MODERATOR: Welcome to Mayo Clinic COVID-19 Expert Insights and Strategies. The following activity is supported in part by an independent medical education grant from Pfizer, Inc. and is in accordance with ACCME guidelines.

ALEX NIVEN: Good morning and welcome to the METRIC-2020 Virtual Critical Care Conference. My name is Alex Niven. I am the education chair for Pulmonary and Critical Care here at Mayo Clinic. And it is my distinct honor to be virtually present with a true panel of experts when it comes to the topic at hand this morning, which is meaningful outcomes in critical care.

What are they? And how do we drive our field towards measuring them more effectively? So I've introduced myself. I'd like the rest of the panel to introduce themselves first. And so I'm going go [INAUDIBLE] screen. Oggie, can you introduce yourself?

OGNJEN GAJIC: Oggie Gajic-- critical care specialist from Mayo Clinic.

ALEX NIVEN: Perfect. Fernando?

FERNANDO ZAMPIERI: Hello, there. I'm a critical care physician at [INAUDIBLE] in Sao Paulo, Brazil.

ALEX NIVEN: Welcome. Arthur? Arthur, just unmute yourself and start over.

ARTHUR KWIZERA: Oh, OK. Hi, everybody. I'm Arthur Kwizera. I'm a anesthesiologist and critical care doctor from Kampala, Uganda, East Africa.

ALEX NIVEN: Excellent. Welcome. Matt?

MATT SIUBA: Hi, I'm Matt Siuba. And I'm a medical intentensivist, our critical care physician at the Cleveland Clinic in Cleveland, Ohio.

ALEX NIVEN: Welcome.

MATT SIUBA: Thanks.

ALEX NIVEN: And Michelle?

MICHELLE BIEHL: Hi, I'm Michelle Biehl. I am one of the pulmonary and critical care physicians at Cleveland Clinic in Cleveland, Ohio-- have the privilege to work with Matt Siuba. I have the privilege to work Ognjen Gajic. And I am originally from Brazil, so I have something in common with most of you.

ALEX NIVEN: Welcome. So I wanted to start with what may be a straightforward question. But I wanted to make sure that we were all on the same common foundation as we start our discussions. Can you all describe some of the common outcomes that we use in critical care? And perhaps describe how you've employed them in your different research areas to this point, so examples are welcome.

MICHELLE BIEHL: So I'm happy to start. So I think when critical care first started in the 1950s, we were a lot more focused on survival. But as the years pass, we started noticing that it's not only important to the patients to be alive, but how are these patients doing after they are discharged from the hospital?

How is their quality of lives? How is their functional styles? Are they able to be independent [INAUDIBLE] daily activities of daily living? Are they dependent on others-- on family members? Have they been able to return to work? Have they been able to go back to their relationships? How is their cognition doing?

Are they having issues of memory loss, difficulty concentration? How is their mental health in terms of anxiety, depression, PTSD? So there's a lot more different outcomes that we have been studying in the last several years that are meaningful to patients, families, and providers. And I think it's a lot more than just survival.

ALEX NIVEN: Yeah, yeah. So Michelle goes straight to open that can of worms. That really is the guts behind this conversation. So suffice it to say that the, quote, "traditional" outcomes that we're very familiar with from many of our critical care studies-- we think about mortality or survival, which ever one you want to do.

We think about the length of stay. We think about use of devices and withdrawal of devices, so ventilation days and the potential reduction in ventilation days, the surrogates with that in terms of ICU length of stay. Those are some of the traditional process measures.

I'll open it up to the group, because I'm sure there's many more that I'm forgetting off the top of my head. But let's just brainstorm on that for a minute. Because what I want to do is highlight some of the challenges with that and then go back to the area that Michelle is talking about.

Because I think from a human standpoint, I think we all recognize that the measuring things that are meaningful for our patients is probably the best way to go. But we also need to ask if it's the only way. So other questions or other examples that people have from their prior work in terms of more traditional critical care outcomes that they've used in the past.

FERNANDO ZAMPIERI: Well, I have seen some personal bias here, because we did a very large trial aimed for mortality in Brazil. So we did a very, very big trial. And the question and my main point here in this conversation is that all end points that aim for quality of life, or [INAUDIBLE] care, et cetera-- they are all conditioned to vital status, so that's an important point.

