Author(s) :

Prakhar Shukla.

Volume/Issue :

Abstract :

While Clinical Decision Support Systems (CDSS) are widely used in clinics and medical settings, machine learning (ML) based systems are only recently being explored, especially in the context of recommending treatments based on previous data. This literature review addresses this gap by comparing the impact of ML-based CDSS to clinician performance in antibiotic treatment in hopes of improving diagnostic accuracy and decreasing chances of antimicrobial resistance (AMR) development. PubMed and Google Scholar were used to identify 11 studies that were categorized based on their diagnostic accuracy, type of ML model used, and the difference in outcomes from physicians. The evidence shows that ML-based CDSS had an average of 16% increased accuracy compared to clinicians. Even though the studies used a variety of models and training sets, the findings indicate that ML can help clinicians in their treatment selection. However, due to the diversity in data sources, model design, and evaluation methods, generalizing these results is difficult. Overall, ML-based CDSS seems like a promising way to reduce overprescription of unnecessary antibiotics and improve diagnostic accuracy.