a near sort of lottery situation.
28:08And so this was a landmark paper that I
28:11think casts a lot of doubts about using
28:12SOFA in the crisis Standards of Care
28:14is that it doesn't work well in this,
28:16in the situation that people are applying.
28:19And on top of that,
28:22the SOFA score would exacerbate
28:24health inequity.
28:25It doesn't incorporate age,
28:26which we'll talk about next,
28:28but it also uses the patient's absolute
28:30value of creatinine to compute a renal score,
28:33right.
28:33So this is problematic for two reasons.
28:36One, some patients end up or show up to
28:38the hospital with chronic kidney disease,
28:40so they have higher creatinines at baseline,
28:42but it's not an acute problem and
28:45they might walk in the door with
28:47like two or three cell phone points
28:49just 'cause they have chronic kidney
28:51disease that's in no way correlated to
28:53their probability of actually dying.
28:55And
28:57the second problem is that certain
29:00populations with higher muscle mass,
29:02particularly those people who
29:03are self identified black,
29:05have higher creatinine bodies.
29:06This is this whole estimated GFR controversy,
29:09why race was used in the
29:11equation to begin with.
29:13And so the same patient with the same
29:15amount of renal function might get who's
29:18black might get two points compared
29:20to one for somebody who's white.
29:24So a lot of people have gone on
29:29to examine the potential bias of
29:32surface core against black patients.
29:34Most notable here at Yale,
29:38where I was very inspired by both
29:40of these papers to replicate your
29:42findings in the EICU data set.
29:44I don't know when they were polished,
29:45but you know I was very they're
29:47all about the same time.
29:48We all were thinking alike,
29:50and we all show that black patients
29:52would have higher SOPA scores than white
29:55patients with the same survival, right.
29:57So instead of giving because of
29:59that chronic kidney disease point
30:01or the OR the creatinine point,
30:03a white person will get a sofa of or be
30:05more likely to allocate a ventilator.
30:08Black person will get a SOFA score of five.
30:10So that's a a form of actual
30:12statistical bias, right,
30:14It's it's miscalibrated.
30:15So it was miscalibrated against
30:18patients who identified as black.
30:21And this is a big,
30:22this is a really nice figure from
30:24Deepishana's version of this paper,
30:28which was using pen and cosmic
30:30fermente data and they show that 10%
30:34of black patients would be assigned
30:36to inappropriate SOFA level, right.
30:38So it would effect on 10% of them
30:40and most of the city's patients
30:42would be shunted into these higher,
30:44higher groups.
30:45And we found the same thing that for
30:48conditional upon their assigned priority,
30:51black patients are much more
30:52likely to survive.
30:53So it's a little confusing,
30:55but basically the score is assigning
30:58a higher level of mortality risk to
31:01black patients than they actually have,
31:03which is a form of bias that leads to both.
31:05Obviously it's discriminatory
31:06and it's black people,
31:08but it's also inefficient because
31:11it's worse at identifying survivors.
31:14So the really we took a a
31:19population with COVID-19,
31:20a lot of the pre prior studies were
31:22pre you know like the pandemic was
31:24still going on so there wasn't
31:25a lot of COVID data.
31:26So this is the same sort of analysis,
31:29but in patients who had COVID-19
31:32required mechanical ventilator.
31:33We also added a little bit more,
31:35met the logic breaker here with
31:37a very simple simulation where
31:38we applied triage rules.
31:42And when we did that, unsurprisingly,
31:44we found that using a silicate
31:47tier system would systematically
31:50disadvantage individuals who identified
31:52as black without improving efficiency.
31:57In fact, it performed substantially
31:59worse than young is first or a
32:03combination model and not as you
32:06can see in the lottery system.
32:08Black and Hispanic people,
32:09although it's not significant actually
32:11have higher survival than white patients.
32:13And that's because white patients
32:15who end up in respiratory failure
32:17with COVID-19 throughout the
32:18pandemic tended to be much older,
32:20which will be the next topic
32:23of the discussion.
32:24And unfortunately because of some actions,
32:26misguided actions,
32:27I believe by, you know,
32:29office civil rights from the Department
32:30of Health and Human Services,
32:31SOFA is now even more dominant in
32:33crisis standards care protocols across
32:35the country than it used to be.
32:36This is a paper from May 2022
32:40and most states now have one.
32:42You know,
32:43remember our first map had a lot more
32:45holes because everyone was scrambling.
32:47Now it's still in some states had no plan.
32:49I don't know.
32:51And the plan is SOFA for the vast
32:54majority of these sofa in various forms
32:57with little other elements of the protocol.
33:00So I think this is really deeply
33:02problematic and one of the things is the
33:04main gap our grant is trying to fail.
33:08So in conclusion you know so I
33:11think I've said all this sofa's
33:13outdated it's not a triage score.
33:15It's less accurate than the Young's
33:17first and statistically diet but
33:19advised means black patients which
33:20makes it even more inaccurate.
33:22So I I think SOFA,
33:24you know should be eliminated in
33:26crisis and secure protocols across the
33:28country and replaced with a better
33:29triage score than one that we're working on.
33:34So that's the first problem.
33:37I obviously have awesome
33:39strong opinionated conclusions.
33:40I'm not asking for someone to defend.
33:42So Mark, I don't know if you want
33:44me to move on to age or if if
33:45you want to have any questions or
33:47feedback just about that this this.
33:49So let me, because I didn't,
33:51I I didn't do my job beautifully
33:54at the beginning,
33:55which was to remind you guys and
33:56let you know that we're going to go,
33:58we're going to go until 6:30 and then
34:01there's going to be a hard stop.
34:02So I apologize if there's something
34:04you really wanted to ask or say
34:05and you didn't get the chance. However,
34:08typically the speaker goes in total 5:50 or
34:115:00 to 6:00 and then we
34:12have questions, but the way
34:13Will's outlined this week and kind
34:15of stop at each of these important
34:16points and have a conversation.
34:18So I would say if someone wants
34:19to speak specifically to the
34:20sofa issue now we can do that.
34:22But I want to tell you one other thing,
34:24since I'm up here and have
34:26the podium, Karen Cold,
34:27who organizes these things so nicely,
34:29is herself out sick.
34:30So we wish Karen a speedy recovery.
34:33She reminds me to remind the people in
34:35Zoom land this number which Karen, please
34:37add it to the chat. Also
34:42203-442-9435, that's the
34:43number to get your CME credit.
34:472O3442, 9435 out of town, please call
34:49collect. No, that's not right. And
34:51the code is
34:56409-624-0962. So that's enough
34:58housekeeping. I think it's
34:59fine. Mark has something
35:00he wants to say about sofa.
35:01So why don't we spend like because
35:03I know, I know will you would
35:04want to get to the other problem.
35:05So let's do, let's spend 5 minutes
35:07talking about sofa and then move
35:08on to the next. Go ahead, Mark.
35:17Yeah, I I I should also say thanks
35:18so much Amir that especially for
35:20the folks who are on the Zoom call,
35:21Please wait till you get the microphone.
35:24I should have thought of that.
35:24Thank you Amir. So, so
35:28yeah, I I think we did a good job so far.
35:31So it can be replaced with
35:32something and I'm excited to hear
35:34about why you're working on.
35:36But pending that you know I, I,
35:39I in parts of some sessions where we
35:41talked about other severity illness
35:43sports and I I thought practically
35:45software was chosen but very simple.
35:47But even if you look at Apache
35:50two and talk to the developers,
35:52they said these were population
35:54statistics to sort of adjust in
35:56large clinical trials and things like
35:58that. They weren't really intended
36:00to be a bed side test. Yeah.