And of course, in terms of reaching and giving power for a study, if you measure quality of life using the ordinary scales or merit scales, is easier to do a large trial or a [INAUDIBLE] powered trial if you go for an American point. But if you do that, ignoring mortality, you might end up with the wrong conclusions.

For example, let's say you apply a very aggressive intervention. Let's give an example like a [INAUDIBLE] recruitment. For example, [INAUDIBLE]. And then eventually, you found out that patients that [INAUDIBLE]. Maybe the intervention can be harmful. But then you can conclude in the future that the intervention was associated with better quality of life. And that may be true. Because intervention was so aggressive that patients that were more frail perished because of the intervention.

And therefore, you only selected or you directed a bias towards more, let's stay, fit survivors. So still, while I appreciate the idea that we should go for quality of life, I believe that somewhat other hard end points should be considered. And maybe they can degroup it. For example, like we are seeing now in the COVID outbreak.

There is a lot of interesting ordinary end points that go from 1 to 7. And 7 is less. And 1 is alive and well at home. So this sounds to me like a very good alternative to do a preliminary trial, especially when you're wanting to know how, let's say, aggressive your intervention can be.

MATT SIUBA: Yeah, if I can jump on to [INAUDIBLE]. I think on a philosophical level, it makes sense to us that we're going to measure mortality and try to have an effect on mortality. Because that's how people end up in the critical care unit, is because we think the risk of dying is higher and is a thing I think probably a lot of us explain to patients. But there is a lot of information lost in a binary outcome like that.

So doing something where you have something ordinal like you described or something that's a little bit more descriptive is probably necessary. Because we're not only learning about function and if you died or not. But I think we should also be learning about, was this a good living? And even on the other hand, was this a good death? If we can capture those kind of things.

ALEX NIVEN: Arthur, thoughts from you?

ARTHUR KWIZERA: Yeah. So from our side of the globe, it's Ugandans been quite philosophical about this, because almost 95% of our intensive care is in the private arena. We did an end of life qualitative study about three years ago, where we interviewed attendants or family members of patients who died in our ICUs in Kampala.

And some of the results are very interesting. We had tended to believe that in many low and middle income countries around the world, not just Africa, but would be most worried about life and death. The first shocking finding was that family members or attendants really, really wanted all the full information about the type of therapy that patients are going-- prognosis, the quality of life after.

One recurring theme we realized was that a lot of families also got financially bankrupted after losing their loved ones in this study. And it was really, really, really-- because at one point, you're like, what? You know how you do a study and you have a finding? You do not want to find-- because you then feel very guilty about [INAUDIBLE] expensive medications that invariably didn't help the patients.

So it was very difficult. So the families were financially crippled, because we only interviewed for those who had passed on. And the other thing we found out was that the concept of some advance directive came up, because of our critical care population health care. I'm the oldest, for example. And I'm just 42.

So we're very eager to do all the best things. If we read that Fernando has done this, we want to do it like Fernando, that sort of thing. We really try to aspire to high-income country. Quality guidance is one. We've gotten to a point where we're trying to make our own-- do our own research.

And so when we're planning-- the next trial that I'm doing is a clinical trial in acute respiratory failure management, high flow [INAUDIBLE] set up as the standard of care in a non-ICU setting. One of the questions now that we are plugging in is to-- I mean, there was the patient comfort surrounding-- is the tolerability of whatever intervention.

But we are also now talking about the concept of, how would you like this to end for you if it's-- so someone talked about-- yeah, we talked about a good quality outcome, whether it's survival or death. We now have to factor that in. Because we are probably a little behind on the rest in terms of research in the high-income level, we've been following what the rest of the world has been doing.

But now, [INAUDIBLE] that study really, really-- it was a very sad study. We almost failed to publish-- failed to write a manuscript, because we took the lessons from it quite personally. And so for example, we are thinking about some sort of community advance directives.

We actually want to tell our country that, look, if you come into ICU, this is what's likely to happen to you. Is it OK if we-- you know how you have organ donation guidelines? Some people have [INAUDIBLE]. Everyone has maybe a will. So [INAUDIBLE] organ donations that [INAUDIBLE] need intensive care.

Should you subject me to this sort of thing? So we are going to do a community study, asking the question, what would you like to-- [INAUDIBLE] of intensive care is this? If this were to happen to you hypothetically [INAUDIBLE]-- but if this were to happen to you, would you want to be admitted to ICU? Yes or no? Just very basic, simple questions.

