Diagnostic Predictive Values of Cognitive Function Tests

Park, Jonghan; Lee, Yang Hyun; Kim, Heecheol

Abstract

OBJECTIVE
The positive and negative predictive values are more important as the diagnostic validity than the sensitivity and specificity in the general population with a low prevalence rate of a disease. The present study is about the diagnostic predictive values of the Korean version of mini-mental state examination and Cognitive Impairment Diagnosing Instrument in an unselected community elderly group. METHOD: The positive and negative diagnostic values were calculated based on the Bayes theorem, using the sensitivity and specificity of the Korean version of mini-mental state examination and Cognitive Impairment Diagnosing Instrument and a previously reported prevalence rate.
RESULTS
When the prevalence of dementia is 10.8% among elderly people living in a community, the positive predictive value, negative predictive value and combined error were estimated as being 56.7%, 99.0%, and 8.4%, respectively, at cutoff score of 23 of the Korean version of mini-mental state examination for diagnosing dementia. In the case of the Cognitive Impairment Diagnosing Instrument, they were 64.2%, 99.1% and 6.3%, respectively, at the point of 57 or less in an unselected community sample with the same prevalence. If the sensitivity and specificity are constant, the positive predictive values of the Korean version of mini-mental state examination and Cognitive Impairment Diagnosing Instrument incresed markedly with the rising prevalence while the negative predictive values and combined errors changed little.
CONCLUSION
If a variety of diagnostic tests available are similar in the diagnostic sensitivity and specificity, a test with higher positive predictive value should be prefered because the negative predictive value and combined error are not significantly influenced by the prevalence. In this regard, the Cognitive Impairment Diagnosing Instrument may be superior to the Korean version of mini-mental state examination.