0:00

Welcome to the Skeptics

0:02

Guide to Emergency Medicine.

0:05

Meet him, Greet him,

0:07

treat him, and to

0:10

the Skeptics Guide to Emergency

0:12

Medicine. Meet 'em, greet 'em, treat 'em,

0:14

and street 'em! and date is October

0:16

1st, 2024, and I'm

0:18

your skeptical host, Dennis title The

0:20

title of today's podcast is... is,

0:22

Like the legend of

0:25

the Phoenix, criteria for

0:27

sepsis. In our our

0:29

guest skeptic today Professor Damien who

0:31

is a consultant at the

0:33

University of Leicester of Leicester NHS and

0:35

honorary professor for the University

0:38

of Leicester of Leicester Sapphire He

0:40

specializes in pediatric emergency

0:42

medicine and is a passionate

0:44

believer a advocate of advocate of

0:46

foam ed. Deming part of

0:48

that wonderful wonderful Don't the Bubbles

0:50

team. back to to the S gem

0:52

Damien. It's All right, it's lovely to

0:54

be back here. It feels like it's

0:56

been too long since I've been able to

0:58

been and discuss things on Ask things on

1:01

Eskam. Well, I think the last

1:03

time you were here, we had

1:05

a very nerdy discussion about sepsis screening.

1:07

Oh, I don't think I'm supposed

1:09

to say that, but I think

1:11

today's episode is also in that

1:13

vein. vein. What's the the case you brought

1:15

us? us? Pretty classical through-year-old

1:19

boy. In your ED, you've got a high

1:21

a high fever. breathing fast, lethargic.

1:24

state that he's had a had a fever

1:26

and cough for the past three days. days, tested

1:28

positive for influenza two days ago, ago, but

1:30

seems to be getting worse, which is why

1:32

they brought them here. they brought You

1:35

look at You look at not even looking up

1:37

at you, up at in the right lung field.

1:39

the right lung field, 88 on room

1:41

air, heart rate is high, rate

1:43

is clinging to his and he's

1:45

just clinging to his parents. you've

1:47

A medical with been working with on the

1:50

shift just comes up and asks, and looks

1:52

really sick. sick. Is Is this pneumonia or

1:54

could this be sepsis? Yeah,

1:56

Yeah, Damien, we see cases like this

1:58

all the time, right? right? have

2:00

a fever in their tachycardic, but

2:03

not all of them are sepsis.

2:05

Well, sepsis is an odd thing

2:08

because I think some of

2:10

us are quite privileged and that

2:12

we see lots of children

2:14

with kind of viral illness, but

2:16

very few of them have

2:18

sepsis. But if we take the

2:21

totality, pediatric sepsis is a major

2:23

global health problem, 3.3 million deaths

2:26

annually, and... We've been struggling with

2:28

this for years. 2005 international

2:30

pediatric sepsis consensus conference. They established

2:32

kind of criteria based on

2:34

expert consensus to use for diagnosis.

2:37

One of the problems was

2:39

though is they weren't particularly specific

2:41

and as I've alluded to there

2:44

are a difference in outcomes between

2:46

countries and so these diagnostic criteria

2:48

weren't adaptable across different resource

2:50

settings. For example, it contains some

2:53

fuzzy terms like SERS, which

2:55

honestly I've never really found at

2:57

all that useful. I mean,

2:59

what is systemic inflammatory response syndrome,

3:02

right? Is it like, is it

3:04

sepsis light, diet sepsis, sepsis zero?

3:06

Like it's just not that useful.

3:09

And the SGM is no

3:11

stranger to sepsis as we've covered

3:13

this topic multiple times that

3:15

we'll list in our show notes.

3:18

And I'm really excited about

3:20

this episode because I was pretty

3:22

downbeat when we did the sepsis

3:24

screening episode about how challenging sepsis

3:27

was but we're now covering the

3:29

newest criteria, the Phoenix criteria

3:31

for the diagnosis of pediatric sepsis

3:34

and sepsic shock. And it

3:36

is more evidence-based and I think

3:38

the interesting thing here is

3:40

it takes data from both high

3:43

and low income settings. So Damien,

3:45

what's the clinical question that we're

3:47

asking today? How actually can a

3:50

new clinical decision rule the

3:52

phoenix sepsis score diagnosed pediatric sepsis

3:54

and septic shock in hospitalized

3:56

children within the first 24 hours?

3:59

In what's our reference? So

4:01

it's Sanchez-Pinto, Eddale, development and validation.

4:03

the Phoenix criteria for pediatric sepsis

4:05

and septic shock published in German

4:08

2024. All right, let's go through

4:10

our Peacot questions. What was

4:12

the population they included in the

4:15

study? So these were children

4:17

less than 18 years of age

4:19

with suspected sepsis and septic

4:21

shock who admitted to one of

4:23

10 hospitals in five different countries.

