A Brief Telephonic Instrument to Screen for Cognitive Impairment in a Managed Care Population

• Objective: To design a brief telephonic survey to screen for cognitive impairment in the Medicare managed care population. • Patients and methods: Eleven questions were pilot tested on 22 patients with Alzheimer’s disease (AD) and 20 normal controls. Analyses of question reliability and discrimination were conducted on 60 AD patients and 44 controls. Logistic regression analysis with forward selection was used to select the questions that best predicted AD status. Two methods for computing screening test scores, the predicted probability of AD derived from logistic regression and an integer-weighted sum of question scores, were compared using receiver-operating-characteristic (ROC) curves. The correlation between test scores and Mini-Mental State Examination (MMSE) scores was computed as a measure of validity. • Results: Ten of the 11 questions had acceptable reliability. Three questions were selected: delayed recall, help remembering medications, and help planning a trip for errands. The ROC curves for the integerweighted scores and predicted probability scores were nearly identical for the 3-question screen, with C-statistics of 0.896 and 0.905. Summary scores were highly correlated with the MMSE (r = ‐0.64, P < 0.001). For patients not receiving medications, 2 questions were selected: delayed recall and help planning a trip for errands. For the 2-question screen, the ROC curves for the integer-weighted scores and predicted probability scores were nearly identical, with C-statistics of 0.853 and 0.854. • Conclusion: A brief telephonic screen for cognitive impairment employing 3 questions with a simple system for scoring compares favorably with longer screening instruments. The screen has promise for integration into existing health risk assessments conducted by Medicare managed care plans.