Septic phenotypes

[lymphocyte]

I thought this was the most exciting article of the year apart from ANDROMEDA-SHOCK. Yes, beware geeks bearing calculators, but still what a doozy of an article in JAMA that used a machine learning algorithm to derive and validate host-response patterns in sepsis from some very large data-sets. 4 major phenotypes -- alpha, beta, gamma, and delta -- associated with wildly different prognoses and predicted responses to treatment when retrospectively dis-aggregated in the trials for Activated Protein C (PROWESS), Eritoran (ACCESS), and all the major EGDT trials.

It honestly feels like Magellen setting off into the New World. What are we even doing in sepsis when there's so much we don't understand? I gotta say I pay much more attention to the overall pattern of organ dysfunction now, more to think about prognosis beyond a generic SOFA score.

Seymour CW, Kennedy JN, Wang S, et al. Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis. JAMA. 2019;321(20):2003-2017. Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis. - PubMed - NCBI

---

Members don't see this ad.

 

[SurfingDoctor]

I thought this was the most exciting article of the year apart from ANDROMEDA-SHOCK. Yes, beware geeks bearing calculators, but still what a doozy of an article in JAMA that used a machine learning algorithm to derive and validate host-response patterns in sepsis from some very large data-sets. 4 major phenotypes -- alpha, beta, gamma, and delta -- associated with wildly different prognoses and predicted responses to treatment when retrospectively dis-aggregated in the trials for Activated Protein C (PROWESS), Eritoran (ACCESS), and all the major EGDT trials.

It honestly feels like Magellen setting off into the New World. What are we even doing in sepsis when there's so much we don't understand? I gotta say I pay much more attention to the overall pattern of organ dysfunction now, more to think about prognosis beyond a generic SOFA score.

Seymour CW, Kennedy JN, Wang S, et al. Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis. JAMA. 2019;321(20):2003-2017. Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis. - PubMed - NCBI

Conclusions and Relevance: In this retrospective analysis of data sets from patients with sepsis, 4 clinical phenotypes were identified that correlated with host-response patterns

Meaning, by the time we identified anything, the patients were either dead from sepsis or already left the hospital.

While I do appreciate this type of study, the applicability to clinical management is far past when I'm long and dead...

 

[lymphocyte]

the applicability to clinical management is far past when I'm long and dead...

As a trainee, (very hopefully) not for me, or you for that matter!

It does give me a lot of pause and humility when it comes to sepsis - so much we don't know, even defining the disease process itself.

 

[SurfingDoctor]

As a trainee, (very hopefully) not for me, or you for that matter!

It does give me a lot of pause and humility when it comes to sepsis - so much we don't know, even defining the disease process itself.

The NIH has a recent committee to try to steer sepsis research. Their basic conclusions were that mechanisms and clinical trials were a waste of time and money because nothing comes from them and the focus should be on massive pooling of genetic data in the hopes they can find some sort of signal buried in the noise. It was kinda disheartening to listen to because basically it was like them saying “we don’t know what we are doing and we are desperate for anything”. Of course, GWAS studies have almost never provided a cause and effect answer for any complex phenotype... but I guess they think this time it will be different.

Maybe machine learning will be employed at some point. It probably will be. But there is a lot of hesitancy from physicians. Algorithms can be wrong too and there is a lot of up front cost. I also have no idea how good AI is in leading to a favorable outcome since a vast majority of septic patients don’t die from the acute phase anymore.

 

[lymphocyte]

I also have no idea how good AI is in leading to a favorable outcome since a vast majority of septic patients don’t die from the acute phase anymore.

Pretty favorable! According to SkyNet. I mean, AI!

AI outperforms human clinicians in treating sepsis

Published in Nature Medicine. Researchers used machine learning to develop models for optimal fluid and vasopressor use. These models, on average, outperformed human clinicians; the further humans deviated from the machine model, the greater the mortality -- see the paper for further details. I...

forums.studentdoctor.net

 

[SurfingDoctor]

Pretty favorable! According to SkyNet. I mean, AI!

AI outperforms human clinicians in treating sepsis

Published in Nature Medicine. Researchers used machine learning to develop models for optimal fluid and vasopressor use. These models, on average, outperformed human clinicians; the further humans deviated from the machine model, the greater the mortality -- see the paper for further details. I...

forums.studentdoctor.net

Yeah, that stimulation tested stimulation. I think I left my opinion pretty clear as to AI in that thread you linked.