So that outcomes thing-- it really got us thinking along those lines, mostly because we are a poor country. And our health care system is free, but we don't have enough. So another finding was that in this study-- in another study that we just completed, we asked just about every question to families. And they were happy to withdraw care [INAUDIBLE] if they felt if we told them the whole truth about-- or if they felt that their loved one was suffering.

Yes, there's the data. Yes, they look comfortable. But if they asked them that question, they say they're happy to actually ask that if there is a single likelihood of a poor prognosis, mostly because of that aspect of financial impoverishment.

And then for patients that we-- we ask patients that were survivors. We tried to do a quality of life study afterwards. But it was a really-- it was a paradox, because we wanted to see if it was feasible. We don't have a very high ICU population, because we are short of ICU beds.

And we asked them. And they said that if they had known there would be in ICU, they would have told us. And if we had given them a list of things that we'd be likely to do, to and for them, that they had a box where they would say no. Many gentlemen said please don't catheterize me.

I know that's generally across the board-- comfortable. Some of the ladies said, I'd rather have a female mastectomy and so on. So in terms of outcomes, those are now some of the things that we are thinking about, the softer things. Because the average mortality in our setting in intensive care is about between 36. And that's all cause unadjusted. It's between 35 to 36.

When we started intensive care-- when we first started taking it seriously about maybe 8 to 10 years ago, it was about 50, 60%. But then we were a bit cautious to celebrate that figure. Because in this, there is a group of people that probably do better or not in this state where they would probably rather have died. So, yeah. That's my take for now.

ALEX NIVEN: Yeah, so thank you very much for making my job very difficult as a group. Some really thoughtful and really high-level conversation. Let me see if I can distill a few points that I took away from that and then turn this to Oggie, because he's been remarkably restrained so far in terms of our conversation.

So I think what I'm hearing-- and I'm going to cue off something that you just said, Arthur, is one of the big challenges that we look to do in critical care is make sure that we're delivering this practice to the people who will potentially benefit the most.

And part of that includes those individuals understanding the potential sequelae of our critical care and the interventions that we do, and making sure that is acceptable, and also benefits them in terms of their quality of life long term. I think what I heard from Fernando is that life and death probably does matter to our patients.

And I would tend to agree with that. And in terms of how we measure the success of the care that we deliver, there has to be some sort of objective outcomes that we use. And to measure success means we have to study our practice. And empowering those studies then for more objective or more easy to stratify outcomes has to be part of the equation.

But it gets complicated really quickly. Because there's a whole long list of other things that are more patient centered and potentially more-- I'll risk putting the air quotes around the word "meaningful." Because I think that means different things to different people. And it's how to measure that. And how to incorporate that within our studies and practice is challenging. Oggie, you're good at boiling things down to make it more simple. Your thoughts here?

OGNJEN GAJIC: I mean, I [INAUDIBLE] you've got to simplify it as much as possible, but not beyond that. So this is a complex problem. That's the perspective that I've shared before. And we had it in our papers from [INAUDIBLE] in general anesthesia.

And really, no one has boiled down or what matters to patients. So what he said was the outcome of good [INAUDIBLE] cannot be just assessed by counting deaths as each. That is a medical failure. It has to be assessed by the quality of lives restored or preserved by the quality of deaths of those who are on death bed, but a quality of relationships of all involved.

So when I have to keep someone on a ventilator two days longer because estranged daughter captive has to come from California to meet the rest of the family, I'm treating not only that patient. I'm treating the [INAUDIBLE]. So I will have-- in my database, it's going to be two days longer on mechanical ventilation. But that's the right thing to do.

So it's very complicated. And question is, can you? As a researcher, always struggling. OK, how you can combine that-- how you could test the mechanistic intervention and look for that complexity, that you would allow that combination of these things-- good life, good death, and good survival.

So practical solutions that we have used was [INAUDIBLE]. So you could even be discharged [INAUDIBLE] in a hospice and die at home. So any days spent at home after critical illness would be good to have. So that's one way. The complexity also arises from the fact that people don't come to the ICU. Most of them don't come from full health. Most often, they have underlying, significant, chronic comorbidities.