36:03The question about how an individual
36:05person was going to do So do we
36:07anticipate that there are any trials
36:08that would actually work well in an or,
36:12you know, measure that you can
36:14use for individual patients?
36:15That would be, yeah,
36:16I know the next, the next topic
36:18when we start talking about age,
36:19I think a score, you know,
36:21kind of giving away when we're talking
36:23about a score that includes age and
36:25several important clinical predictors.
36:26Like is the patient in shock,
36:27the degree of their hypoxia,
36:28respiratory failure in combination
36:30with perhaps having a four hour trial
36:34period on life support to collect
36:36more data that if we fit a score,
36:39a multigradable prediction model
36:40of that Haitian population.
36:42I think we can get something that's
36:44parsimonious that doesn't require a lot
36:46of heavy duty calculation trying to
36:48avoid sort of deep learning AI approaches,
36:50which I'm always very excited about.
36:52But it I think in practice like you said,
36:55SOFA was chosen because it's practical.
36:57We can sort of see how someone
36:59can calculate the bedside.
37:00Although if you ever look at those
37:02SOFA scores that are epic and then
37:04you look at the actual numbers,
37:06they're very discordant.
37:07So I think SOFA is actually
37:10fairly complicated to calculate.
37:13So you know personality is
37:15not a partners for.
37:16I mean the related thing is one of the things
37:20we working on this system is trajectory.
37:24You know,
37:25you you see somebody getting
37:26better and somebody getting worse.
37:27And that's.
37:28Yeah,
37:29no,
37:29that's why we really should try
37:30to get to the platform because
37:32then that's that's exactly right.
37:34I think one of a lot of the
37:36thought experiments around this,
37:38imagine a bunch of patients in
37:39a room with the one ventilator,
37:41and that's not the situation at all.
37:43You have is population of ICU patients.
37:45And once they're in the ICU,
37:47you can actually run much more
37:49complicated prediction models.
37:50You have a lot more information.
37:52You might be able to know very
37:54specifically what their survival's
37:55gonna be with a lot more certainty
37:57than that person who just showed up.
37:59Right.
38:01Good. Can't we have someone here?
38:02Fight for the sofa. Come on. Anybody.
38:05Nobody wants to do that. No expense, OK.
38:08The Ben's not going to have time.
38:10Yeah, it's gone.
38:10Move it on. Right.
38:12OK How about how old is somebody?
38:15Can we can we use, can we use that?
38:18Obviously a fair innings.
38:18A ******** fair innings.
38:20Prudential Lifespan Equity
38:21person would say yes.
38:23But we live in America,
38:26so it's a little bit more complicated.
38:29This was Utah's triage
38:31score before the pandemic.
38:32They actually were one of the rare
38:34states that had like something
38:35written down like New York did.
38:36New York was just all based on sofa.
38:39I don't know if everyone knows that story,
38:41but most studies never activating it.
38:42But
38:46Utah's career school score
38:47has estimated survival,
38:49so saving lives is protocolized,
38:52explicitly right?
38:533 bins and sort of equally as
38:56important as how old someone is.
38:58So are they. They're less than 30 years old.
39:01They get only one point.
39:02If they're over the over 60,
39:03they get three points.
39:05So being over 60 is the same as having
39:08less than a 10% chance of survival.
39:10So this is a very large, I would argue,
39:14fair innings weight in this protocol.
39:18Not that this was none of this is
39:21explicitly argued from bio in perspective.
39:23Like it just sort of somebody writes
39:25it down and then you can kind of
39:27see which is what I think is so
39:29interesting about quantitative biotics.
39:30But then this mid category is
39:32kind of problematic too, right?
39:34It's, it's an ASA score,
39:35so it's capturing the patient's
39:37chronic disease state,
39:38but it's a different access
39:40than estimated survival, right.
39:42So the idea is that people who are,
39:44the problem potentially with this
39:46is that people who have disease are
39:49somehow less deserving of the resource,
39:51right?
39:52That's what this is kind of implying,
39:55because if these factors matter
39:57for their Bible to discharge,
40:00they would be incorporated in
40:01this bottom column, right?
40:02And if these factors are
40:05about life expectancy, OK,
40:07And then you can sort of see how
40:08these would be combined together.
40:10It's still a fair innings argument,
40:11potentially.
40:11Not really, though,
40:12because what if you're a child?
40:14This is yours.
40:17This is a little muddled both
40:19bioethically and practically.
40:20And so protocols like this cause a
40:25lot of action over the summer after
40:28our initial waves by the Department
40:31of Health and Human Services Office
40:33of Civil Rights where they sort
40:35of went through all the CSCS and
40:37stripped out mention of age or
40:39disability in a primary score and
40:41even sometimes in the secondary score,
40:43a tiebreaker.
40:44So This is why that map is
40:47all sofa only sofa,
40:49because all considerations of age or
40:51disability were essentially removed.
40:53I think Doug White was able to keep
40:55like his tiebreaker in there somehow.
40:57But you know, in general,
40:59age was dramatically deprioritized
41:01from the OR removed from these
41:03protocols where using age to
41:05decide how you're going to triage
41:07was essentially from a regulation
41:10standpoint made impossible.
41:11So they did this in like 10
41:13different States and this is the
41:14type of language they would use,
41:16move on to life expectancy,
41:18categorical exclusion based on age,
41:20disability and functional impairment.
41:21There's a lot of concern in the
41:23disability community that there would
41:25be explicit discrimination against
41:26patients with chronic physical
41:29or neurological disabilities.
41:31Impairment and like would take
41:32ventilators away from people
41:33who are chronically ventilated,
41:35for example,
41:37and make sure that people with
41:39disabilities are valued based
41:40on their actual mortality risk,
41:42not the value of their life or their,
41:45you know, sort of qualities remaining.
41:46Right. And so apparently they changed.
41:50Utah, changed their plan.
41:52But when I clicked on the link,
41:53it's broken. I did a lot of
41:54searching last night. I'm like,
41:55oh, what did they change it to?
41:56But it's probably just this bottom,
42:01the bottom one. Now this is
42:02kind of like well payment,
42:04so we're going to have to
42:05really worry about it.
42:05But I assume state of Utah is
42:08just about estimated survival
42:10and throwing all these these
42:12other considerations out. So I
42:16want to talk about
42:16the two potential ethical justifications
42:18for using age, and this is a good
42:20time to have some discussion.
42:22The first idea is that the value
42:24of younger lives is higher.
42:26This of course has been sort of explicitly
42:29rejected by the previous administration's
42:31Health and Human Services department.
42:33But, you know, this is justified.
42:35And this fits into the idea that
42:37younger lives in general, not always,
42:39but have more like years to gain, right?
42:42If you're like like a previous example,
42:44if you're 40 years old,
42:45even if you have a higher
42:47probability of short term mortality,
42:49you're much more likely to gain
42:51more life years with treatment
42:53than some others in their 80s.
42:55And then the second idea,
42:56as we discussed 4,
42:57is that younger lives really are
42:59higher in terms of that they haven't
43:01got to play in their 90s at baseball.
43:04So we owe them because they're worse off.
43:08But there's another reason to
43:10use age in a triage war.
43:12And that age is a strong independent
43:14predictor of short term survival.
43:16Who was most likely to die
43:18from COVID the elderly?
43:19Who did we allocate COVID vaccines to?
43:221st the elderly?
43:25We used age because it was a
43:28tremendous predictor of benefit
43:31from COVID-19 vaccination.
