Online Program

Will Data Save Us? Predicting the Future of Antibiotic Resistance (309578)

Keywords: Antibiotics, Antibiotic Resistance, Infection, Public Health, Hospital-associated Infections, Geographic Patterns of Antibiotic Use, Machine Learning, Prescribing Behavior

The problem of antibiotic resistance is emerging as a major public health threat in the US. While we expect a clear relationship between antibiotic use and antibiotic resistance, demonstrating this link remains notoriously difficult. Using data garnered from multiple sources, including the CDC’s Antibiotic Resistance Portal and the CMS Medicare Prescriber Data, we investigate the link between usage and resistance across geographic regions in the U.S. Employing a combination of data mining and analytic techniques (multiple regression, machine learning), we determine the relative importance of age structure, provider density, patient age, disease state and pathogen on the prevalence of antibiotic resistance. Our results suggest that frontline antibiotics are vastly overprescribed in relation to older, broad-range antibiotics, although this pattern varies from region to region. We are constructing a model to visualize the relationship between use and resistance that incorporates the relevant confounding variables. This model will be useful in predicting the rise of antibiotic resistance in the U.S., and should inform potential interventions that will address this important threat.