And there is a so-called [INAUDIBLE]. The question is, can we get them back to the baseline in the trajectory? Is [INAUDIBLE] just like one of the [INAUDIBLE] on the way? So does sepsis really make any difference in patients in these or it's just by phenomena that happens on a way to demise as it happens?

So really, I'm sorry, Alex. This is not going to be distilled in one or two words. It's very, very complicated. And I think we just-- this is beautiful to hear the perspectives from all sides. And we don't have to have a unit form solution for every single study. I think it has to be divided into domains. And then for each domain, whatever you study in intervention, should have a domain-specific outcome. I think that's more important than trying to find out a very global solution.

ALEX NIVEN: So for me, that resonates in terms of thinking about the quality of life, the quality of death, and the quality of relationships. I think that's also pretty challenging to measure. And so I want to turn to the group to talk a little bit about the challenges that they've faced in terms of the outcomes that they've used in past clinical trials.

We've already talked about this a little bit, but I'd like to try to enumerate that a little bit. And then I'd like to turn to Michelle, because I think she's probably spent the most time thinking about this space to perhaps use those examples to give us a little bit of her wisdom and a practical approach to potentially address some of these challenges.

So, Matt, Fernando, Arthur, thoughts in terms of challenges that you faced as you've analyzed your previous data, thinking about, again, those quality of life, quality of death, and quality of your relationships. Not an easy question, I know.

ARTHUR KWIZERA: Yeah. Well, of course, as you said before, it is very cultural and context specific, for example. So when we tried to do a small study once, where we just asked the physicians that were in the ICU when a patient was dying, what actually killed this patient? So what was the disease or the process that ultimately killed this patient? And the agreement was very low.

So some physicians used to say, well, this guy-- this patient is dying off his chronic diseases. So this acute process is part of a chronic process, et cetera. And other colleagues, of course, would say no, no, no, these are the [INAUDIBLE] acute process that is unrelated or not very much related to their baseline stature.

So of course, I agree. I fully agree with Oggie when he said that we should target at avoidable deaths when we talk about clinical trials or order of processes. But of course, defining what's avoidable is a little bit cumbersome and complicated. So when you randomize, of course, you expect that these would be evenly distributed between groups, so you'd be OK.

But all the times we try to do quality of life assessment after ICU discharge. We're faced with a lot of problems. For example, how to talk to patient that is severally ill or debilitated at home. Who is going to answer the questionnaire? Or is the patient and the family-- do they feel the same?

Or even worse, how does the patient feel about himself over time? Because people tend to get to used to their condition. So if you ask someone that has an acute disability, how he feels, he's probably going to feel very bad in the first months. And then he gets over it. And even if his actual performance is OK in terms of what you can measure, the patient might just feel better. Because it's human nature to become used to a situation.

So I believe we lack proper tools. So every time we try to use SF-36, 5Q3L, or 5L, we face a number of problems. And we never actually obtain a meaningful result. So I don't know if the other group or people have a better experience. Because all of my experience were pretty bad with that.

ALEX NIVEN: No, challenging process and hard questions. Matt? Arthur?

MATT SIUBA: Yes, I think that really resonated with me, a lot of what Fernando just said. I think the challenge is a lot of what we get concerned about and complain amongst ourselves with clinical trial design or some of the research that we produce is that we're dealing with really heterogeneous groups of patients in terms of the clinical side.

But I think to echo Fernando's point, I think this exists on the value side, too. So if we're going to try to measure a patient-centered outcome, a meaningful outcome for me is different than it is for you than for somebody else. So that part-- I'm not really sure how to reconcile that. But I think it's going to make any of the more subjective data points really difficult to collect and to see if you can make a meaningful difference in.

FERNANDO ZAMPIERI: Yeah, so perhaps we should consider very objective, as Oggie said and Alex and Arthur said. So perhaps a very practical, pragmatic answer. So for example, for patients that were employed before the disease-- they got back to work. Yes or no?

So it's not a perfect end point, but it's measurable at least. Maybe. But that doesn't make sense for retired people when their options and their desires are much different from a 40-year-old to a 70-year-old. So how to couple off of that-- that's the problem.

ALEX NIVEN: Yeah. Arthur?

ARTHUR KWIZERA: Yeah, like Fernando said, so it might be a different culture. So when we started doing research, we-- and one of my mentors told me to start with simple descriptive studies before and build up some capacity. And also, I will say that over the last 10 years, the landscape of our critical ill patients has also changed.

