Differentiating Dementia From Normal

To the Editor.—  I read with interest the article by Drs Eslinger et al¹ entitled "Neuropsychological Detection of Abnormal Mental Decline in Older Persons." The article described a brief neuropsychological test battery that correctly classified 87% of normal controls and dementias in a validation study, and 89% in a subsequent cross-validation study.The method the authors used to discriminate between the normals and dementias was a stepwise linear discriminant-function analysis that used scores from eight neuropsychological tests as potential predictors. This statistical method maximizes the identification of true-positive and true-negative cases, with a minimum number of predictors, and thus optimizes for a parameter called "test efficiency."^2,3 However, my colleagues and I have questioned the value of test efficiency as the appropriate characteristic to be maximized in diagnostic screening tests for dementia.⁴Galen and Gambino³ stress that high test efficiency is desirable when a disease is serious