Thanks for articulating the contradiction that I have felt for the last decade (or more) as sepsis bundle metrics seem to have become more important than an accurate diagnosis, and actual patient outcomes.
I didn't know that. I see from your newsletter you have some pretty deep knowledge on the payment side of things -- the not so hidden hand that is influencing the whole machine. Thanks for writing.
I think the value of the evidence on sepsis is that we can still say that people treated it with this syndrome, whatever you call it, respond or don't respond to the interventions in the trials.
This is an excellent post looking deeper into the state of the science.
Dan, if we look at the results over the past 35 years (since Bone’s methylpred trial in 1989) I’m not sure there has been any value from the sepsis RCT. They are not reproducible and generally lead to reversed guidelines.
Here is a discussion of the PettyBone RCT which is the standard type of RCT performed.
It seems that we have not progressed in sepsis diagnosis and initiation of specific treatment in decades. However, I believe that one or more of the many emerging technologies are poised to revolutionize this issue and provide us with a nearly instantaneous bacterial (yes/no and what) diagnosis and AI-based recommendations that will help especially residents and peripheral hospitals.
I hope you are right! I have been looking at these new tests and so far I have not found a major game changer yet. Sepsis is just too diverse a syndrome for any one test to have the high performance most of us would need to see before relying on them. But maybe AI or new technologies will change that. Thank you for writing.
Disease definitions (and re-definitions) are a major issue in modern medicine, including sepsis
(Sneak peek-- I am working on an article on 'disease definitions" to be published in a couple of months on www.pcplens.com).
Since sepsis cannot be clearly defined, prediction algorithms in EMRs meant to detect "early-onset sepsis" don’t work. They produce excessive false positives with no clear guidance on management, increasing hospitalists' workload and malpractice risk.
The same problem applies to AI tools for sepsis. Despite numerous articles and ongoing work in Artificial Intelligence to identify sepsis early, they all ignore the core issue: there is no standard definition of sepsis.
You said it. But since we can’t clearly say what sepsis “actually is” — but have claimed to be able to , or gone along with what others have claimed (in the guise of “definitions” and clinical guidelines), we set the trap for ourselves.
This is great! The problem with sepsis is that it can easily take you by surprise and then kill your patient, and it will be too late. My own daughter was killed by septic shock, and it was very subtle…until the very end when was she was in irreversible multiorgan failure. So, if we “smell” sepsis, we need to treat it. That’s the problem. And…insurance companies love to deny sepsis DRGs because of this inherent uncertainty. I know we are still reeling from the UHC’s CEO murder and it has to be said.
Your comment took my breath away--sending you a private message. Yes: recognizing sepsis early is both difficult and essential. And so is standing up for what's right. We and our families are all vulnerable. Thank you for writing.
Thanks for articulating the contradiction that I have felt for the last decade (or more) as sepsis bundle metrics seem to have become more important than an accurate diagnosis, and actual patient outcomes.
Thank you for seeing and pointing out the through line from the limitations of clinical research to problematic policy / admin / operations.
The DRG for sepsis is higher--this is why the bundle has become very important to hospital systems.
E.g., For UTI with sepsis DRG, the payment rate for the hospital is higher than plain old UTI.
I didn't know that. I see from your newsletter you have some pretty deep knowledge on the payment side of things -- the not so hidden hand that is influencing the whole machine. Thanks for writing.
I think the value of the evidence on sepsis is that we can still say that people treated it with this syndrome, whatever you call it, respond or don't respond to the interventions in the trials.
I absolutely agree. I’m not arguing the trials don’t have value. As I said, “guesses have value and they’re what we have to work with.”
This is an excellent post looking deeper into the state of the science.
Dan, if we look at the results over the past 35 years (since Bone’s methylpred trial in 1989) I’m not sure there has been any value from the sepsis RCT. They are not reproducible and generally lead to reversed guidelines.
Here is a discussion of the PettyBone RCT which is the standard type of RCT performed.
Please also join the discussion there.
https://discourse.datamethods.org/t/the-petty-bone-rct/22077/17
It seems that we have not progressed in sepsis diagnosis and initiation of specific treatment in decades. However, I believe that one or more of the many emerging technologies are poised to revolutionize this issue and provide us with a nearly instantaneous bacterial (yes/no and what) diagnosis and AI-based recommendations that will help especially residents and peripheral hospitals.
I hope you are right! I have been looking at these new tests and so far I have not found a major game changer yet. Sepsis is just too diverse a syndrome for any one test to have the high performance most of us would need to see before relying on them. But maybe AI or new technologies will change that. Thank you for writing.
Disease definitions (and re-definitions) are a major issue in modern medicine, including sepsis
(Sneak peek-- I am working on an article on 'disease definitions" to be published in a couple of months on www.pcplens.com).
Since sepsis cannot be clearly defined, prediction algorithms in EMRs meant to detect "early-onset sepsis" don’t work. They produce excessive false positives with no clear guidance on management, increasing hospitalists' workload and malpractice risk.
The same problem applies to AI tools for sepsis. Despite numerous articles and ongoing work in Artificial Intelligence to identify sepsis early, they all ignore the core issue: there is no standard definition of sepsis.
IMHO, we need to divorce what CMS tells us what sepsis is and what it actually is.
You said it. But since we can’t clearly say what sepsis “actually is” — but have claimed to be able to , or gone along with what others have claimed (in the guise of “definitions” and clinical guidelines), we set the trap for ourselves.
The point is: what makes sepsis distinct from infection? That shall be the “substance” of sepsis
This is great! The problem with sepsis is that it can easily take you by surprise and then kill your patient, and it will be too late. My own daughter was killed by septic shock, and it was very subtle…until the very end when was she was in irreversible multiorgan failure. So, if we “smell” sepsis, we need to treat it. That’s the problem. And…insurance companies love to deny sepsis DRGs because of this inherent uncertainty. I know we are still reeling from the UHC’s CEO murder and it has to be said.
Your comment took my breath away--sending you a private message. Yes: recognizing sepsis early is both difficult and essential. And so is standing up for what's right. We and our families are all vulnerable. Thank you for writing.