Duke leverages AI to identify patients with early stage sepsis
The early warning system uses a machine learning model, a custom dashboard to present risk scores, and a rapid response team to monitor patients at-risk of sepsis and deliver appropriate treatment.
Time is of the essence when it comes to detecting sepsis—the deadly condition has no clear time of onset and no clear biomarker, making it difficult to diagnose.
To address the problem, the Duke Institute for Health Innovation has developed an artificial intelligence system to help clinicians identify patients in the early stages of sepsis so that they can intervene before it’s too late.
The early warning system leverages a machine learning model, a custom dashboard to present risk scores, and a rapid response team to monitor patients at-risk of sepsis and deliver appropriate treatment.
The AI system, called Sepsis Watch, was initially implemented in November 2018 at Duke University Hospital’s emergency department—this past May, the pilot phase was completed.
Will Ratliff, innovation program manager at the Duke Institute for Health Innovation, reported the pilot’s preliminary bundle compliance results for patients with sepsis at last month’s Machine Learning for Health Care conference in Ann Arbor, Mich.
The pilot’s primary outcome measures were the rate of Centers for Medicare and Medicaid Services bundle completion for patients with sepsis (within 96 hours of emergency department arrival), and the proportion of patients with sepsis that complete CMS treatment bundle.
“In aggregate, bundle compliance from 2016 through (the third quarter) of 2018 on average was 28 percent, and between (the fourth quarter) of 2018 and (the first quarter) of 2019, we saw in aggregate 63 percent,” Ratliff told the Machine Learning for Health Care conference, adding that these are “pretty exciting results so far.”
With the success of the pilot, Duke University Hospital is working to optimize Sepsis Watch and to extend the system to “other care settings” within the hospital, according to Mark Sendak, MD, population health and data science lead at the Duke Institute for Health Innovation.
“We don’t want to just build Sepsis Watch at one university—we want to scale this model beyond the walls of Duke,” adds Michael Gao, senior data scientist at the Duke Institute for Health Innovation.
“We’re bringing the sepsis model from our main hospital—Duke University Hospital—to Duke Regional Hospital and Duke Raleigh Hospital as well,” Gao notes. “And we’re currently exploring partnerships with places like NYU (Langone) and a few other sites.”
To address the problem, the Duke Institute for Health Innovation has developed an artificial intelligence system to help clinicians identify patients in the early stages of sepsis so that they can intervene before it’s too late.
The early warning system leverages a machine learning model, a custom dashboard to present risk scores, and a rapid response team to monitor patients at-risk of sepsis and deliver appropriate treatment.
The AI system, called Sepsis Watch, was initially implemented in November 2018 at Duke University Hospital’s emergency department—this past May, the pilot phase was completed.
Will Ratliff, innovation program manager at the Duke Institute for Health Innovation, reported the pilot’s preliminary bundle compliance results for patients with sepsis at last month’s Machine Learning for Health Care conference in Ann Arbor, Mich.
The pilot’s primary outcome measures were the rate of Centers for Medicare and Medicaid Services bundle completion for patients with sepsis (within 96 hours of emergency department arrival), and the proportion of patients with sepsis that complete CMS treatment bundle.
“In aggregate, bundle compliance from 2016 through (the third quarter) of 2018 on average was 28 percent, and between (the fourth quarter) of 2018 and (the first quarter) of 2019, we saw in aggregate 63 percent,” Ratliff told the Machine Learning for Health Care conference, adding that these are “pretty exciting results so far.”
With the success of the pilot, Duke University Hospital is working to optimize Sepsis Watch and to extend the system to “other care settings” within the hospital, according to Mark Sendak, MD, population health and data science lead at the Duke Institute for Health Innovation.
“We don’t want to just build Sepsis Watch at one university—we want to scale this model beyond the walls of Duke,” adds Michael Gao, senior data scientist at the Duke Institute for Health Innovation.
“We’re bringing the sepsis model from our main hospital—Duke University Hospital—to Duke Regional Hospital and Duke Raleigh Hospital as well,” Gao notes. “And we’re currently exploring partnerships with places like NYU (Langone) and a few other sites.”
More for you
Loading data for hdm_tax_topic #better-outcomes...