***Descriptive Statistics*** In 2009, across 400 Canadian communities, physicians dispensed an average of log 10(2.35) (95% CI: log 10(2.33) – log 10(2.37)) defined daily doses of azithromycin per 10 000 people. The descriptive statistics for the socio-economic, demographic, and flu-related variables from these 400 communities are included in [Table 1][1] and [Table 2][2]. There were no missing observations, low frequencies, or category percents within these predictor variables. ***Univariable Analysis*** The results of the univariable analysis of the unconditional associations between these predictor variables and azithromycin consumption are presented in [Table 3][3]. As all predictor variables contained a statistically significant association with azithromycin consumption, no variables were dropped following the univariable analysis. The variable measuring the number of health workers per 10 000 had a significant linear relationship with azithromycin consumption. However, the variable for percentage of the community that is low income did not have a significant linear or quadratic relationship with azithromycin consumption, so it was categorized using quartiles, creating four percentage low income categories. Pearson correlation analyses of the predictor variables found no predictor variable correlations above 0.8. ***Multivariable Analysis*** In order to model the association of azithromycin consumption with the community-level predictor variables, we created a multivariable model. When combined in a main effects model, all predictor variables had a statistically significant association with azithromycin consumption, with the exception of the high percentage of adults over 65 (p = 0.672) and high vaccination rate (p = 0.448). The F-test for removing the high percentage of adults over 65 variable was not significant (p = 0.67), and the greatest percent difference in significant variable coefficients following removal of the variable was 11%, below the confounding cut-off of ≥ 20%. The F-test for removing high vaccination rate variable was not significant (p = 0.45), and the greatest percent difference in significant variable coefficients following removal of the variable was 5%, below the confounding cut-off of ≥ 20%. Therefore, the high percentage of adults over 65 variable and the high vaccination rate variable were dropped from the model. A statistically significant interaction between the high percentage of children under 5 variable and the high percentage of low income variable (p < 0.001) within our main effects model, and was included in the model. A contrast was conducted to determine the difference between the multivariate model with and without the interaction between High % Under 5*Low Income(categorized). The resulting difference was 0.401 with a p value of <0.001 (see [Table 5][4]). The results of our final model describing the associations between azithromycin consumption and socio-economic, demographic, and flu-related predictor variables are included in [Table 4][5]. ***Outlier and Residual Analysis*** The final model met the assumptions of linear regression. The residuals were homoscedastic, and their distribution was visually not deviating from normality, despite a significant Shapiro-Wilks test (p = 0.001), which was not unexpected due to the large sample size. Qladder functions illustrated that the current state of the model was preferable to other transformations which increased heteroskedasticity, so the normality assumption was judged to be sufficiently met. Two outliers were identified (community ids 300 and 301), and they were not recording errors. When removed from the model, the direction of the coefficients remained the same as with the outliers present. Additionally, removal of the outliers had a consistently small impact on the magnitude of the coefficients. The greatest change in coefficient magnitude was for the level 4 of the income variable, which changed from -0.1766 to -0.1739 following removal of the outlying communities. The significance of the coefficients did not change. [1]: https://osf.io/3kthw/ "Table 1" [2]: https://osf.io/ewy9p/ "Table 2" [3]: https://osf.io/8kmy5/ "Table 3" [4]: https://osf.io/kzxah/ "Table 5" [5]: https://osf.io/2vtws/ "Table 4"