Using Retention Scores to Improve the Clinical Utility of the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) Word List

Daniel Weitzner, James A. Haley Veterans' Hospital, Tampa, United States
Emily Vanderbleek, James A. Haley Veterans' Hospital, Tampa, United States
Joel Kamper, James A. Haley Veterans' Hospital, Tampa, United States


Retention scores have shown promise as a marker to help distinguish between clinical groups. In particular, retention scores improve the diagnostic accuracy of patients with Alzheimer’s disease, who demonstrate hallmark difficulty of consolidation of information, versus other etiologies in which learning and free recall are more affected. However, not all studies have found that retention scores provide incremental validity beyond other memory scores. This discrepancy may be related to overinflated retention scores due to floor effects on learning trials. For example, if a participant retains the only word that they recalled on the final learning trial, it is unclear whether that retention score (i.e., 100%) should be interpreted. The goals of the current study were to a.) examine the diagnostic utility of retention scores in those with and without Alzheimer’s disease and b.) determine if interpretation of retention scores is dependent upon the number of words learned on the final learning trial.

Participants and Methods:

Data were obtained from 120 male Veterans (mean age=79.9 years, SD = 6.84) who underwent comprehensive neuropsychological evaluation in a VA memory disorders outpatient clinic from 2016-2022. Data from cognitively intact (n=27) patients and those diagnosed with Major Neurocognitive Disorder due to Alzheimer’s disease (AD; n=41) or vascular disease (VaD; n=52) were used. The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) word list, a 3-trial list learning task with short- and long-delay recall trials, was used in analyses. Logistic regression analyses were run to determine the ability of retention scores to distinguish AD from VaD based on different scores on trial 3 of the CERAD. After the optimal trial 3 score was identified, regression was used to analyze the incremental validity of the retention score above the CERAD total learning score. Sensitivity and specificity analyses were then conducted to identify retention score cut-offs for distinguishing AD from VaD.


CERAD retention scores significantly distinguished AD and VaD at various trial 3 cut-points (p’s < .05), but were best able to distinguish when a cut-off of >4 words on trial 3 was used (OR = 1.051, p = .014). The retention score (p = .020) added incremental validity over total learning (p > .05) in distinguishing AD from VaD. A retention cut score of 50% when using a cut-off of >4 words on trial 3 had a.) sensitivity of 100% and specificity of 33.33% when distinguishing AD from VaD, b.) 100% sensitivity and 62.96% specificity when distinguishing AD from cognitively intact, and c.) 66.67% sensitivity and 62.96% specificity when distinguishing VaD from cognitively intact.


The present study supports interpretation of the retention score on the CERAD. The retention score adds incremental validity above other memory scores in distinguishing between clinical groups, similar to previous findings on other measures. There is improved interpretability of retention scores on the CERAD when patients recall at least 5 words on trial 3 of the learning trials.

Category: Dementia (Alzheimer's Disease)

Keyword 1: memory disorders
Keyword 2: neuropsychological assessment