Poor diagnostic reliability, the Null-Bayes Logic Model, and their implications for sexually violent predator evaluations.

Abstract

Forensic psychologists have approached sexually violent predator (SVP) civil commitment evaluations from the position that respondents must be positive for a condition from the Diagnostic and Statistical Manual of Mental Disorders (DSM) of the American Psychiatric Association to be classified as SVPs. The only research on DSM diagnostic reliability in SVP cases has been undertaken by J. S. Levenson (2004a) and R. L. Packard and J. Levenson (2006). Although Packard and Levenson claimed that diagnostic evaluations in SVP cases were highly reliable, a reanalysis of their data indicated otherwise. Further, high levels of diagnostic uncertainty were found for a proposed paraphilia referred to as paraphilia not otherwise specified‐ nonconsent. Diagnostic criteria used to identify paraphilias among SVP respondents are therefore characterized by poor reliability. Logic models that were previously used to determine diagnostic confidence are also obsolete. Recommendations for improving diagnostic reliability are discussed, and the Null-Bayes Logic Model (NBLM) is proposed as a method for reaching certainty opinions that is superior to past models based on unrestrained clinical judgment. The implications of the present results and the NBLM for future practice, research, and policy directions are discussed.