4:26

And they excluded newborns and children

4:28

with post-conceptional age less than 37

4:31

weeks? And what was the

4:33

intervention? Or the intervention was the

4:35

Phoenix criteria and this is

4:37

essentially a scoring system based on

4:39

clinical science, symptoms and laboratory

4:41

values. And the comparison? The current

4:44

international pediatric sepsis consensus conference, that's

4:46

the IPCC criteria. Let's talk about

4:49

their outcomes now, what was the

4:51

primary outcome? Well this is

4:53

the thing that people fear the

4:55

most of it was in

4:57

hospital mortality as a single primary

5:00

outcome. Yeah, I would say

5:02

that's pretty patient oriented. I care

5:04

about that. And what about the

5:07

secondary outcomes? So they had a

5:09

composite of early death within 72

5:11

hours of presenting at the

5:13

hospital and a requirement for ECMO,

5:16

it's extracorporeal membrane oxygenation. For

5:18

the Phoenix scoring system, they used

5:20

the area under the receiver,

5:22

operating characteristic here, if that's the

5:25

orc, and for binary criteria, positive

5:27

predictive value, the PPV, and sensitivity.

5:30

And finally, what kind of trial

5:32

was this? Well this ends

5:34

up being a multi-center retrospective cohort

5:36

study. We are joined by

5:38

extra special guest today. We have

5:41

two of the guest authors

5:43

on this paper. Dr. Elizabeth Alpern

5:45

is a professor of emergency medicine

5:48

and chief of emergency medicine at

5:50

Lurie Children's Hospital of Chicago and

5:52

vice chair of pediatrics at

5:54

Northwestern University Fineberg School of Medicine.

5:57

She is also a clinical

5:59

epidemiologist and expert in large databases,

6:01

including the Pecorn Registry. Dr.

6:03

Halden Scott is an associate professor

6:06

of pediatricians. at the University of

6:08

Colorado School of Medicine, and in

6:10

attending physician at Children's Hospital, Colorado,

6:13

where she also serves as

6:15

director of research. Her research interest

6:17

includes the diagnosis and treatment

6:19

of sepsis across the emergency care

6:22

continuum. Dr. Alpern, Dr. Scott,

6:24

welcome to SGMPeds. Thanks so much

6:26

for having us. Great to be

6:28

here. Thank you so much. Thank

6:31

you so much. I have a

6:33

warm-up question for you all,

6:35

and that's around the name. Phoenix.

6:38

How was this decided? This

6:40

was presented at Phoenix Arizona at

6:42

the Society for Critical Care

6:44

Medicine Conference, but it's also the

6:46

name of a mythical bird. What

6:49

was the inspiration? You might have

6:51

thought you were giving us a

6:54

softball question there, but like

6:56

everything to do with sepsis criteria

6:58

and definitions. Nothing is that

7:00

simple? So certainly it was inspired

7:02

by the city in Arizona

7:05

where the definitions were presented and

7:07

also probably being named in relation

7:09

to that city was enhanced by

7:12

it being related to the mythical

7:14

bird, the phoenix, symbolizing, you

7:16

know, ongoing life from potential throes

7:19

of death and that's what

7:21

we like to do when we

7:23

treat patients with sepsis is...

7:25

capture and reanimate them and bring

7:27

them back to life from a

7:30

potential downward trajectory. Maybe a bit

7:32

of a stretch there. I don't

7:35

know. Libby, would you want

7:37

to jump in? I think you

7:39

captured it well, Halden. Of

7:41

course, as you implied, there was

7:43

a lot of discussion and

7:45

some other dissenting views, but we

7:48

came together in consensus section. Or

7:50

Dr. Alperm, would you mind giving

7:53

us your conclusions from the paper?

7:55

So the paper conclusions were

7:57

that the novel Phoenix sepsis criteria

7:59

were derived in value. using

8:01

data from higher and lower resource

8:04

settings had improved performance for

8:06

the diagnosis of pediatric sepsis and

8:08

septic shock compared with the existing

8:11

IP SEC criteria. All right, let's

8:13

go through our quality checklist for

8:15

clinical decision tools. Did the

8:17

study population include or focus on

8:20

those in the emergency department?

8:22

Yep, the study involves children presenting

8:24

to hospital, which could include

8:26

the emergency department. Were the patients

8:29

representative of those with the problem?

8:31

Yeah, the population included a broad

8:33

spectrum of pediatric patients with suspected

8:36

sepsets. I mean, we'll come

8:38

back to the suspected elements because

8:40

that could be challenged a

8:42

little bit. Were all important predictor

8:45

variables and outcomes explicitly specified?

8:47

Yeah, I mean, this was a

8:49

big, big data study. Variables like

8:52

organ dysfunction, lactemia, and vital signs

8:54

were explicitly detailed. And there wasn't

8:56

as much missing data as

8:58

you may have thought. Was this

9:01

a prospective multi-centered study including

9:03

a broad spectrum of patients and

9:05

clinicians? No, because it wasn't

9:07

prospective. That doesn't necessarily mean it's

9:10

of poor quality. It was retrospective,

9:12

but it was multi-centered and the

9:14

data collected was pretty thorough. Were

9:17

the clinicians interpretation of individual

9:19

predictor variables and scores in the

9:21

clinical decision rule reliable and

9:23

accurate? No, we can't say this.

