Saving Hospital Patients’ Lives with Machine Learning

Sepsis or septicaemia is a life-threatening illness.

If you had to guess the most expensive condition treatment in the US, what would come to mind? Most people tend to think of chronic conditions like heart disease and cancer, so you might be surprised to hear that the culprit is actually sepsis, a medical emergency triggered by an extreme immune response to infection.

Many people may not even know what sepsis is, but it causes 20 to 30 percent of all U.S. hospital deaths. It is killing more Americans than AIDS, breast cancer and prostate cancer combined together. The good news is that sepsis is generally treatable, but many people die simply because it hasn’t been detected in time.

Predicting Septic Shock with Machine Learning

Early diagnosis is the key to prevent septic shock and organ failure, which eventually leads to the life-threatening condition. Major advancements in this have recently been made by Dr. Suchi Saria and her team  who built a predictive algorithm based on electronic health records of 16,234 patients admitted to intensive care units at Boston’s Beth Israel Deaconess Medical Center.

Their algorithm was hugely successful, predicting septic shock in 85 percent of cases without making a false prediction, and more than two-thirds of the time to predict septic shock before any organ dysfunction. Now, Dr. Saria’s algorithm is currently being used in a real-time surveillance tool at Howard County General Hospital.

With the increasing adoption rate of electronic health records in hospitals and in outpatient clinics, the utility of this algorithm will only continue to grow. Dr. Saria and her team’s advances in predicting septic shock is just one example of how machine learning algorithms are starting to play an important role in clinics. There still remains a great deal of work to leverage analytics for prevention of early patient deaths, but the outlook is clearly promising.


Users who have LIKED this post:

  • avatar
  • avatar