Psy 475_Week 2_Psychological Measure Paper

Journal of Substance Abuse Treatment 37 (2009) 25–31

Regular article

Beck Depression Inventory for depression screening in substance-abusing adolescents

Geetha Subramaniam, (M.D.)a,⁎, Paul Harrell, (M.A.)b, Edward Huntley, (M.A.)b, Melinda Tracy, (M.S.W.)a

aDepartment of Psychiatry, Johns Hopkins University, Baltimore, MD 21287, USA bDepartment of Psychology, American University, Washington, DC 20016, USA

Received 2 May 2008; received in revised form 10 September 2008; accepted 19 September 2008

Abstract

Co-occurring major depressive disorder (MDD) in adolescents with substance use disorders (SUD) has been linked to poor treatment outcomes. Use of validated depression screens in adolescent SUD populations may improve the detection of depression. In this study, we evaluated the diagnostic efficiency of the Beck Depression Inventory (BDI) in detecting MDD, as assessed by psychiatrists administering the Diagnostic Interview for Children and Adolescents, and its factor structure, internal consistency, and discriminant validity in a clinical sample of adolescents with SUD (n = 145). Results indicate that BDI scores of 12 and higher had the most optimal sensitivity (73%), whereas BDI scores of 17 and higher, the most optimal specificity (75%). Five factors accounted for approximately 56% of the variance. Overall, internal consistency was high, and the BDI adequately discriminated MDD from non-MDD cases. Results support the use of BDI as a screen for MDD with moderate to high psychometric properties in an adolescent SUD sample. © 2009 Elsevier Inc. All rights reserved.

Keywords: Depressive symptoms; Adolescent substance abuse; Depression screening

1. Introduction

Adolescent depression is a treatable psychiatric disorder with serious consequences when untreated (American Academy of Child and Adolescent Psychiatry, 1998; Brent & Birmaher, 2002). Of clinical samples of adolescents with substance use disorders (SUD), 15% to 50% present with co- occurring depressive disorders (Bukstein, Glancy, & Kami- ner, 1992; Clark et al., 1997; Deykin, Buka, & Zeena, 1992; Kashani, D'Souza, Reid, & Neal, 1985; Riggs, Baker, Mikulich, Young, & Crowley, 1995). When depression is present at the time of entry to substance abuse treatment, it has been linked to poorer posttreatment outcomes (Cornelius et al., 2004; Hasin, Nunes, & Meydan, 2004). Based on these

⁎ Corresponding author. Department of Psychiatry, Johns Hopkins University, C/O Mountain Manor Treatment Center, 3800 Frederick Ave, Baltimore, MD 21229, USA. Tel.: +1 410 233 1400; fax: +1 410 233 1666.

E-mail address: [email protected] (G. Subramaniam).

0740-5472/08/$ – see front matter © 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.jsat.2008.09.008

features, published treatment guidelines (American Acad- emy of Child and Adolescent Psychiatry, 2004; Center for Substance Abuse Treatment, 2005) recommend that sub- stance treatment programs identify and manage co-occurring psychiatric disorders. In practice, however, very few programs routinely assess for psychiatric disorders such as depression because of limited resources. For example, psychiatric diagnoses are obtained after lengthy clinical interviews performed either by a psychiatrist or by a licensed mental health provider, which these programs often cannot afford. It may be useful to have validated self-reported psychiatric screens that can supplement a counselor's intake assessment, which can lead to identification of cases that can be referred for the management of their co-occurring disorders. Although a number of screening instruments exist for depression, few, if any, have been tested among adolescent samples with SUD. Psychometric properties for assessment tools in this population may be unique, particularly because they may have depressive disorders of mixed origins (substance-related, preexisting, or even

26 G. Subramaniam et al. / Journal of Substance Abuse Treatment 37 (2009) 25–31

coexisting mood states) and because vegetative symptoms of depression can appear similar to symptoms of substance use and withdrawal. It is especially important to investigate this in adolescents because substance use and depression are associated and predictive over time (Chinet et al., 2006). Therefore, it is important to establish the use of psychometric tools developed for adults in adolescent populations.