9:26

A study relied on electronic

9:28

health records to extract and automatically

9:30

collect the necessary variables. So this

9:32

clearly helps ensuring standardization and it

9:35

probably enhances reliability of the interpretation

9:37

of some of the predictive

9:39

variables, but you have to remember

9:42

you don't know how those

9:44

variables were put into the EHR

9:46

in the first place. And

9:48

so we don't know how accurate

9:50

that was. And was this an

9:53

impact analysis of a previously validated

9:55

clinical decision rule? No, this is

9:58

the first validation study of

10:00

criteria. Was the impact the impact

10:02

on clinician behavior and patient

10:04

-centric outcomes reported? No, just by the

10:06

just by the nature of the study

10:08

design, the study did not assess changes

10:10

in clinical behavior or long -term patient outcomes,

10:13

making it a level one study. study. Do

10:15

you you think the follow -up was sufficiently long

10:17

and complete? and complete? follow -up

10:19

focused on on sepsis that critical

10:21

24 -hour window, window. but longer follow

10:23

-up for mortality was was less

10:25

So So possible, for example, example,

10:27

child may have been severely

10:29

affected by organ dysfunction, but

10:31

may have only died six

10:33

months later. six and that may have been

10:35

that may have been the effect large

10:38

enough and precise enough

10:40

to be clinically significant. Yeah, the

10:42

study reports study reports a clinically

10:44

significant diagnostic tool, I I think

10:46

it's powered to do that with

10:48

a focus on sensitivity and precision

10:50

due to the low the mortality rates. And

10:52

finally, the the study funding here

10:54

will list that, but of there

10:57

any conflicts of interest in

10:59

there? some foundational was

11:01

some foundational support, but I don't think

11:03

anything make people believe make people believe that

11:05

there was conflicts of interest here. here. Let's

11:07

talk a bit about their

11:09

results now. They They collected data

11:12

from over 3 million emergency departments in

11:14

and intensive care unit encounters

11:16

from 10 systems in the United

11:18

States. States, Colombia. Bangladesh,

11:20

China China, and

11:22

Kenya. in the United in

11:24

the United States were settings settings, and

11:26

the two sites in Bangladesh and

11:29

Colombia were deemed low resource settings, and

11:31

and those were included in the

11:33

derivation. And then then one from

11:35

the United States, China and Kenya,

11:37

were included in the validation. in the validation.

11:39

There are main systems

11:41

included included in the Phoenix Epsis

11:43

score, respiratory, cardiovascular,

11:46

coagulation and neurologic. and they recommend that

11:48

sepsis should be considered for any

11:50

child with suspected infection and

11:52

a score of at least two

11:54

points. least two points. And shock was

11:56

identified by at least one point

11:58

in the cardiobacteria. component. Do you

12:01

mean what was the key

12:03

result? Essentially they showed that

12:05

their Phoenix criteria demonstrated strong

12:07

diagnostic performance with high sensitivity

12:09

and improved position over previous

12:11

sepsis criteria. They found that

12:13

children with suspected infection in

12:15

the first 24 hours of

12:17

presentation had in hospital mortality

12:19

of 0.7% in higher resource

12:21

settings and 3.6% in lower

12:23

resource settings. And children with

12:25

sepsis had a mortality of

12:27

7.1% in higher resource settings

12:29

and 28.5% in lower resource

12:31

settings. So that's a big

12:34

difference. And then children with

12:36

septic shock had a mortality

12:38

of 10.8% in higher resource

12:40

settings, so we see quite

12:42

a disparity there. Yeah, I

12:44

mean these are big figures

12:46

and the Phoenix school had

12:48

a higher positive predictive value

12:50

and higher or comparable sensitivity

12:52

for in hospital mortality and

12:54

early death or ECMO compared

12:56

to the IPSCC definition. And

12:58

I think this is really

13:00

important. It's that improvement in

13:02

the predictive value that makes

13:04

this a useful tool. All

13:06

right doctors. It is time

13:08

for my favorite section. Are

13:10

you all ready to talk

13:12

nerdy? Always. Oh, we're ready

13:14

Dennis. Our first point is

13:17

around the scoring tool development.

13:19

And I want to acknowledge

13:21

that this was a huge

13:23

effort amongst a multidisciplinary group.

13:25

You all reviewed a bunch

13:27

of organ dysfunction scoring systems.

13:29

You ran them through multiple

13:31

models that includes stacked regression,

13:33

lasso, logistic regression. And then

13:35

you voted through a modified

13:37

delfi process on thresholds for

13:39

the score. The final Phoenix

13:41

score did not include kidney

13:43

function, which may leave some

13:45

nephrologist feeling a bit salty.

13:47

But with such a large

13:49

team in so many processes...

13:51

Were there any big areas

13:53

of agreement or disagreement? You

13:55

can probably guess the answer

13:58

there, and I think we've

14:00

highlighted some of them in

14:02

some of the supplements that

14:04

came out with the paper

14:06

and some of the presentations,

14:08

but we can touch on

14:10

the major ones. Maybe Libby

14:12

and I can take turns

14:14

talking about them. So I

14:16

think that one of the

14:18

key components that came for

14:20

discussion as we looked at

14:22

what would be the final

14:24

criteria was a concept of

14:26

parsimony. We wanted to balance

14:28

the ability of having a

14:30

score and criteria that would

14:32

be able to be done

14:34

throughout differing settings and an

14:36

understanding that if we could

14:39

achieve the same and identification

14:41

of the patient with sepsis

14:43

through a parsimonious score, that

14:45

there was a real benefit

14:47

to that. So when we

14:49

looked at the score that

14:51

came out, that had four

14:53

organ systems within it and

14:55

compared it. to the score

14:57

that had potentially eight organ

14:59

systems in it. As you

15:01

can see in the article,

15:03

there is the ability to

15:05

identify those patients that had

15:07

high mortality and the criteria

15:09

of sepsis using the four.