43:33The converse is true here that
43:35younger patients are much more likely
43:38to benefit to survive from life
43:41support if they develop respiratory
43:43failure or chronic respiratory
43:45failure or chronic failure.
43:47So you need to use age if you
43:49want to save the most lives.
43:51We don't have an alternative number.
43:53That's the practical thing
43:54that we can do on the bedside.
43:56And this is some data that
43:58we have under review.
44:00We we presented ATS,
44:03the American Thrust Society
44:04conference last spring,
44:05so I'll walk you through it.
44:07The X axis is how old the person went and
44:10this is the population of like 90% COVID,
44:1210%.
44:13Others supposed to simulate a pandemic surge.
44:15And then the black bars are what
44:18percentage of them actually died.
44:19So as you can see yes,
44:21people get older.
44:22The probability of death goes up the
44:25the red bars are their predicted
44:27mortality by sofa score of all.
44:29And remember we've defined this as
44:31a crisis standard care population.
44:33So they're all quite they're pretty
44:34sick and they have higher sofa scores
44:36and it just all is pretty much the same.
44:39So the red bars are all the same.
44:41But if you make a new model that
44:43incorporates both sofa and age,
44:45you're much more accurate.
44:46You're actually predicting
44:47who's going to survive.
44:48And any critical care physician
44:49in the room would say,
44:50I'd much rather have a patient who's
44:5240 with a Silva of eight than an
44:5480 year old with a Silva of three.
44:56Right.
44:57That age tells you so much
45:00clinically about someone's ability
45:01to survive critical fullness.
45:04This is nothing new.
45:04It's why I'm trying to get this
45:06published cause the critical care
45:07journal's like this is obvious.
45:08This is why age is in Apache
45:10and all those
45:11other scores and then you know like what's
45:13all this ethics stuff in the discussion.
45:15But we'll we'll get there.
45:16We'll get there.
45:18Why are you talking about law?
45:20Like what's what is happening
45:21in this paper Bud?
45:23I think you know it sort of jumps off
45:25the the page to me that you know if
45:27your if your goal is to save the most lives,
45:30you have to use age,
45:31just like we use age to distribute vaccines.
45:34So I think there's a robust
45:37ethical justification.
45:38And even Dan Salmaisy who
45:40used to be in Chicago,
45:41who's really against fair
45:43innings and saving life years,
45:45concedes his first point,
45:46that using age as one,
45:47as one variable among many to save
45:49lives as a robust justification.
45:51If you remove age from
45:52life support allocation,
45:53I would say that's like anti young ages.
45:56I'm almost like you're penalizing,
45:58you're you're saying the lives of younger
46:00people are less valuable than older people.
46:02I would argue that's what our
46:04current trans protocols would do.
46:05And then finally, you know,
46:08all these ideas, fair things,
46:10parental lifespan,
46:10equity saving lives have broad appeal.
46:13And I would argue that CSCS ignoring
46:14these ideas are problematic.
46:16And the nice thing is if you
46:18just use it to save lives,
46:20you get, you know,
46:21kind of knock on benefits across
46:22these other principles, right?
46:23They tend to go together.
46:26So even though your objective with the
46:28protocol could be to save the most lives,
46:30there will be sort of secondary
46:32benefits for the other balance.
46:35So that's age.
46:37I'd like to hear people's thoughts
46:39and comments about using agency.
46:42S ES Ben. Oh yeah. Sorry.
46:45Wait probably wait for that.
46:48So, so I I
46:49strongly agree with the
46:52argument for using a based on
46:56predictive value when when we turned
46:59away from sofa aid was was definitely,
47:03you know the the comparator
47:05we were looking at was more
47:08accurate in in our community.
47:11The white patients were
47:14just fortunately older than
47:16David Doss. Yeah same thing.
47:18So it would have been you
47:20know the perspective of
47:22racial equity would have
47:23been better than sofa.
47:27Yeah and so it's and and
47:29also it was much easier.
47:31We didn't we couldn't put
47:33together triage things just
47:35from a feasibility perspective.
47:36Age would have been needed to be
47:39right and age is of course not chronological
47:41age is surrogate for biological age.
47:44There's like they're you know it's
47:46imperfect right. But it's something
47:47that's verifiable and easy. Yeah.
47:49I was hoping that age plus sofa
47:51score would debias it. It doesn't.
47:53You have to do something else.
47:54I'll show you later on we get there.
47:56But so there's still a sofa.
48:00Score's bias is so severe even if you
48:02account for the fact that black and
48:04Hispanic patients are younger and and
48:06in the in the predictive score you
48:07still have to over with the disparity.
48:09So that's but I think as we said
48:11the most logical thing to do
48:13is throw soap out completely.
48:15You can build a new score.
48:16We're trying out SEPA severity
48:18illness plus age 'cause we don't
48:20want to start with age that like
48:22trigger the anti ageist people like.
48:24So that's what we're starting.
48:28One other point,
48:30even with one national triage that I'm,
48:32I'm aware of that we've done
48:33recently with vaccines,
48:36age was universally accepted, right?
48:39It's bizarre to me that it was so
48:43widely accepted and uncontroversial
48:44in the allocation of vaccines
48:46that which has been so,
48:48so controversial in ICU allocation.
48:52Yeah, I think part of it is that
48:53if you don't allocate someone
48:54life support and needs it,
48:55they just will die immediately.
48:57Whereas young people, you know,
48:59most of them just were able to wait.
49:01You guys can wait and get their
49:02vaccine later on and they survive
49:04except for the ones who didn't, right.
49:05And and I think there were you
49:08know there was there was trade-offs
49:10there with that decision of 65 plus,
49:13right for for vaccines obviously I think
49:15they were justified because we saved
49:17a lot more lives by vaccinating the
49:19elderly people than people under 65.
49:20But make no mistake that was a
49:22choice and there were a lot of
49:24people who were sixty with diabetes
49:26would Incarnate settings who died of
49:28COVID and waited for their vaccine.
49:32So any other comments on age just
49:35just I I haven't not in the past I
49:38might not see you next could you just
49:40clarify for us now so so you you make
49:42a good argument for using age and
49:44I I I agree with that too but can
49:47you just clarify for us where the
49:49federal government stands on this now.
49:51Well, it's a new administration,
49:53presumably there's been some shake up
49:56this hasn't this actually never went
49:58to court And then Scoben's explained
49:59this to me like 5 times with the law
50:01of stuff and then it was screwed up.
50:02But it's never been litigated.
50:04So it's not like it's gone to court,
50:05federal court and they've said the
50:07using age in the CSC violates the age,
50:10just anti Age Discrimination Act
50:12of 1976 or whatever.
50:14And then also like from a
50:17constitutional perspective,
50:18age is not a protected class in the
50:20same way as race and ethnicity is.
50:22So a, a,
50:23a state could presumably pass a
50:25law that says we care about saving
50:28life years and that would hold up,
50:31although none of this he has like
50:32a huge law review article on that
50:34can't make sense on this. So.
50:35So yeah, that's where it is now.
50:37I I don't think.
50:38I think the first step from a
50:40research perspective and bioethical
50:42perspective is just to kind of like
50:45hammer this home in the literature,
50:46right, And just show like, hey,
50:48this is if you're making a triage
50:50for how old someone is,
50:51is critically important.
50:52And hopefully the weight of that evidence
50:56will effect policy down the line.
50:57But we're really far away from having
50:59sensible CSC policy these days.
51:01And just a quick note,
51:02because I think a lot of people
51:03in the room know,
51:03certainly other people who are
51:05working on our policy here
51:06because in the pediatric world,
51:07the sofa isn't really for the kids.
51:10So we used a different store called
51:11the PLA Two and for newborns
51:13there was nothing available.