The Beck Depression Inventory (BDI) is one of the most commonly used self-reported screens for major depressive disorder (MDD) and has been well validated with well- established psychometric properties (Beck, Steer, & Gabin, 1988). It was found to be highly sensitive (i.e., ability to correctly identify those with depression, range = 68%–87%) and specific (i.e., the ability to identify those without depression, range = 70%–82%) in detecting MDD among adolescents diagnosed as using standardized structured psychiatric interviews such as the Kiddie Schedule for Affective Disorders and Schizophrenia (Ambrosini, Metz, Bianchi, Rabinovich, & Undie, 1991; Bennett et al., 1997; Kaufman, Birmaher, Brent, Rao, & Ryan, 1996) and the Diagnostic Interview of Children and Adolescents (Marton, Churchard, Kutcher, & Korenblum, 1991; Reich, 2000). The BDI has not yet been examined for its diagnostic use among adolescents with SUD. However, data exist for substance- dependent adult patients where studies have shown that the BDI offered the best validity as a screening tool for depression when compared to other depression screens (Rounsaville, Weissman, Rosenberger, Wilber, & Kleber, 1979; Weiss, Griffin, & Mirin, 1989). In these samples, the sensitivity ranged from 65% to 100% for cutoff scores of BDI ≥11–15, whereas for the same cutoff scores, the specificity was much lower ranging from 39% to 61%. The BDI also merits further evaluation because in one study, BDI scores ≥11 robustly predicted postresidential substance use outcomes across 3, 6, 9, and 12 months in an adolescent sample (Subramaniam, Stitzer, Clemmey, Kolodner, & Fishman, 2007), suggesting the role of affective distress negatively impacting outcomes.

The BDI is psychometrically sound and has been shown to have high levels of internal consistency in psychiatric populations (coefficient alphas ranging from .76 to .95, M = .86). Three to seven factor solutions have been proposed for the BDI (Beck et al., 1988), with four factors (negative self- attitude, performance difficulty, somatic symptoms, and physical worry) being able to discriminate between depressed and nondepressed adolescents (Bennett et al., 1997). However, these properties of the BDI have not been evaluated in adolescents with SUD. It is possible that the factor structure may be unique in this population. Therefore, it is necessary to identify and characterize the latent variables underlying the depressive symptomatology to use the instrument efficiently and to further understand the under- lying nature of depression in this population. The objectives of this study are as follows: (a) to assess the diagnostic efficiency of the BDI in detecting MDD in a treatment- seeking sample of adolescents with SUD and (b) to assess the

BDI for internal consistency, factor structure, and discrimi- nant validity in this sample.

2. Methods

2.1. Participants

This is a secondary analysis of data obtained from a cross- sectional study designed to compare the clinical characteristics of treatment-seeking adolescents with opioid use disorder (OUD) with a matched sample of adolescents with cannabis/ alcohol use disorders (Subramaniam, Stitzer, Woody, Fish- man, & Kolodner, 2009). For this study, 184 participants with either OUD or cannabis/alcohol use disorders were recruited from patients aged 14 to 18 years seeking either residential or outpatient treatment at an adolescent substance abuse treatment program in Baltimore, MD.

2.2. Procedures

A total of 184 participants and their guardians provided assent/informed consent if they met all study eligibility criteria including less than 2 weeks of abstinence or confinement at time of study entry. This criterion was included to reduce variability in rates of depressive symptoms in relation to duration of abstinence. All participants were assessed, typically within 2 weeks of treatment entry, using a demographic instrument, structured interviews Diagnostic Instrument for Children and Adoles- cents-IV (DICA-IV) for psychiatric disorders and Composite International Diagnosis Interview for Diagnostic and Statistical Manual of Mental Disorders (DSM) SUD and the BDI (instruments described below). Typically, the patients completed the battery of assessments in two sessions and were paid $25 for their time and effort. Further details of the parent study methods are described in Subramaniam et al. (2009). The study protocol, instruments, and consent forms were approved by the Western Institutional Review Board (a Johns Hopkins Institutional Review Board designee).

2.3. Study Instruments

2.3.1. Diagnostic Instrument for Children and Adolescents-IV This structured psychiatric interview for children/adoles-

cents (Reich, 2000; Welner, Reich, Herjanic, Jung, & Amado, 1987) provided information on several lifetime and current DSM-IV Axis I diagnoses. We selected seven current psychiatric disorders commonly reported in the adolescent SUD literature (Hovens, Cantwell, & Kiriakos, 1994; Kandel et al., 1997; Stowell & Estroff, 1992) for assessment: attention- deficit/hyperactivity disorder (ADHD), major depressive episode (MDE), manic episode, generalized anxiety disorder (GAD), posttraumatic stress disorder (PTSD), oppositional defiant disorder (ODD), and conduct disorder (CD). The first author (G. S.) administered 70% of DICA interviews; the

27G. Subramaniam et al. / Journal of Substance Abuse Treatment 37 (2009) 25–31

others were administered by two psychiatrists who were trained by the first author (G. S.) to 100% agreement on observed interviews prior to beginning the study.

2.3.2. Beck Depression Inventory This 21-item self-report measure assessed depressive

symptoms in the past 7 days (Beck et al., 1988). Each item is scored from 0 to 3 with a maximum score of 63. Participants self-administered the BDI under the supervision of a research staff who read the questions aloud for those who requested help or had difficulty reading (at the time of initial informed consent process).

