Can Early Symptom Reports Improve Risk Prediction in Pediatric Leukemia

Kids with leukemia can experience different side effects to treatments, so this project can baseline and week-one PRO, mixed with basic clinical information, could find higher-risk cases during the following 30 days. I created 3,000 simulated pediatric leukemia records and compared logistic regression, random forest and gradient boosting models. On 600 simulated records that were not used, logistic regression did the best with a ROC-AUC of 0.915 and 92% sensitivity, showing the potential of PRO-based risk prediction but still needing validation with real patient data.