Evidence-Based Medicine. GRADE Approach. Systematic Reviews as Topic. Machine Learning. Deep Learning.

Medical practice involves continuous decision-making processes. To enhance the precision of these actions, a conscious utilization of literature is imperative. However, the exponential growth of published scientific articles surpasses human capacity for comprehensive interpretation, necessitating innovative approaches. This study aimed to investigate the utilization of Artificial Intelligence (AI) in the analysis of medical literature and to design and evaluate a system for the semi-automated execution of the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach, for the categorization of systematic reviews by levels of evidence. Primarily, a scoping review was conducted to establish the state-of-the-art in the use of AI for medical literature analysis. Subsequently, an AI-based functional system was developed to identify the domains imprecision, risk of bias, inconsistency and methodological quality of the review and subsequent classification by levels of evidence in real scenarios. The system's performance evaluation was executed on a sample of Cochrane systematic reviews. As outcomes, this endeavor yielded a published scoping review, a digital product, and an ongoing secondary article. The system achieved an overall accuracy of 63.3% and a Cohen’s Kappa of 0.47

(moderate) for GRADE classification. These results exceeded those found in the existing literature, showcasing the potential of the system to semi-automate the GRADE approach as a complementary tool to human evaluators, thus aiming to enhance expediency and mitigate inconsistencies. Subsequent efforts should be

undertaken to extend implementation across diverse scenarios and to tailor dimensions and criteria for the GRADE approach based on the outcomes. The incorporation of cutting-edge AI technologies will be indispensable on this evolving trajectory.