2.3.3. Composite International Diagnostic Interview The substance abuse module of this structured interview

was used to obtain current DSM-IV diagnoses of SUD (i.e., substance dependence/abuse) for 9 substance categories: opioids, cannabis, alcohol, cocaine, sedative, other stimu- lants, phencyclidine, inhalants, and any other substance (Cottler, Robins, & Helzer, 1989). It was administered by trained research staff.

2.4. Statistical analyses

2.4.1. Diagnostic efficiency Area under the curve (AUC) using the receiver operating

curve (ROC) function was calculated for BDI total scores for the total sample and separately for males and females.

The validity of the BDI in detecting current MDD is assessed by calculating the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), for a range of cutoff scores on the BDI as follows:

MDD

BDI scores

Yes

No

≥ cutoff

True pos

False pos

b cutoff

False neg

True neg

Sensitivity = true pos / true pos + false neg (i.e., accurate identification of those with MDD). Specificity = true neg / true neg + false pos (i.e., accurate identification of those with no MDD). Positive predictive value = true pos / true pos + false pos (i.e., accurate identification of MDD among screen positives). Negative predictive value = true neg / true neg + false neg (i.e., accurate identifications of those with no MDD among screen negatives).

Existing literature in adolescent psychiatric samples suggested cutoff scores of BDI ≥11–13 for optimal sensitivity and BDI ≥16 for specificity. We examined these and other cutoff scores in detecting MDD, based on the sample ROC characteristics. All statistical analyses were performed using SPSS version 15 (SPSS, 2006).

2.4.2. Internal consistency and factor structure analyses Validity measures were calculated using crosstabs, and

internal consistency of the BDI in this sample was measured by calculating the Cronbach's alpha (Cronbach, Gleser,

Nanda, & Rajaratnam, 1971). BDI's factor structure was determined by conducting a principal components analysis. Factors were retained if they had an Eigenvalue greater than 1.0, accounted for at least 5% of the variance, and were acceptable by scree plot criteria of Cattell (1966). The Kaiser– Meyer–Olkin measure (0.86) indicated that the sample was adequate for conducting a principal components analysis. Both orthogonal (Varimax) and oblique rotation (Oblimin) methods were used to maximize loading on the factors. Because both methods provided relatively similar results and the orthogonal rotation is considered to be the more parsimonious method of the two (Thompson, 2004), the results of orthogonal rotation are presented. To examine whether each factor was useful in discriminating depressed from nondepressed cases, we conducted analyses of variance.

3. Results

3.1. Sample characteristics

A total of 184 participants were recruited, of which 145 with no missing BDI or DICA data were retained for this study. Participants included (n = 145) were not different from those who were not (n = 39) on age, gender, race, school or graduated status, residential level of treatment, or court- ordered to treatment status. The study sample was mostly male (58%) and Caucasian (73%); 75% were in residential level of treatment, and 22% were court-ordered to treatment. The distribution of past-year DSM-IV SUD diagnoses were as follows: cannabis 71%, alcohol 58%, opioid 56%, cocaine 41%, sedative 24%; other stimulant, phencyclidine, halluci- nogen, and inhalant use disorders were less than 10% each; 52% had three or more SUD diagnoses.

This sample had high rates of co-occurring disorders, with 82% meeting criteria for any Axis I psychiatric disorder and 51% with a current Axis I MDE diagnosis. The prevalence of other current DSMI-IV Axis I psychiatric disorders were as follows: 52% had CD, 36% had GAD, 35% had ADHD, 20% had manic/hypomanic episode, and 18% had ODD. The mean BDI score for the study sample was 16.0 (SD = 9.9); 65% had BDI scores ≥11, and 47% reported BDI scores ≥16.

Table 1 displays characteristics of those with and without a current MDD diagnosis. Females versus males (66% vs. 39%, p = .002) and Caucasians versus African Americans (55% vs. 39%, approaching significance, p = .072) were more likely to have current MDD. Having a diagnosis of MDD was associated with having any DSM Axis I psychiatric disorder and each of the other psychiatric diagnoses assessed. The groups did not differ on rates of DSM SUD diagnoses or duration of current abstinence. The MDD group was also more likely to report current life stressors and be admitted on at least one psychotropic medication. As expected, the MDD group had higher mean BDI scores (18.3 vs. 13.6, p b .01) and greater proportion with BDI scores ≥11 and ≥16.

