Multiple diagnostic approaches are available for Clostridium difficile infection (CDI); current guidelines support two-step testing (2ST) as the preferred approach. We retrospectively evaluated the impact of switching from toxin enzyme immunoassay (EIA) to 2ST, and then to polymerase chain reaction (PCR), on CDI rates, test utilization and CDI treatment at a 900-bed tertiary care community teaching hospital. All inpatients tested for CDI between December 2008 and February 2011 were included. A positive toxin EIA or PCR was diagnostic of CDI; 2ST was performed using glutamate dehydrogenase EIA, followed by PCR if positive. Repeat tests within 8 weeks on the same patient were considered part of the same testing episode. Data were collected electronically and studied in aggregate from 9725 unique inpatients tested for CDI, representing 20 836 individual tests. PCR detected 41% more patients with CDI than toxin EIA (p <0.0001), and 15% more than 2ST (p 0.02), corresponding to higher hospital-onset and community-onset CDI rates. The number of CDI tests performed per patient decreased by 48% with PCR (p <0.0001) compared with toxin EIA. For patients with CDI, time to the first positive test result was shortest with PCR. For patients without CDI, a negative PCR, but not 2ST, was associated with 22% fewer CDI treatment days, compared with toxin EIA (p <0.0001). Compared with both toxin EIA and 2ST, PCR detected more CDI patients faster and with less frequent testing, and negative PCR results were associated with less empirical CDI treatment.