15:11

It was acknowledged within the

15:13

group that those other systems

15:15

are very important and there

15:17

would be times that we

15:19

may want to use the

15:22

criteria that were put forward.

15:24

in the Phoenix sepsis score

15:26

of eight in other avenues.

15:28

So for example, Dennis, as

15:30

you mentioned, we all know

15:32

that acute kidney injury related

15:34

with sepsis is something that

15:36

we care a lot about.

15:38

And just because it is

15:40

not included in the Phoenix

15:42

sepsis criteria, that it doesn't

15:44

mean that it's not an

15:46

important outcome for patients who

15:48

may have sepsis. So in

15:50

the Phoenix 8 organ dysfunction

15:52

score, you can see that

15:54

renal is... a component of

15:56

the score that it has

15:58

been determined to have the

16:00

differing levels of that criteria

16:03

and for say research particularly

16:05

related to kidney injury, we

16:07

can certainly implement the Phoenix

16:09

renal criteria and bring those

16:11

in. I think this was

16:13

one of the surprising and

16:15

interesting findings when we really

16:17

got into the data. Many

16:19

of us had been familiar

16:21

with the older definitions of

16:23

sepsis, which included multiple organ

16:25

systems, and so we're used

16:27

to looking for AKI as

16:29

a marker of early sepsis.

16:31

And it was really interesting

16:33

to learn from the data

16:35

that only those four systems

16:37

were sufficient on their own

16:39

if weighted properly to identify

16:41

children at the highest risk

16:43

of mortality, to identify the

16:46

most severe sepsis. In many,

16:48

many cases, these children also

16:50

had AKI, had these other

16:52

markers of illness that are

16:54

important. And they're really important

16:56

to managing their care. You

16:58

need to know if they

17:00

have AKI for the sake

17:02

of how you manage nefrotoxins

17:04

and fluids and resuscitate them.

17:06

But it turned out there

17:08

were few children with this

17:10

severity of illness who had

17:12

AKI without having the other

17:14

things. And when we pulled

17:16

out just something like the

17:18

children with kidney injury alone

17:20

while they were sicker than

17:22

children without kidney injury were

17:24

not nearly as sick as

17:27

that full cohort. So in

17:29

terms of daily operational uses,

17:31

you know, this whole project

17:33

to derive the criteria that

17:35

became the Phoenix criteria before

17:37

we even got into the

17:39

data and the modeling, it

17:41

started with a survey of

17:43

clinicians worldwide in a systematic

17:45

review of all existing literature

17:47

to think like how do

17:49

we conceptualize sepsis and how

17:51

do we use it? What

17:53

are its purposes? So from

17:55

a quality improvement purpose and

17:57

minimizing testing. when appropriate,

17:59

having fewer

18:01

criteria is easier to work with.

18:04

An An interesting controversy question

18:06

the question of whether organ

18:08

dysfunction needed to be remote

18:10

from the site of infection.

18:12

or if organ or if limited to the

18:14

to the site of infection.

18:16

defined sepsis. sepsis. a For

18:19

example, a patient with pneumonia

18:21

can have ARDS, truly life

18:23

-threatening organ dysfunction. all in the lungs

18:25

in the lungs where the site of

18:27

infection is. of or they can

18:30

be obtunded have shock and have

18:32

systematic organ dysfunction remote from the

18:34

site of infection. site of infection.

18:36

think from the the frontline department,

18:39

pediatrician perspective. to

18:41

me, I don't me. organ I

18:43

don't need the organ dysfunction to be remote

18:46

for me to think of it as they are high risk

18:48

of are high risk of mortality.

