Introduction

Drug use and drug abuse are increasingly common terms in modern society, generating concerns and demands in the area of public health (^{Merz, 2018}). The pathological extreme of drug use concerns drug addiction. According to the World Health Organization (WHO, 2015), substance dependence refers to behavioral, cognitive and physiological phenomena that are caused by the repeated use of substances, including a desire for use, difficulties in controlling it even when facing the consequences, a higher priority given to drug use than to other activities and obligations, tolerance, and physical abstinence. The etiology of substance dependence is related to social (e.g., socioeconomic status), biological (e.g., genetic), interpersonal relationships (e.g., family conflicts), and psychological characteristics (e.g., personality traits) (^{Frisher, Crome, Macleod, Bloor, & Hickman, 2007}; ^{Goode, 2007}; ^{Nakhaee & Jadidi, 2009}).

The literature indicates specific personality traits typically increased in people diagnosed with substance dependence (^{Littlefield & Sher, 2016}). For instance, people with this diagnosis are prone to present high levels of hopelessness, depression, superiority, dominance, callousness, anxiety, impulsivity, and manipulation traits (^{APA, 2013}; ^{Bornovalova, Lejuez, Daughters, Rosenthal, & Lynch, 2005}; ^{Kotov et al., 2017}). Among these traits, impulsivity and anxiety seem to be the most commonly observed pathological traits in people with substance dependence (^{Chapman & Cellucci, 2007}; ^{Feingold, Weiser, Rehm, & Lev-Ran, 2016}; ^{Grant et al., 2004}; ^{Kotov, Gamez, Schmidt, & Watson, 2010}; ^{Loree, Lundahl, & Ledgerwood, 2015}; ^{Marín-Navarrete et al., 2018}; ^{Verdejo-García, Lawrence, & Clark, 2008}).

Assessment measures that evaluate pathological traits should encompass traits typically identified in people with substance dependence. The Clinical Dimensional Personality Inventory (^{Carvalho & Primi, 2015}), currently in its second version (IDCP-2; Carvalho & Primi, in press) and the focus of this study, is a self-report scale for assessing pathological traits, composed of 206 items that must be answered on a four-point Likert scale. The scale was developed to be used in the clinical context, and its recommended for establishing the personality profile of the patient (^{Carvalho, 2019}). The IDCP-2 is composed of 12 dimensions, namely: dependence, aggressiveness, mood instability, eccentricity, need for attention, grandiosity, distrust, isolation, criticism avoidance, self-sacrifice, conscientiousness, and inconsequence, which are divided into 47 factors. The psychometric properties of the factors of these dimensions suggested the suitability of the scale for assessing pathological traits (e.g., ^{Carvalho, Sette & Ferrari, 2016}; ^{Carvalho & Sette, 2017}; ^{Carvalho & Silva, 2016}).

Although studies have been conducted to investigate the discriminant ability of IDCP-2 for specific conditions (e.g., borderline personality disorder, dependent personality disorder, bipolar disorder; ^{Carvalho & Pianowski, 2019}a; ^2019b; ^{Carvalho, Painwoski, & Gonçalves, 2018}), no studies were found investigating its discriminant capacity for people with substance dependence. However, based on the literature (^{APA, 2013}; ^{Bornovalova et al., 2005}; ^{Kotov et al., 2017}; ^{Moraleda-Barreno, Díaz-Batanero, Pérez-Moreno, Gómez-Bujedo, & Lozano, 2018}), it is possible to identify the conceptual relevant IDCP-2 factors for this purpose: Anxious Concern, Hopelessness, Superiority, Dominance, Indifference, Anxiety, Depression, Risk-taking, Deceitfulness, and Impulsivity. Although these factors lack empirical evidence on substance dependence, previous studies found sound psychometric properties for them (^{Carvalho, 2018}; ^{Carvalho & Sette, 2017}; ²⁰¹⁵; ^{Carvalho et al., 2016}; Carvalho & Silva, 2016).

