In high-income countries, about one in 100 children require intensive care unit (ICU) support due to life-threatening illness, trauma or surgery. In the United States alone, over 500,000 neonates and children are admitted to ICUs every year. Although survival of critically ill children has continuously improved, with current mortality rates as low as 2.18%, one in three ICU survivors suffers from physical, cognitive, mental or psychomotor impairment after discharge.
Today, there is still a fundamental knowledge gap in accurate and timely prediction of outcomes, and a lack of effective early warning systems that can flag deteriorating pediatric patients. Most artificial intelligence (AI)-based studies in pediatric patients have used retrospective datasets for static prediction models and have not led to the implementation of assisted decision-making systems for the sickest patients. Consequently, the traditional approaches still being used to recognize and treat deteriorating patients do not adequately address the complexity, acuity and dynamics of ICU patients.
Due to the unique physiology, epidemiology and host response of neonates and children, it is essential to accelerate our understanding of the factors and learn to promptly identify previously unknown patterns that may lead to the rapid deterioration of children’s health. To improve the prevention and treatment of life-threatening events such as shock, sepsis, respiratory failure and cardiac arrest in the pediatric ICU, we urgently need embedded decision systems that integrate multimodal information using AI in a dynamic way, leading to real-time solutions at the bedside.
Through state-of-the-art integration of the multiple data sources around the patients in the ICU, the project Using Artificial Intelligence to Improve Early Recognition and Treatment of Critically Ill Children (AI in Pediatric ICUs) aims to derive and validate prediction algorithms to identify critically ill neonates and children who are likely to develop a major life-threatening event. The researchers hope to develop novel decision support tools that can enable personalized interventions in the future.
The project is being led by Luregn Schlapbach at the University Children’s Hospital Zurich (Switzerland).
