Objective: Examine the psychometric properties and classification accuracy of a digit symbol substitution task (DSST) to screen for mild cognitive impairment (MCI) in older adults. Methods: Twenty-seven older adults with MCI and 50 demographically-matched healthy controls (HC) over age 60 completed the MMSE, Quick MCI (Qmci), and DSST, the latter of which was administered twice to examine the added utility of practice effects. Results: Patients with MCI performed significantly worse than their HC counterparts on the MMSE (η2 = 0.191), Qmci (η2 = 0.153), and the first (η2 = 0.337) and second (η2 = 0.393) administration of the DSST. HC participants also demonstrated significantly greater practice effects than those with MCI who did not appear to benefit from task repetition (mean score increases of 7.4 and 0.5, respectively). The DSST exhibited excellent within-session test–retest reliability (R2 = 0.774) and was more strongly correlated with Qmci (R2 = 0.311) than MMSE (R2 = 0.112) scores. ROC analyses indicated that a cutoff score of 23.5 for the first administration (AUC = 0.860) and 27.5 for the second administration (AUC = 0.900) of the DSST were associated with superior classification accuracy to that of the MMSE (AUC = 0.759) and Qmci (AUC = 0.777). DSST difference scores (AUC = 0.764) reflecting within-session practice effects were slightly superior to the MMSE but not the Qmci. In logistic regression analyses, scores from the first administration of the DSST significantly predicted group membership and the addition of practice effects significantly improved model fit (p < 0.001) resulting in an overall classification accuracy of 85.1%. Conclusions: Findings suggest that the DSST is a valid and reliable rapid screening measure which may prove useful for clinicians and researchers alike.