Toxin enzyme immunoassays (EIAs) are inadequate for the diagnosis of Clostridium difficile infection (CDI) when used alone. In September 2010 we replaced toxin EIA with a two-step algorithm, testing first with glutamate dehydrogenase and confirming with polymerase chain reaction for toxin B gene. We compared this to the gold standard of toxigenic culture, observing a positive predictive value of 96% (laboratory prevalence of 4.7%). There was no deterioration in turnaround time but there was a decrease of 11% in repeat specimens sent from the same patients. The improved performance of the algorithm increased the laboratory positivity rate from 2.2% to 5.6%. This led to an increase in our Trust CDI rate reported under the Health Protection Agency's mandatory surveillance scheme. We investigated whether the change was due to increasing nosocomial transmission, environmental contamination or consumption of antimicrobials, but found no evidence of this. We conclude that it probably resulted from the change in testing algorithm. Although we have improved testing and enhanced patient safety, we are likely to be unfairly financially penalised because of our apparent (but not real) increase in CDI rates. Assessment of CDI rates should take testing methodology into account and national policies should be revised to reflect this. (C) 2011 The British Infection Association. Published by Elsevier Ltd. All rights reserved.