Let's click on it, here we go. I do have a financial disclosure related to Intellects Healthcare Systems, which looks into artificial intelligence, and I'm a clinical advisor. This is not pertinent to this presentation. This regulated immune response is what sepsis is, by definition. It is very challenging at the bedside to figure it out. It is a global problem, as you all know. Nearly 80% of the patients presenting with sepsis come through the emergency room to be admitted. At the bedside, it's very challenging, despite all the efforts and all the work that's being done in that aspect. That's mainly related to the variability of the presentation, and really the lack of a gold standard for what sepsis is. Early intervention is a requirement, otherwise a patient is going to die. Over-diagnosis is also a problem, because it's leading to overuse of antibiotics and secondary antimicrobial resistance. Thank you. If I'm an emergency room healthcare provider, and I see a patient that I'm suspecting sepsis in, I have two things in my mind. This person is at a high risk for dying in front of my eyes, and within every five minutes I delay the care, there is somebody who's going to die from it. This is one of the oldest studies that you have showing the rapidity by starting antibiotics and the rate of deterioration if it's not happening. It's an amazing visual presentation. Ken Remy showed earlier what the trajectories of these patients are, and all the different chemicals that get produced in our bodies, but the emergency room physicians only have a couple of hours to figure this out. It's a little bit different than when you have all the labs and all the data that you have in front of you in order to make a clinical decision, but for the ER presentation, it's a very short period of time. So what do we have to have in order to make such a diagnosis? There has to be an infection. There has to be either a documented infection or a suspected infection, in addition to a presence of inflammation, in addition to the organ dysfunction. The tools that we have at the bedside are some clinical scoring tools. Some of them have been useful, some of them have not been useful, but lately there has been more automated alerts to get people to the bedside to try to address the management issues. There has been several alerts in the system, and I only show you this as a kind of multiple studies that have put alerts in their emergency departments. Many did not really improve mortalities, actually only one perhaps did, but they all document the importance of the process of improvement in care. Decreasing the time to antibiotic use, decreasing the time to admissions to the hospital. So these processes of alerts are helpful, but not necessarily the answer to every single question. All of you know what the national early warning score is, the news that we use it, and that has been implemented in the UK for several years now. This is when it started off in some part of the UK, and each line here represents a hospital, and there is an overall crude mortality reduction despite the fact that the diagnosis of substance has increased significantly. So we can impact the outcome in these situations. Unfortunately, the scoring systems that we have are not very accurate in one, and they are not necessarily very specific. We look at vital signs, we look at some lab work, and we always look at it as a single entity and compare it to each other, and you end up with a ROC curve, and you figure out if there is any helpful events. In one late study in the emergency department, it seems like news was a bit better than the QSOFA and SIRS, and obviously SIRS and QSOFA are still being used just because there are numbers that we can tackle. They are not really very helpful in the establishment of a diagnosis of sepsis. The price that you pay is as you increase the threshold for news, you will have a better specificity, but you lose the sensitivity, and the problem in those situations is you may miss some patients that are going to be deteriorating. PIC is one of the major healthcare electronic records, and it has created a sepsis prediction model through which it has really got machine learning on a lot of prior cases, and in one of the recent establishments tried to compare it to how SIRS does and how QSOFA and SOFA does, and it fared pretty well. The problem with it, it has not been validated well across multiple data sets, so so far it's been kind of, I would say, being told to every hospital that uses the EPIC system that this is the way to go, but it needs to be validated before we continue on that. The one prospective real-time intervention happened in Maryland, actually, where five hospitals or a little bit more used machine learning, particularly in supervised machine learning, in determining a targeted real early warning sign machine, and they had a training set and then they applied it at the bedside for a treatment group and a comparison group, and this system that has at least like 50 or 53 data points was able to predict the presence of sepsis before sepsis by about three hours or so. This would be preventive if this is applied or if we have it available to all of us and it truly is validated and outside these hospitals, it would be great that you can identify people who are going to get sick before they get sick, and in reality, we will go out of business and retire. The one thing that is very difficult for the ER doctors is to, they don't have a biomarker, they go by clinical grounds. Clinical grounds are okay, but can you imagine anybody taking care of a myocardial infarction without an EKG? It's a bit odd that we take care of very sick syndrome, very problematic syndrome, without a biomarker. Biomarkers, there are so many of them, there are actually like 200 plus of them, but none of them has been validated to be the absolute truth in determination of sepsis. These biomarkers, by definition, are a measurable indicator of biological state, whether it's normal or pathologic, and in reality, for it to be ideal, it should be affordable, sensitive, specific, user-friendly, rapid, and you don't need any extra machines because it won't be available to the world. The CBC, we all use. There is one element in all the flow cytometers that occurs nowadays, which is the monocyte diameter width, which is actually tossed in the garbage, it's not even recorded in the electronic health records, but when you add it to the white blood cell count, it becomes a little bit better than the white blood cell count alone. As I said, biomarkers are increasing and every lab decides to do one and they prove it one way or the other, but so far, none of them has reached clinical medicine in a good way. We have to take into consideration that the one that we commonly look into, which is the CRP, has never been validated, and it's the most nonspecific, yet every single study compares it, any new thing, to CRP, which is literally not helpful. Same for procalcitonin. Procalcitonin had made its way for now probably 15 years or so. It's a good tool, but it's not a tool for diagnosis. It's a tool for perhaps de-escalating therapy if people are improving. The newest one that seems to have some promise in Europe, that's being applied in Europe well, is the pancreatic stone protein. I never knew this, but I learned that the pancreas can tell if organs are failing, and if it is, it starts to increase, and if it's overwhelming, it circulates in the body, and it looks like it has a similar benefit. Now, supposedly its results are going to be, they're investigating whether it can be available within 10 minutes or so, but it still has a little gadget that has to be able to, as a point of care testing, for us to do. Other advances are occurring as well, and these other advances have gone into the genetic world, genetics and transcriptomics, and there are currently two systems that are available in the U.S. They're not, they just started last year, I believe, they got their FDA approval, and they look into the upregulation of some genes, and they come up with a clinical scoring system that discriminates infection from sepsis, and the research world is where the actual assessment of functional immune assessment has to happen, as at some point, each, whether the lymphocytes are producing something or not producing something. All of these have pitfalls, and we have to be very careful in how we use them, if we use them. To summarize, the world of sepsis is growing tremendously, actually, when you look at the literature, it's becoming difficult and complicated. Every single area is growing, but they are growing outside, and they are exploding outside, but they don't communicate with each other, and it's becoming very difficult for clinicians to have a good integrated view of what should happen. In conclusion, sepsis is challenging. We still are clinicians that have to see patients at the bedside and come up with the right diagnosis. We have to be on the alert that some of the scoring systems that are there have to apply with a clinical judgment. Alerts may be helpful, but they cannot be used just commercially. One has to go to each hospital and figure out what is the land like, what are the resources and how we can put the resources together in order to reach the bedside. Predictive sepsis models are good. Once we add more biomarkers to them, perhaps some biology to them, some science to them, with a field of informatics, we can integrate it all, regardless, it has to be validated before we start applying it. One has to look at the environment that they're in, the population that they are dealing with, and come up with a plan. With that, I end, and thank you very much.

sepsis diagnosis

emergency treatment

antibiotic resistance

machine learning

biomarkers