Discussion

In this nationwide spatiotemporal analysis using data from the VHA, we identified increasing prevalence of community-acquired phenotypic ESBL E. coli as well as heterogenous spatial distribution of resistant isolates. The slowly increasing rates of phenotypic ESBL E. coli found in our study are consistent with recent data from the United States. ^{Reference Kaye, Gupta and Mulgirigama8}

We identified 11 clusters of increased community-acquired ESBL E. coli prevalence across the county. We theorize that the overall higher antibiotic usage in large population centers is responsible for the clusters of ESBL E. coli observed in the highly populated areas of southern California, the northeastern United States, and the Gulf Coast states. However, several clusters of resistance occurred in regions without major urban centers. This raises the question of which community-level factors outside human antibiotic consumption are contributing to antibiotic resistance, for example, antibiotic use in livestock. ESBL rates in European Union livestock have been previously reported as 1.6% (poultry), 2.3% (hogs), and 8.5% (cattle). ^{Reference Seiffert, Hilty, Perreten and Endimiani9} These rates have been attributed to the liberal use of third- and fourth-generation cephalosporins in animals.

The clustering of ESBL E. coli infections in rural areas in the United States suggests that antibiotic usage in livestock might be contributing to human ESBL infections. For example, the high-risk cluster in southwest Minnesota–northwest Iowa is the largest hog-farming region in the country, and hogs are also substantially present in eastern North Carolina, which exhibited ESBL resistance prevalence of >20% during several years of our study. Our study findings highlight the need to further explore the connection between humans, animals, and their shared environment, an approach consistent with the ‘One Health’ movement. ¹⁰

A key strength of this study is the large, national data set of VHA patients spanning a 10-year period, which led to a robust data set at the county level which is otherwise challenging to obtain in the United States. Secondly, inclusion of all E. coli isolates increased our ability to capture the spectrum of E. coli disease activity.

This study had several limitations. Large areas in the central United States had no data points due to the small patient population in those areas. This factor may have masked relevant spatial relationships. Cluster mapping is an exploratory method of data analysis and is not designed to make causal inferences. Also, genetic ESBL testing would have improved our understanding of antibiotic resistance patterns, but these data were not available for our study. Finally, we looked exclusively at outpatient ESBL isolates, which may not have reflected inpatient ESBL infection trends.

These findings highlight the increasing prevalence of community-acquired phenotypic ESBL E. coli over time as well as the spatiotemporal distribution of this organism. Cluster analyses showed areas of significant clustering in both urban and rural locations, suggesting that further research is needed to elucidate the population and environmental drivers of resistance, such as time-series analyses of temperature, climate, and population trends.