An algorithm that detects sepsis reduces deaths by nearly 20 percent
Sepsis, a condition that can lead to death within hours and causes one in three hospital inpatient deaths in the United States. The electronic health record research community has developed automated systems that send reminders to doctors to check patients for sepsis. To ensure that these alerts are not ignored,
Sepsis, a condition that can lead to death within hours and causes one in three hospital inpatient deaths in the United States. The electronic health record research community has developed automated systems that send reminders to doctors to check patients for sepsis. To ensure that these alerts are not ignored, machine learning researchers are working to customize programs to reduce the number of disabled notifications.
Now the AI will not make decisions about sepsis on behalf of health care providers. It will check boxes in the patient's electronic medical record so that doctors or nurses checking it will see a record that the patient is at risk for sepsis, as well as a list of reasons for it.
Such a machine-learning model is as important in treating sepsis as an electrocardiogram is in diagnosing a heart attack. It will allow the clinician to go from the computer, which takes 15 years of information into account, to the patient's bedside and reassess their condition more quickly.