Table 1 Characteristics of substance-abusing adolescents with and without MDD

Characteristics MDD (n = 73) No MDD (n = 72) χ2/t value

Mean age (SD) 16.8 (1.1) 17.0 (1.0) −1.56 % Caucasian race 79.5 66.2 3.22 % African American 20.5 33.8 3.22 % Female 54.8 29.2 9.77 ⁎⁎

% city residence 39.4 35.8 0.19 % in residential treatment 72.6 76.4 0.27 % in school or graduated 56.5 58.3 0.05 Psychiatric Mean BDI scores (SD) 18.3 (9.2) 13.6 (10.0) 2.95 ⁎⁎

% BDI ≥11 76.7 52.8 9.11 ⁎⁎ % BDI ≥16 63.0 30.6 15.34 ⁎⁎⁎ % any DSM-IV psychiatric disorder 100.0 62.3 33.26 ⁎⁎⁎

% DSM-IV CD 63.0 41.4 6.68 ⁎

% ADHD 47.9 21.7 10.68 ⁎⁎

% mania 30.1 10.0 8.96 ⁎⁎

% GAD 56.2 14.3 27.31 ⁎⁎⁎

% PTSD 44.4 17.1 12.37 ⁎⁎

% with past suicide attempt 29.9 11.1 7.58 ⁎⁎

% reporting current life stressors 79.1 37.5 24.59 ⁎⁎⁎

% currently on psychotropic medications 46.6 25.0 7.34 ⁎⁎

Substance abuse % DSM-IV cannabis use disorder 69.9 72.2 0.10 % DSM-IV alcohol use disorder 56.2 59.2 0.19 % DSM-IV OUD 58.9 52.8 0.55 % DSM-IV cocaine use disorder 37.1 45.1 0.92 Mean current abstinence in days (SD) 8.8 (8.9) 6.9 (4.2) 1.62

⁎ Indicates p ≤ .05. ⁎⁎ Indicates p ≤ .01. ⁎⁎⁎ Indicates p ≤ .001.

Table 2 Sensitivity, Specificity, and PPV, and NPVof a range of BDI cutoff scores in detecting MDD

BDI cutoff Sensitivity (%) Specificity (%) PPV (%) NPV (%)

≥11 a 77 47 60 67 ≥12 73 56 62 67 ≥13 67 58 62 64 ≥14 66 63 64 64 ≥15 66 65 66 65 ≥16 63 69 68 65 ≥17 60 75 71 65 ≥18 56 75 70 63 ≥19 55 75 69 62

a BDI ≥11 retained in table because it was reported as having the most optimal sensitivity in the psychiatric literature.

28 G. Subramaniam et al. / Journal of Substance Abuse Treatment 37 (2009) 25–31

3.2. Diagnostic efficiency

The AUC obtained from the ROC was of moderate size (0.67, p = .001) in detecting current MDD diagnosis in the total sample; AUC for males was 0.69 (p = .003) and for females was 0.55 (p = .544).

3.2.1. Validity Table 2 shows the results of validity measures (sensitivity,

specificity, PPVs, and NPVs for a range of cutoff scores). We present only the sensitivity/specificity values exceeding 50% (i.e., greater than chance) in detecting positive or negative cases. (BDI ≥11 was retained because it had been reported in the literature to have optimal sensitivity). In the range of cutoff scores examined, sensitivity ranged from 55% to 77%, specificity from 56% to 75%, PPV from 62% to 71%, and NPV from 62% to 67%. In our sample, BDI ≥12 had the most optimal sensitivity (73%) with specificity of 56%. Optimal specificity in this sample was seen for BDI ≥17 (75%) with a corresponding sensitivity of 60%. Similarly, BDI ≥12 and BDI ≥17 also had the highest NPV and PPV (67% and 71%, respectively) in this range. Although scores less than 12 and more than 17 had either higher sensitivity or higher specificity values, respectively, they were not optimal in relation to other validity measures.

We conducted post hoc analyses to determine the demographic and clinical characteristics of those who scored

BDI ≥11, BDI ≥12, BDI ≥6, and BDI ≥17 (cutoff selected based on reports in the literature and most optimal in this dataset). Female gender, higher mean BDI scores, having a current MDD diagnosis, and history of past suicide attempt were consistently and significantly associated across each of the cutoff scores. On the other hand, mean age, Baltimore city residence, percentage in residential level of care, percentage in school or graduated, percentage reporting current life stressors, percentage currently being prescribed at least one psychotropic medication (latter approaching statistical significance for BDI ≥17), DSM cannabis,

Fig. 1. Average factor score (±SE) of no MDD (n = 72) and MDD diagnosed adolescents (n = 73).

29G. Subramaniam et al. / Journal of Substance Abuse Treatment 37 (2009) 25–31

alcohol, cocaine, OUDs (latter approaching statistical significance for BDI ≥ 17), and duration of current abstinence were not associated with any of the cutoff scores (results not shown in tables).

