Machine learning has been introduced and continuously developed, enhancing its decision-making capabilities as it has found its way into the medical field, where it is being applied to gain further information about COVID-19. The study analyzed the use of machine learning in interpreting data collected from studies related to COVID-19 patients, focusing on identifying the most effective machine learning model and the critical factors that exist along the process. The researchers conducted a thorough literature search to gather and retrieve potentially relevant materials. The gathered literature was carefully screened and was selected. Most studies showed that random forest performed best among the other algorithms, with a prediction accuracy range of 85–100%. Extreme gradient boosting, or XGBoosting, is the second-best performer in predicting patient outcomes, with accuracy percentages ranging from 92% to 97%. The least machine learning algorithms that were used were the support vector machine (SVM), Naive Bayes and neural network, bagging algorithms, and hybrid CNN-LSTM, and logistic regression has 81.1%, 85.71%, 83.55%, 96.34%, and 81.46% accuracy, respectively. Supervised machine learning has helped predict COVID-19 patients’ outcomes, particularly disease severity and mortality rate. Supervised machine learning on COVID-19 data utilized various factors, including demographic data, clinical parameters, symptoms, comorbidities and underlying conditions, and imaging studies. The collective use of these factors enabled the development of predictive models that could aid in understanding disease progression, predicting patient outcomes, predicting mortality rate, and, most importantly, guiding health personnel to make an efficient and smart decision in managing COVID-19 patients.