Predictive Modelling of Patient Outcomes Using Machine Learning: A Comparative Analysis of Algorithmic Approaches in Healthcare

The study addresses the problem of prediction of patient outcomes in healthcare with machine learning methods, including such options like logistic regression, decision trees, random forests and support vector machines. Comparative modelling of these approaches also was done to assess their effectiveness in predicting outcomes like disease worsening, treatment response, and mortality rates among other clinical delineations. The study employed a data collection policy of diverse data set which encompassed the demographic information, clinical measures and treatment methods from an electronic health record system of tertiary care hospital. This reveals a fact that Randon Forests performed better than rest of the algorithms (0.88 for accuracy and 0.93 for AUC-ROC score). On the contrary, logistic regression and SVM also showed good results with an accuracy reaching 0.85 and 0.87 respectively. Decision trees acquire a small performance, and such accuracy is 0.78. It is the results of these studies that provide information about the abilities and the difficulties of different approaches in the historical analysis, guiding the healthcare workers and data scientists in the selection of appropriate methods for improving patient care and the process of medical decision-making.