Application of statistical and machine learning methods to predict antimicrobial susceptibility to blood borne microorganisms to support administrative and policy level decision making

Background Antibiotic resistance is an overwhelming issue in humans, animals, and the environment. To combat AMR, a global approach has been recognized and implemented as a main strategy in many countries. Global strategies for strategic action plans have been developed by the WHO, and individual nationals have also made national strategic plans for the respective countries worldwide. Methods In this study, we aimed to apply machine learning and statistical methods to estimate the incidence of antimicrobial resistance in blood-borne infections for selected pathogens with socioeconomic and public health parameters based on computer-based models (Atlas dataset), to determine the correlation between phenotypic resistant profiles and risk factors such as sex and age, and to formulate a model to forecast the antimicrobial resistant prolife in the given blood-borne infection in humans. Results and Discussion Non-metric Multidimensional Scaling analysis identified significant differences between country groups, developed nations, and less developed nations. Notable antimicrobials contributing to resistance disparities were identified, which can be used to guide policy implementations. The study emphasized the importance of collaborative efforts among countries when implementing AMR combatting strategies. Conclusions Statistical analysis showed that sex was not significantly associated with resistance to gram-positive infections, but age was significantly associated with resistance to both Gram-positive and Gram-negative infections. Logistic regression models underscored that age is a critical factor in non-empirical antimicrobial therapy, with resistance increasing with age, particularly in gram-negative infections.