51:14So we actually sort of jury rigged
51:16something for the purposes of our.
51:17Yeah, our here. Yeah, You know, they don't.
51:21So there we go first.
51:22Yeah, favouring the young person,
51:25the old becomes hugely important
51:27when for example here in Yale it's
51:29essentially the same ventilators that
51:30we use for the 80 year olds and two
51:32year olds and the 23 week preterm baby.
51:34Now that may be changed by the time the next.
51:36Sure. Yeah it's a little,
51:37I guess it's a little U-shaped
51:39in the sense that a 23 week old,
51:42you know may may have a you
51:44know 50% mortality or something.
51:45So that that that those types of
51:48considerations would happen and obviously
51:50with with COVID since infecting you
51:53know 99% adults then in terms of
51:56causing critical illness anyway,
51:58we kind of got a free pass on that
52:00but that's another issue with the
52:01age that we have to deal with right.
52:02Thanks. Here's another question.
52:06So I'm a first year my student never
52:08heard of sofa before this but I would just
52:10hopefully you'll never hear again.
52:11No, it's going to be around.
52:12It's been around for 30 years.
52:13People like people would go stick or sofa for
52:16This is why?
52:24Well this is that's a great question.
52:26This is a cohort defined
52:28as critically ill people.
52:30So everyone here needed a
52:32ventilator or needed basolactin
52:34medications to treat their shock.
52:36So the by construction
52:37this is a sick population,
52:39the population that you would
52:41be running crisis standard care
52:43protocol like you have to have,
52:44you have to have at least like
52:46a SOFA by construction of three
52:47or four if you think about the
52:50score in order to get in there.
52:51But yes, there isn't as much correlation
52:54between age and sofa as you'd expect.
52:56But remember this is part of the problem of
52:58it just it's before the life support starts.
53:02So you're just using like how how bad was
53:06there pulse oximetry to its own problems too,
53:10right before they started on the ventilator.
53:12And so everyone's was bad.
53:14The old people and young people
53:16are about the same right there.
53:18If you recalculated everything,
53:20couple days into the ICU stay age and
53:23silk would start to be more better
53:25correlating and like you would see this red,
53:27these red lines kind of
53:28**** a little like that.
53:29Does that make sense? Yeah.
53:31Thank you.
53:32All right,
53:32great question.
53:32So,
53:41so with age, the issue is so I
53:43work in the emergency department,
53:45they tell me about 30 patients of theirs
53:46and stuff like this is an 80 year old,
53:48blah blah. And I said stop.
53:49Is this an 80 year old who was playing
53:52golf earlier today or is this an 80 year
53:54old who scrolled up in a ball in the
53:55nursing home with three times of cancer?
53:57Because it's very different.
53:59But then immediately we're interested
54:00in ableism, you know, And so you know,
54:04in medicine we rarely think about
54:06age in any other situation.
54:08We're always talking about functional status.
54:09For 20 or 30 years,
54:11it's all about functional status.
54:13But then I just get, you know,
54:15twirled up and and and stuck.
54:17Yeah, age doesn't even want his own.
54:19But it it should matter.
54:21Yeah, I think it's obviously there's for
54:24any given numerical age someone is 80,
54:27let's say that that has there's a
54:30distribution of what that means
54:32for surviving critical illness.
54:34Some 80 year olds probably are
54:37actually like the average 70
54:39year old or 65 year old, right.
54:42But that being said,
54:44I think you using the average
54:46value for the average 80 year old,
54:48so including your two extreme examples
54:51in the middle is part of the triage
54:54score is ethically justified because
54:56our goal is to save the most lives.
54:58And you know if you look at
55:00the relationship between COVID
55:05anti pneumonia or critical illness
55:08in general and survival or mortality,
55:11it just stopped like that.
55:13So particularly after 80 is
55:16when things really skyrocket.
55:18But no, your point is well taken.
55:21Chronological age is an imperfect
55:23variable here but I would
55:25argue it's one complicated or
55:29question I have is you
55:30guys so much more did that
55:31was the type of the cycle plan
55:32as well as the subdivsines
55:35that's there. Yeah.
55:36The sofa score does not have any history.
55:39The the sofa score is just based
55:41on lab values and mild signs and
55:44medications that the patient's receiving.
55:46So the sofa score does not you know,
55:49which is nice.
55:50This is why he's appealing, right.
55:51It's like this kind of objective
55:54descrip description the patient's
55:56physiological state without any
55:58stigmatizing points for their past medical
56:00history or other medical conditions
56:05And So what
56:05if you are fit with other than that
56:08is is the fact that we do have other
56:10than that the which the substance use
56:13as well as many on the site that we're
56:17dealing with after that the COVID.
56:19So I always call people to play into
56:22evaluating persons coming into the
56:23hospital where they can save them
56:25because they're younger and they got it,
56:28it might go out and do something with the
56:31guidance, right.
56:32This gets into what benefit
56:33are you trying to maximize?
56:35Is there should you think about
56:37other things than just who's
56:39alive at the end of the day,
56:41I think it's really tricky when you
56:43start to do quality adjusted life years
56:45calculation and cost effectiveness.
56:46People love to do right,
56:48Discounting certain types of life,
56:51you know, how do you even assign that value
56:53if someone with substance use disorder,
56:56should they have like the
56:58priority lower about 20%?
56:59I don't think that that's hard to build
57:01that and justification to that nice thing
57:03about lives just sort of objectively,
57:06I mean treats back to
57:07treating people equally.
57:08Everyone's a person even if you
57:10have chronic medical conditions.
57:13One thing you mentioned
57:14right at the beginning of your
57:17talk and wondering if this might be
57:20accurate is is user regression model.
57:23Instead of having a triage store where
57:27you with a triage store you're making
57:30arbitrary decisions about what categories
57:33can predict mortality, and with a
57:38regression model you find out what.
57:41You find out what factors predict mortality.
57:44And it may be that in certain
57:48cases age is important.
57:50But you know, I guess is if
57:54somebody comes in short of breath,
57:57it's going to be way more important
57:59that they have in your renal failure
58:01than if they're 70 years old.
58:03I mean the the age is going to be,
58:07well, maybe,
58:08but that's what the regression will test.
58:10So that's exactly right and that's
58:11exactly the approach we're taking.
58:13We're developing a development data
58:14set where we're fitting a multi
58:16variable prediction model which will
58:17probably just be a simple regression.
58:19From that regression,
58:20we'll make a triage score.
58:23All it does is all you do is
58:24convert the predictions from the
58:26model to numbers that's what.
58:28And then propose that and the relative
58:31weight of age to an urog renal
58:34failure will be something I make up.
58:37It'll be based on the you know
58:39Cisco relationship between
58:40those variables and the outcome.
58:41So yeah thanks for that comment.
58:43That's that's perfect.
58:44That's the plan.
58:45And that makes this prevents us
58:47from being anti ageist, right.
58:48Because that's just what the,
58:51you know the the fiscal relationship
58:53between age and ICU survival,
58:55controlling a pot for all the
58:57other important medical variables
58:59that we can measure at the time.
59:00It's not age alone.
59:02So I I I want.
59:04I'm worried that in in my mind that
59:06maybe and some others it's easy to
59:08complete 22 important but separate issues.
59:11One is that the age is going
59:13to predict survival.
59:14But there's the separate,
59:15the fair.
59:16The fair eatings argument is really
59:18a separate discussion isn't it?
