Symptom Validity Indicators Embedded in the Controlled Oral Word Association Test

Embedded symptom validity measures facilitate the detection of below-capacity performance in neuropsychological assessment. A number of such indicators have been proposed for the Controlled Oral Word Association Test (COWAT), a widely used test of word generation. However, several of these embedded indicators have not been cross-validated and it is currently unclear which represent the optimal combination of predictors. This study used Bayesian Model Averaging (BMA) to determine the set of predictors that best differentiate between patients presenting with (n = 46) and without (n = 55) malingered neurocognitive dysfunction (MND). Mild traumatic brain injury was the most common diagnosis in the MND group (96%). BMA selected the COWAT total score and a measure of change in output over time. A logistic regression model combining these variables yielded good discriminability, with an AUC of .774, (95% confidence interval = .679 to .869), 78% of cases were classified correctly, with 67% sensitivity and 88% specificity. Two alternative models performed similarly, but the variables involved required slightly longer administration and/or calculation time, making them somewhat less desirable. These results support the use of a weighted combination of embedded symptom validity measures in the COWAT.