UC San Diego

Kawasaki disease (KD) is an idiopathic febrile disease primarily affecting children younger than five years of age that is characterized by a persistent fever and five clinical signs: rash, bilateral conjunctival erythema, cervical lymphadenopathy, changes in the lips and oral cavity, and changes in the extremities. If left untreated, KD patients may develop coronary artery aneurysms which could lead to myocardial infarction and sudden death. The overall aim of this dissertation was to improve outcomes in children with suspected risk or confirmed diagnosis for KD. I first addressed KD from a screening perspective by developing deep learning algorithms to determine KD risk based on the positive classification of two or more clinical signs and integrating those algorithms within a screening tool for parents of at-risk children to assess whether to seek medical attention. I then introduced a clinical decision support tool called KIDMATCH to distinguish KD from similar febrile illnesses in the emergency department and conducted an implementation study to assess its utility in the clinical workflow at Rady Children’s Hospital. Finally, I investigated the feasibility of whether resistance to the standard treatment for KD, intravenous immunoglobulin (IVIG), can be predicted using clinical, laboratory, and echocardiographic features. I demonstrated that the continued development of IVIG risk scores is ineffective until novel biomarkers or other specialized features are incorporated.