59:19It's not just about how age predictions,
59:21even if two individuals
59:23with the exact same likelihood of
59:25surviving COVID or whatever it is,
59:26one is 80 and one is 30, Those of us,
59:29and I'm with those who advocate
59:31for the fair eatings argument,
59:32we still say that then we should
59:34favour the 30 year old over the 80
59:36year old regardless of the predicted,
59:39the predicted mortality is the same.
59:42Yeah, I mean that's what I was trying to
59:45do with this slide is sort of separate
59:47those two out and prevent that conflation.
59:49You know, I think that for for
59:51those of us who are closet fairings,
59:53people like Mossad, we,
59:56I think we just make this argument right.
59:58The second one,
59:59the one that is much harder to
01:00:01push back against because we can
01:00:03forget we can fit regression models.
01:00:05We can.
01:00:06Isolate the independent prediction
01:00:08of a effective age,
01:00:10and I should have said controlling for
01:00:12all other measurable clinical variables
01:00:15that we can gather and just say all
01:00:18we're trying to do is save lives here.
01:00:20We love old people.
01:00:21Their value,
01:00:22the life of an old person and
01:00:24a young person is the same.
01:00:25We're not going to do fair ending stuff,
01:00:27and in practice,
01:00:28though you will have life years and fairness
01:00:31benefits when you put that in place,
01:00:33if that makes sense.
01:00:35Even though you're not building that explicit
01:00:38tiebreaker mechanism like you described,
01:00:41Mark into your score.
01:00:42At the end of the day,
01:00:43at the end of the simulation,
01:00:44you're going to save a ton more life
01:00:47years if you use age in this way.
01:00:49Does that make sense?
01:00:53I was wondering I guess like on
01:00:55a slightly different note in the
01:00:57context of the COVID vaccines,
01:00:59when you said that the elderly got confers
01:01:02because there was a stronger benefit, Yeah.
01:01:04To what extent is, I'm not familiar as
01:01:07familiar with a lot of these models.
01:01:09To what extent is therapeutic benefit
01:01:11included in these models or is that like a
01:01:13case specific thing or disease specific?
01:01:15Well, yeah, I mean the for for the vaccines,
01:01:18it's you basically say who's most likely
01:01:20to die from COVID, who's on vaccinate.
01:01:23It's the oldest people, right?
01:01:25So by protecting them with the vaccine,
01:01:28it's like the the exact opposite
01:01:30of this situation.
01:01:31Then you dramatically lower their
01:01:33risk of death from COVID and you
01:01:35save more lives here.
01:01:36Everyone who doesn't get treated
01:01:39with life support dies by definition
01:01:41because they're in respiratory failure,
01:01:43cardiac failure, right.
01:01:45And so then you need to identify
01:01:48the people most likely to survive
01:01:50to save the most lives.
01:01:52I should move on to the slack ones
01:01:54or if it gets harder, harder for me.
01:01:56But Mark, do you want to say something or.
01:01:58Yeah,
01:01:59so one thing that seems like there's
01:02:01like certain effort to remove
01:02:03judgment a lot of these metrics.
01:02:07So, so for example,
01:02:09we're taking something that's
01:02:11incontrovertible like how old somebody is.
01:02:13But,
01:02:13but as you sort of applied before,
01:02:15you know,
01:02:16you get some divisions in the room
01:02:18and they can probably
01:02:19predict a pretty algorithm too,
01:02:21like who's calling? And so I'm
01:02:24wondering, have you thought about using
01:02:25and they probably use the person's H or
01:02:28they're pure H but a lot of things too,
01:02:30like like Karen is standing out
01:02:32like frailty or *****. Yeah.
01:02:33So have you thought at all about
01:02:36it taking a Bayesian statistical
01:02:38approach where somebody says, well,
01:02:39I've got a pretest probability of XYZ
01:02:42and now, you know, like some data.
01:02:44So actually having the for your absolute
01:02:47position state their prior belief,
01:02:49I mean just to exhibit doesn't,
01:02:51because I didn't mean that that
01:02:53statistic doesn't really stand
01:02:54alone in the absence of other.
01:02:55That's cool. That's a really cool idea.
01:02:59I try to.
01:03:00I'm trying now to keep things simpler,
01:03:02but I really like that.
01:03:03So you would need of course
01:03:05a data set of predictions,
01:03:06which would be hard to obtain,
01:03:08of subjective predictions.
01:03:09You would need a data set perspectively
01:03:11collected of prediction from the ER,
01:03:14for example, before they debated some other
01:03:16like what's the probability of survival?
01:03:19That's cool.
01:03:19All right.
01:03:20So now this one is a really big problem,
01:03:23very perhaps the most contentious thing I
01:03:27think in in the current biological debate,
01:03:30and that's how to address
01:03:31structural inequity.
01:03:31I showed you this earlier,
01:03:32right?
01:03:32Where people died in Chicago
01:03:34was based on structural factors,
01:03:36based on a history of redlining.
01:03:39People with disadvantaged communities
01:03:40were much more likely to acquire COVID-19
01:03:43because of where they were living,
01:03:46because of where they had to work.
01:03:49They had to be.
01:03:50They're essential workers.
01:03:50They're out acquiring COVID-19
01:03:52living in congregate living settings.
01:03:55They didn't have the luxury of
01:03:56locking themselves in their
01:03:58room and zooming all the time.
01:03:59They had to be out of that in the world.
01:04:01And all of this is because the
01:04:03city is designed on purpose,
01:04:05or was designed on purpose,
01:04:06I should say,
01:04:07by the federal government to look like that,
01:04:09right?
01:04:09That's what redlining is,
01:04:11a systematic investment disinvestment
01:04:14campaign that was explicitly racist.
01:04:16If you haven't read this Mapping
01:04:20Inequality website,
01:04:21I strongly encourage you to see it.
01:04:23I The words are repugnant,
01:04:26but it makes it quite clear that
01:04:28our cities were designed by the
01:04:30federal government to be racially
01:04:33segregated on purpose, right?
01:04:34And we have to deal with this
01:04:37in sort of everything we're
01:04:39addressing from clinical medical
01:04:41ethics and bioethics perspective.
01:04:43But the question is how to handle this,
01:04:47this history of structural racism,
01:04:50this history of disadvantaging
01:04:53certain populations on purpose when
01:04:55we're making a triage score for,
01:04:57like,
01:04:58crisis care.
01:04:58And what I'm gonna go through
01:05:01is 4 different ideas I have.
01:05:03Kind of taken from the machine
01:05:05learning literature actually,
01:05:07about different goals you could
01:05:08have when you're making a protocol,
01:05:11and I'll go through these one by one.
01:05:13The 1st is demographic parity,
01:05:16which is each member of any
01:05:19racial ethnic group has the same
01:05:21probability of receiving truth, right?
01:05:24Probably the only way you can do that
01:05:26in practice mathematically is a lottery,
01:05:28a random assignment.
01:05:29It turns out that works pretty well,
01:05:31right? Almost as well as using sofa in terms
01:05:34of saving lives because of sofa's bias.
01:05:37But if you it, it's far from the optimal
01:05:39solution in terms of maximizing benefits.
01:05:42So a lottery while we achieve
01:05:45equal allocation does not
01:05:48respect maximizing benefits.
01:05:50So then the next idea is non discrimination.
01:05:53But make sure your SOFA is not biased
01:05:55against the racial and ethnic group.
01:05:56Be very sensitive that certain groups,
01:05:58particularly for black patients for example,
01:06:00have been structurally disadvantaged
01:06:02by our society and we have to be extra
01:06:05careful to not make things worse when
01:06:07we're allocating scarce resources.
01:06:10Right.
