The study developed and internally validated a nomogram to predict the risk of medication non-adherence in patients after heart valve surgery. It included 244 patients operated between March 2025 and July 2025, of whom 112 were non-adherent according to the ARMS scale (score >16). Predictors were selected using LASSO regression and multivariable logistic regression, with an event-per-variable ratio of 22.4. The final nomogram includes five factors: warfarin use, accompanying children, frequency of daily dosing, level of education, and distance to hospital. The model achieved a C-index of 0.839 (95% CI: 0.808–0.870) in the training cohort and 0.833 in internal validation with 1000 bootstrap resamples. The AUC was 0.817 (95% CI: 0.784–0.845), calibration plots showed good agreement, and the decision curve confirmed clinical utility at a threshold of 12–68%. The nomogram is used for individualized prediction of the risk of non-adherence.