This study presents a comprehensive analysis of patient cohorts admitted to Intensive Care Units (ICUs), with the objective of evaluating, comparing, and optimizing the metrics employed for mortality prediction in critically ill patients. Given the inherent complexity of clinical prognosis in these settings, widely adopted scoring systems such as APACHE, SOFA, and SAPS are critically examined, highlighting their principal strengths and limitations. To enhance predictive performance, the study proposes the integration of advanced Machine Learning and Data Science methodologies. The approach includes the implementation of machine learning algorithms, systematic variable selection, cross-validation, and performance assessment using metrics such as AUC-ROC, accuracy, sensitivity, and specificity. Model development will be based on real ICU patient data, ensuring strict adherence to ethical standards and data confidentiality. The best-performing model will subsequently be benchmarked against traditional tools to evaluate its capacity to improve mortality prediction and support clinical decision-making. Beyond predictive accuracy, the study emphasizes the importance of feasibility and applicability in real-world clinical environments. Overall, this research seeks to contribute to the development of innovative technological solutions that enable more personalized, efficient, and evidence-based medical care in high-complexity hospital units, thereby supporting optimized management of resources, healthcare personnel, and clinical protocols within the ICU.