The Value of Bayes’ Theorem for Interpreting Abnormal Test Scores in Cognitively Healthy and Clinical Samples

Abstract

The base rates of abnormal test scores in cognitively normal samples have been a focus of recent research. The goal of the current study is to illustrate how Bayes' theorem uses these base rates--along with the same base rates in cognitively impaired samples and prevalence rates of cognitive impairment--to yield probability values that are more useful for making judgments about the absence or presence of cognitive impairment. Correlation matrices, means, and standard deviations were obtained from the Wechsler Memory Scale--4th Edition (WMS-IV) Technical and Interpretive Manual and used in Monte Carlo simulations to estimate the base rates of abnormal test scores in the standardization and special groups (mixed clinical) samples. Bayes' theorem was applied to these estimates to identify probabilities of normal cognition based on the number of abnormal test scores observed. Abnormal scores were common in the standardization sample (65.4% scoring below a scaled score of 7 on at least one subtest) and more common in the mixed clinical sample (85.6% scoring below a scaled score of 7 on at least one subtest). Probabilities varied according to the number of abnormal test scores, base rates of normal cognition, and cutoff scores. The results suggest that interpretation of base rates obtained from cognitively healthy samples must also account for data from cognitively impaired samples. Bayes' theorem can help neuropsychologists answer questions about the probability that an individual examinee is cognitively healthy based on the number of abnormal test scores observed.

Psichikos sveikatos sutrikimų turinčių asmenų paprastos ir sudėtingos informacijos apdorojimo greitis

“…We used common thresholds that are used in neuropsychology for identifying abnormal scores: 1, 1.5 and 2 standard deviations below the mean (Gavett, 2015). The relationships between variables were tested using χ² statistic.…”

“…Although, without a doubt, psychiatric disorders are etiologically and clinically extremely heterogeneous, statistically worse performance on almost any cognitive test is a feature in most clinical samples (Gavett, 2015). General slowing of information processing is also related to psychia tric disorders, such as depression (Reppermund et al, 2007;Tsourtos, Thompson, & Stough, 2002), schizophrenia (Brebion et al, 2015;Cella & Wykes, 2013), schizoaffective disorder (Fryar-Williams & Strobel, 2015;Simonsen et al, 2011) and others.…”

“…It has been demonstrated that acetylcholinergic (Thompson, Stough, Ames, Ritchie, & Nathan, 2000), dopaminergic (Backman et al, 2000), and serotonergic (Herrera-Guzmán et al, 2009) neurotransmission; posterior EEG measures (Gazzaley et al, 2008) and BOLD variability between cognitive states (Garrett, Kovacevic, McIntosh, & Grady, 2013); regional brain volumes (e.g., thalamic volume (Van Der Werf et al, 2001), and even overall anatomical feature of the brain (e.g., grey matter volume and ventricular size (Coffey et al, 2001)) are all linked to information processing slowing. Also, some of the most consistent results come from studies of IPS relationship with the white matter structure (Kerchner et al, 2012) and volume (Magistro et al, 2015).…”

“…First of all, it has been suggested that IPS might not be a unitary construct (Chiaravalloti, Christodoulou, Demaree, & DeLuca, 2003;Jurkuvėnas, 2015). Secondly, there is evidence to indicate that IPS tasks of different complexity show different patterns of deficit at least in some clinical samples (Gerritsen, Berg, Deelman, Visser-Keizer, & Meyboom-de Jong, 2003;Gualtieri & Johnson, 2006).…”

“…Importance of both simple and complex IPS tasks in differentiating mild cognitive impairment and Alzheimer's disease (Gorus, De Raedt, Lambert, Lemper, & Mets, 2008) unilateral stroke patients (Gerritsen et al, 2003), ADHD, post-concussion syndrome and depression (Gualtieri & Johnson, 2006) has been demonstrated. However, studies about clinically significant as opposed to statistically significant differences in mental disorders are still lacking.…”