Outpatient antibiotic stewardship interventions: geographic scale and associations between antibiotic use and resistance

Abstract Background Antibiotic stewardship interventions aim to combat antibiotic resistance by reducing inappropriate antibiotic use. One obstacle to the rational design of outpatient stewardship programs is that small-scale pilot experiments that aim to reduce antibiotic resistance by reducing antibiotic use may produce results that are systematically different from results observed in larger-scale implementations. Here, we investigate the relationship between geographic scale and the effect of reductions in antibiotic use. Methods and findings First, we show that dynamical models of antibiotic resistance exhibit “spillover”, such that resistance in an intervention population is partly due to antibiotic use in surrounding populations, which attenuates the intervention’s effect size. Second, using observational antibiotic use and resistance data from US states and European countries for 3 pathogen-antibiotic combinations, we show that use-resistance associations are robust to aggregation above the level of US states or European countries. Finally, we did not detect differences in the strength of use-resistance associations measured between pairs of adjacent states or countries, which presumably have stronger spillover, compared to the associations among non-adjacent pairs. Conclusions These results imply that interventions at the level of US states will yield effect sizes that can be used to estimate the effects of regional or national interventions.


Introduction
In this study, we aim to determine whether outpatient stewardship experiments at the 116 level of US states or European countries can be expected to provide accurate estimates 117 of the effect that the same reduction in antibiotic use would have if applied over a larger 118 area, indicating that interventions in smaller populations can be used to predict the 119 effect of an intervention in a larger population. First, we show that the spillover effect 120 does occur in mathematical models of antibiotic use and resistance, and we measure 121 how interactions between theoretical populations attenuate use-resistance associations. 122 Second, we look for empirical evidence that spillover has a measurably different effect 123 at scales above US states or European countries. Methods. In the simulations, antibiotic use as monthly treatments per capita and 137 resistance as the proportion of colonized hosts carrying resistant strains. 138 . CC-BY 4.0 International license It is made available under a was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which . http://dx.doi.org/10.1101/536714 doi: bioRxiv preprint first posted online Feb. 1, 2019; was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which . http://dx.doi.org/10.1101/536714 doi: bioRxiv preprint first posted online Feb. 1, 2019; US states within a region will interact more strongly with one another than with states in 162 other regions, for example. 163

164
To measure the effect of population structure on use-resistance associations, we set 165 super = sub = 4 and varied three parameters, setting sub to 0.00, 0.01, 0.10, and 0.50; 166 super to the same values, subject to super ≤ sub ; and setting 7 to a range of values 167 between 0.05 and 0.20 treatments per person per month (Supplemental Table 1 was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which . http://dx.doi.org/10.1101/536714 doi: bioRxiv preprint first posted online Feb. 1, 2019; To test the idea that the same difference in antibiotic use will be associated with smaller 208 differences in antibiotic resistance when the two populations have stronger interactions, 209 we tested whether the use-resistance association is weaker for geographic units (US 210 states or European countries) that are physically adjacent to one another. Two units 211 were considered adjacent if they share a land or river border (Supplemental Methods).

229
. CC-BY 4.0 International license It is made available under a was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which . http://dx.doi.org/10.1101/536714 doi: bioRxiv preprint first posted online Feb. 1, 2019; In simulations of two populations, representing an intervention and control group, 230 interactions between the two groups attenuated the effect of the intervention (Figure 1). 231 With increasing interaction strength, the same difference in antibiotic use between the 232 populations was associated with a smaller difference in antibiotic resistance. We used theoretical models to show that interactions between a control and intervention 267 group can attenuate the reduction in antibiotic resistance expected from an antibiotic 268 stewardship intervention. However, consistent with at least one previous study (40), 269 empirical data did not provide robust evidence that aggregating US states or European 270 countries into regions yielded stronger use-resistance associations. Furthermore, the 271 same difference in antibiotic use between a pair of US states or European countries 272 was associated with similar differences in antibiotic resistance between the units in the 273 pair regardless of whether the units were physically adjacent or not. These results 274 suggest that spillover at the level of US states and European countries is not 275 . CC-BY 4.0 International license It is made available under a was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which . http://dx.doi.org/10.1101/536714 doi: bioRxiv preprint first posted online Feb. 1, 2019; substantially stronger than spillover at regional scales. interact more often with other residents of that state than with residents of other states, 298 . CC-BY 4.0 International license It is made available under a was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. shown that spillover is important for individuals (17-20), and the results of this study 307 suggest that US states and European countries do not have substantially stronger 308 spillover than larger regions, but the importance of spillover at smaller scales remains 309 unclear. Depending on the epidemiology of bacterial transmission and the distribution of 310 antibiotic use within the targeted populations, it may be that cities, daycares, schools, 311 workplaces, or even families represent the optimal trade-off between logistical feasibility 312 and the accuracy of measured effect size for a particular pathogen and antibiotic. was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.  was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

Disclaimers 555
The views and opinions of the authors expressed herein do not necessarily state or 556 reflect those of the ECDC. The accuracy of the authors' statistical analysis and the 557 findings they report are not the responsibility of ECDC. ECDC is not responsible for 558 conclusions or opinions drawn from the data provided. ECDC is not responsible for the 559 correctness of the data and for data management, data merging and data collation after 560 was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.  was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.  was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.  CC-BY 4.0 International license It is made available under a was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which . http://dx.doi.org/10.1101/536714 doi: bioRxiv preprint first posted online Feb. 1, 2019; . CC-BY 4.0 International license It is made available under a was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which . http://dx.doi.org/10.1101/536714 doi: bioRxiv preprint first posted online Feb. 1, 2019; . CC-BY 4.0 International license It is made available under a was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.