In the rapidly evolving domain of Healthcare 4.0, sustainable decision-making is essential to achieving high-quality and long-term medical outcomes. As healthcare systems become increasingly complex due to the integration of electronic health records (EHRs), diagnostic precision tools, and resource optimization strategies, machine learning (ML) has emerged as a powerful means of enhancing clinical intelligence. This study presents a novel and comprehensive framework that systematically categorizes and evaluates thirteen widely adopted ML algorithms, offering a taxonomic perspective on their design, interpretability, and diagnostic potential. To address the limitations of conventional feature selection methods, the proposed framework integrates Association Rule (AR) mining with the Apriori algorithm to identify clinically significant features while effectively reducing dimensionality. Proof-of-concept experiments are conducted on a real-world liver disease dataset, demonstrating that the proposed framework improves predictive accuracy, achieving 83% with Decision Tree and Extra Trees classifiers. It also enhances model interpretability and computational efficiency. This approach supports sustainable diagnostic practices by lowering computational costs and focusing on the most informative features. While liver disease serves as the focal case study, the framework is generalizable to a wide range of diagnostic applications in the context of Healthcare 4.0. The study concludes with a strategic roadmap for future research, highlighting the importance of high-quality data, ethical artificial intelligence, and the translational deployment of ML models to fully realize their potential in sustainable healthcare decision-making.