This study aimed to investigate the capability of IDCP-2 factors to identify people with substance dependence, as well as to examine their discriminant capacity in comparison to the Personality Inventory for DSM - 5 (PID-5; ^{Krueger, Derringer, Markon, Watson & Skodol, 2011}), a well-recognized scale in the field.

Three hypotheses were formulated : H1) the IDCP-2 factors should be able to discriminate the participants according to the level of substance dependence, especially the extreme groups (i.e., clinical group and general population group); H2) the factors representing the impulsiveness and anxiety traits should be the most discriminant (^{Bornovalova et al., 2005}; ^{Loree et al., 2015}; ^{Verdejo-García et al., 2008}); and H3) the IDCP-2 factors should present a discriminant capacity for substance dependence superior or similar to the facets of PID-5.

Method

Results

The assumptions for conducting ANOVA, i.e., independent samples, the similar variance between groups, and distribution close to normal were achieved, according to the criteria presented in the literature (^{George & Mallery, 2010}). The Hopelessness factor presented kurtosis slightly higher than the threshold (kurtosis = 1.25). Levene's test indicated that most of the factors in this study did not show homogeneity of variance. Only Anxious Concern, Dominance, Indifference, and Anxiety factors showed homogeneity of variance. Empirical evidence indicates that, even in these cases, it is possible to use ANOVA (^{Blanca Mena et al., 2017}). However, it was chosen to use Welch's ANOVA, a variation of the classic ANOVA, that can be employed for cases in which homogeneity is violated (^{Moder, 2007}; ²⁰¹⁰).

Analyzes of variance were performed to verify the differences between the sample groups in the IDCP-2 factors. All factors showed statistically significant differences: Anxious Worry [F (2, 67) = 16.448; p < .001], Hopelessness [F (2, 61) = 18.330; p < .001], Superiority [F (2, 59) = 9,200; p < .001], Dominance [F (2.63) = 4.704; p = .01], Indifference [F (2, 63) = 10,198; p < .001], Anxiety [F (2.69) = 11.631; p < .001], Depression [F (2, 65) = 7.838; p < .001], Risk-taking [F (2, 57) = 27.358;p < .001], and Deceitfulness [F (2.61) = 20.003; p < .001]. In order to find out which groups differed, Table 2 presents Tukey's post hoc analysis.

The groups were discriminated with the Anxious Worry, Risk-taking, and Deceitfulness factors, with the clinical group presenting higher means and the general population group with lower means. The factors Hopelessness and Anxiety were able to discriminate the clinical and subclinical group from the general population. The factors Superiority and Indifference differentiated the clinical group, which had higher means, the subclinical group, and the general population. In the Dominance and Depression factors, two groups were separated, general population and clinical groups, with the latter presenting higher means.

In order to verify the predictive capacity of the IDCP-2 factors for substance dependence, logistic regression was employed with only the general population (n = 110) and clinical (n = 24) groups, as can be seen in Table 3. In the first model, the IDCP-2 factors were used as independent variables and a dichotomous variable, presence or no presence of substance dependence, was the dependent variable. A second regression model was conducted only with the independent variables that presented a significant single contribution in the first model.

Model 1 was able to predict 77% of the group variable. The variables with statistically significant contributions were Sex and the Hopelessness factor of IDCP-2. This model was able to correctly identify 82.6% of people in the clinical group and 100% of people in the general population group. Model 2, including variables with a single contribution in the first model, was able to predict 56% of the group variable. This model was able to identify 58.3% of the clinical group and 97.3% of the non-clinical group, indicating that the first model provides information that improves the ability to identify the groups. Table 4 presents the logistic regression analysis with the PID-5.

The model with the PID-5 facets as independent variables was able to predict 76% of the group variable, similar to what was observed with the IDCP-2 factors. In this model, sex was also significant, in addition to the Anxiety facet. This model was able to correctly identify 78.3% of people in the clinical group, and 100% of people in the general population group. Model 2, including only variables significant in the first model, was able to predict 50% of the group variable. This model was able to correctly identify 54.2% of people in the clinical group and 95.5% in the non-clinical group.

