How Artificial Intelligence Can Help Your Sepsis

2.8 seconds. It’s a fascinating amount of time. It is the time it takes you to sneeze. It is the time it takes Americans to chow down on 1,300 Hamburgers. It is the time it takes a Porsche 911 Turbo S to go from 0 miles per hour to 60 miles per hour! Unfortunately, it is also the time it takes for another person to die as a result of sepsis around the world.

Many people have heard of sepsis. Many of you wouldn’t bat an eye at that fact. Yet if you are reading this, most likely an American, you probably have never met anyone that has or ever had sepsis. As a matter of fact, people are extremely surprised that even in a premier First World Country like the United States of America, there are 1.7 million adults that get sepsis every year. Of those 1.7 million, 270,000 never survive the sepsis. That is almost a 16% mortality rate for disease in modern-day America. In contrast, the WHO’s world’s largest killer, Ischaemic Heart Disease, also had a mortality rate of 16%. Let’s bring light to this unrecognized killer and talk more about sepsis.

Sepsis is your body’s extreme immune response to an infection or bodily injury. When you get sepsis, your immune system spirals into an uncontrollable chain reaction that ends up causing organ failure and even death. In fact, Sepsis is the leading cause of death in hospitals; about 1 in every 5 deaths in a hospital setting is caused by sepsis! Bacterial infections cause most Sepsis cases. The immune system detects the infection, and the proteins and chemicals are released into the body to combat it. If these inputs are unchecked, the body might go through a vicious chain reaction that causes blood clots, leaky blood vessels, and the blockage of oxygen and nutrients from critical organs. Some of Sepsis’ symptoms include fever, chills, rapid breathing and heart rate, rash, confusion, and disorientation. This is why Sepsis is so dangerous. The symptoms of Sepsis are so generic that it takes time to diagnose the condition. If the diagnosis is too late, there is little to no chance of surviving. Treatment for sepsis simply consists of trying to contain the symptoms and treat the initial infection with antibiotics before further damage. Therefore, the most critical part of sepsis treatment is an early diagnosis.

This is where Artificial Intelligence comes in. Doctors have always struggled with diagnosing sepsis fast because the symptoms are too generic. Sepsis isn't their first guess unless they know of a recent wound or infection. This delay of diagnosis is what ultimately kills the patient. Therefore, the goal of Sepsis treatment should be to diagnose fast and accurately. This is what a new AI system developed at Johns Hopkins University does. In fact, the AI is so successful that its use resulted in 20% more survivors than just human staff. The AI developed at Johns Hopkins takes an input of current symptoms, patient’s recent medical history, and lab results to output the binary answer or if they have sepsis or not. The AI, called the Targeted Real-Time Early Warning System, also provides suggested protocols. In other words, it analyzes the input and suggests what steps to take next. An example protocol output would be to start a course of antibiotics if the AI suspected sepsis.

To train the AI, more than 4000 clinicians from five hospitals used the AI in treating 590,000 patients. The system also reviewed 173,931 previous patient cases. The results showed an astounding success. In 82% of sepsis cases, the AI was accurate for nearly 40% of the time. This is a massive increase from the 2% to 3% success rates that other electronic tools currently have. Adding more hope is the fact that the AI eventually guessed all cases correctly. As you know, however, the 40% was the percentage of cases where the diagnosis was early enough to save the patient’s life. The other 60% was the percentage where the AI was too late, resulting in prolonged recovery or death.

This new technology is a monumental breakthrough. When the algorithm of this AI is improved, which is bound to happen due to the continual improvement of AI development and technology in the world due to innovations, it will serve as a simple way to save millions of lives worldwide. While 40% might seem too low for comfort, which it is for mass use, it is only the success rate for the first trial. Now that this technology has gained recognition, it is inevitable that we master the algorithm to such a point where we can diagnose 90%+ of sepsis cases accurately and early enough. Now, we must wait as the AI encounters more data points and slowly adapts its algorithm to finally solve the curse of sepsis that humanity has faced since our dawn!


Featured Posts