Which Domain of the PSY–5 Is Most Relevant to Substance Use?

ABSTRACT

Substance use has generally been related to lower levels of agreeableness and conscientiousness, but has evidenced relationships with other personality trait domains as well, including impulsivity. This study was conducted to determine which trait domain of personality is most related to substance use from the perspective of the Personality Psychopathology Five model (Harkness & McNulty, 1994). Archival data were used from 2 clinical settings: 1 outpatient community mental health center and 1 inpatient Veteran's Affairs hospital. The outpatient sample was mostly female (58.9%), White (78.7%), and had a mean age of 33.01 (SD = 10.26). The inpatient sample was entirely male, predominantly White (91.5%), and had a mean age of 48.03 (SD = 13.88). Correlations were conducted to identify the Minnesota Multiphasic Personality Inventory–2–Restructured Form Personality Psychopathology Five scales (Ben-Porath & Tellegen, 2008; Harkness et al., 2013) that were significantly related to variables measuring substance use. Subsequent regression analyses revealed that Disconstraint was the only significant predictor in the majority (89%) of analyses. In the event Disconstraint was not the only significant predictor, it continued to exhibit the largest unique predictive impact in the regression models. These results suggest that Disconstraint is the Personality Psychopathology-Five dimension most relevant to substance use.

Acknowledgments

The authors would like to thank Allan R. Harkness and the Personality Assessment Laboratory at the University of Tulsa for general guidance and professional support in preparation of the article. Additionally, thanks are owed to Yossef S. Ben-Porath, Paul A. Arbisi, and John R. Graham for allowing access to the data used in this study. A previous version of this study was presented as a paper in May 2014 at the 49th Annual MMPI–2/MMPI–2–RF/MMPI–A Symposium in Scottsdale, AZ.

Notes

¹ w is described by Cohen (1988) as a measure of effect size for chi-square analyses and interpretive guidelines suggest a small effect is w = .10, medium is w = .30, and large is w = .50.

² Past 6 months heroin abuse was excluded from subsequent analyses as only one participant was rated as having abused this substance in the past 6 months.

³ The ϕ² coefficient was used for the logistic regression effect size instead of w, which was used earlier in comparing the valid profiles to the invalid profiles (and those without an MMPI–2 for the outpatient sample). This is because ϕ is only appropriate for use with χ² of a 2 × 2 design, whereas w can be used with contingency tables of greater dimensions (as is the case with many of the preliminary analyses such as race in the inpatient VA sample). The ϕ² coefficient was also chosen to describe the logistic regression models as interpretation of the magnitude for ϕ² is the same as R² (Cohen, 1988).

⁴ The results for the analyses presented in Tables 5 and 6 are similar to analyses reported in the MMPI–2–RF technical manual by gender (Tellegen & Ben-Porath, 2008). The exclusion criteria used to build the samples reported in the technical manual differed from those used in this study. This resulted in slightly different sample sizes. For example, the analyses of the inpatient VA sample in this study differ from those in Table A.36 in Tellegen and Ben-Porath (2008) with at least 207 fewer sample participants in this study.