Dr. Josh Lowe: Shock Index as a Predictor of Mortality in GI Bleeds (8/9/24)

Yeah. Now that we're at the top of the hour. So most of us in emergency medicine and ems, we've heard of the shock index. Right. You take the heart rate, you divide it by the systolic blood pressure. And the idea is the higher that value is, the higher the likelihood that your patient is in shock. And in the setting of trauma, an elevated shock index has been shown to be a more sensitive marker of hemorrhagic shock and need for intervention following trauma in comparison to standard vital signs. Right. Than looking at systolic blood pressure thresholds alone or heart rate alone. And as a result of this, organizations like the American College of Surgeons have incorporated shock index into trauma triage guidelines. Right. So this is a measure that's already in use out there in the real world. But today we're going to be joined by Dr. Josh Lowe. He's an EM and EMS physician who serves as the Associate Medical Director of the San Antonio Fire Department as well as the Deputy Director of Joint Base San Antonio. And he's going to be telling us about a study that he led evaluating the prognostic value of shock index in a different population. So today we're going to be talking about patients who present to EMS with GI bleeds. So with that said, Dr. Lowe, the floor is yours.

Yeah, absolutely.

So let me just start with the disclaimer. These are my views based off of the research that I've done. This doesn't represent San Antonio Fire Department of Defense, Department of the Army.

So this is the paper that we started with.

Or that we published. It's shock index as a predictor of mortality and hospital admission and pre hospital GI bleeding. It was a cohort study where we looked at the ESO data for a two year period between 2020 and 2021. And really though, it kind of went back to a question and a conversation.

When should we give prehospital blood to a GI blade? And the reason that this came up was I have a brother who is a paramedic in a volunteer fire department in rural Virginia. And I was talking to them about possibly getting a blood program started. And what they said to me was, well, we don't have a lot of trauma out here. Most of what we would see would be like GI bleeding. And the data just doesn't support it.

So I kind of disagreed. But then when looking into the data, there wasn't a lot of data.

You know, so when should we give prehospital blood to a GI bleed? In the ER we have things like a hemoglobin list 7, we have the AIM65, we have the Glasgow Blatchford score, but pre hospital we don't have access to those labs. So started thinking about what data we do have in the pre hospital setting, which brought us to the shock index.

But then started thinking, is shock index alone an indicator to give blood to a gip?

So we wanted to try to identify what we meant by a severe gi.

And instead of just pulling random numbers out, we decided to look at other risk stratifying GI bleed.

Tools that are out there, specifically the AIM65 and the Glasgow Blast first board. And what did they use to identify what was a severe GI bleed? So the aim 65 used mortality and hospital length of stay.

For their definition of a severe GI bleed, while the Glasgow Blacher scored looked at the need for intervention. Based off of the data that I was able to get from ESO and image trend, I couldn't find where we could find intervention. But we did have access to mortality data and hospital length of stay data. So given the focus of our question was trying to determine the efficacy of shock index on determining the need of early blood for transfusion, we chose to use the Same endpoints as aim 65.

So our methods were this is a retrospective cohort study. We used the ESO data collaborative EMS records from 2021 and 2022.

We started with over 11 million records. So based off of our inclusion criteria of you had to have a either primary secondary diagnosis of GI bleed and you had to have all the vital sign data points actually documented. That actually brought us down to about 2,419.

Records that we deemed complete. And then after missing vital signs, missing age, we finally found 1525 that had all the metrics that we felt were able to be analyzed. These were people that diagnosed with GI bleed and had all the vital signs that we were looking for.

So the analysis that we used depended on what we were looking at. We used linear regression looking at the relationship between shock index and continuous variables such as hospital length of stay and honestly only hospital length of stay. We used independent T tests to compare the variables between two groups looking at the vital signs and then survivors, non survivors. We use box plots because it actually shows visually kind of the same thing.

Using shock index across mortality and hospital infants stay and whether or not people got admitted. And then we used sensitivity, specificity and receiver operating curve when looking at using shock index itself as a predictor for mortality. Is this a tool that we can actually use? And we used area under the curve looking at that, you know, 0.5 to 0.65 is not very good, 0.65 is, 0.8 is fair, and then above 0.8 is a pretty good tool to use.

