In partnership with the UCLA Center for SMART Health, Hearst Health is honored to amplify the achievements of Mount Sinai Health System by naming them the winner of the 2024 #HearstHealthPrize! Learn how the team at Mount Sinai Health System has implemented #machinelearning to improve staff efficiency and health outcomes across six hospitals.
Transcript
Being malnourished greatly affects your recovery from a health episode. If you're malnourished, it affects your whole entire hospital course and your recovery afterwards, whether you're in for surgery or whether you're in for just a medical admission. So it's an important patient problem. We underdiagnosed malnutrition terribly in hospitals. There has been for many decades of focus on nutrition support in hospitals, but figuring out who those patients are is not as simple as it might seem. Our hospital, our health system was below our peers in terms of diagnosing and treating malnutrition and the traditional screening tools that our dietitians use to visit a patient weren't really finding enough patients. The dietitians have a problem, they're not seeing the patients they need to see are diagnosis of malnutrition is kind of scattershot and that's harming our patients. We came up with the idea that we started clinical data science team, sort of a modern data stack to Mount Sinai. We developed neutral scan AI. And ability to detect patients with malnutrition and send our dietitians to confirm a diagnosis. What the use of AI in these deep learning models let us do is take a huge historical cohort of data and build a real model that used AD or 90 variables and looked at time series data to build a model that really performed much, much better than the classic just simple rule based model. We don't want AI to make a diagnosis. We want AI to identify patients in whom a diagnosis is very likely to be present. So that we can get the right team to the right patient at the right time. Of the patients that were predicted upon that had a positive prediction, how many were then successfully seen by the nutritionist out within the first two days of being in the hospital? For those that weren't seen, what was the lag in seeing them? But within a short period of being deployed, we were getting the nutritionist to see Haiti or 90% of the patients who were predicted positive. And then of that, maybe 60% were actually ultimately diagnosed. So that's a very high what's called positive predictive value, meaning that our algorithm is really good at it. Unifying the patients who really are malnourished and that really improves the efficiency of the dietitian of the team. And once we had that performance tuned at the main hospital, then we could start to deploy at the other hospitals. It worked at Mount Sinai W, Mount Sinai Morningside, Mount Sinai Brooklyn, Mount Sinai Beth Israel. We are approximately 2 1/2 to three times more likely now to diagnose malnutrition than before we started. It actually results in more reimbursement for the hospital and it also affects the way we are. Rated by other organizations on observed to expected outcomes because it appropriately indicates that the patient is at higher risk. So that really helps when you're running a hospital, which is nonprofit and a very slim margins. And it's it's it's expensive to take care of sick patients. We have a limited resource. There are only so many dietitians. And so getting them to the correct patients and making the diagnosis and letting them do what brings them joy in their clinical work, which is treating malnutrition and helping patients. Is a positive all the way around. I just love the fact that we can take cutting edge technology and apply it to something that really improves a patients life. We're helping our staff get to the right patient at the right time and that is what makes for better medical care.To view or add a comment, sign in
Jefferson Senior Perfusion student.
1moLove seeing MSH performing such innovative work in nutritional science.