ABSTRACT Mach-LETSGO leverages machine learning to improve the accuracy of chronic health condition classification in the Childhood Cancer Survivor Study (CCSS; U24 CA055727; PI: Armstrong), addressing limitations of its self- reported data. By leveraging reference-standard clinical assessments performed in the St. Jude Lifetime Cohort (SJLIFE; U01 CA195547; CA301480; MPI: Hudson/Armstrong) among 2,436 survivors who have participated in both CCSS and SJLIFE, in addition to germline whole genome sequencing, and detailed childhood cancer treatment data, this proposal aims to refine chronic health condition classification for chronic health conditions such as diabetes, hypertension, and cardiomyopathy. Machine learning methods will identify patterns in misclassification, leveraging predictors such as treatment exposures, genetic risk scores, demographic factors, and complex dependencies among survey responses. With training data from 2,000 survivors participating in both CCSS and SJLIFE, along with 25,735 CCSS participants, the study will develop robust predictive models of CCSS participants’ chronic health condition classifications, evaluated in the training dataset through advanced cross-validation techniques along with regularization, ensemble methods, and interpretability tools (SHAP, LIME) to ensure avoidance of overfitting, followed by an independent validation in the remaining 436 survivors participating in both CCSS and SJLIFE. This transformative approach will enhance the accuracy of chronic health condition outcomes in the 25,735 CCSS survivors, strengthen epidemiological analyses, and ensure the continued global impact of CCSS, the largest resource for survivorship research. Findings from this pilot will provide methodological insights to inform future CCSS analyses.