A recent study revealed that a newly trained artificial intelligence (AI) may assist identify children at risk of developing sepsis within 48 hours.
According to research published in JAMA Pediatrics, AI can identify children who are at risk of sepsis.
The AI was trained on more than 1.6 million medical records, the AI flagged children are expected to face sepsis—a fatal response to infection causing organ dysfunction.
Children’s Hospital of Chicago lead researcher Elizabeth Alpern of Ann & Robert H. Lurie stated that the discovery marks a major step forward towards precision medicine for pediatric sepsis.
The AI was able to detect high-risk children in emergency departments (ED) without overpredicting, assisting avoid unnecessary treatment.
Each year, more than 75,000 children across the US got sickened for sepsis, with mortality rates reaching up to 20%.
Current tools can predict sepsis outcomes but not who will develop it. To acknowledge this issue, scientists trained the AI on data from children aged 2 months to 18 years who visited EDs at five major US health systems between 2016 and 2020—excluding those already diagnosed with sepsis.
Researchers stress combining AI with clinical judgment for best results.
The AI assessed patients using vital signs, pre-existing conditions, and data. It was later confirmed that using 720,000 ER visits from 2021–2022 and successfully identified those at risk.
However, it showed a reduced positive predictive value, underscoring the challenge of predicting rare outcomes.
