External Validation of a Machine Learning Model to Diagnose Kawasaki Disease.
Jonathan Y Lam, Chisato Shimizu, Michael A Gardiner...
https://pubmed.ncbi.nlm.nih.gov/40122277/Actively Recruiting
Led by University of California, San Diego · Updated on 2025-12-18
200
Participants Needed
1
Research Sites
N/A
Total Duration
Researchers are evaluating the impact of a machine-learning clinical decision support tool called Kawasaki MATCH on how healthcare providers diagnose and manage patients with Kawasaki Disease KD and other febrile illnesses. The study aims to assess the accuracy of Kawasaki MATCH when used at the point of care and its influence on clinical decisions such as additional testing and hospital admissions. Participants healthcare providers will be randomly assigned to one of two groups one group will use the Kawasaki MATCH decision support tool to help identify KD by entering patient information including age, laboratory results, and clinical features to generate a risk score the other group will follow standard care without the decision support tool. This setup will allow researchers to compare usual care with care guided by the Kawasaki MATCH tool. During the study, providers will record assessments and decisions about patients presenting with fever and possible KD. Researchers will track outcomes including the time to treatment for KD patients, hospital admission rates, additional interventions, consultations, and return visits to the emergency department. The study will also gather feedback from providers on the usefulness of the decision support tool. Participation will involve follow-up over 90 days to monitor treatment timing and outcomes.
CONDITIONS
Kawasaki MATCH Trial
M
Michael A Gardiner, MD
Explore thousands of other clinical trials that might be a better match.
Sign up to get personalized trial recommendations delivered to your inbox.
Already have an account? Log in here
University of California, San Diego
Lead Sponsor
Gordon and Marilyn Macklin Foundation
Collaborating Sponsor
This table compares participation in this trial with other options available for this condition. The information is drawn from published research and the trial’s study details and is provided for educational purposes only.
| Factors | Kawasaki MATCH tool | Routine care | AHA standard diagnostic pathway |
|---|---|---|---|
| What it is | A machine‑learning decision support tool that generates a risk score using patient age, 18 lab features, and 5 clinical signs to aid in Kawasaki Disease diagnosis | Standard diagnostic evaluation without using the MATCH algorithm | Clinical diagnosis based on AHA criteria using fever plus principal clinical features, laboratory findings, and echocardiography when indicated |
| Role in trial | Used to generate a risk score to guide diagnosis for randomized patients | Assigned diagnostic pathway without algorithm influence | Represents current standard clinical diagnostic pathway for Kawasaki Disease in practice |
| Primary use | Aid clinicians in diagnosing Kawasaki Disease during emergency or clinical evaluation | Provide standard clinician‑led diagnostic evaluation based on signs and routine tests | Guide diagnosis through established AHA clinical criteria with optional lab and echo support |
| How it is performed | Clinician inputs patient data into the MATCH tool and receives a risk score output | Clinician evaluates patients using history, examination, and standard testing without algorithm support | Diagnosis based on clinical signs plus supportive labs and echocardiography per guideline algorithm |
| Evidence status | Evidence comes from this trial; external validation exists supporting algorithm accuracy in prior studies | Represents usual diagnostic evaluation; standard clinical performance varies by clinician experience | Established diagnostic criteria endorsed by AHA with demonstrated clinical utility |
| Important risks | Algorithmic misclassification risk if input data are incomplete or inaccurate | Potential delayed diagnosis if clinical signs are subtle or overlooked by provider | May miss incomplete KD cases that lack full criteria without lab or echo support |
| Important limitations | Performance depends on input quality; requires clinician entry of multiple data points | Relies heavily on clinician judgment and may vary between providers | Requires multiple clinical signs; incomplete KD or young infants may not meet criteria early |
Disclaimer: This comparison is created with the help of artificial intelligence(AI) and may contain errors or missing information. It is not medical advice and does not mean one option is better or safer than another. Clinical trials are still being studied, and joining a trial is your choice. Please consult your healthcare professional before making any decisions.
You may qualify if you...
You will not qualify if you...
Duration - 2 to 4 weeks
Participants are screened for eligibility to participate in the trial.
1 visit (in-person)
Duration - Day 1 (day of enrollment)
Participants undergo laboratory evaluation and clinical assessment to identify features of Kawasaki Disease and to determine risk using the Kawasaki MATCH decision support tool or routine care.
1 visit (in-person)
Duration - Up to 90 days
Participants are monitored for clinical outcomes including treatment timing, hospital admission, consultation rates, and emergency department return visits.
Follow-up within 7 days and other assessments up to 90 days
Study Type
INTERVENTIONAL
Masking
NONE
Allocation
RANDOMIZED
Model
PARALLEL
Primary Purpose
DIAGNOSTIC
Number of Arms
2
Have more questions? Get in touch with our team for quick support
Jonathan Y Lam, Chisato Shimizu, Michael A Gardiner...
https://pubmed.ncbi.nlm.nih.gov/40122277/Jonathan Y Lam, Chisato Shimizu, Adriana H Tremoulet...
https://pubmed.ncbi.nlm.nih.gov/36150781/University of California, San Diego
Lead Sponsor
Gordon and Marilyn Macklin Foundation
Collaborating Sponsor