Derivation of embedded symptom validity indicators for the Brief Symptom Inventory-18

Symptom validity assessment is an important component of the evaluation of psychological sequelae of traumatic brain injury (TBI). Many personality inventories have embedded symptom validity scales (SVTs). A growing body of research reports derivation of embedded SVTs in brief symptom and self-report measures. We sought to develop embedded SVT indicators based on the Brief Symptom Inventory-18 (BSI-18).

This project involved secondary analysis of TRACK-TBI data obtained from the Federal Interagency Traumatic Brain Injury Research Informatics System repository. Participants (N = 1,694) sustained TBI (n = 1,515) and or non-head orthopedic injury (n = 179). Independent SVTs were used to identify TBI cases with questionable validity (QV). Two novel SVTs were examined: BSI-V, which was developed using Bayesian model averaging (BMA) to determine a subset of BSI-18 items that predict QV, and BSI GSI, which is a total score cutoff based on the BSI-18 Global Severity Index. Receiver operating characteristic (ROC) analysis was used to compare classification accuracy of BSI-V and BSI GSI.

Four BSI-18 items were determined to be optimal predictors of QV. The BSI-V and BSI GSI both showed outstanding discrimination with area under the ROC curve (AUC) of .95. BSI-V (≥5) and BSI GSI (≥64 T) cutoffs showed 83% sensitivity and 90% specificity.

Findings provide preliminary evidence of novel embedded SVTs in the BSI-18 with comparable support for BSI-V and BSI GSI. These findings warrant independent cross-validation to provide additional evidence of classification accuracy and refine cutoffs prior to implementation in practice.