18:50

Their are the same. the same. I wanna

18:52

give them great early treatment for

18:54

infection regardless. regardless. From the perspective

18:56

of someone in the ICU, the ICU of

18:58

their patients have life -threatening organ dysfunction

19:00

of some kind, that distinction

19:02

maybe matters more and certainly

19:04

matters in terms of precision

19:06

therapies. and certainly And that's a

19:08

really, really vital point that's one of

19:10

the things that we've not had up

19:12

until you guys have done this

19:15

work. not had up until you guys

19:17

have done this work is this understanding

19:20

that concept that is not

19:22

only fluid and we know

19:24

that because kids go from go

19:26

from zero to very septic very

19:28

very quickly, but also

19:30

is a concept that is

19:32

done on environment. The emergency experience

19:35

of sepsis is not the sepsis

19:37

experience the is not the family and is

19:39

And one of the battles, and

19:41

I still think I've got a

19:43

bit of this the battles and I Is

19:45

it possible for a

19:47

child to die it possible for

19:49

a lower to tract infection? of

19:51

a lower If you just have have

19:54

and you die, die, then what

19:56

have you died you died of? I love

19:58

love that question. and think

20:00

if you asked 10 different

20:02

doctors, you'd get 10 different

20:04

answers. answers about whether a child can

20:06

die with a respiratory infection alone. I

20:09

know for sure And I

20:11

you asked the asked the group

20:13

of the Phoenix criteria, you you

20:15

would get different answers. And

20:17

we know when we surveyed

20:20

globally, there were different answers

20:22

to that question around the

20:24

world. That's This was where people's of

20:26

sepsis differed. The the

20:28

treatment may be. similar or

20:30

the same the same of how

20:33

of how you answer

20:35

that question. interesting. interesting, and

20:37

I don't know, if you if

20:39

you intended to add a pun

20:41

there with the word when it

20:43

it comes our sepsis talk, but I talk,

20:46

but I like it regardless. point

20:48

So our nerdy two is is about

20:50

the actual chosen criteria. We've

20:53

We've mentioned before that

20:55

the that the final focuses on

20:57

four organ systems, systems, respiratory

20:59

coagulation, and neurological. and

21:02

Some criteria, lab criteria, which certain centers

21:04

may not have or may not be

21:06

able to get results in a timely

21:08

fashion. in a And to be

21:11

honest, I'm probably not I'm kind

21:13

of calculating kind of PAO2 over SPO2, ratios

21:15

in the emergency department unless unless unless

21:17

the sole person looking after

21:19

the most sick kid there. And

21:21

I'm literally handing over to

21:23

the intensive care team. care team,

21:25

data on the use of lactate of

21:27

a bit mixed, a bit and the

21:29

the interracial of GCS scores can also

21:32

be quite variable. variable. So how do you choose

21:34

you choose which components under each each system

21:36

to include? Did that all come out

21:38

of the data data? Or there a lot

21:40

of of about those points? those points?

21:42

Don't get me started on the data

21:44

for data but lactate, but a separate question. Another

21:47

episode, perhaps? perhaps. Oh, I I

21:49

would love to. The idea was

21:51

to work after idea was to

21:53

work after this systematic

21:55

review was done, after the

21:57

survey was done, and three

21:59

for adults... existed. All the data

22:01

and conceptual understanding of sepsis

22:03

worldwide still seemed centered on

22:05

organ dysfunction. So the way

22:07

the data-driven process worked was

22:09

to take components of previously

22:11

derived and validated organ dysfunction

22:13

scoring systems. and choose the

22:15

best one of each of

22:17

them in each system. So

22:19

every organ dysfunction score that

22:21

existed for cardiovascular competed against

22:23

each other to select the

22:25

best, and then respiratory, etc.

22:27

And that is where this

22:29

very high threshold for lactate,

22:31

a lactate of five. It's

22:33

brilliant that threshold. I mean,

22:35

don't get me wrong. For

22:37

those of us who are

22:39

frustrated. by the retrospective criticism

22:42

you can get by not

22:44

acting on a lactate of

22:46

2.3, I think that's a

22:48

really reasonable threshold to go

22:50

for, because you are dog

22:52

sick at that point. Yeah,

22:54

we're using it to define

22:56

a life-threatening condition, a lactate

22:58

of five, came out in

23:00

the data, this definition that

23:02

included a lactate of five,

23:04

also included hypotension, use of

23:06

vasopressors, came out as the

23:08

cardiovascular definition. There are limitations

23:10

to all of them. The

23:12

pragmatic aspect of the way

23:14

these scores were applied to

23:16

existing data as it's currently

23:18

captured in the electronic health

23:20

record should mean that noise

23:22

such as differences in assessing

23:24

GCS or differences in the

23:26

threshold came out and the

23:28

best of existing scores was

23:30

chosen. So the initial choice

23:32

of organ systems was based

23:34

on existing sepsis scores. And

23:36

we did not look at

23:38

the data de novo and

23:40

come up, say, in a

23:42

predictive modeling way, which would

23:44

be out of all possible

23:46

components, which would come forward

23:48

in this score. But we

23:50

did bring in, think it

23:52

was five different existing scores

23:54

and utilized their components. And

23:56

then we went to the

23:58

data to help us look

24:00

and see which were most

24:02

associated with the outcome. And

24:04

I think if you did

24:06

what you suggested, you'd end

24:08

up back with SERS. Hopefully

24:10

not. Which is what we

24:12

did, which is what we

24:14

did. We don't want that

24:16

anymore. Our third nerdy point

24:18

is about suspected infection. And

24:20

the inclusion of patients was

24:22

based on suspicion of sepsis,

24:25

and that might lead to

24:27

some selection bias affecting generalizability.

24:29

So the child with the

24:31

severe traumatic brain injury from

24:33

a motor vehicle accident, fixed

24:35

dilated pupils, is not septic

24:37

because he's got a score

24:39

of two based on the

24:41

Phoenix criteria. But suspected infection

24:43

was defined as the receipt

24:45

of systemic antimicrobials or antimicrobial

24:47

testing. And this still creates

24:49

quite a degree of subjectivity

24:51

and possible variation amongst providers.

