This study examined agreement (concordance or convergent validity) between self-report and birth certificate for gestational diabetes. Study population was 2,854 women who had live births 2-6 months earlier and responded to a questionnaire from the New York State Pregnancy Risk Assessment Monitoring System (PRAMS) survey, 2004-2006. Agreement between self-report and birth certificate was assessed for the study population overall, and for subgroups defined by race, age, education, marital status, number of previous live births, time of first prenatal care, and birth weight of the newborn. A total of 258 women self-reported gestational diabetes, while birth certificates indicated that 138 women had gestational diabetes. For the study population overall, percent agreement was 93.8% and Kappa was 0.53. Due to the moderate bias index (68.2% overall, ranged from 33.3 to 100% in subgroups) and the high skewed prevalence index (91.8% overall, ranged from 70.7 to 97.5% in subgroups), we determined Prevalence-Adjusted and Bias-Adjusted Kappa (PABAK) was a better measure of agreement. PABAK was 0.88 overall, indicating very good agreement. PABAK was uniformly high in all subgroups. The highest PABAK was found among women aged 25 years and younger (0.93), and the lowest PABAK was among Asian women (0.79). Although the absence of a gold standard for gestational diabetes hinders assessment of criterion validity, high PABAK measures suggest that self-reporting by PRAMS respondents is feasible for identifying cases of gestational diabetes for surveillance and population-based epidemiologic research.