So what did we find when looking at. We looked at both initial and highest shock index, because sometimes all you have is that initial vital sign to go off of. So when looking at initial shock index.

By mortality, the people who died obviously had a higher initial shock index. But even more important than that, the average was above that 0.9 threshold that's so commonly used, whereas the average for those who lived was below the 0.9 threshold. As you can see, there were a lot of outliers, especially with those who lived.

Which just goes to show how rezonia a lot of people are. But for the most part, the average shock index of those who died was significantly higher than those who lived. Then we looked at it based off of ED disposition. Although this was statistically significant, higher in those admitted than discharged, honestly, the separation didn't cross that 0.9 threshold that's commonly used. And it wasn't really clinically significant. There was just way too much overlap to be able to say, yes, if you have a higher shock index, you're going to be more likely to be admitted than discharged. It trended that way. Again, it was statistically significant, but it didn't appear to be clinically significant.

Then we looked at the highest shock index. What was the highest recorded shock index during that transport? And when comparing those who lived versus those who died, it was even more pronounced than with the initial shock index, which is what you would expect, right? Whatever the worst shock index is, is going to be much higher than those who live. Of note, if you look at the people that live, there was two people who had a shock index at one point over three.

Those triggered me a little bit. I didn't understand how you could have a shock index more than three and live. So I looked specifically at those two cases. GI bleed was the secondary diagnosis in both of them. Primary diagnosis was actually AFIB and rvr. So once they got appropriate treatment for their afib, their shock index improved. And so we included them in the data because we wanted to be completely transparent and include all comers. But that's why there were 2 people with a shock index over 3 who live because GI bleed was their secondary diagnosis.

And then the highest shock index by ED disposition. Again, people who were admitted had a statistically significant higher shock index compared to those who were discharged. But it wasn't as clearly delineated there was still a lot of overlap. However this one the average for those admitted was over that 0.9 threshold commonly used and the average for those discharge was below the 0.9 threshold commonly used. So maybe there is some potential clinical significance but again.

It too much overlap to really hang your hat on it. And so we talked about that in our paper. That can't really hang your hat on saying if you have a positive shock index that you're more likely to be admitted.

So then looking at the receiver operator curve using that 0.9 cutoff for mortality, it provided fair utility. Right? The.

R value for this came in at 0.73 if you're looking at the highest recorded shock index and 0.65 for that initial shock index. So. So not great. But honestly somewhere around there similar to the.

Point of care strep test that we all love and hate in the er.

So another data point that you can use, but I wouldn't use this as the only indicator when looking at admission. It's not a good cutoff for determining if someone's going to be admitted or not.

Both of them were below 0.6. So pretty poor tool to use to determine pre hospital if someone's likely to be admitted.

Then we started looking at initial shock index by hospital length of stay and honestly there's no correlation at all.

The R squared was less than 0.001. So very weak correlation doesn't really relate at all for both initial and highest reported shock index. So while there may be slight indication for higher shock index means you're more likely to be admitted. There's absolutely no correlation with this day.

So what are the implications for this? When using that cutoff of 0.9 the sensitivity was 74%, specificity was 55%. The positive predictive value was 6.3. But that fits with established literature. Right?

If you look at all comers for GI bleed presenting to the hospital, the mortality is going to be around 2%. So an odds ratio increase of 3.5 given the incidence in the natural population is going to bring our predictive positive predictive value up to 6.3. It's pretty good as a negative predictive value though. If that shock index is less than 0.9.

These are people that are probably going to be okay and don't necessarily need pre hospital blood or pre hospital much pre hospital intervention.