24:53

Do you have any tips

24:55

on how to deal with

24:57

this aspect? I think a

24:59

lot of our work is

25:01

done. The first part of

25:03

our job in emergency medicine

25:05

is done by the time

25:07

we decide to do infectious

25:09

testing or give a child

25:11

systemic antimicrobials, right? We see

25:13

a sea of self-limited viral

25:15

illness. filling our waiting room

25:17

and a lot of our

25:19

job is discerning the child

25:21

who needs more of a

25:23

workup and more treatment. It

25:25

makes sense for the definition

25:27

of sepsis to start with

25:29

some clinical suspicion. We have

25:31

not taken humans out of

25:33

this loop yet. So clinicians

25:35

are discerning that someone needs

25:37

treatment for infection, and then

25:39

the sepsis criteria help identify

25:41

those with life-threatening infection from

25:43

among those patients. It also

25:45

highlights the clinical work still

25:47

needs to be done prior

25:49

to deciding someone has sepsis.

25:51

Also, there's plenty of research

25:53

work still to be done.

25:55

determining who

25:57

this population is

25:59

to whom to

26:01

criteria were

26:03

applied. I just

26:06

think that I has really

26:08

highlighted has of the next steps.

26:10

Now, next as we have

26:12

with the the for sepsis,

26:14

we can identify those

26:16

kids that have sepsis

26:18

in a unified and

26:20

rigorous fashion. rigorous that we

26:22

can we can have a better

26:25

ability in our. our upstream effects to

26:27

to do predictive modeling

26:29

and other identification so that

26:31

we can actually have

26:33

a better chance to identify

26:35

these patients and treat

26:37

them for their infection in

26:39

a consistent manner. 30 point is

26:41

about our fourth nerdy point

26:43

is about generalizability. And this study

26:45

used data from many institutions across

26:48

the world. They had an They had

26:50

an international sample that included high

26:52

and low resource settings. settings. You You

26:54

also included children with complex health

26:56

care needs and all of these things

26:58

make the results more generalizable. generalizable.

27:01

was some missing data from one

27:03

of the lower resource sites,

27:05

and the higher resource sites were

27:07

all all children's hospitals. hospitals. So when

27:09

you undertake the next iteration of

27:11

this project, are are there plans

27:13

to recruit from an even

27:16

wider array of sites, including more

27:18

community hospitals and higher higher settings?

27:20

settings? Yeah, I Yeah, I think that

27:22

there's the balancing factors of of

27:24

and where data can

27:27

be obtained. be obtained. It was a

27:29

real step real step forward

27:31

that lower resource settings were

27:33

included and contributed data that

27:35

that we were able

27:37

to test the criteria and

27:39

their performance across settings,

27:41

even though to some extent,

27:43

it It performed well across

27:45

settings, but certainly in one setting.

27:48

it performed less well than in others.

27:50

And so And so that to be taken into

27:52

consideration when it's... Applied more

27:54

broadly. applied more the truth is, the

27:57

the lowest resource settings have

27:59

have the. least ability to contribute data.

28:01

And even the settings where

28:03

this data came from did have

28:05

like large organized systems of

28:07

data collection that is not available

28:09

worldwide. So these still were

28:11

lower, but not truly low resource

28:13

settings that could contribute all

28:15

of the data. Similarly,

28:18

any one community hospital by

28:20

itself does not see that many

28:22

children with sepsis in a

28:24

year getting data from that kind

28:26

of setting as an individual

28:28

hospital can be really challenging. And

28:31

yet in total, that's where

28:33

most of the children with sepsis

28:35

present. So we work with

28:37

the data we have while constantly

28:39

striving to find mechanisms to

28:41

study the care that's happening in

28:44

the most real world settings. And

28:47

one future direction for research

28:49

that we're involved in now

28:51

is trying to come up

28:53

with code to allow for

28:55

identification of patients that make

28:57

the phoenix sepsis criteria within

29:00

the EHR. I was surprised

29:02

that it was so homogenous

29:04

despite the huge differences in

29:06

patient type and quality of

29:08

care received. They've alluded slightly

29:10

harder to the fact that

29:12

maybe one of the centers

29:14

was slightly different, but it

29:16

does interest me. If you

29:19

take someone who's lived in

29:21

a non -deprived area who

29:23

ends up with sepsis goes

29:25

to a very high functioning

29:27

high tech ICU in the

29:29

States and you compare that

29:31

with a child who's from

29:33

a perhaps deprived background who

29:35

ends up in a low

29:38

resource country without some of

29:40

the care needs that you

29:42

would expect or you can

29:44

give in some North American

29:46

sites. Yes, I know that

29:48

you were looking at some

29:50

of the baseline data, but

29:52

it interests me that actually

29:54

you were able to bring

29:57

all these sites together so

29:59

effectively. you surprised by that? by

30:01

that? I I mean, this is

30:03

really credit to the and

30:05

and their collaborators on the

30:07

data research aspect of this

30:10

project. So Tell Bennett and Nelson

30:12

Sanchez, Pinto led the building

30:14

of these datasets with

30:16

their partners globally And

30:18

it was really incredible.

