(a) Predicted percent change in incidence (points) with the 95% confidence interval (bars) of all 3GCR and CR infections for every doubling of antibiotic use, connectivity (estimated no. of patients infected with the same variant entering the ward), ward size (no. of beds); and ward type, based on quasi-Poisson regression models of the pooled incidence of 3GCR and CR infections in each ward (n = 357) that included the incidence control covariate (see Materials and methods). (b) Predicted incidence and 95% confidence bands of infections with 3GCR and CR pathogen variants depending on the consumption of CTX/CRO, IPM/MEM, and TZP, in models that included connectivity, the incidence control, and the consumption of 8 other antibiotic groups as covariates. (c) Consumption patterns of CTX/CRO, IPM/MEM, and TZP per ward in the hospital network. Variants classified as 3GCR were 3GCREC, 3GCRKP, CRKP, 3GCREB, CREB, PA, CRPA, AB, CRAB, EF, VREF, and MRSA; the CR category included CREC, CRKP, CREB, CRPA, CRAB, EF, VREF, and MRSA. Acronyms of pathogen variants and antibiotics are listed in Tables 1 and 2, respectively.