So some commonly asked questions since I've been talking about this paper is did the analysis of our admission length of stay include patients who died? Yeah, it did and I completely Understand that, you know, dying artificially shortens your length of stay. We also analyzed the data removing those patients who died and there really wasn't any significant difference to the R value. And given the word count limitations for the journal, we decided to present the data that included all patients because that's what we do pre hospital, we take care of all patients. So that's why we included that. Does the trend of the shock index matter? Probably that's actually what we're currently looking at now is the delta shock index using that same ESO data set. Unfortunately, given the nature of that data set, there's a lot of things that aren't necessarily documented. So when we look at it, we're seeing a lot of patients with an initial shock index above 0.9 whose shock index improves in route and then they end up dying anyways. What we think that we're seeing is that these were people who are probably given crystalloid enroute, which does improve your numbers but doesn't really improve your mortality. You know, looking at the DOD and the Joint Trauma Service or Systems Research, crystalloid probably, although improving your heart rate and your blood pressure temporarily at best, doesn't touch mortality at worst can worsen mortality due to increased acidosis and dilution of the clotting factors, specifically in hemorrhagic shock. So we think that it's likely that these patients received IB crystalloid, but I have no proof. That's pure conjecture on my part. What we do know is that if we look at just the patients whose shock index worsened over time, there was definitely an association with both hospital admission and mortality. And then what do I think people should probably use as a protocol? I would continue using the cutoff in 0.9, but more specifically I wouldn't use it on its own. I think that this needs to be part of a protocol that uses multiple parameters entitled CO2 less than 25, which Folger et al showed is associated with increased mortality and need for blood and hemorrhagic shock, altered mentation and agitation, which Bokenkamp et al showed his three time likelihood increase in needing massive blood transfusion. These are all things that we can see pre hospital that can help us determine whether or not we should give blood in the field. Now there's no study to date that has looked at if these factors are additive or multiplicative or if they cover all the same information. But I do think that if you have blood with your services and these people have a GI bleed that is currently ongoing and have one of These parameters should at least consider it, give them a direction, a call, you know, to talk about it. And if they have two, or even all three, I think we might be remiss in not giving them blood again, if we carry blood. Because, you know, data has shown that for GI bleeders that come to the ED already in shock, up to 25% mortality in those folks. So anything we can do to improve mortality, if we've got the resources, I think is probably a good thing.

All right, so went through that pretty quick because I want to answer questions. Here's the references that I used for this presentation. What question do you guys have? And I'm going to open up the chat so I can see it.

Oh, and sorry, Dr. Pepe, to answer your question, in San Antonio, the last that I looked, I think that we're seeing close to 30% of our blood transfusions have been medical. We've had a couple that have gained notoriety because they were OB related, and the news really likes to hang their hat on those, but most of them have been. Most of our medical ones have been GI bleeds.

That's exactly. Your latest data are that there's about 30%. So, yeah, that's the point. It's a large number. And then, of course, we've had, for example, in our own area here in Palm beach county, we've had cases of, you know, tonsillectomy. And it's a. I realize that that's actually a bigger problem with, you know, in children, even that that's not an uncommon problem is what I'm saying. Okay. So here, if you don't mind, I'm going to make a comment because I have to go over and get on to another meeting here in a few minutes.

But.

But. Okay. Were you there when I gave the talk at the National Whole Blood Summit back in June? Were you there by any chance?

I was not. I was in the Philippines in June. Wow.

Okay. Got it. Okay. No, the reason why I was asking, because I was going to relate to it. So one of the things that Don Jenkins had me do because he's rethinking the whole triage thing for trauma, let alone, you know, flip the GI bleed things, was the fact that the shock index has been great. You know, it's like when it's positive, it's pretty positive. The problem is, is that you could have someone who normally has a blood pressure of 160, and now they're down to 110, and so they're horrible vital signs. Or you have the person that's crashing. Who has the blood pressure of 75, but their heart rate is 75, either because they are crashing or they're on beta blockers and all that. And that's what disturbs the shock index because there's a lot of other factors that you need to go into decision making. And with GI bleeds, it's going to be even more complicated because of the other medication histories and all that. So I'm sure that was going to be difficult because you can't the, what do you call it, records aren't going to grab those. Anyway. I just wanted to throw that out as these, some of these confounding variables that make it really difficult to say. So we need to work through this. But let me not.