30:20

I think we were

30:22

all surprised and impressed how

30:24

well well the score

30:26

calibrated across settings. settings. You know,

30:28

we have to say at any

30:30

single Phoenix score, the The outcomes were

30:32

worse in lower you settings,

30:34

where the where the global

30:36

burden of disease and mortality

30:38

is worse. we saw But what

30:40

we saw and included in the

30:42

manuscript are these calibration plots. plots. And so

30:44

additional point, a score of of

30:47

three had a a higher mortality than

30:49

a score of two. score of

30:51

four had a higher mortality. Each

30:53

increment that the score went up,

30:55

the mortality was worse. was And that

30:57

held true in higher and lower

30:59

resource settings. settings, even though the

31:01

absolute mortality was highest. And

31:03

there's there's redundancy in the

31:05

score. The threshold of of two the

31:07

diagnosis of sepsis meant that it

31:09

could be be met. through clinical scores

31:11

alone like hypotension. with settings So

31:14

helped with be that might not

31:16

be measuring a lactate could meet

31:18

their cardiovascular criteria a by having

31:20

a very low blood pressure. that

31:22

didn't that didn't have or didn't

31:25

or didn't have more than one

31:27

vasopressor could still meet the

31:29

criteria by having an even lower

31:31

blood pressure. So there are

31:33

multiple ways to account for treatments

31:35

that might not be available

31:37

everywhere or measurements that weren't available

31:39

everywhere. everywhere. Very Very cool.

31:41

Very cool that that redundancy honestly

31:43

adds and strengthens the generalizability of

31:45

the study. of the study.

31:47

Our last point, and and Damian, you have

31:50

to be the one to tackle this, is

31:52

this is about this is is not

31:54

a screening tool. is really This

31:56

is really important really was really

31:58

important to me when I when I this paper

32:00

for the first time, because I

32:03

had two reactions to it. One

32:05

is brilliant. We now have something

32:07

that tells us from a clinical

32:10

point point, for an emergency clinician

32:12

standpoint for some time, that children

32:14

with sepsis, when you've got it,

32:16

you are sick and that's what

32:19

the school shows. You have to

32:21

have some level of organ dysfunction

32:23

and you are going to present

32:26

a well. And then my second

32:28

thought is... how are people going

32:30

to misuse this? And the real

32:33

danger and the problem that we've

32:35

got ourselves into is that as

32:37

a child health professional, I want

32:40

to improve outcomes for children, but

32:42

I'm not going to be able

32:44

to improve outcomes for children if

32:46

we set up an almost regulatory

32:49

setting that penalizes clinicians for making

32:51

good judgment. and one of the

32:53

problems, and this may not have

32:56

been a problem perhaps in North

32:58

America, but was certain a problem

33:00

in the UK, is that because

33:03

of the application of SERS criteria,

33:05

on a decision-making basis, if you

33:07

unfortunately decided for completely the correct

33:10

reasons to not treat a child

33:12

who was a febrile tachycardia, and

33:14

then four days later they came

33:17

back and were severely unwell, you

33:19

would be penalised for that decision.

33:21

because that initial presentation said this

33:23

kid could have had sepsis. You've

33:26

made it very clear and that

33:28

maybe I should read out your

33:30

sentence because I think this was

33:33

a really important one. Phoenix criteria

33:35

for sepsis, septic shock were not

33:37

intended to identify life-threatening organ dysfunction

33:40

due to infection in children. They

33:42

were not designed for screening in

33:44

children at risk of developing sepsis

33:47

or early identification of children with

33:49

suspected sepsis. But the question that

33:51

leaves this is, what is going

33:53

to be that? and the holy

33:56

grail remains. I think we all

33:58

agree, while there's many controversies, I

34:00

think you've summed it up, we

34:03

need screening tools. More screening tools

34:05

are needed. better screening tools are

34:07

needed. And I also don't know

34:10

if we'll ever find the perfectly

34:12

sensitive and specific screening tool. Some

34:14

tradeoffs are needed in how we

34:17

apply them. Like you said, judgment

34:19

is needed. The advance of the

34:21

Phoenix criteria are we can all

34:24

agree what sepsis is now and

34:26

work to predict this outcome. And

34:28

another real step forward is You

34:30

know, papers have been showing for

34:33

years the limitation of SERS criteria.

34:35

They were neither sensitive and at

34:37

the same time they created too

34:40

many false positives. So this really

34:42

is codifying the end of SERS.

34:44

SERS does not define sepsis in

34:47

any way anymore, not even on

34:49

paper. And we can move forward

34:51

to try to find a better

34:54

screening tool that has slightly fewer

34:56

limitations than SERS did. as has

34:58

been stated throughout the manuscript, this

35:01

is not a predictive model. This

35:03

does not show which patient needs

35:05

treatment for sepsis. This shows which

35:07

patient has sepsis. And can you

35:10

set the record straight for us?

35:12

If a patient who scores zero

35:14

on the Phoenix score, does that

35:17

mean like, I'm in the clear?

