ObjectiveTo examine the temporal changes in the likelihood of dementia and mild cognitive impairment (MCI) between 1993 and 2012 using a short battery of cognitive tests.MethodsA cohort of 10,342 participants underwent a short battery of cognitive tests collected during triennial in-home interviews with 2,794 of those evaluated for the clinical diagnosis of dementia and MCI. We used a generalized logit regression model to estimate the likelihood of dementia and MCI, and a quasibinomial regression model to examine the temporal changes in those likelihood scores.ResultsA short battery of cognitive tests—delayed story recall test, Symbol Digit Modalities Test, and the Mini-Mental State Examination—were associated with the clinical diagnosis of dementia and MCI. The classification accuracy of likelihood scores was 0.92 for dementia and 0.85 for MCI. After adjusting for age, race/ethnicity, and education, the likelihood of dementia in the population decreased from 21.6% (95% confidence interval [CI] 20.9%–22.3%) to 18.9% (95% CI 18.1%–19.7%) between 1993–1996 and 2000–2002 and showed no significant decline between 2000–2002 and 2009–2012 (−0.2%, 95% CI −1.1% to 0.7%). The estimated likelihood of MCI remained similar between 1993–1996 and 2009–2012 (29.0%, 95% CI 27.9%–30.1%), but showed a nonsignificant decrease in 2000–2002.ConclusionThe likelihood scores based on a short battery of cognitive tests can serve as a measure of dementia and MCI in epidemiologic studies. The decline in the likelihood of dementia and MCI over earlier years was not sustained in later years.