Purpose: The purpose is to examine predictors of intervention non-compliance and develop a risk stratification score. Design: Prospective cohort. Setting: Early care and education (ECE). Subjects: Early care and education programs (n = 3883) randomly allocated (3:1) to a development (n = 2909) or validation (n = 974) sample. Intervention: Go NAPSACC provides a structured, web-based process to help improve the health of children around 7 modules (nutrition, physical activity, oral health, breast/infant feeding, farm to ECE, outdoor play, and screen time). Measures: Program characteristics and participation data are collected via Go NAPSACC tool. Analysis: Multivariable Lasso logistic regression was used to identify predictors. Discriminative ability was based on area under the ROC curve (AUC). Results: Overall, ECE program non-compliance (lack of valid pre-/post self-assessment) was 65.5%. Six predictors were retained in the final development model: type of program (P =.002), Child and Adult Care Food Program (CACFP) participation (P =.065), acceptance of subsidies (P <.001), past modules attempted (P <.001), past modules completed (P <.001), and action plans created (P <.001). These factors generated a non-compliance risk score which showed good discrimination in the validation sample (AUC:.922, 95% CI:.903–.940). Conclusion: Lack of qualitative data limits the ability to fully understand the context of non-compliance; however, this study demonstrates readily available data captured by Go NAPSACC are strong predictors of future success. Early identification of high-risk programs will inform targets for future implementation strategies geared toward improving program success. [ABSTRACT FROM AUTHOR]