01:06:10And I hopefully have made the argument
01:06:13and convinced you that SOFA would violate
01:06:16this principle of non discrimination
01:06:18and it would exacerbate the disparities
01:06:21that we've already seen in the COVID-19
01:06:24pandemic if implemented to triage.
01:06:26So that's the the second principle,
01:06:28which is these are kind of in,
01:06:30you know,
01:06:31oriented in terms of more and
01:06:33more equity potentially.
01:06:35So how do you debias the score that's biased
01:06:38against a particular racial ethnic group?
01:06:40Well it turns out that using race
01:06:44ethnicity directly to fix SOFA
01:06:47like -1 if the person's black for
01:06:50example to sort of correct the
01:06:52bias I described earlier is very
01:06:54challenging for multiple dimensions.
01:06:57The state of Minnesota tried to do this.
01:06:59They they ran a regression model and
01:07:01they put all the clinical variables
01:07:04including H for probability of death
01:07:07from COVID-19 and and they also
01:07:09included a term for that was Bipoc.
01:07:12So basically non white.
01:07:13Anybody who identified as non white
01:07:15and that term statistically and
01:07:17independently predicted COVID-19
01:07:19mortality because it's capturing,
01:07:22even though it's a social construct,
01:07:23correlated with other unmeasured
01:07:26clinical variables.
01:07:27So they put that into their score.
01:07:29If you were, you're the same person,
01:07:31the same age, same medical comorbidities.
01:07:33If you identified as Bipoc,
01:07:35you'd be more likely to get monoclonal
01:07:37antibody treatment if you got COVID.
01:07:39This of course was grossly
01:07:41misinterpreted by certain people
01:07:43and manipulated for political gain.
01:07:46That's a completely erroneous statement.
01:07:49But this is the political challenge
01:07:50that we have to deal with these people.
01:07:52There are people like that in our
01:07:54country that we have to handle.
01:07:56And also from a constitutional perspective,
01:07:59with the recent affirmative action
01:08:02decision explicitly using someone's race,
01:08:04it's like one of, you know the in
01:08:06in general to allocate anything,
01:08:08maybe run, it's a legal challenge.
01:08:10And finally,
01:08:11there's the practical one where,
01:08:13you know, if it's like you're
01:08:15trying to give ventilators,
01:08:16and if somebody who looks to your
01:08:18eye that you would racialize them,
01:08:21as White says,
01:08:22oh,
01:08:22I'm black and I know your score
01:08:24gives me higher priority,
01:08:25How do you handle that 'cause this
01:08:27is a life or death situation.
01:08:28And I think that practical issue
01:08:31of are you actually counting on
01:08:33triage teams to racialize people
01:08:35and to socially constructed groups,
01:08:37that seems very problematic.
01:08:40So how do we get it on 'cause we have
01:08:43to one the what?
01:08:45What people have done is
01:08:46just modify the sofa score.
01:08:48That's what state of Colorado's done.
01:08:50So get rid of the renal component.
01:08:52I think it's best to just throw
01:08:54it out all together and come up
01:08:56with a new score that perhaps much
01:08:58better captures acute renal failure.
01:09:00It's the extent that we can measure
01:09:02them in triage scenario compared
01:09:04to this score which rolls in acute
01:09:06and chronic renal failure together.
01:09:08But in the pulmonary data for the the grant,
01:09:11which I think I took out,
01:09:12'cause I have way too many slides already,
01:09:14we used area deprivation index,
01:09:16which I'll talk about in a second,
01:09:17where someone lives as a way to
01:09:20achieve the outcome that Minnesota was
01:09:22going for without explicitly using
01:09:25someone's racial or ethnic identity.
01:09:29So the next idea,
01:09:32aside from non discrimination,
01:09:34is to actually look at that map and say like,
01:09:37can we even the playing field here,
01:09:40right? Can we spread?
01:09:41Can we mitigate the severe inequity of the
01:09:45pandemic by how we're allocating scarce
01:09:48life support treatments, and should we?
01:09:52There's tools, objective tools to do this.
01:09:54This is the area of deformation
01:09:57index as you see this map of Chicago.
01:09:59I don't know, I didn't explain that.
01:10:02This is where Druryville,
01:10:04it's like the wealthiest area
01:10:05city is right by Navy Pier.
01:10:07This is like a park like way very wealthy.
01:10:10Here's Hyde Park sort of an island that's
01:10:13where Chicago is wealth and privilege.
01:10:16And then here's the South and West side
01:10:18structured disadvantaged neighborhoods,
01:10:19right.
01:10:19So the homeowner,
01:10:20you can sort of see in that homeowner
01:10:23or the mapping inequality website
01:10:25how Hyde Park was constructed
01:10:27literally by the federal government
01:10:29to be to be blue on this map.
01:10:32And so you can,
01:10:33you can take someone's home address,
01:10:35map it to this area of information index.
01:10:38And what people like Doug White
01:10:39have suggested is that you literally
01:10:42would subtract points because
01:10:44they're coming from a structurally
01:10:47disadvantaged neighbourhood.
01:10:48And the idea is that we're trying
01:10:51to correct the structural inequity
01:10:53in the present day crisis.
01:10:55We recognize that things are way worse
01:10:57for certain communities than others.
01:10:59And we're taking one point off for that.
01:11:01And it turns out that there's an
01:11:04implicit ethical happening here,
01:11:05which is, you know,
01:11:06not really argued for in the paper.
01:11:08But correcting this,
01:11:09correcting that map,
01:11:11making it the spreading the burden
01:11:13of COVID around is about 1/4 of
01:11:15as important as saving most lives,
01:11:17which I think is interesting.
01:11:18This is an example where one of these
01:11:20protocols can reveal the underlying ethics.
01:11:22Here's the narrative description they
01:11:24use in the paper about how sofa based
01:11:27only system would prioritize this patient.
01:11:30The second patient will be prioritized
01:11:33in their novel system and they,
01:11:36you know,
01:11:36hand kudos to Doug White and and Pittsburgh.
01:11:40They actually did this when they were
01:11:42allocating their monoclonal antibodies.
01:11:43They got around that problem with
01:11:46you can't use race and ethnicity and
01:11:48they actually used where someone was,
01:11:52where someone lived.
01:11:53Calculate their ADI and give them
01:11:55twice the chance if they came
01:11:56from a high ADI neighborhood.
01:11:58And that led to higher rates of allocation
01:12:00than people who identified as black,
01:12:02which was their goal.
01:12:05So why?
01:12:07What are the potential
01:12:08criticisms of this approach?
01:12:09Well, you know,
01:12:11there's you're using these
01:12:12narrative descriptions.
01:12:13They didn't like the thesis.
01:12:14Hickett,
01:12:15Hickett handling 2 guys who
01:12:16were involved with the National
01:12:18Academy of Medicine and you know,
01:12:19defining what crisis standards of care were,
01:12:21they really didn't like the
01:12:22narrative description of the patient,
01:12:23right.
01:12:24You're making one patient really
01:12:25appealing based on like being
01:12:27a bus driver or whatever it was
01:12:29an essential worker and another
01:12:31person you're really painting as a
01:12:337 year old who's had been able
01:12:34to live their whole life.
01:12:35They're kind of like bleeding and
01:12:37fair innings there too. And of course,
01:12:40the triage team is not supposed to,
01:12:42you know, think about those
01:12:43social factors in triage, right.
01:12:45And that's what's the thrust
01:12:47of their main argument.
01:12:48They also talk about ADI not being granular
01:12:50enough to identify with disadvantaged.
01:12:52So one story about this is we very
01:12:54explicitly allocated our vaccine
01:12:56to our primary service area first,
01:12:58like around the University of Chicago.