We used to have average age 28 years old, young, mostly male patient who has heart or traumatic brain injury following an accident. Because we've got those motorcycles that are the most common form of transportation. And people didn't wear helmets.

And therefore, that cut across all ICUs. And therefore, the outcomes were mostly related to severity of brain injury. When we got that started, we noticed that over there [INAUDIBLE] changing to a more mixed population, where you have your average age now in the mid-40s. And then you have mixed conditions now.

A lot of our knowledge of sepsis identification [INAUDIBLE]-- this identification had improved. Some of the resources-- better ICU beds, better equipment, better education, more numbers. But interestingly, it also brought a challenge. Because when you are analyzing data in a data set, let's that one line you would see-- things that may not vary-- that showed you that there was a knowledge gap.

And sometimes some of these data points would maybe affect your baseline characteristics when you do randomizations. For example, you'd find that because one particular hospital maybe had very poor-- it's a blood pressure septic shock management. And you would have a situation where, for example, when you look at the data, you will see in one arm of a trial, there's a statistically significant difference in, let's say, [INAUDIBLE] pressure, something like that.

So some of this is related to human resources and knowledge and skills and of course, logistics. Because at that point, we couldn't do combined studies in public and private hospitals. Because in public hospitals, they tended not to have these drugs and so on. And therefore, that was key outcome. That was usually the beginning.

And now, we've seen a change where even the public hospitals-- there's a little more knowledge. There's a little more acceptance of that ICU special. And it needs its own kind of investment. So we now regularly have [INAUDIBLE]. Some of us face it. Sometimes not. We'll have central lines all the time. That kind of thing

So I don't know if that-- so sometimes when you're writing a manuscript to a journal, the person on the other side not quite understand what you're going through and would not be that statistically astute as, say, Fernando with complex algorithms, because stuff like that.

But they're now lately. Because we are a bit more stable. Of course, we are still learning. The one thing that we notice is we have-- when we have-- the little work we've done that, that if you say you want to do a 90-day follow-up, we have challenges there. Because when you phone the family-- and let's say at consent, you'd say, we're just going to call you to see how you're doing afterwards.

This was a big lesson for one of my recent studies. Then you call. And the person says, yeah, we are alive. That's all you need to know. So it became very difficult. So my lesson now is that we want to call you, or we want to visit you. We want to call you, find out how you are doing, if you're back at work and so on.

So like I said, we are not as modern as the other one, but we are still learning. So I hope I've answered the question.

ALEX NIVEN: So Arthur always is making things more complicated, but I think he's right on point in terms of a couple of things here. So what I'm hearing is, again, another complexity when it comes to outcomes management. We're out measuring outcomes in the ICU, which is process and process variation.

So I'm going to hold that question as the next question in terms of looking at process outcomes is worthwhile and things that we should be doing or not. You also raised the issue of follow-up, especially when it comes to some of these patient-centered outcomes and some of the challenges that go with that.

And then [INAUDIBLE] you had has been raising an understanding in our patients in terms of what they potentially face as they go into these critically ill situations. So, Michelle, I'm confused. And I'm thrilled that you're here to answer all of these questions and set us straight [INAUDIBLE], so you have the comm.

MICHELLE BIEHL: So I think everything that was mentioned there, extremely valid points on challenges of how to measure outcomes. And I just want to point out something that was already mentioned, but just continuing on your point of loss of follow-up.

So how can we bring these patients back after a long ICU or hospitalization and have them come to do a battery of screening tools and spending hours with us at length of follow-up at three months, six months? So first, I think there is a problem of follow-up. Patients who are doing too well-- they don't want to come. They are fine at home, right?

Patients who are not doing well at all-- they are either readmitted to the hospital. And they're not coming to your follow-up or to your new research time point. Or they [INAUDIBLE] at home, that there is no way they can come, or their families cannot bring them. And then there is the issue of trying to use families as proxies and the issues of proxies and no reporting outcomes.

That is another point with studies having a selection bias. For example, tertiary [INAUDIBLE] your studies. There's going to be a selection bias of maybe the sickest or, again, the selection bias of the youngest population coming back to your studies versus when you look at population-based study, as I did with the [INAUDIBLE] in [INAUDIBLE] County when you see a population and your data is a lot more real in terms of elderly population, versus that they're sharing centers which are younger.

