Objective: Cognitively healthy individuals commonly produce one or more “abnormal” test scores during a neuropsychological evaluation, especially when multiple tests are administered, when test scores are correlated, and when abnormality is identified with a high (sensitive) cutoff score. As such, appreciation of the base rates of abnormal test scores within both healthy and clinical samples is pertinent for accurately interpreting the scores for the individual patient whose cognitive status is unknown. Method: Data from the National Alzheimer's Coordinating Center (NACC) were used to estimate base rates of abnormal test scores separately for cognitively healthy participants and those with all-cause dementia. Monte Carlo simulation was used to estimate prevalence rates of impaired test scores and Bayes' theorem was used to calculate post-test probabilities of cognitive impairment at common cutoff scores (z < −1, < −1.5, and < −2). Results. Both cognitively healthy people and those with dementia frequently produced abnormal test scores. For example, 65.8% of controls and 97.9% of demented participants produced at least one abnormal score (z < −1) across the 10-test battery of the NACC Uniform Data Set. When cutoff scores are low (specific) and the expected prevalence of cognitive impairment is high (e.g., >70%), even one or two abnormal test scores should not be interpreted as benign. Conclusion. Given the prevalence of abnormal test scores within both cognitively healthy and all-cause dementia samples, these analyses provide information that may prevent the over- or under-interpretation of impairment in an individual's performance. Consideration of these analyses may improve accuracy of diagnostic interpretation.
PROFESSIONAL ISSUES: TEST DEVELOPMENT AND METHODS
Rethinking Abnormal Test Scores in Normal People: Application of Base Rate Data to the Individual Patient
B Gavett, S John; PROFESSIONAL ISSUES: TEST DEVELOPMENT AND METHODS
Rethinking Abnormal Test Scores in Normal People: Application of Base Rate Data to the Individual Patient. Arch Clin Neuropsychol 2015; 30 (6): 594-595. doi: 10.1093/arclin/acv047.289
Download citation file:
© 2017 Oxford University Press×