OBJECTIVES: This study examined the reliability of the first positive HIV test date reported in the U.S. HIV/AIDS Reporting System (HARS). This date is essential to determine case counts for resource allocation for HIV treatment and prevention efforts. METHODS: The dates of first positive HIV tests reported by individuals with HIV in an interview survey conducted in 16 states (n=16,394, interviewed 1995-2002) were compared with the dates of HIV diagnosis reported to HARS. The percentage of agreement for the year of diagnosis and the weighed kappa (k) with 95% confidence intervals (CIs) was calculated. RESULTS: Self-reported year of diagnosis agreed with the year of diagnosis in HARS for 56% of date pairs (k=0.69; 95% CI 0.68, 0.70); 30% reported an earlier diagnosis year. Agreement differed by sex, age, race, exposure, and reason or place of testing (p<.01). Lower agreement was found when the self-reported diagnostic test was anonymous (k=0.57; 95% CI 0.52, 0.62) compared with confidential tests (k=0.66; 95% CI 0.64, 0.68). Lower agreement was also found for cases first reported with AIDS (k=0.58; 95% CI 0.55, 0.62) compared with cases first reported with HIV not AIDS (k=0.71; 95% CI 0.70, 0.73) as well as for participants interviewed three years or more after their HARS diagnosis date (k=0.55; 95% CI 0.52, 0.57) compared with those interviewed within one year (k=0.62; 95% CI 0.61, 0.63). More than 20% of participants in almost all groups, however, reported earlier diagnosis years than those recorded in HARS. CONCLUSION: As many as 30% of HIV diagnoses may have occurred earlier than recorded in HARS. Additional studies need to determine mechanisms to adequately capture diagnosis dates in HARS.