3.3. Factor analyses, discriminant validity, and internal consistency of the BDI

Our data revealed a five-factor solution (displayed in Table 3), which together accounted for approximately 60.63% of the total variance, that is, 60.63% of the common variance among the 21 items could be accounted for by the five factors. Item 21, “loss of interest in sex,” failed to load onto a single coherent construct and therefore was not included as a factor but is displayed for reference.

Internal consistency was high overall for the BDI (Cronbach's α = .87), similar to studies cited above. For the five factors, reliability ranged from low for Pessimism (α = .49) to moderate for Nutrition (α = .66), Negative Affect (α = .65), and Negative Self-perception (α = .63) and high for Negative Cognition (α = .85).

The mean scale scores between groups were signifi- cantly different for factors Negative Cognition (1), Negative Affect (2), and Negative Self-perception (4) with effect sizes ranging from small (d = 0.28) for

Table 3 Factor loadings from the rotated component matrix: principal component analysis

1 2 3 4 5 6

Factor 1: Negative Cognition, 30.4% of variance 8. Blame 0.743 0.048 0.161 0.258 0.037 0.001 4. Satisfaction 0.689 0.156 0.102 0.009 0.052 0.265 3. Failure 0.666 0.136 0.108 0.358 0.157 0.164 13. Decision 0.662 0.122 0.021 0.272 0.114 0.072 17. Tired 0.562 0.065 0.001 0.007 0.269 0.343 15. Effort 0.530 0.281 0.218 0.199 0.249 0.209 7. Myself 0.509 0.224 0.181 0.265 0.381 0.134 1. Sad 0.498 0.410 0.049 0.349 0.082 0.029 2. Future 0.455 0.318 0.122 0.052 0.240 0.300 Factor 2: Negative Affect, 7.4% of variance 10. Cry 0.151 0.676 0.070 0.422 0.068 0.088 6. Punished 0.220 0.676 0.263 0.025 0.220 0.200 16. Sleep 0.259 0.573 0.423 0.130 0.074 0.273 12. Interest 0.388 0.552 0.144 0.068 0.160 0.168 11. Irritated 0.220 0.444 0.087 0.177 0.188 0.039 Factor 3: Nutrition, 6.2% of variance 18. Appetite 0.118 0.117 0.834 0.123 0.097 0.001 19. Weight 0.134 0.020 0.778 0.082 0.156 0.005 Factor 4: Negative Self-perception, 5.9% of variance 14. Looks 0.244 0.098 0.028 0.720 0.069 0.198 5. Guilt 0.363 0.221 0.217 0.619 0.069 0.063 Factor 5: Pessimism, 5.8% of variance 20. Worry 0.122 0.168 0.075 0.069 0.842 0.032 9. Suicide 0.131 0.094 0.238 0.369 0.632 0.083 21. Sex 0.175 0.075 0.029 0.146 0.093 0.816

Note. Extraction method: principal component analysis. Oblique rotation method: Oblimin with Kaiser normalization.

Negative Self-perception to moderate (d = 0.51) for Negative Cognition (Fig. 1).

4. Discussion/Conclusions

This is the first study to document the diagnostic utility of an intake BDI in a clinical sample of U.S. adolescents presenting with a variety of SUD. The BDI was found to be useful as a self-reported depression screen for adolescents with SUD. It showed relatively adequate sensitivity/specifi- city values comparable, albeit less robust, to the cutoff scores reported in adolescent psychiatric samples (Marton et al., 1991). We speculate that this is due to the heterogeneity in psychiatric diagnoses in our sample. Having a MDE diagnosis was significantly associated with having another psychiatric disorder, and the MDE in some cases could have been substance-related. In this sample of adolescents with SUD, a cutoff score of BDI ≥12 was more optimal as a clinical screening tool in identifying more than 70% of those with diagnoses of MDD while reducing the likelihood of missing those with a true MDD in less than 50% of the cases. The BDI was also moderately useful in predicting correctly those with MDD among those who screened positive (PPV), with PPV highest for BDI cutoffs of ≥16 and ≥17. In select circumstances that require optimal specificity, for example, determining eligibility for research interventions or to conserve resources, BDI cutoff of ≥17 would be optimal. It correctly identified 75% of those without a diagnosis of MDD (accurately excluding those who do not meet criteria) while identifying those with MDD in 60% of the cases. The BDI showed moderate discrimination between those with and without MDE as shown by an AUC 0.67 for this sample. The higher AUC for males (0.69) and the lower and not significant AUC for females (0.55) are most likely due to two thirds of females meeting having a current MDE.

