OBJECTIVES: To cross-validate the Diabetes Knowledge Test (DKT) on an independent sample of Medicaid-eligible patients with type 2 diabetes using both English-language and Spanish-language versions of the instrument. METHODS: Using baseline data from a randomized control trial (RCT) of pharmacist-led patient treatment vs. standard care, psychometric analysis is conducted to evaluate internal consistency reliability using Cronbach’s alpha. Comparisons in item difficulty are made with the samples used for the original test creation. Construct validity is assessed using educational attainment and clinical measures.  Sources of variation in baseline patient knowledge are examined using regression analysis. RESULTS: A total of 150 patients (29 Spanish-language, 121 English-language) are analyzed. The Cronbach’s alpha for the population is 0.58 (0.64 English-language, 0.67 Spanish-language). The average score is 56.2% for English-language and 49.0% for Spanish-language compared to 64.8% for the original validation study. Education is positively associated with score (t=2.07 p<0.05).  Age is negatively correlated with score (t=-2.36, p<0.05). Years since diagnosis, language, gender, and race do not show associations that are statistically significant.  Correlations between HbA1c and score are 0.01 for English-language (p>0.05) and 0.21 for Spanish-language (p<0.05). CONCLUSIONS: Scores for the test and Cronbach’s alpha are lower than in past literature. Differences in the score between languages are not statistically significant indicating consistency between forms. The impact of education on score suggests construct validity.  However, scores are uncorrelated with HbA1c for those completing English-language forms, and are weakly positively correlated with HbA1c for those completing Spanish-language forms.  Further research is needed to determine if increases in diabetes knowledge are associated with improvements (decreases) in HbA1c.