Geisinger and IBM Create Machine-Learning Model to Detect Sepsis

Samara Rosenfeld
SEPTEMBER 12, 2019
tech

Experts from Geisinger and the IBM Data Science Elite team created a predictive model to detect sepsis, the organizations announced this week.
 
The predictive model uses data from the integrated health system’s electronic health record. The model, making use of IBM Watson technology, makes a challenging diagnosis potentially a bit easier, Geisinger said.
 
“For clinicians, making a sepsis diagnosis can be very difficult, as the symptoms overlap with many other common illnesses,” said Donna Wolk, Ph.D., director of the molecular and microbial diagnostics and development division at Geisinger. “If we can identify patients more quickly and more accurately, we can administer the right treatments early and increase the chances of a positive outcome.”
 
A six-person team of scientists developed a model to predict sepsis mortality, and a tool to help the team stay on top of the latest research in the sepsis field.
 
Using IBM Watson Studio tools, the team built a predictive model that ingested nearly 10,600 de-identified files for patients diagnosed with sepsis between 2006 and 2016. Geisinger and IBM broke the data into 199 separate features for each patient — age, infection type, surgery and treatments, medical history and lifestyle.
 
The researchers developed a scalable machine-learning algorithm to predict patient all-cause mortality during the hospitalization period or during the 90 days following a patient’s hospital stay.
 
The machine-learning model identified 1,190 true positive and 2,087 true negatives, according to the researchers.
 
“The results of the predictive model were highly encouraging,” Wolk said. “The tools available in IBM Watson Studio provided the high performance and speed we needed for analysis, and the support of the IBM Data Science Elite team ensured the project ran smoothly.”
 
IBM and Geisinger then used IBM Watson Explorer software to create an index of thousands of medical publications so the researchers could mine the archive to uncover relevant content. The searchable index, which incorporates natural language processing, has accelerated the time taken to locate useful research studies, Wolk said.
 
Geisinger hopes the model can help the health system develop more personalized care plans for at-risk patients. The health system also wants to identify key factors linked to sepsis death to increase a patient’s chance of recovery.
 
“Our experience using machine learning and data science has been very positive, and we see huge potential to continue its use in the medical field,” Wolk said. “Thanks to IBM, we are well on our way to breaking new ground in clinical care for sepsis and achieving more positive outcomes for our patients.”

IBM’s technology has been used by health systems and organizations across the country to improve patient outcomes. Using IBM Watson Cloud, ERGO Aktiv developed a virtual assistant to help stroke patients navigate the rehabilitation process. 

IBM Watson Health also found that artificial intelligence improves medical decision-making for cancer treatment.

What’s more, the U.S. Food and Drug Administration tapped IBM, along with other industry powerhouses, to participate in a project using blockchain to build a digital, interoperable network to monitor the country’s prescription drug supply chain.

Get the best insights inside digital health. Register for our newsletter.

Related
IBM, Aetna, Anthem & Others Announce Collab to Create a Blockchain Network
What IBM Watson Health Hopes to Accomplish Through $50M in Deals with Brigham & Women’s, Vanderbilt
IBM Watson Health Releases Diabetes App, Touts Power of Data

SHARE THIS SHARE THIS
25
Become a contributor