Higher scores on any of the cutoffs were associated with female gender, having a current MDD diagnosis and a past history of suicide attempt suggest a risk for affective illness and a female gender vulnerability. The higher scores on lower cutoffs were more likely to be associated with having a

30 G. Subramaniam et al. / Journal of Substance Abuse Treatment 37 (2009) 25–31

cannabis use disorder and any Axis I psychiatric disorder, perhaps accounted for largely by CD suggesting the possibility of withdrawal phenomenon, global distress, or demoralization. Those who scored higher than BDI ≥16 and 17 were more likely to have anxiety disorders (GAD and PTSD). There was also a trend toward higher rates of being prescribed a psychotropic medication, which may suggest a more selective and severe mixed affective state requiring the use of psychotropic prescriptions, as reported in the literature (Marton et al., 1991). We were surprised to find that high scores on any of the cutoffs was not associated with most of the substance abuse variables examined including OUD, as reported before (Subramaniam et al., 2007). This may be because more than 50% of the sample had multiple (three or more) SUD diagnoses.

This study also extends the literature by providing data on psychometric properties of the BDI in adolescent substance- abusing populations. The BDI was found to have good internal consistency (Cronbach's α = 0.87) comparable to values reported in adolescent psychiatric samples (0.76– 0.95; Beck et al., 1988). Five BDI factors accounted for 60% of the variance, with one representing Negative Cognition (Blame, Satisfaction, Failure, Decision, Tired, Effort Myself, Sad, and Future) alone accounted for 30%. These factors were similar to those reported by Bennett et al. (1997) and Beck and Lester (1973). In addition, three of these factors: Negative Cognition, Negative Affect, and Negative Self- perception, correctly discriminated between those with and without MDD. Therefore, higher scores on these items (particularly Negative Cognition items) may warrant closer attention when screening for MDD. The two factors Nutrition (appetite, weight) and Pessimism (worry and suicidal ideation) did not discriminate between MDD and no-MDD and had lower internal consistency, suggesting that their clinical significance may be nonspecific and indepen- dent of MDD. However, it should be noted that this is an exploratory analysis, and the temporal stability of the measure is not yet established within this population. Thus, conducting a Confirmatory Factor Analysis with a new sample, as well as test–retest data, would be helpful in further validating the psychometric properties of the BDI in substance-abusing adolescents.

4.1. Limitations

The prevalence of MDD in this sample was high (51%) and represents the higher end of the 25% to 61% range prevalence of depressive disorders reported in the literature (Clark et al., 1997; Deykin et al., 1992; Stowell & Estroff, 1992). This was a diagnostically heterogeneous sample (drawn from residential and outpatient treatment settings; presenting with a variety of psychiatric and SUD; and in some cases mood disorders being substance induced while not in others). However, study data showing that the mean age of onset of diagnosis of MDD was approximately 2 years prior to the earliest mean age of onset of SUD diagnosis suggest that

for many the diagnoses of MDD was not substance induced. It is likely that diagnostic efficiency of BDI may be different if it was repeated after 3 or more weeks of abstinence because depressive symptoms remit, for many, after protracted abstinence (Subramaniam, Lewis, Stitzer, & Fishman, 2004).

A newer version of the BDI (Beck, Steer, & Brown, 1996) has been developed and found to have good psychometric properties among psychiatric samples of adolescents. In the revised BDI-II, symptoms are assessed for a period of 2 weeks instead of past week to correspond with DSM criteria for MDD; also, 4 new items (agitation, worthlessness, loss of energy, and concentration difficulty, respectively) have replaced 4 of 21 existing items (look ugly, cannot do any work, weight loss, and worried about physical problems, respectively). It is not known how these changes may impact psychometrics in adolescents with SUD, which warrants further examination. Likewise, it is important to test other depression screens, such as the Child Depression Inventory, to establish convergent validity and compare their efficiency in detecting cases with MDD. Our results may not generalize to other samples given the predominance of Caucasians and patients in a residential level of care. False-positive and false-negative results may have risen because of the differences in the time frames examined (i.e., past 7 days for BDI and past 30 days for MDE) and also because the BDI may have detected other mood disorders apart from MDE such as depressive disorder not otherwise specified, dysthymic disorders, adjustment disorder with depressed mood, and so on.

4.2. Clinical implications

This is the first study to demonstrate the use of the BDI as a valid self-reported screen for depression in a clinical sample of adolescents with SUD. Because it can be administered in 5 to 10 minutes, it may easily be added to existing intake assessments at substance abuse treatment settings without additional time burden on clinical staff. Cutoff (≥12) on BDI may be useful as a screen for identifying adolescents in clinical populations of adolescents with SUD. Those patients with scores higher than the cutoff of ≥16 or ≥17 may represent a higher risk psychiatric sample that may benefit from specialist psychiatric assess- ment because of the added risk of association with anxiety disorders and potential need for psychotropic medications. However, patients who endorse suicidal ideation, especially in the context of a plan or serious recent attempt, would warrant additional attention, independent of their total BDI scores. Therefore, we recommend retaining all 21 items of the BDI to comprehensively evaluate the extent of depressive symptomotology and guide clinical interventions.

