Previous research has shown that the Mini-Mental State Examination (MMS) is biased as a measure of cognitive impairment in minority and low-education patients. The purpose of this study was to (1) develop a statistical correction for effects of age and education and (2) test the efficacy of the statistically adjusted MMS (MMSAdj) as a screening test for dementia using different ethnic groups and education levels. We used a population-based community survey sample (n = 590) composed of 46.6% Hispanics and 53.4% non- Hispanics to derive the statistical correction, defined as: MMSAdj = Raw MMS · (0.471 x [Education-12]) + (0.131 x [Age-70]). Ethnicity and language of test administration were not significantly related to MMSAdj in the community survey sample, but the raw MMS was strongly influenced by these factors. We used an independent sample (n = 2,983) of patients evaluated through the California Alzheimer's Disease Diagnostic and Treatment Centers to test the diagnostic accuracy of the MMS and the MMSAdj across low- and high-education groups and across whites, Hispanics, and blacks. Results showed greater stability of sensitivity and specificity across education levels and ethnic groups for the MMSAdj than for the raw MMS and suggest that the MMSAdj is a preferable measure of cognitive impairment for low-education and minority individuals.