Construct validity of the abbreviated mental test in older medical inpatients

Objectives: To evaluate validity and internal structure of the Abbreviated Mental Test (AMT), and to assess the dependence of the internal structure upon the characteristics of the patients examined. Design: Cross-sectional examination using data from the Italian Group of Pharmacoepidemiology in the Elderly (GIFA) database. Setting: Twenty-four acute care wards of Geriatrics or General Medicine. Participants: Two thousand eight hundred and eight patients consecutively admitted over a 4-month period. Measurements: Demographic characteristics, functional status, medical conditions and performance on AMT were collected at discharge. Sensitivity, specificity and predictive values of the AMT 80 years), gender, education (5 years) and presence of congestive heart failure (CHF). Results: AMT achieved high sensitivity (81%), specificity (84%) and negative predictive value (99%), but a low positive predictive value of 25%. The principal component analysis isolated two components: the former component represents the orientation to time and space and explains 45% of AMT variance; the latter is linked to memory and attention and explains 13% of variance. Comparable results were obtained after stratification by age, gender or education. In patients with CHF, only 48.3% of the cumulative variance was explained; the factor accounting for most (34.6%) of the variance explained was mainly related to the three items assessing memory. Conclusion: AMT >6 rules out dementia very reliably, whereas AMT