Abstract
In civil and forensic evaluations of psychological damage, depression is one of the most commonly identified disorders, and also one of the most frequently feigned. Thus, practitioners are often confronted with situations in which they must assess whether the symptomatology presented by a patient is genuine or being feigned for secondary gains. While effective, traditional feigning detection instruments generate a high number of false positives—especially among patients presenting with severe symptomatology. The current study aimed at equipping forensic specialists with an empirical decision-making strategy for evaluating patient credibility on the basis of test results. In total, 315 participants were administered the Beck Depression Inventory-II (BDI-II) and SIMS Affective Disorders (SIMS AF) scales. Response patterns across the experimental groups (i.e., Honest, Simulators, Honest with Depressive Symptoms) were analyzed. A machine learning decision tree model (i.e., J48), considering performance on both measures, was built to effectively distinguish Honest with Depressive Symptoms subjects from Simulators. A forward logistic regression model was run to determine which SIMS AF items best identified Simulators, in comparison with Honest with Depressive Symptoms subjects. The results showed that the combination of feigning detection instruments and clinical tests generated incremental specificity, thereby reducing the risk of misclassifying Honest with Depressive Symptoms subjects as feigners. Furthermore, the performance analysis of SIMS AF items showed that Simulators were more likely to endorse three specific items. Thus, computational models may provide effective support to forensic practitioners, who must make complex decisions on the basis of multiple elements. Future research should revise the content of SIMS AF items to achieve better accuracy in the discrimination between feigners and honest subjects with depressive symptoms.