Current issues of ACP Journal Club are published in Annals of Internal Medicine


Review: Test characteristics and pretest probability of disease need to be considered in screening for dementia

ACP J Club. 1991 Nov-Dec;115:88. doi:10.7326/ACPJC-1991-115-3-088

Source Citation

Siu AL. Screening for dementia and investigating its causes. Ann Intern Med. 1991 Jul 15;115:122-32.



To evaluate the diagnostic properties of tests used in screening or the preliminary work-up for adults with dementia.

Data sources

Articles on the diagnosis of dementia were identified from MEDLINE and from reference lists of pertinent articles.

Study selection

Studies that reported on specific diagnostic findings or tests that primary care physicians could use in diagnosing dementia were selected. Studies were included if they had an adequate diagnostic standard for diagnosis, an appropriate spectrum of patients with suspected dementia, and adequately detailed data on findings or tests. 49 studies that examined formal screening tests for cognitive impairment were selected.

Data extraction

Sensitivity and specificity of tests and likelihood ratios were extracted and recalculated for additional cut-points.

Main results

Quick screening methods, formal screening tests, and tests for determining the cause of dementia were reviewed. The probability of dementia was greatly reduced with normal serial 7s, 7-digit span, recall of 3 items, or clock drawing (likelihood ratios [LR] 0.06 to 0.2). Abnormal clock drawing increased the likelihood of Alzheimer disease as distinguished from other dementias (LR 3.7; 95% CI 2.4 to 5.9) and normal drawing decreased the likelihood (LR 0.03; CI 0.01 to 0.07). Other brief screening questions for cognitive impairment (e.g, orientation to calendar, simple arithmetic) were less informative. The Mini-Mental State Examination (a 10-min test) had a sensitivity of 87% and a specificity of 82% for dementia or delirium for scores < 24. {The likelihood ratios of a positive and negative test result were 4.8 and 0.16}.* Scores ≥ 8.2) and ≥ 26 further decreased (LR 0.06 to 0.1) the probability of disease. The Short Test of Mental Status, the Set Test, the Mental Status Questionnaire, and the Cognitive Capacity Screening Examination were less well evaluated. Combining 5 studies of patients with neuropathologic diagnoses, a Hachinski Ischemic Score of > 5 increased the likelihood of multi-infarct dementia (LR 5.0, CI 2.8 to 8.8) and a score of < 3 decreased the likelihood (LR 0.3, CI 0.2 to 0.5).

The informative value of tests for diagnosing specific causes of secondary dementia (cobalamin, Venereal Disease Research Laboratory tests for neurosyphilis, cerebral imaging, erythrocyte sedimentation rate) depended on the pretest probability of the specific disease and was generally low when the pretest probability was low.


Screening tests and preliminary investigations for dementia need to be interpreted based on the test characteristics and on the pretest probability of disease.

Source of funding: National Institute on Aging.

Address for reprint: Dr. A.L. Siu, UCLA Department of Medicine, Los Angeles, CA 90024-1687, USA.

*Numbers calculated from data in article.


Although primary care physicians should be able to screen for and evaluate dementia, many are unsure of appropriate screening and diagnostic methods. The purpose of diagnostic testing is to identify "treatable" causes of disease and to reduce diagnostic uncertainty. However, fewer than 5% of all dementias improve with treatment. One half to two thirds are associated with Alzheimer disease, the prevalence of which approaches 50% among adults over the age of 80. Adherence to the standard laboratory work-up recommended by the National Institute of Neurological and Communicative Disorders and Stroke would alter the management of few patients, whereas the total laboratory costs per improved case would be enormous.

This article focuses on reliable, commonly used tests for screening and the standard diagnostic battery. Siu provides valuable guidance for a rational approach to test selection. The elegant demonstration of how LRs modify pretest probabilities encourages the clinician to consider disease prevalence and physical findings before ordering a test and aids in the interpretation of the result. Because a brief, reliable cognitive screen is needed by clinicians, it is unfortunate that the section on screening does not provide an assessment of combinations of individual questions. When orientation to month and year is combined with subtraction of serial 7s to 79 and recall of 2 of 3 items, a sensitivity and specificity of 88% (LR 7.3) are achieved (1), comparable with the Mini-Mental State Examination. Siu provides important evidence that routine ordering of cerebral imaging, serum vitamin B12 and folate levels, and syphilitic serologic testing is unwarranted. Omitted from the discussion are HIV testing, which should be considered in patients with positive risk factors, and screening for depression, which may be recognized by the primary care physician through the application of depression scales designed specifically for dementia (2).

Calvin H. Hirsch, MD
University of California, DavisSacramento, California, USA


1. Klein LE, Roca RP, McArthur J, et al. Diagnosing dementia: univariate and multivariate analyses of the mental status examination. J Am Geriatr Soc. 1985;33:483-8.

2. Alexopoulos GS, Abrams RC, Young RC, et al. Cornell scale for depression in dementia. Biol Psychiatry. 1988;23:271-84.