35:19

There is no possibility that they

35:21

are septic. Is that true? They

35:24

don't have sepsis. They might have

35:26

it soon. We still need to

35:28

figure out how to get better

35:31

at predicting trajectories. That is very

35:33

well put. Wow, that was some

35:35

great nerdy discussion, but can you

35:37

comment on the author's conclusion compared

35:40

to the SGM conclusion? Yeah, I

35:42

mean, I agree with the author's

35:44

conclusion. I think they've developed a

35:47

great tool. and I think despite

35:49

some of the limitations that we

35:51

may have discussed, this is a

35:54

real way forward in the management

35:56

of sepsis. And give us the

35:58

SGM bottom line. The Phoenix Sepsis

36:00

score offers a new and improved

36:02

evidence -based method of defining sepsis and

36:05

septic shock in children. shock in that

36:07

case at the beginning with that kid

36:09

that maybe has a pneumonia but really doesn't

36:11

look that well. Can you resolve that

36:13

for us? that for us? Well, I I think

36:15

this is interesting and this is

36:17

where is where comes comes own as

36:19

a research tool, but we probably

36:21

haven't answered this question yet. So we can

36:23

can tell the trainee the patient

36:25

looks very ill. ill. We suspect

36:28

the possibility of bacterial pneumonia on

36:30

top of the influenza. We know

36:32

that happens and sleepiness is concerning given

36:34

how long he's been unwell for. And

36:36

actually as as it happens,

36:38

he ends up being placed on

36:40

bipap getting antibiotics, and and his

36:42

mental status continues to decline. He

36:44

eventually is intubated and transferred

36:46

to the intensive care unit where

36:48

his where his shows that the

36:50

high risk of mortality he eventually

36:52

had. had. Now I

36:54

Now, I have three brilliant minds

36:56

in this conversation. I'm excluding

36:58

myself from that. But when

37:01

it comes to actually applying

37:03

this clinically in the emergency

37:05

department in the have you found

37:07

ways to incorporate it? you

37:10

found ways I think we

37:12

don't talk and think

37:14

and terms of organ

37:16

dysfunction enough in the

37:18

emergency setting. emergency setting. So

37:20

Phoenix. highlights four that

37:22

are sufficient for

37:24

defining sepsis. spend a lot of spend

37:26

a lot of time in

37:28

conversations about the appropriate disposition for

37:30

a patient. Are they safe

37:32

for the Do Do they require

37:34

transfer to our center or not?

37:37

If of a enough of a question

37:39

that we are having a

37:41

large conversation about this, we really

37:43

should be thinking in terms

37:46

of organ dysfunction scores. Phoenix some

37:48

of them. think think out to

37:50

the eight criteria is really helpful

37:52

because the data that went

37:54

into this project and many Many

37:56

other studies have shown the number of

37:58

organ systems that are present

38:00

by by the time of ICU

38:02

admission, determine outcomes, in and in that

38:04

sense. to encourage really try to

38:06

encourage residents, when medical students, fellows

38:09

when we're having these conversations about

38:11

patients to try to be explicit

38:13

in counting the organ systems that

38:15

are present even even thinking about the

38:17

ones they've tested for. tested for.

38:19

Yes, and and hopefully that's

38:21

really what Phoenix does. And

38:23

I think it's gonna

38:25

be very important for all

38:27

of us to recognize

38:29

that work that comes forward

38:31

now, utilizing Phoenix as

38:33

the the criteria, may show

38:35

differences in therapeutics or outcomes

38:37

did when work did when

38:39

sepsis was identified through a

38:41

different criteria. going to be it's

38:43

gonna be important for us to

38:46

continue to recognize that so that

38:48

we don't make false conclusions, but

38:50

but that we understand that

38:52

Phoenix really does identify a very

38:54

ill population. and

38:56

that if you're comparing that,

38:58

say, to to P sofa or the

39:00

the international criteria, that

39:03

you may have differing outcome

39:05

distributions. So Damien what are what are

39:07

you telling your learners and those students in

39:09

this setting in they're setting when you think

39:11

this person or this patient might

39:13

be or this patient might so what

39:15

I'm telling my I'm telling my is

39:17

still an there to medicine to

39:19

it's not about it's not about numbers

39:21

and You still need to be

39:23

guided by patients you are of are

39:25

are going to pick up

39:27

and treat. be very think we'd

39:29

be very clear. You don't

39:31

apply the every child on every

39:33

your department. That That is not

39:35

how it works. But what

39:37

we need to do is

39:39

pick up those patients are bat bells,

39:41

are ringing. We're worried. So ringing, be we're

39:43

worried. it's ill appearance and his ill makes

39:45

me makes me concerned that there's

39:47

something else going on. and that I

39:49

that I am going to start treating him

39:52

and as part of that that work up he may

39:54

may well end up getting a a

39:56

Phoenix Sepsis score and on on the basis

39:58

of that, we can start moving. forward to

40:00

making sure that this child has

40:02

a definitive care that they need. care

40:04

was talking about as being

40:06

able to define about,

40:08

dysfunction. to define a clinical measure, a

40:10

something that we should be doing

40:13

in the way we're talking between specialties,

40:15

I think it's gonna be really

40:17

helpful moving forward. specialties, I think is to be

40:19

really Scott, and moving thank you

40:21

very much for joining us on

40:24

Scott, and Thank you for having

40:26

us. very much been a pleasure, it's been

40:28

a great discussion. Thank you. you. I learned

40:30

so much from doing these

40:32

episodes with cool guest skeptics and

40:34

very knowledgeable authors. authors.

40:36

before we go, go, do you mind giving

40:39

us? us the SGM tagline? Remember

40:41

to to be skeptical of anything

40:43

you you learned. if you heard it.

40:45

on the Skeptics Guide to

40:47

Emergency Medicine. Talked everyone next

40:50

time.