In Clinical rheumatology ; h5-index 39.0
INTRODUCTION : In Kawasaki disease (KD), accurate prediction of intravenous immunoglobulin (IVIG) resistance is crucial to reduce a risk for developing coronary artery lesions.
OBJECTIVE : To establish a simple scoring model predicting IVIG resistance in KD patients based on the machine learning model.
METHODS : A retrospective cohort study of 1002 KD patients diagnosed at 12 facilities for 10 years, in which 22.7% were resistant to initial IVIG treatment. We performed machine learning with diverse models using 30 clinical variables at diagnosis in 801 and 201 cases for training and test datasets, respectively. SHAP was applied to identify the variables that influenced the prediction model. A scoring model was designed using the influential clinical variables based on the Shapley additive explanation results.
RESULTS : Light gradient boosting machine model accurately predicted IVIG resistance (area under the receiver operating characteristic curve (AUC), 0.78; sensitivity, 0.50; specificity, 0.88). Next, using top three influential features (days of illness at initial therapy, serum levels of C-reactive protein, and total cholesterol), we designed a simple scoring system. In spite of its simplicity, it predicted IVIG resistance (AUC, 0.72; sensitivity, 0.49; specificity, 0.82) as accurately as machine learning models. Moreover, accuracy of our scoring system with three clinical features was almost identical to that of Gunma score with seven clinical features (AUC, 0.73; sensitivity, 0.53; specificity, 0.83), a well-known logistic regression scoring model.
CONCLUSION : A simple scoring system based on the findings in machine learning seems to be a useful tool to accurately predict IVIG resistance in KD patients.
Sunaga Yuto, Watanabe Atsushi, Katsumata Nobuyuki, Toda Takako, Yoshizawa Masashi, Kono Yosuke, Hasebe Yohei, Koizumi Keiichi, Hoshiai Minako, Kawakami Eiryo, Inukai Takeshi
2023-Jan-11
Intravenous immunoglobulin resistance, Kawasaki disease, Machine learning, Shapley additive explanation
