Predicting Increased Incidence of Common Antibiotic-Resistant and Antibiotic-Associated Pathogens Using Ensemble Species Distribution Modeling

Abstract

The Centers for Disease Control estimates antibiotic-associated pathogens result in 2.8 million infections and 38,000 deaths annually in the United States. This study applies species distribution modeling to elucidate the impact of environmental determinants of human infectious disease in an era of rapid global change. We modeled methicillin-resistant Staphylococcus aureus and Clostridioides difficile using 31 publicly accessible bioclimatic, healthcare, and sociodemographic variables. Ensemble models were created from 8 unique statistical and machine learning algorithms. Using International Classification of Diseases, 10th Edition codes, we identified 305,528 diagnoses of methicillin-resistant S.aureus and 302,001 diagnoses of C.difficile presence. Three environmental factors - average maximum temperature, specific humidity, and agricultural land density - emerged as major predictors of increased methicillin-resistant S.aureus and C.difficile presence; variables representing healthcare availability were less important. Species distribution modeling may be a powerful tool for identifying areas at increased risk for disease presence and have important implications for disease surveillance systems.

Keywords: Clostridioides difficile; Disease Surveillance; Ecological Niche Modeling; methicillin-resistant Staphylococcus aureus.