EVALUATING THE APPLICATION OF MACHINE LEARNING ALGORITHMS IN PREDICTING DISEASE OUTCOMES AND ENHANCING DIAGNOSTIC ACCURACY IN HEALTHCARE SYSTEMS

Abstract

Background: The integration of machine learning (ML) into healthcare systems offers a transformative potential for enhancing disease prediction and diagnostic accuracy. Despite a surge in primary research, a comprehensive synthesis of the current evidence on the efficacy of ML algorithms across diverse clinical settings is lacking.

Objective: This systematic review aims to evaluate the application of machine learning algorithms in predicting disease outcomes and improving diagnostic accuracy within modern healthcare systems.

Methods: A systematic review was conducted following PRISMA guidelines. Electronic databases (PubMed, Scopus, Web of Science, Cochrane Library) were searched for studies published between 2019 and 2024. Included studies evaluated ML models for diagnosis or prognosis in human patients compared to standard care. Two reviewers independently screened studies, extracted data, and assessed risk of bias using appropriate tools (QUADAS-2, ROBINS-I).

Results: From 2,847 initial records, 8 studies were included. The studies encompassed oncology, ophthalmology, and cardiology, utilizing data from medical imaging and electronic health records. ML models, particularly deep learning algorithms, demonstrated high performance, frequently matching or surpassing clinical expert accuracy. Key metrics included area under the receiver operating characteristic curve (AUC-ROC) values often exceeding 0.90. Common limitations included retrospective design and risks to generalizability.

Conclusion: Machine learning algorithms show significant promise in enhancing diagnostic and prognostic precision, potentially supporting clinical decision-making. However, the current evidence is primarily derived from retrospective studies. Future research requires robust prospective validation and standardized reporting to ensure reliability and facilitate successful clinical integration.