EP142 Diagnosis of pain deception using MMPI-2 based on XGBoost machine learning algorithm: a single-blinded randomized controlled trial

Background and Aims

Assessing pain deception is challenging due to its subjective nature. This study explores using Minnesota Multiphasic Personality Inventory-2 (MMPI-2) analysis with machine learning (ML) to detect malingering. We hypothesize that ML analysis of MMPI-2 can detect pain deception. The main goal of this study was to evaluate the diagnostic value for pain deception using ML analysis with MMPI-2 scales, considering accuracy, precision, recall, and f1-score as diagnostic parameters.

Methods

We conducted a single-blinded, randomized controlled trial to evaluate the diagnostic value of the MMPI-2, Waddell's sign, and salivary alpha amylase (SAA). We grouped the non-deception (ND) group and the deception (D) group randomly.

Results

Of the total of 96 participants, 46 were assigned to group D and 50 to group ND. In the logistic regression analysis, pain and MMPI-2 did not show diagnostic value, however in ML analysis, values of selected MMPI-2 (sMMPI-2) which is related to malingering showed accuracy 0.684, precision 0.667, recall 0.800, and f1-score came out as 0.727. When performed with whole MMPI-2(wMMPI-2), accuracy 0.621, precision 0.692, recall 0.562, and f1-score 0.651 was showed. The f1-score was higher in sMMPI-2.

Conclusions

We suggest that the diagnosis of pain deception through the pattern changes of MMPI-2 scales using ML could be valuable. It could be a benefit to clinicians to detect deception exactly and objectively in various situations. Further large-scale studies would be needed to screen and predict more precisely Institutional Review Board