Let'S say, denigrate this at all. I'm saying this is fabulous work that you're doing and it's really a good cause. We just got to do a better job of it. And we're, you know, or figure out. One of the things, for example, I suggest in the talk was that we know that people with normal blood pressures can be bleeding to death. And based on lactate levels that predicted they were going to go to surgery and all that. And also in general, we're not going to know. But it may be that we'll have some more objective data in the future that we could use like a point of care, lactate test or something we could do in the streets. So anyways, but I just, this was just that's my initial reactions to everything. But so, number one, great job that you're doing this, that you're bringing to the forefront, that it's not all trauma. That's number one. Number two is that this is a big problem. Those are more vulnerable patients oftentimes and they don't tolerate as well. So I think so. And I think that those are going to be some of the things to keep in mind. And then the other thing is that you don't want to miss that one person. It's the sensitivity and specificity issues that really loom large in this one. So, but thank you for doing this, though. This is fantastic. So fantastic work.

Thanks, sir. And I agree, you know, that's one of the reasons that towards the end there I talked about the end tidal and the altered mentation, because especially in those patients that are already on beta blockers where we don't know what their baseline blood pressure is and they have a relative decrease, but we can't see it, sometimes you have to rely on those other factors. Those HMP things that we can get.

Hey, one thing I wanted to ask you about was that you. Maybe I missed it or wasn't paying attention enough as I was transitioning in that you said something about the age of 65. What? What tell me about that issue of trying to understand that discriminator is that.

I didn't use an age of 65 discriminator. We initially used an age of 18 discriminator just because of. I'll be honest, there's not a lot of GI bleeds in people under the age of 18. And the IRB that I was working with, their requirements for me to include those patients was too prohibited.

That was it. I just thought, for some reason I picked up. I thought I heard the number 65 thrown around. That's all.

Aim 65 was the.

Score that we use.

Okay.

Oh, yeah, we use the AIM65, their endpoints for severe GI bleed because it was data that we could get pre hospital.

Yeah. Okay.

It's already been established in the literature that that's a good indicator for severe GI bleed.

No, now I absolutely understand. I just. Sorry, I missed. I misunderstood that. That husband. Why are they doing that? That's great. Well, thanks a lot. Anybody else have any other questions out there, please?

Hey, Paul, it's Brett Greenfield. Good morning. So, Josh, great job. I just have a quick question for you. Right. So we're hearing more and more about using the modified shock index, right? Which is heart rate over MAP as opposed to systolic blood pressure, which may be a little better in identifying especially some of the medical patients that may be in trouble. What are your thoughts about that? And have you ever thought about going back and looking at all your data and applying the modified shock index as opposed to the shock index and see what comes out of that? Yeah, that's one of the things that we're looking for next right now. Like I said, we're trying to look at the delta shock index and the issues that we're running into our with these people whose numbers are magically improving and we can't really tell why.

That'S probably where we're going to go next. Just because again with the delta shock index, we can't account for those co founders. But looking at the map, since the data does show, especially in the more prolonged bleeder versus the acute traumatic leader, that you're right, the MAP seems to be something that is the modified shock index, using the MAP does seem to be more sensitive. So it's coming up. That's one of the things that we're looking at with the ESO data. I just have to get back with Ramli Crow and make sure that she's okay with me using it for a third reason. Yeah. Awesome. Thanks. Appreciate it.

Great work.

Yes. I had one question, Dr. Lowe. So in terms, if I'm understanding this correctly, you identified patients based on ICD10 codes that were recorded either the emergency department or at hospital discharge. And so I'm just curious, in terms of kind of the. When we're looking at provider impression, which is available in the ESO dataset, you know, what proportion of patients that have a provider impression of GI bleed also were diagnosed in terms of their ICD10 code with GI bleeds in the emergency department or during their hospital stay?

So I didn't look at it that way. I did look at it the other way to see if there was any. And I'll be honest, the people that ended up being diagnosed with GI bleed, the initial impression was all over the place. We had syncope, we had chest pain, we had weakness, a lot of weakness.

Sometimes just sick person. There was not a lot of correlation there because I figured I was looking at the data anyways and maybe I could get, like a second publication out of it, but there. There's. There's nothing there. At least not looking that way. That initial impression is all over the place.