Also, the issue of different tools being measured like different studies measuring different outcomes. For example, quality of life. Some are doing SF-12, some SF-36, others EQ-5D, in terms of, how do you then compare all of this data when you're measuring, when you're realizing different tools, different instruments?

So I think there has been a consensus using a delphi way of trying to get most of these studies utilizing the same type of instruments or tools to follow up on quality of life, to follow up on functional status, mental health, what screening tools we use to use for anxiety, depression, PTSD, for cognition, and so forth.

So I think utilizing the same tools helps [INAUDIBLE] for us to compare the data. And lastly, but not least, I think very important is, how are we going to put into context the pre-morbid state of these patients? How are these patients doing before they are in the ICU? [INAUDIBLE] trajectory before they're admitted. How are you going to incorporate that into our data analysis of how they are doing after?

So most of these studies-- they evaluate patients after they are discharged, because we don't have data, how these patients are doing before. And there's very few studies that we have data. Depending on where you are, there may be a way of [INAUDIBLE]. We had a way of looking back and seeing how these patients are doing in terms of their functioning status, because that was corrected, at least yearly basis.

But most of these studies-- we don't have that. And then we're saying these patients are declining in their cognitive function. But maybe they already had cognitive impairment before, right? And the cognitive impairment may have been a risk factor for them to be admitted to the hospital into the ICU. So I think we face a lot of challenges with research in terms of outcomes and long-term outcomes after ICU, and that you have to acknowledge that.

ALEX NIVEN: So just reflect back. And then I'm going to open it back up to the group. So the challenges we have, first of all, are barriers to follow-up, barriers to understanding baseline so that we can have adequate comparison between the two.

And then it sounds like we're at least starting to migrate as a critical care community towards some consensus in terms of the domains that are worth-- that are a good starting point, at least, for measurements and meaningful patient-centered outcomes. I'm going to hazard a guess that the tools to measure those outcomes effectively are still an area of quite active research.

So, gentlemen, how does that help you in the short term? And I guess I'm going to-- well, actually, let me interject one thing that is perhaps a little bit different, but certainly topical. So one of the things that I've been doing over the course of the last [INAUDIBLE] or so is reviewing with a committee-- our hospital's triage criteria, should we reach a crisis situation with the current COVID pandemic.

And I think that that highlights a lot of the challenges that we face here in terms of trying to identify the individuals who would best benefit from critical care and critical care resources if we found ourselves in a situation where those resources were inadequate to care for the population of patients presenting to us, the limitations of a largely population-based physiologic scoring systems that we have in terms of effectively stratifying people, and then how best to quantify, again, the baseline medical data that we have on people to make thoughtful decisions about, do we focus our priorities on saving as many lives as we can, as many life years as we can, using the potential limitations from comorbidities as a tool for that?

Or what we'd love to do is focus on saving the best quality of life years as we can, recognizing that that's probably the most meaningful patient-centered things that we can do. So I guess I'll just open it up to the group reflections on those comments or Michelle's comments. And then I want to talk a little bit about ICU processes before we close this conversation. Oggie, I haven't heard from you for a while. Your thoughts?

OGNJEN GAJIC: I mean-- yeah, it is a challenge. I think from the research perspective, what I'm hearing from everyone is probably there is never going to be a final outcome and final study design that you can follow that would capture a meaningful outcome.

I think it's very contextual and depending on a mechanism or an intervention that you are interested in big educational intervention, [INAUDIBLE] intervention, process of care, specific medications, specific intervention, equipment. Whatever you're studying, I think the outcomes have to be twofold, have to be really key to the mechanism that is being tested, so some kind of bereavement therapy.

A strategy for the families will have to have the outcome that's related to how the families relate to their loved ones-- that, for example. Or if you are studying the cardiovascular intervention to prevent or improve cardiac function without any harm, that intervention needs to have an outcome that measures the mechanism as well.

And then it's not enough just to have that. You also have to have some hints or some surrogate-- that's a terrible word-- or some way to capture the ultimate effect, if any, of that intervention and meaningful outcomes. And that when we talk about-- the second thing that comes is that you are talking about [INAUDIBLE] for pandemics and what is the benefit of critical care.

