Clinical Decision Support Systems (CDSS) based on machine learning offer a promising approach to improve decision-making in nephrology by enabling predictive insights from complex clinical data. In this work, we present a modular, web-based CDSS that automates the machine learning pipeline, including data ingestion, preprocessing, model training, feature selection, and evaluation. The system supports both categorical and continuous outcomes and dynamically adapts the workflow based on the input data structure and selected target variable. The platform was applied to a real-world dataset of 854 hemodialysis patients from Uruguay, with the objective of predicting treatment discontinuation outcomes—specifically distinguishing between patients who continued treatment and those who died. Following class balancing and variable selection, a total of 36 model configurations were evaluated using cross-validation. Performance was assessed using metrics like accuracy, F1 score, Area Under the Curve (AUC), confusion matrices, and ROC analysis. Models such as logistic regression and random forests demonstrated robust classification performance, with several configurations achieving AUC values above 0.72. Consistently selected predictive features included vascular access type, pre- and post-dialysis body weight, and specific clinical markers. The system also enabled interactive prediction on new input instances, ensuring reproducibility and transparency in the decision process. The results demonstrate the applicability of the proposed CDSS in nephrology, supporting early identification of high-risk patients and contributing to improved care planning in dialysis programs.