01:12:59And so that meant our wealthier patients who
01:13:02live in the suburbs had to wait their turn.
01:13:05And that's not something
01:13:06they're used to doing.
01:13:07So once they found out the allocation system,
01:13:09they say, well,
01:13:10if I buy an apartment in Inglewood,
01:13:12which is one of the nearby
01:13:13disadvantaged neighbourhoods,
01:13:14can I get them by vaccine?
01:13:16So not a lot of them are really
01:13:19nice people who care about such.
01:13:22I don't describe all of
01:13:23our our patients that way,
01:13:24but you know of course the
01:13:26the bad apples and the ones,
01:13:27the emails that you remember
01:13:29and so we said no,
01:13:31you have to just stay in your
01:13:33stay in your house for one more
01:13:35week and you'll get it.
01:13:36So you know but I think in in practice
01:13:38aside from those extreme examples
01:13:40it would be it's very granular.
01:13:42This is a census block like you
01:13:44could look around you should play
01:13:45go on the website and look around
01:13:47and you can you know neighborhoods
01:13:48that you know are systematically
01:13:50worse off will be red on there.
01:13:52It's pretty good.
01:13:53And there's always this possibility of
01:13:56introducing social factors in triage
01:13:58of unintended consequences downstream
01:14:00the facts that you haven't anticipated.
01:14:03So these guys are OK with allocating
01:14:05vaccine and preventative medications
01:14:07based on error deprivation index or
01:14:10where someone lives as a way to address
01:14:12structural inequity but not life support.
01:14:14So here's what people think.
01:14:16And then finally,
01:14:17I think the the last idea,
01:14:19which is perhaps the most controversial
01:14:22and often is the criticism of efforts to
01:14:25correct the present day structural equity,
01:14:27is that you're really trying to correct,
01:14:29like, you know,
01:14:30hundreds of years of wrongs on
01:14:32a particular population.
01:14:34And is that really the best place to do that?
01:14:36And so that's the criticism
01:14:38of a reparations argument.
01:14:40But it's distinct from trying to make
01:14:44things more fair in the current crisis,
01:14:47if that makes sense.
01:14:49All right, So with that,
01:14:51I want to make sure we have some,
01:14:53some at least 10 minutes
01:14:55for discussion on this.
01:14:56Or maybe I can, I can just,
01:14:59why don't I just keep talking and we'll and
01:15:02we'll talk about the last two together.
01:15:04Because I always,
01:15:05never, never this one.
01:15:07And I think this is the perhaps
01:15:09the approximate.
01:15:11You know,
01:15:12the Bob Trude wrote this article
01:15:14in the Hastings report very early
01:15:16on the pandemic and pointed out
01:15:18that essentially all the thought
01:15:19experiments people
01:15:20were using were incorrect, right.
01:15:21The way a pandemic would work
01:15:23is that the ICU would fill up,
01:15:25then a new patient would show up,
01:15:27be in respiratory failure,
01:15:29and your decision would be to treat
01:15:31that person and withdraw life support
01:15:33from someone already receiving it.
01:15:35You very rarely would you
01:15:37have this three patients,
01:15:38one validator and you know,
01:15:41this is sort of an example, right?
01:15:44The one thing I don't think I wrote
01:15:46here is that this person who's in the
01:15:48ICU to sit his patient in the ICU,
01:15:50you would know with a great much higher
01:15:52degree of certainty that they're what
01:15:54their probability of survival is
01:15:55than this person who just showed up.
01:15:57You know, you don't know much about them,
01:16:00that's whether they're 5050, right?
01:16:02Whereas where you can have a lot more
01:16:05confidence but I think that confidence
01:16:07around their survival function is
01:16:09much smaller and this is way so.
01:16:12Despite these crisis standards of care
01:16:15being enormously long documents full
01:16:17of they're very hard to parse through.
01:16:19Almost none of them like really
01:16:20get into the weeds on this except
01:16:22for the New York plan which has an
01:16:24incredibly strict sofa based system.
01:16:26Like if your sofa doesn't go down,
01:16:29ventilator's out,
01:16:30so that's not been tested or validated.
01:16:33Whereas Maryland would have a very
01:16:35high barrier to withdraw off the
01:16:37the patient surrogates like said
01:16:39they don't withdraw life support,
01:16:40then they would have this chance to appeal,
01:16:43which of course would probably undermine
01:16:45any active reallocation in practice.
01:16:48So what we're doing in the grant
01:16:50is actually building a simulation
01:16:52model of sufficient complexity and
01:16:54depth to simulate what would happen.
01:16:57And one of my main hypotheses is
01:17:00that without some withdrawal rule,
01:17:02without some mechanism to remove
01:17:03life support and reallocate it
01:17:05to the waiting list,
01:17:05it's going to be first and first serve.
01:17:08So you can make this fancy triage
01:17:12store and it's not going to matter
01:17:14because it's just going to be who
01:17:16showed up first and then there's
01:17:18going to be very and with sort of
01:17:20randomness as people die if there's
01:17:22an available event when you arrive.
01:17:27All right. So with that,
01:17:31let's we can spend the rest
01:17:33of the time on discussion.
01:17:34These are my big conclusions.
01:17:36I think life support triage protocols
01:17:39across the US remain poorly defined.
01:17:42Well, the practical ethical perspective
01:17:44get rid of sofa triage scores,
01:17:46to use age, but only with the intention
01:17:48of saving more lives in the short term,
01:17:50just like we did for vaccines.
01:17:52Not not necessarily for any fair innings
01:17:54purpose and structural inequities
01:17:55need to be directly addressed,
01:17:57but exactly what the correction link
01:17:59should be needs to be determined
01:18:02and then withdraw of life support.
01:18:04Maybe the critical triage process
01:18:06should not be ignored.
01:18:07And before we go to questions,
01:18:08I just want to thank you to all
01:18:11my collaborators and mentors.
01:18:13You know,
01:18:13Govind's like this guy whose
01:18:14papers who've been reading forever
01:18:15and then he finally,
01:18:16he's a real person and will talk to you,
01:18:18which was like an incredible experience.
01:18:20And then Monica Pete,
01:18:22who's a HealthEquity scholar and my
01:18:25main mentor for all of this work.
01:18:27And Robert Gibbons is my PhD advisor
01:18:30and Elvis Long and a simulation
01:18:32model expert at the University
01:18:33of Chicago who's my KO8 mentor.
01:18:35So
01:18:40yeah, QR code is my, it's my Google
01:18:42stock page if it's not broken.
01:18:44So you can see some of the other
01:18:45things they've written and thank you.
01:18:46Let's let's talk for them.
01:18:51That was that was fantastic.
01:18:53I'm actually having my friend task
01:18:55trying to stay away a little bit.
01:18:58This was, this was really wonderful,
01:19:00you know, in terms of trying to deal
01:19:02with the issue of structural inequity,
01:19:04how to address them.
01:19:05I mean Mike, who's here,
01:19:07Mike and and you know Williams,
01:19:09they led the group that consisted
01:19:10of some of our folks who built our,
01:19:12our protocol as well as some members
01:19:14of the community all working together.
01:19:16Is it a fair, Stephen,
01:19:17Doctor Ivy that we never
01:19:19really cracked that nut?
01:19:21If you did,
01:19:22I would love to know what you decide.
