Assessment of Paraphilic Disorders
All but one study investigating the psychometric properties of assessment instruments for paraphilias or paraphilic disorders were conducted with sexual offender populations. Concerning sexual offender evaluations, it can be assumed that the vast majority of clinicians primarily rely on their clinical experience and judgment when diagnosing a paraphilic disorder.
Perillo et al. evaluated the interrater reliability of mental disorder diagnoses (DSM-IV) in 375 sexual offenders who were evaluated under the Sexually Violent Predator (SVP) Act 2004 in the USA. Data from 128 mental health professionals were analyzed (PhD-level clinicians, psychologists, M.D., and Master's-level clinicians) and it was found that clinicians had the highest agreement concerning a pedophilia diagnosis, with a probability of 78% that two or more clinicians would agree on the presence of that diagnosis. Concerning paraphilias not otherwise specified, there was a probability of 52% that clinicians would agree upon the diagnosis and a probability of 33% concerning a sexual sadism diagnosis.
However, it was suggested that clinical assessments are susceptible to socially desirable answering and impression management limiting the diagnostic validity of clinical judgments. In contrast, assessment instruments measuring implicit cognitions and implicit attentional processes were described as more objective, less vulnerable to deception, and could thus be considered as one useful addition to support clinical diagnostics. During the last 18 months, studies assessing the psychometric properties of implicit association tasks (IATs) or viewing time measures concerning the assessment of pedophilic sexual interests (not pedophilia diagnosis) were published. Reliability of implicit measure is usually reported as Cronbach's alpha. Classification validity of implicit measures is usually assessed as the area under the curve of the receiver operating characteristics curve (AUC of the ROC curve). Thereby, an AUC of 1 would represent a perfect distinction between individuals with and individuals without a paraphilia, whereas an AUC of 0.5 would refer to a prediction at chance level. According to the commonly used classification, AUCs below 0.63 are considered as small, AUCs between 0.64 and 0.71 as moderate, and AUCs greater than 0.72 as large effect sizes.
Hempel et al. found a reliability of α = 0.67 for their IAT in a sample of 118 men [65 child sexual abusers (CSAs), 13 rapists, and 40 nonoffenders]. Furthermore, the IAT could significantly distinguish between CSAs and nonoffenders (AUC = 0.65), meaning that CSAs had stronger associations concerning child stimuli.
Mokros et al. evaluated the psychometric properties of a viewing time measure and an explicit attractiveness rating of pictures showing children or adults using a sample of 42 pedophilic CSAs, 27 nonsexual offenders, and 95 men from the community. Both measures produced highly acceptable reliability scores (viewing time α = 0.79–0.89; explicit rating: α = 0.90–0.98). Classification accuracy between CSAs and community men produced significant AUCs of 0.62 for viewing time, 0.64 for the explicit ratings, and 0.80 for a combined measure (viewing time and explicit rating). However, a sensitivity of 90.5% in the combined measure would go along with a rather low specificity of 50.5%. Comparably, Babchishin et al. assessed the classification accuracy of an IAT, a viewing time measure, and an explicit rating in 35 CSAs in comparison to 21 nonsexual offenders. They found that a combined measure consisting of viewing time, IAT, and the explicit rating had a higher predictive accuracy (AUC = 0.88) in distinguishing CSAs from nonsexual offenders concerning their intensity of pedophilic sexual interests compared with the IAT (AUC = 0.60) and the explicit rating (AUC = 0.73) alone but not compared to the viewing time measure (AUC = 0.82). Furthermore, a combined IAT, viewing time, and explicit rating could also distinguish between different CSA subgroups, with intrafamilial CSAs (n = 19) having lower sexual deviance scores than extrafamilial CSAs (n = 35) and child pornography offenders (n = 18).
Also in men not previously convicted because of a sexual offence, the combination of two implicit tasks [IAT and picture association task (PAT)] revealed a higher classification accuracy (AUC = 0.97) than the IAT (AUC = 0.89) or the PAT (AUC = 0.84) alone. Both tasks intended to assess pedophilic sexual interests in 20 self-reported pedophilic men (only 1 man was previously convicted for a sexual offence against a child) and 20 self-reported heterosexual men from the community.
Although implicit measures can be used as a valid addition in clinical diagnostics in men who have committed a sexual offence, such a multimethod approach also requires greater institutional resources, greater assessment time, and more personal expertise. Furthermore, a clinical diagnosis should not be solely based on the results of these implicit measures and always requires information from the patient. It should also be noted that if these so-called objective measurements are used to detect information about sexual interests that the patient wants to hold back, they may also cause negative consequences in the therapeutic processes.