Kid CBC Checker: Adaptive boosting decision tree-based prototype digital health platform for the classification of red blood cell disorders in children

This is the authors' abstract. We don't add key points for this paper.

Red blood cell (RBC) disorders indicate a significant global health burden among pediatric populations. Despite their prevalence, early and accurate diagnosis of pediatric RBC disorders remains a challenge. This is largely due to the complexity of hematological interpretation and limited access to specialized diagnostic expertise in resource-constrained settings. To address this gap, we developed a prototype digital health platform based on a machine learning model for the automated classification of RBC disorders in children using age and complete blood count (CBC) data. This retrospective study analyzed 4183 pediatric cases stratified into normal and abnormal RBC profiles. Five machine learning algorithms were systematically compared: binomial logistic regression, Gaussian naïve Bayes, decision tree, random forest, and adaptive boosting decision tree. All models were trained in age and routine CBC. The adaptive boosting decision tree demonstrated excellent classifier performance with an overall accuracy of 98.81%. With the abnormal class treated as the positive class, the model yielded a sensitivity of 97.85% and a precision of 99.66% for identifying abnormal cases, together with a specificity of 99.69% and a negative predictive value of 98.04% for identifying normal cases. The analysis of feature importance identified age and mean corpuscular volume as the most influential predictors, followed by hemoglobin and hematocrit. The best-performing model was subsequently integrated into the publicly accessible Kid CBC Checker ( https://kidcbc.work/medservice/public/index.php ), a prototype digital health platform powered by the adaptive boosting decision tree algorithm. This platform offers a reliable digital solution for the screening of pediatric RBC disorders to support evidence-based clinical decision-making for general physicians.