Geisinger, IBM develop predictive model to identify at-risk sepsis patients
Geisinger has teamed with Big Blue to better diagnose and treat patients with sepsis, a potentially life-threatening condition, using machine learning.
Geisinger has teamed with Big Blue to better diagnose and treat patients with sepsis, a potentially life-threatening condition, using machine learning.
By leveraging de-identified electronic health record data for more than 10,500 sepsis patients, Geisinger and IBM scientists developed a model to predict mortality by identifying “descriptive and clinical features,” such as age, prior cancer diagnosis, decreased blood pressure, number of hospital transfers, time spent on vasopressor medicines, as well as the type of pathogen—all critical factors associated with sepsis deaths, according to researchers.
“For clinicians, making a sepsis diagnosis can be very difficult, as the symptoms overlap with many other common illnesses,” says Donna Wolk, division director of molecular and microbial diagnostics and development at Geisinger, a Danville, Penn.-based integrated health system. “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.”
Using the open-source XGBoost library and the Python programming language in IBM’s Watson Studio solution, a scalable machine learning algorithm was created based on gradient-boosted decision trees to analyze Geisinger’s clinical data—with the goal of predicting patient all-cause mortality during the hospitalization period or during the 90 days post-discharge.
The predictive model was able to accurately predict death for patients in the test data who actually did die, as well as those patients who survived and were successfully treated.
Geisinger contends that the predictive model will enable the health system to pay more attention to the key factors linked to sepsis mortality and develop more personalized clinical care plans for at-risk patients, increasing their chances for 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,” adds Vida Abedi, staff scientist in Geisinger’s Department of Molecular and Functional Genomics. “We are well on our way to breaking new ground in clinical care for sepsis and achieving more positive outcomes for our patients.”
Going forward, Geisinger says it will leverage more recent data on sepsis hospitalizations—along with additional information on patients’ socio-economic backgrounds—to increase the granularity of the model for identifying factors that influence mortality rates.
In addition to building the predictive model, a tool was created to keep clinicians and researchers abreast of the latest sepsis research by using IBM Watson Explorer software to create a searchable index of thousands of medical publications.
By leveraging de-identified electronic health record data for more than 10,500 sepsis patients, Geisinger and IBM scientists developed a model to predict mortality by identifying “descriptive and clinical features,” such as age, prior cancer diagnosis, decreased blood pressure, number of hospital transfers, time spent on vasopressor medicines, as well as the type of pathogen—all critical factors associated with sepsis deaths, according to researchers.
“For clinicians, making a sepsis diagnosis can be very difficult, as the symptoms overlap with many other common illnesses,” says Donna Wolk, division director of molecular and microbial diagnostics and development at Geisinger, a Danville, Penn.-based integrated health system. “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.”
Using the open-source XGBoost library and the Python programming language in IBM’s Watson Studio solution, a scalable machine learning algorithm was created based on gradient-boosted decision trees to analyze Geisinger’s clinical data—with the goal of predicting patient all-cause mortality during the hospitalization period or during the 90 days post-discharge.
The predictive model was able to accurately predict death for patients in the test data who actually did die, as well as those patients who survived and were successfully treated.
Geisinger contends that the predictive model will enable the health system to pay more attention to the key factors linked to sepsis mortality and develop more personalized clinical care plans for at-risk patients, increasing their chances for 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,” adds Vida Abedi, staff scientist in Geisinger’s Department of Molecular and Functional Genomics. “We are well on our way to breaking new ground in clinical care for sepsis and achieving more positive outcomes for our patients.”
Going forward, Geisinger says it will leverage more recent data on sepsis hospitalizations—along with additional information on patients’ socio-economic backgrounds—to increase the granularity of the model for identifying factors that influence mortality rates.
In addition to building the predictive model, a tool was created to keep clinicians and researchers abreast of the latest sepsis research by using IBM Watson Explorer software to create a searchable index of thousands of medical publications.
More for you
Loading data for hdm_tax_topic #better-outcomes...