Acknowledgments

We thank the National Institute of Drug Abuse (NIDA)— American Academy of Child and Adolescent Psychiatry

31G. Subramaniam et al. / Journal of Substance Abuse Treatment 37 (2009) 25–31

(AACAP) K-12 Physician Scientist Career Development Award for funding this research. We also thank Dr. Stitzer, Dr. dos-Reis, and Dr. Ostrander for their assistance in preparing this manuscript. Some of the findings in this study were presented as an abstract at the 2006 Annual Meeting of the American Academy of Child and Adolescent Psychiatry. Financial Support: NIDA grants: AACAP-K12 DA 000357 (Subramaniam, P.I.).

References

Ambrosini, P. J., Metz, C., Bianchi, M. D., Rabinovich, H., & Undie, A. (1991). Concurrent validity and psychometric properties of the Beck Depression Inventory in outpatient adolescents. Journal of the American Academy of Child and Adolescent Psychiatry, 30, 51−57.

American Academy of Child and Adolescent Psychiatry. (1998). Practice parameters for the assessment and treatment of children and adolescents with depressive disorders [Electronic Version]. Journal of the American Academy of Child and Adolescent Psychiatry, 37, 63S−83S Retrieved April 7, 2008 from http://www.jaacap.com/pt/re/jaacap/ abstract.00004583-199810001-00005.htm;jsessionid=LMybpFjvhvm- S7QL7vRsJRKcqLZVpPTnS5K1tRTKCQ0GR3T62F2KL! 670793751!181195628!8091!-1.

American Academy of Child and Adolescent Psychiatry. (2004). Practice parameters for the assessment and treatment of children and adolescents with depressive disorders [Electronic Version]. Journal of the American Academy of Child and Adolescent Psychiatry, 37 Retrieved April 7, 2008 from http://www.aacap.org/galleries/PracticeParameters/ substanceUseDisorder.pdf.

Beck, A. T., & Lester, D. (1973). Components of depression in attempted suicides. Journal of Psychology, 85, 257−260.

Beck, A. T., Steer, R., & Gabin, M. (1988). Psychometric properties of the BDI: Twenty-five years of evaluation. Clinical Psychological Review, 8, 77−100.

Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual for the Beck Depression Inventory-II. San Antonio, TX: Psychological Corporation.

Bennett, D. S., Ambrosini, P. J., Bianchi, M. D., Barnett, D., Metz, C., & Rabinovich, H. (1997). Relationship of Beck Depression Inventory factors to depression among adolescents. Journal of Affective Disorders, 45, 127−134.

Brent, D. A., & Birmaher, B. (2002). Clinical practice. Adolescent depression. New England Journal of Medicine, 347, 667−671.

Bukstein, O. G., Glancy, L. J., & Kaminer, Y. (1992). Patterns of affective comorbidity in a clinical population of dually diagnosed adolescent substance abusers. Journal of the American Academy of Child and Adolescent Psychiatry, 31, 1041−1045.

Cattell, R. B. (1966). The scree test for the number of factors. Multivariate Behavioral Research, 1, 245−276.

Center for Substance Abuse Treatment. (2005). Substance abuse treatment for persons with co-occurring disorders. Rockville, MD: Substance Abuse and Mental Health Services Administration (No. Treatment Improvement Protocol (TIP) Series 42. DHHS Publication No. (SMA) 05-3922).

Chinet, L., Plancherel, B., Bolognini, M., Bernard, M., Laget, J., Daniele, G., et al. (2006). Substance use and depression. Comparative Course in Adolescents. European Child & Adolescent Psychiatry, 15, 149−155.

Clark, D. B., Pollock, N., Bukstein, O. G., Mezzich, A. C., Bromberger, J. T., & Donovan, J. E. (1997). Gender and comorbid psychopathology in adolescents with alcohol dependence. Journal of the American Academy of Child and Adolescent Psychiatry, 36, 1195−1203.

Cornelius, J. R., Bukstein, O. G., Salloum, I. M., Kelly, T. M., Wood, D. S., & Clark, D. B. (2004). Fluoxetine in depressed AUD adolescents: A 1- year follow-up evaluation. Journal of Child and Adolescent Psycho- pharmacology, 14, 33−38.

