Between 1999 and 2000, new diagnoses of HIV in Victoria (Australia) rose by 41%, from 140 to 197. In this time period, sera from new HIV diagnoses were tested using the Organon Teknika "detuned" enzyme immunoassay (EIA). We compared the results of the detuned EIA with incident infections defined by surveillance (on the basis of a previous negative or indeterminate HIV test and/or a seroconversion illness within the 12 months preceding HIV diagnosis). Of 317 specimens, 97 (31%) incident infections and 114 (36%) recent infections were detected using surveillance and detuned EIA, respectively. The detuned assay misclassified 11 cases with AIDS and 2 cases with CD4 counts < or = 200 micro3 (probable long-standing infections) as recent infections and was unable to identify 31 (32%) of 97 cases previously classified as incident cases by surveillance. The assay detected an extra 35 recent infections that were previously classified as nonincident by surveillance. By combining the detuned assay and surveillance, 132 (42%) incident infections were identified from 317 specimens, 36% more than surveillance alone. We recommend that a detuned assay or similar test become part of the routine strategy to identify incident infections in Victoria. Incident infections provide important information for targeting prevention strategies and the opportunity to interrupt ongoing viral transmission.