DIFFERENTIATING between the diagnoses of Kawasaki disease and multisystem inflammatory syndrome in children (MIS-C) could be achieved using artificial intelligence (AI), according to new research from the University of California (UC) San Diego School of Medicine, USA.
The team, led by Jane Burns, Rady Children’s Hospital-San Diego, California, USA, and Director of the Kawasaki Disease Research Center at UC San Diego School of Medicine, developed an AI model called KIDMATCH, in order to help support clinicians in differentiating between MIS-C and Kawasaki disease using test results and five physical examination features.
To achieve this a two-stage process was performed. Initially to internally validate the model, the authors enrolled a total of 1,538 children with MIS–C from three different hospitals between 7th May 2020 and 20th July 2021 or a diagnosis of Kawasaki disease between 1st January 2009 and 31st December 2019 for the internal validation process. The KIDMATCH AI model was trained to differentiate MIS-C from other paediatric febrile illnesses. Following this, the model was then trained to differentiate between Kawasaki disease and other paediatric febrile conditions.
During internal validation, the training resulted in a median area under the receiver operating characteristic curve of 98.8% and 96.0% in stages one and two, respectively. Regarding external validation, the model was able to successfully identify MIS-C in 94% of cases.
This success is promising and highlights how AI models could potentially be used to aid clinicians in differentiating between Kawasaki disease and MIS-C, as well as other paediatric febrile illnesses, which can be challenging. Burns highlighted the significance of this research, stating that “for 40 years, the Kawasaki research community has tried to create a diagnostic test for [Kawasaki disease] and failed.” Implementing AI models like KIDMATCH could enable earlier detection of diseases, resulting in earlier intervention, preventing complications, and ultimately improving patient outcomes. This highlights how digital health solutions could be used in disease-specific pathways to improve healthcare.