Cottler, L. B., Robins, L. N., & Helzer, J. E. (1989). The reliability of the CIDI-SAM: A comprehensive substance abuse interview. British Journal of Addiction, 84, 801−814.

Cronbach, L. J., Gleser, G. C., Nanda, H., & Rajaratnam, N. (1971). The dependability of behavioral measurements. New York: Wiley.

Deykin, E. Y., Buka, S. L., & Zeena, T. H. (1992). Depressive illness among chemically dependent adolescents. American Journal of Psychiatry, 149, 1341−1347.

Hasin, D., Nunes, E., & Meydan, J. (2004). Comorbidity of alcohol, drug and psychiatric disorders: Epidemiology. In H. R. Kranzler, & J. A. Tinsley (Eds.), Dual diagnosis and psychiatric treatment: Substance abuse and comorbid disorders, Vol. 1. (pp. 1–34). New York: Marcle Dekker, Inc.

Hovens, J. G., Cantwell, D. P., & Kiriakos, R. (1994). Psychiatric comorbidity in hospitalized adolescent substance abusers. Journal of the American Academy of Child and Adolescent Psychiatry, 33, 476−483.

Kandel, D. B., Johnson, J. G., Bird, H. R., Canino, G., Goodman, S. H., Lahey, B. B., et al. (1997). Psychiatric disorders associated with substance use among children and adolescents: Findings from the Methods for the Epidemiology of Child and Adolescent Mental Disorders (MECA) Study. Journal of Abnormal Child Psychology, 25, 121−132.

Kashani, J. H., D'Souza, A. B., Reid, J. C., & Neal, J. G. (1985). Affective disorders in an inpatient community mental health center. Missouri Medicine, 82, 654−657.

Kaufman, J., Birmaher, B., Brent, D., Rao, U., & Ryan, N. (1996). Diagnostic Interview Kiddie-Sads-Present and Lifetime Version (K- SADS-PL). Version 1.0 of October 1996. Retrieved April 9, 2008, from http://www.wpic.pitt.edu/ksads/ksads-pl.pdf.

Marton, P., Churchard, M., Kutcher, S., & Korenblum, M. (1991). Diagnostic utility of the Beck Depression Inventory with adolescent psychiatric outpatients and inpatients. Canadian Journal of Psychiatry, 36, 428−431.

Reich, W. (2000). Diagnostic Interview for Children and Adolescents (DICA). Journal of the American Academy of Child and Adolescent Psychiatry, 39, 59−66.

Riggs, P. D., Baker, S., Mikulich, S. K., Young, S. E., & Crowley, T. J. (1995). Depression in substance-dependent delinquents. Journal of the American Academy of Child and Adolescent Psychiatry, 34, 764−771.

Rounsaville, B. J., Weissman, M. M., Rosenberger, P. H., Wilber, C. H., & Kleber, H. D. (1979). Detecting depressive disorders in drug abusers: A comparison of screening instruments. Journal of Affective Disorders, 1, 255−267.

SPSS. (2006). SPSS 15.0 for Windows, Release 15.0.1. Chicago: SPSS Inc. Stowell, R. J., & Estroff, T. W. (1992). Psychiatric disorders in substance-

abusing adolescent inpatients: A pilot study. Journal of the American Academy of Child and Adolescent Psychiatry, 31, 1036−1040.

Subramaniam, G. A., Lewis, L. L., Stitzer, M. L., & Fishman, M. J. (2004). Depressive symptoms in adolescents during residential treatment for substance use disorders. American Journal of Addiction, 13, 256−267.

Subramaniam, G. A., Stitzer, M. L., Clemmey, P., Kolodner, K., & Fishman, M. J. (2007). Baseline depressive symptoms predict poor substance use outcome following adolescent residential treatment. Journal of the American Academy of Child and Adolescent Psychiatry, 46, 1062−1069.

Subramaniam, G. A., Stitzer, M. L., Woody, G. E., Fishman, M. J., & Kolodner, K. (2009). Clinical characteristics of treatment-seeking adolescents with opioid versus cannabis/alcohol use disorders. Drug and Alcohol Dependence, 99, 141−149.

Thompson, B. (2004). Exploratory and confirmatory factor analysis: Understanding concepts and applications (1st ed.) Washington, DC: American Psychological Association.

Weiss, R. D., Griffin, M. L., & Mirin, S. M. (1989). Diagnosing major depression in cocaine abusers: The use of depression rating scales. Psychiatry Research, 28, 335−343.

Welner, Z., Reich, W., Herjanic, B., Jung, K. G., & Amado, H. (1987). Reliability, validity, and parent–child agreement studies of the Diagnostic Interview for Children and Adolescents (DICA). Journal of the American Academy of Child and Adolescent Psychiatry, 26, 649−653.