01:19:30Thanks Martin. So so the health
01:19:32system was very concerned about the
01:19:35perception of the draft or public
01:19:38development community so it bans
01:19:41suggestion system and and members
01:19:43of the committee put together the
01:19:45transmitting we by members of the
01:19:48community like we intentionally reached
01:19:51out to people with local media the
01:19:55disabled community community staff
01:19:57took the New Haven but in British
01:20:00Portland you landed in Greenwich a
01:20:02number of ministers and and rabbis
01:20:04and you know so we really tried to
01:20:07intentionally reach a large number of
01:20:09people to break it wasn't that people
01:20:14but to explain what and it's not easy to
01:20:18explain necessarily what you're doing.
01:20:21To reunite people but
01:20:22it it seemed to go well.
01:20:24I I don't think we cracked the code
01:20:27of how to address Yeah I mean we had
01:20:29I had we had a similar experience
01:20:32presenting our trash for to our
01:20:35community Advisory Council for our
01:20:37hospital and what they were very
01:20:40forceful about is removing all the major.
01:20:42I didn't really go into this,
01:20:43but there were a lot of original
01:20:45plans that if you had major chronic
01:20:47conditions like you were on dialysis,
01:20:48that huge deprioritization and they're like,
01:20:51no, that's good, that's out.
01:20:52And so that was a very useful ex expe.
01:20:55Every time I presented them,
01:20:57I learned so much.
01:20:57I mean, I really do think that
01:20:59that should be part of what health
01:21:01system I think they should.
01:21:02But I do worry about you have these
01:21:04councils and groups and people leaders
01:21:06in the community that you collect.
01:21:08But it's somewhat arbitrary,
01:21:09like these are just people, you know,
01:21:11they're also usually people who
01:21:14are in social, socio,
01:21:15economic status positions
01:21:17that are pretty high, right.
01:21:19Like we have the guy who runs Howard
01:21:21Brown Clinic on the South side on ours.
01:21:23And yes,
01:21:24they may have the right race,
01:21:26ethnicity, diversity,
01:21:27make up that you want to
01:21:29represent the community,
01:21:30but do they really represent the socio
01:21:32economic spread or the community overall?
01:21:34You know,
01:21:35it's just like these groups you put together.
01:21:37So that's the that's the
01:21:38one problem with that.
01:21:38But I agree you for for thinking about
01:21:41ideas that you hadn't thought of.
01:21:43It's so helpful.
01:21:43I present all the time,
01:21:44although,
01:21:48yeah, so just a big comment if you would and
01:21:50then Ben will be the last common question.
01:21:52So it's up real quick and I'll move to Ben.
01:21:54Sure. It says run 4 minutes. I was 6 thirds.
01:21:57I think that's fast, but yeah.
01:21:59Yeah, please. Thank you.
01:22:01I I was just curious if you can describe the,
01:22:04the process that goes into choosing the
01:22:07data set used to build a regression model.
01:22:10Yeah, yeah. So I took all the
01:22:13clinical informatics slides out of
01:22:14here because it's an ethics talk.
01:22:16But we are constructing A collaborative
01:22:19networks from based on where my
01:22:21people are trained by one of my
01:22:23old mentors across the country,
01:22:24ICU doctors who like are data
01:22:27scientists too generally.
01:22:28And we're all clearing our data
01:22:29in the same format.
01:22:30So what will happen is we'll collect,
01:22:33we'll collect all all the observation
01:22:36electronic healthcare record that
01:22:38would be relevant for a critically
01:22:40I'll person and build a regression
01:22:42model based on the data from their
01:22:45like 42 hours before they start
01:22:47life support OR and then the first
01:22:50six hours afterwards with the idea
01:22:52that like the ER would have this
01:22:54temporary supply to stabilize patients.
01:22:56Because my hypothesis is that
01:22:58that would dramatically improve
01:22:59the accuracy of the triage car.
01:23:02But the nice thing is we can
01:23:02track both of those.
01:23:03And So what we're setting up with the
01:23:06collaborative network is like develop
01:23:07the data in the University of Chicago.
01:23:09Northwestern tested it.
01:23:10John Hopkins for example.
01:23:12We're about other collaborators
01:23:13and that adds a lot more.
01:23:15Whenever you make a model,
01:23:16you gotta keep your test data set separately.
01:23:20So that's the plan.
01:23:21Final question is Doctor Solch but you're
01:23:23just you know one with respect
01:23:26to the the community for the
01:23:28the night measures we actually
01:23:32sort of presented to them the
01:23:34possibility of using the area
01:23:35deprivation index as a modifier
01:23:37of. So that's where we were at the time
01:23:40and they we're we're not enthusiastic
01:23:43about that and and the more I thought
01:23:45about it the less enthusiastic I've
01:23:47I've become overtime you know I I
01:23:50I do I am concerned that that that
01:23:54bringing in you know non clinical
01:23:57factors really opens the triad
01:23:58vertical up to legitimate criticism
01:24:02and and also illegitimate criticism
01:24:04and and undermines the entire project
01:24:08baby out with the bathwater I guess right.
01:24:10You know, is the idea like sofa?
01:24:12Getting rid of sofa is sort of step
01:24:14one that will do most of the work.
01:24:15But if we try to do both at the same time,
01:24:18then you know, I I agree and I worry
01:24:25about just very arbitrary weights
01:24:27to like this mapping, right? Why?
01:24:30Where does that 4th come from?
01:24:32Why twice as many chances
01:24:34to get monoclonal antibody?
01:24:36Like I think that has to be
01:24:38really well justified.
01:24:38Harold Schmidt from Penn is thinking about,
01:24:41you know, you look at the map and
01:24:43see how the pandemic's hidden that
01:24:44the communities and then you design
01:24:46the weights proportional to that.
01:24:48So that's an idea.
01:24:49But I think the nice thing about
01:24:51having a simulation model is
01:24:52you can just try like
01:24:56see what's ethical after you look
01:24:57at your results. That's not the
01:24:59way you're supposed to do it.
01:25:01No, no. We said that sort of where
01:25:03where we ended up in Omicron when when
01:25:05we actually had our our most severe
01:25:09shortages were allowing 2 positions. They
01:25:16have a a lower threshold to to to
01:25:22with Cold War withdrawal and was
01:25:26usually withdrawal interventions and
01:25:28and so that sort of incorporated
01:25:31something that that Mark mentioned
01:25:33you know allowing clinicians
01:25:36to to use their clinical judgement.
01:25:38And and also you know your your point
01:25:40that it's actually less about
01:25:43allocating 11 ventilator among
01:25:44three patients than having some
01:25:46kind of mechanism to to discontinue
01:25:50intervention where where seeing
01:25:52that they're not beneficial. Right.
01:25:54I I think if you don't have this then just
01:25:56for comfort that's but we'll have that's
01:25:58the nice thing about having this a mod
01:26:00you can actually test that hypothesis.
01:26:02So I completely agree. OK.
01:26:04Well, thank you so much, Will.
01:26:06And please
01:26:09please join me in thanking
01:26:10Will, first of all.
01:26:16But, but so let's you know to realize that,
01:26:18I mean this is the program for biomedical
01:26:20ethics and we need to approach this with
01:26:22some ethical principles in mind and we
01:26:23have to agree on those first. But to have
01:26:25somebody here who's got really the
01:26:27ethical expertise as well as the clinical
01:26:29expertise as well as the quantitative
01:26:31public health expertise in an individual
01:26:34and also give some marvelous presentations,
01:26:35this was a real treat.
01:26:36But thank you so much. I
01:26:37think this is going to be
01:26:39helpful. And I do hope to
01:26:40the ones who are leading the charge
01:26:41here and the ones who are going to
01:26:42lead the charge someday soon, I do hope
01:26:45this would keep this going.
01:26:47We'll keep this going.
01:26:48Thank you all very much. Good night.
01:26:56OK, good.