I think, Matt, and I, and you, and many of us have realized over the years that critical care is not necessarily the life saving or quality of life saving intervention. It can literally increase your chances not to do well. And in fragile patients, the aggressiveness of care that we provide can actually be just tip, something to tip them over and never go back to life. And if they would be somewhere where there is no intensive care, they might do even better.

So it is very contextual. And you need to provide an informed consent of, is this beneficial, not assuming that critical care or just coming to the ICU is beneficial? It's certainly not. It's very, very difficult and aggressiveness of care that is natural, the way it is dangerous.

And we are living through it nowadays where there is probably in addition to fighting against pandemics of the virus, there is a pandemic of [INAUDIBLE] going all around with the different constraint that we have. So it's a very, very difficult issue.

So ICU is not always lifesaving. It could be the other way around. So that's what we need to capture as well. And the third point is just from a research point as a researcher, I have to say. It just-- now, this is the intensive case, like a place in your life.

It's not something that is always-- we do that. It's a great job. It's fantastic job. But you have a life. You have the disease here. You have leukemia. And you die of leukemia. And maybe you have-- seven times get [INAUDIBLE], can live longer with each chemotherapy. But you cannot really put that aside.

So a related discipline in anesthesiology is, I think, very insightful. They have tremendously-- they focused on processes. And I think you're going to get to it later. They kept tremendous improvement over the last 40 years with really eliminating preventable death to almost zero. And they can measure it. Because if you come to elective surgery, you shouldn't die from anesthesia, OK?

So they had a good outcome with what we don't. But they had. And they've done process improvement, quality improvement. They did only one clinical trial, was 20,000 patients, randomized the positive symmetry and [INAUDIBLE], which was negative, despite lack of any large clinical trials, that this clinic anesthesiology has made phenomenal progress focusing on economics, how to make it easy to do the right things and difficult to make bad things.

And from that example, which is a very related discipline, it's basically keeping a person alive during the [INAUDIBLE] of homeostasis caused by surgery. They were able to get to the-- because they have called outcome. They have deaths. You shouldn't be dying after anesthesia. And they measure that.

And over 40 years, they had tremendous progress with really never studying that in a formal way. So I just have to throw this away. And maybe a bit before you get to the process, maybe if you just have different opinion or another way of how. Fernando, and Arthur, and Matt, and Michelle, say [INAUDIBLE]. But I'm almost thinking, OK, let's just focus on small studies for mechanisms and forget about the infinite goal to see something works for mortality.

ALEX NIVEN: Yeah. So I'm going to throw it to Fernando first, because he was bouncing around in his chair during that conversation.

FERNANDO ZAMPIERI: So first of all, I fully agree [INAUDIBLE] when he said ICU care has a very narrow therapeutic dose. And the more frail-- when the older you are, the closer the lethal dose, the therapeutic dose of ICU care. Yeah, so there's a very, very important point.

So for quality of care, and quality improvement, and other interventions, of course there are some interventions that are cheap enough and crystal clear enough that they will be useful for physicians like boost oximetry and other process of care.

And of course, although I recognize that I like trials, I fully believe that for some very simple measurements, you don't need a trial for that, besides the fact that the tendency-- if you look in the long term, the tendency is that everyone is improving. So the tendency is that care overall is improving.

So of course, pulse oximetry might be a little bit of a simple example, but it's a valid example. Regarding mechanistic trials, I believe they are useful. And I have some-- of course, [INAUDIBLE] are a little bit different from Oggie's here, that I believe that they should be embedded inside a larger trial and not be run independently. Because of course, you will be under-powered to see other meaningful effects.

So for example, if you do a trial and then you concluded that an intervention reduces the need of pressures in six hours, for example. So that's perhaps a meaningful end point. So you can even say that, well, using [INAUDIBLE] is bad, so this intervention might be useful in the future.

But of course, if you don't look in the long term and if you don't look at the other points, perhaps you can miss something that may be unexpected. And of course, things are not also crystal clear. And finally, the problem with mechanistic trials is that not always you are aware of all the mechanisms that are happening behind your intervention.

So one point that I always make-- for example, dissociation between chloride and worse outcomes in critical care. That's a clear conundrum in my opinion. Because we don't have actually a real proof or a mechanistic proof that chloride is toxic.

So if you dig deeper in the literature, there are some animal models or something like that. And of course, you'll have several studies saying that, well, perhaps higher chloride levels are related towards outcomes. And we have this hypothesis, et cetera.

