The following section, therefore, examines the classification accuracy of the Stroop scores at a range of cut-offs. Note that because the Interference score was not significantly different between groups for the ANOVA and ROC analyses, it will not

be included in this analysis. The relevant indices of classification accuracy are Sensitivity, False Positive Error Rate (FP rate), and LR (Gouvier, Hayes, & Smiroldo, 1998; Hennekens & Buring, 1987). Sensitivity represents the percentage of malingerers correctly classified (true positives). The FP rate reflects the proportion of non-malingerers selleck kinase inhibitor whose scores fell in the malingering range according to a specified cut-off. The LR indicates the likelihood that a score was produced by a malingerer relative to non-malingerers (sensitivity/FP rate). An LR value of 1.0 indicates that a given score does not differentiate between groups, whereas a higher LR value indicates a higher probability that the observed www.selleckchem.com/products/Maraviroc.html score was produced by a malingerer (Grimes & Schulz, 2005). LRs from 2 to 5, 5 to 10, and greater than 10 yield small, moderate, and large increases in the post-test probability, respectively (Grimes & Schulz,

2005). The LRs were calculated using only the mild TBI groups. The data for the moderate–severe TBI and mixed-diagnosis cases are presented for comparison. Clinical diagnoses other than CVA were combined into one group due to small individual sample sizes. As can be seen in Table 5, sensitivity ranged from 12% to 32% for WR, 12% to 24% for CR, and 0% to 24% for CWR at cut-off scores associated

with a 0% to 9% FP rate. LRs for cut-off scores associated with a 5% FP rate MCE were 6.32 (95% CI = 1.48–26.96) for WR, 3.79 (95% CI = 0.82–17.62) for CR, and 1.26 (95% CI = 0.19–8.52) for CWR. Although CWR did not differentiate between groups at a 5% FP rate, post-test probability was greater at cut-off scores associated with a 9% FP rate (LR = 2.53; 95% CI = 0.83–7.70). WR was most effective at differentiating MND and non-MND TBI patients, detecting 29% of MND patients when the cut-off (−47) is associated with a 5% FP rate. There was, however, a high FP rate (19%) associated with this cut-off in patients with CVA. Based on these data, the indicators provide small-to-moderate probability that a score reflects invalid performance, depending on the indicator used. The following section examines the classification accuracy of the three Stroop variables in combination, rather than individually. Joint classification accuracy was examined to (a) determine whether sensitivity is increased by using a combined set of the indicators and (b) ensure that FP error rates are not increased when using the indicators in combination. Based on the score distributions, cut-offs were established to correspond to FP rates of 0%, 5%, and 9%, and the number of hits at each FP rate was combined (i.e.