Abstract
Purpose:
Standard tests of colorblindness screen for organic disease. We have developed a method that attempts to recognize malingerers of colorblindness. The purpose of this study is to validate this test in a sample population.
Methods:
An online survey was distributed to 84 self-reported and verified colorblind participants and 131 participants instructed to simulate colorblind malingering. The survey contained three sets of twelve color-adjusted versions of the standard Ishihara color plates, as well as one set of twelve unmodified plates. Participants were asked to identify numbers on these 48 plates. Two tests were then created, each using six unique plates. A "balanced test" emphasizing both sensitivity and specificity was assembled by prioritizing test plates that colorblind participants most often correctly identified, as well as those test plates which malingering simulators most often incorrectly identified. A second set of six test plates comprised the "specific test," which maximized test specificity. This was assembled by prioritizing only those plates that colorblind participants most often correctly identified. Statistical measures of both the "balanced test" and "specific test" (sensitivity, specificity, and Youden indexes) were assessed at each possible cut-off threshold, and a receiver operating characteristic (ROC) function with its area under the curve (AUC) charted.
Results:
For the "balanced test," colorblind and colorblind simulating participants had a difference of means of 60.1% (CI: 53.6% to 66.5%). Statistical measures showed an optimal cut-off of at least 1 missed "balanced test" plate to recognize a colorblind malingerer (Youden index: 0.81, sensitivity: 88.6%, specificity: 92.9%), with an AUC of the ROC of 0.93. For the "specific test," colorblind and colorblind simulating participants had a difference of means of 41.6% (CI: 34.8% to 48.4%). Statistical measures showed a cut-off of at least 2 missed "specific test" plates to identify a colorblind malingerer with 100% specificity (sensitivity 51.9%, Youden index 0.52, AUC of 0.85).
Conclusions:
Our method for recognizing colorblind malingering demonstrates a high degree of reliability in a large population, and can be used to both screen for colorblind malingerers and to identify them with 100% specificity.