Task performance to discriminate mild cognitive impairment from healthy aging

Mild cognitive impairment (MCI) is clinical diagnosis that refers to individuals whose performance is below average on standardized cognitive tests, but who otherwise function independently in instrumental activities of daily living. Few prior studies have addressed the problem of selecting the optimal combination of behavioral instruments and cutoff scores for detecting MCI in an outpatient setting. The aim of this work is to provide insight into two related questions: (1) What is the relative sensitivity and specificity of a battery of standardized tests frequently used to assess for MCI, as defined using receiver operating characteristic (ROC)-based analysis? (2) What are the optimal "cut point" scores for distinguishing patients mildly impaired performance based on these instruments? Two hundred forty outpatient behavioral neurology evaluations were retrospectively analyzed. All work was conducted with the formal approval of the Johns Hopkins University School of Medicine Institutional Review Board. All instruments that were evaluated performed very well in the detection of dementia (mean AUC = 0.8). However, fewer tasks performed acceptably in the detection of MCI (mean AUC = 0.7). Instruments that performed best in the detection of MCI tended to have higher total possible scores or not to reflect a score out of a total number possible. Cognitive screening tools, like the MMSE, did not perform well in the detection of MCI, raising important considerations for their interpretation. No one task in isolation is sufficient to detect MCI, and behavioral performance is not the only relevant consideration in differential diagnosis. However, these findings highlight the relative weakness of many assessments when used to build a comprehensive profile of a very large portion of outpatients presenting at clinic, those whose deficits are more subtle.