And I guess it's kind of a related question. You know, I'd imagine as. As ems. Right. As kind of pre hospital systems of care, the good saves that we want to get out of this are the GI bleeds are going to exsanguinate and have a cardiac arrest in the emergency department. So. So. And again, I just in case you have a sense of it, do you know what proportion of the patients that experienced mortality experienced mortality in the ED versus in the hospital?

So I should clarify. All the mortality that we did was ED mortality.

Okay.

Once they got into the hospital, it was. They were admitted. But then also it was a different mortality code.

Okay, perfect.

Yeah. So the mortality that we're looking at is ED mortality.

Perfect.

Which is why it's lower that 2% versus the 25% all mortality during state.

So, Josh, one of the obvious things here is that GI bleeds are this generic GI bleeds. Okay. So. Yeah. So is it a peptic ulcer that just blew on the top of the bottom or whatever? Or is it a person with, you know, the. Their. You know, their various problems from alcoholism. Right. And they. They blow those things, but they already have clotting problems and they are at higher risk for obvious reasons, or they may not work as well in those patients. So will you, or did you by any chance sort of sort out those people in terms of their formal diagnoses? You know, if it was more of a. A localized true bleed from a vessel there, here or there, or was it something that's linked to a situation where they didn't have very good clotting parameters?

We haven't looked into that. This was again looking at all GI bleed ICD 10 codes, pretty much the entire K92 series, K25 series, K26 series, and then the K62 series. So all the ICD 10 codes that we're dealing with. Upper, lower duodenal, gastric, jejunal, esophageal, and then otherwise unknown. Okay, we looked at all of them, but we didn't, we didn't differentiate and do any sub analysis.

Probably worthwhile. The reason why I bring that up is that, for example, with our TXA protocols we. I would probably give TXA to somebody who is, you know, somebody who has got, you know, gross hemopsis, gross hemogem or whatever, bright red blood rectum or whatever it may be. Whereas a lot of us may back off on giving the TXA in situations because the GI bleeds might be more chronic under those circumstances. And that's been some of the. I mean, what's your protocol there under those circumstances for txa?

So we were actually recently changing it. I don't know if we've pushed out our new cogs yet. I honestly have to look. But we've started to pull away from using TXA and GI bleeds just based off of the crash trials in general. That being said, we do leave it up to our paramedics because it's not going to hurt. And so I think a lot of the times when I'm on the receiving end working in the community hospital, I noticed that they've still given it quite a bit of times. But.

Unless our new cogs have come out, we haven't specified one way or the other yet. We've just deemphasized it.

Cool.

Tanner, anything else from your end or anybody else have any other questions?

Nothing else for me. Yeah, wonderful job, Dr.

Le.

We are all anticipating what you come out with next with delta and the modified shock index, et cetera.

Editor answered Jordan in the chat. Just looking at our local statistics here in San Antonio, it's about 30% that we're giving pre hospital blood for GI bleeds.

And where your brother is, it's like 80% right. Yeah.

Like, they just. The trauma that they're getting.

Those good old boys are taking themselves to the hospital, so.

Yeah, that's funny. Okay, then. I don't know. Does anybody take care of the. It says Sam was asking us, can anyone put in the chat the link for the 11:30 call? So that would be. I don't do that. So. But I. Maybe. Maybe I can do that. Sorry about that. Maybe I can do that myself. I'll see if I can get to one of my emails and then throw it in the chat here in a second, so. All right, well, this is a good job. And I want. Josh, want to thank you because this was nice because you got it done, like in 10 minutes so we could have interactions. I really like that. So that. That's really, really appreciated. So thanks a lot. Tanner, did you take care of that, then? So is that right?

It's in the chat.

Okay, so listen, everyone, just want to say that thank you very much for doing this. And remember, your CME for today would be 19419, 9419. And with that. Right, Tanner, you want to close it out?

Absolutely, yeah. Thank you all for coming. Appreciate it.

See you next week. Right. Thanks again, Josh. Great job.