Discussion

This study aimed to verify the contribution of the IDCP-2 self-report scale in the identification of people with substance dependence and to compare their predictive capacity with PID-5. The findings of this study confirmed the initial hypotheses: H1) the IDCP-2 factors were capable of differentiating the groups, especially the extreme groups, i.e., general population and clinical group; H2) the Risk-taking factor, representing the impulsiveness trait, and the two factors related to anxiety (Anxious Worry and Anxiety factors) were those with the best performance in discriminating groups; and, H3) similar results (r 2) were observed between IDCP-2 and PID-5 in the predictive capacity of substance dependence. The hypotheses are discussed in the following paragraphs.

The IDCP-2 factors were able to discriminate between the general population and the clinical group. These findings are in agreement with the literature that lists hopelessness, superiority, dominance, indifference, anxiety, depression, impulsivity, and manipulation as characteristics of drug addicts (APA, 2013; ^{Kotov et al., 2017}; ^{Moraleda-Barreno et al., 2018}). It was also observed that the Risk-taking factor, representing the impulsiveness trait in this study, and the factors Anxiety and Anxious Worry, showed greater differences between the general population and clinical groups. This finding confirms previous evidence indicating that Impulsiveness and Anxiety traits as the most relevant for discriminating people with substance dependence (^{Belcher et al., 2014}; ^{Ersche, Turton, Pradhan, Bullmore, & Robbins, 2010}; ^{Grant et al., 2004}; ^{Lai, Cleary, Sitharthan, & Hunt, 2015}; ^{Mitchell & Potenza, 2014}; ^{Moraleda-Barreno et al., 2018};). These results confirm the H1 and H2 of our study.

As for the third hypothesis of this study (H3), the predictive capacity for substance dependence on the IDCP-2 factors was similar to the capacity of the PID-5 facets. The findings suggest that the scores of IDCP-2 and PID-5 are capable of predicting, respectively, 77% and 76% of the sample groups. In IDCP-2, the only significant factor in the model was Hopelessness. This factor concerns feelings of sadness, discouragement, and suicidal thoughts (Carvalho & Primi, in press), representing traits frequently identified in drug addicts (^{Baines, Jones, & Christiansen, 2016}; ^{Karamat & Ahmed, 2015}). Although coherent, it was expected that other factors of IDCP-2 would show discriminant capacity with a similar or higher weight to the Hopelessness factor. Future studies should verify whether the findings of the present research are replicated. The PID-5 presented only the Anxiety facet as significant in the regression analysis. This facet represents the exaggerated concern about the possibility of uncomfortable events in the future that may or may not happen (^{Krueger et al., 2011}). Studies have found that groups of drug addicts have higher levels of anxiety compared to the general population (^{Belcher et al., 2014}, ^{Grant et al., 2004}; ^{Lai et al., 2015}).

The findings of this study are preliminary evidence indicating that the IDCP-2 factors are capable of identifying people with substance dependence. It was observed that Risk-taking, Anxious Worry, and Anxiety factors were the best for discriminating people with substance dependence from individuals who reported not having this diagnosis. These factors were also discriminant to correctly identify people without a diagnosis of substance dependence but were characterized as a risk group by the AUDIT and ASSIST scales.

The findings should be weighted according to the major methodological limitations of this study. First, the sample size of the groups was very distinct, the distribution of demographic variables was uneven (e.g., sex), and information was not accessible on the most used type of chemical substance and the form of administration (e.g., injectable, inhaled). Second, most of the clinical sample was being treated at a rehabilitation clinic, which may have mitigated responses to the scales. Third, attention is called on the use of the Risk-taking factor of IDCP-2, as representative of the impulsivity trait. Future studies should investigate the application of the IDCP-2 Impulsivity factor for this purpose. Although the findings of this study are promising, they should be considered as preliminary to the use of specific factors from IDCP-2 to identify people with substance dependence. Future studies must verify whether these findings are replicated in samples, especially designs with larger clinical samples