Despite the end of the COVID-19 pandemic, its impact on strained hospitals underscores the enduring relevance of timely prognosis in strengthening healthcare systems. This paper proposes to build a predictive framework for COVID-19 patient outcomes through the integration of machine learning algorithms with clinical information. Data were obtained from 2,291 patients at An Giang General Hospital and used to construct and compare six models. Visualization techniques were also applied to identify patterns in clinical features and assess their prognostic relevance. The results indicate that Random Forest and XGBoost consistently achieved superior performance compared to other algorithms. Specifically, Random Forest obtained 95.61% accuracy, 92.24% precision, and an F1 score of 90.04%, while XGBoost reached 95.09% accuracy, 90.51% precision, and an F1 score of 89.90%. The study highlights the advantages of integrating clinical data visualization with machine learning to improve prognostic accuracy, ultimately enhancing healthcare quality and resource management.