But again, it's hard to say what came first. So why is chloride toxic, for example? So we could do a mechanistic trial and say, well, we will try to keep chloride [INAUDIBLE] in this patient, using whatever. That could be a point. But of course, you require a lot of medications, drugs, and interventions.

And when you mix it all together, I don't know if you can find something that's unexpected. So I have problems with mechanistic trials. But where should they be placed? And so should they be placed inside a quality improvement program? Perhaps. Should they be embedded in a randomized clinical trial or a platform trial? Probably. Should they be run independently? I'm not that sure. So this is my point on that. But of course, there's a lot of room for discussion on this.

ALEX NIVEN: Yeah. Yeah. So I guess we've just got a few minutes left. And I think I'm going to cue off some of that to talk about-- I mean, what resonated with me from your description is just the heterogeneity of our clinical practice, the multitude of interventions that each of our critical care patients receive.

And that's superimposed on the fact that our critical care patients themselves are heterogeneous. They come with a whole set of comorbidities. And those comorbidities are not [INAUDIBLE]. So Oggie has raised an issue of a process in systematic delivery and quality improvements.

And Fernando has reflected on that a little bit Arthur, one of the things that I wrote down from your conversation before was the dramatic reduction in mortality that you've seen in your region through simple interventions, education, and process improvements over time.

How do we factor this all into the outcomes discussion? Is examining process outcomes or using rapid cycling quality improvement methods as an outcomes measure-- how do we fit that into the conversation? And I'll turn to anybody on the panel for this.

ARTHUR KWIZERA: We're short on time. To be honest, I don't know. I [INAUDIBLE] to try and build on [INAUDIBLE] in Uganda. Now, it's become an African thing. We have lots of group of 40 African countries that I created three years ago, four years ago.

And I think that we agree. For example, one of the discussions we've been having has been that in our setting, we have to fast scan our landscape. There are certain studies we cannot do, because [INAUDIBLE] of practice, skills, and so on, but then small quality improvements.

When I said do improvements, [INAUDIBLE] measuring. This was just we looked at one manuscript for 2012. We looked at all the ICUs. And we looked at one manuscript from last year. And that's when we saw, oh, this is the jump. It's mostly been about just education and some improvement in equipment, triaging process, and so on.

I don't know if I'm going to answer this question very well. But I'm more of a quality improvement kind of person, mostly because our setting does not allow us to control. When we hear the word randomized controlled trial, how much are we going to control? Because like Fernando said from the correct aspect.

So it's more about quality improvements and some sort of pragmatism in the trial. Because I want to imitate. Here, I'm about telling people, this is how we can change. If I run that clinical trial, what you need pragmatically cited can mirror, for example, everyday practice. I don't know if I've answered your question, but [INAUDIBLE].

ALEX NIVEN: [INAUDIBLE] sure if that's unique to your setting. I'm not really sure how I can control chloride dose in my ICU. And there's plenty of data out there that would suggest-- even the well-established practices that are well evidence based get delivered very inconsistently, regardless of the practice setting you're in. So I think that's a real challenge that really ups the noise to signal ratio when we talk about studying anything in terms of meaningful outcomes in critical care. Go ahead.

ARTHUR KWIZERA: I guess one more thing. I think one of the ways that maybe we can-- one of the new things that's coming is I've seen coming across the board, is if when we start to understand. One way of controlling is when you start to understand and identify phenotypes and efficient phenotypes. That's one way.

And one of the things we are doing-- we have a population that's young, has HIV/malaria. So if I were to do a fluid study, I want to do a comparison of-- we've now realized anecdotally that that's a totally different phenotype when it comes to sepsis and septic shock. So then I will do a fluid study comparing [INAUDIBLE] and saline, normal saline, in that population alone. That's just an example.

ALEX NIVEN: Yeah, but I think it's a fantastic one. So I think we started just a little bit late at the beginning of the hour. So I think we've probably got about five minutes left. So I'll open for the group in terms of other reflections on this vein of conversation. And then we'll move to a few closing remarks. Matt?

MATT SIUBA: Yeah, I think the process issue is a major issue. When you look at the few very strong data points that we have about things that make a difference, even if we're just going to talk about the mortality outcome, it all comes down to doing less to people. It doesn't come down to doing more to people-- lower [INAUDIBLE] lines, less sedation, more liberation from sedation.