New AI System Developed That Can Quickly Pinpoint Sepsis Occurrence

It is normal for the body’s immune system to respond to the invasion of foreign bodies via inflammation, such as during an infection. However, during sepsis, this response to infections goes overboard leading to an increase in the inflammatory processes in the body. This is accompanied by fever, an increase in heartbeat, and difficulty breathing, among other symptoms.

Sepsis

Sepsis is a serious condition that needs immediate medical attention. It can lead to septic shock which is associated with hypotension, organ failure, and possibly death. Sepsis (also known as blood poisoning) is often caused by bacterial infections; viral and fungal infections may also be possible causes. Sepsis is life-threatening because it is seldom diagnosed at the early stage in most cases. Every year, approximately 1.7 million adults in the United States develop sepsis and more than 250,000 deaths are recorded.

The fact that sepsis is not diagnosed early is a huge contributor to the high number of deaths due to this condition. Research has been ongoing to discover a way to detect sepsis at the early stage, and thanks to the researcher at John Hopkins, Suchi Saria, a new AI system that can detect the symptoms of sepsis at the early stage has been created.

The AI system

Table of Contents

1 The AI system
2 Clinical significance
3 Conclusion
- - 3.0.1 References

Saria and the team of other researchers at John Hopkins hospital built the Targeted Real-Time Early Warning System – the AI system that can detect the symptoms of sepsis early in patients by combining their medical history with existing symptoms, plus their lab results and revealing to doctors if there is a possible risk for sepsis. This machine-learning system also suggests treatment procedures, which may be administering antibiotics.

They designed the system to be able to track patients right from their arrival at the hospital through to when they are being discharged. This allows for complete monitoring of patients (even when they are moved from one department to another) to ensure that no vital information is left out while treatment is ongoing. This study was very much personal to Saria because she lost her young nephew to sepsis.

To test the system, over 4000 clinicians treated 590,000 patients with it. The system detected both sepsis and previous cases in patients: it reviewed 173,931 previous cases; in 82% of the sepsis case, the system was correct almost 40% of the time. On the contrary, former attempts to detect sepsis using electronic tools revealed less than half that number and were correct only 2% – 5% of the time. The new AI system is so powerful that in severe sepsis cases where a delay of one hour becomes the difference between life and death, it was able to detect it six hours earlier as opposed to traditional methods.

The team partnered with: Bayesian Health, a subsidiary of John Hopkins, which led and managed the implementation of the system across all sites where it was been tested; Epic and Cerner, the two largest electronic health record system providers, to make sure that the system can be used in other hospitals.

Clinical significance

Sepsis develops very quickly but is not detected by doctors as quickly as it develops leading to increased mortality. This innovation by Saria and her team will go a long way to stop this, as the new AI system can detect sepsis early enough to save the patient.

Conclusion

Since sepsis is a condition that requires immediate medical attention, the development of the AI system will make this possible as the system can detect sepsis early enough and even suggest possible treatment procedures for doctors. This would ensure that more lives are saved, and the number of deaths recorded each year is greatly reduced.