OBJECTIVE The aim of this study was to characterize the evolution of gram-negative antibiotic resistance during a study of empiric antibiotic rotation. METHODS We showed previously that quarterly rotation of a single antibiotic class is inferior to cycling two antibiotics per quarter for empiric treatment of gram-negative rod (GNR) infections, as evidenced by increased incidence of antibiotic-resistant GNR (rGNR) infections. Resistance patterns were examined by quantifying GNRs resistant to one or more of the following drug classes: Aminoglycosides, cephalosporins, carbapenems, fluoroquinolones, or piperacillin-tazobactam. For all rGNR isolates, the mean number of antibiotic classes to which an organism was resistant was calculated per quarter, as was the number of rGNR species. RESULTS Single-antibiotic rotation (SAR) was associated with significant increases in the incidence of piperacillin-tazobactam (p < 0.0005) and cephalosporin (p = 0.003) resistance, reaching nearly 25% and 30% of rGNR isolates respectively, most notably during the quarter of designated cephalosporin use (VI). Multi-drug resistance emerged over time; resistant classes/resistant GNR isolates ranged from 1.2 in the dual-antibiotic rotation (DAR) to 1.9 in the SAR period (p = 0.02). Resistance was evident in an increasing number of unique GNR species. On average, 1.3 species were isolated per month in the DAR period and 3.0/month in the SAR period (p = 0.004), but proportionally, no single GNR species became significantly more resistant across time. Compared to only 5.8% in the DAR period, 29% noncompliance was observed in the SAR, with a six-fold increase in the use of nonscheduled empiric antibiotics due to the presence of an organism resistant to the scheduled rotation drug. CONCLUSIONS A single-antibiotic rotation is associated with increased incidence and heterogeneity of resistant GNR isolates, as well as increased multiple-drug-class resistance. The attenuation of resistance observed in the single-antibiotic rotation may reflect the effect of unintended antibiotic heterogeneity driven by increasing resistance to the antibiotic class recommended for use each quarter. This suggests that reliance on a single antibiotic class for empiric treatment of GNR infection exerts sufficient pressure within the environment to encourage the development of diversified resistance, as well as cross-resistance over antibiotic classes, thus narrowing the availability of effective antibiotic treatment.