Normative Data on the Beck Depression Inventory – Second Edition (BDI-II) in College Students Mark A. Whisman and Emily D. Richardson University of Colorado Boulder

Objective:

This study was conducted to provide normative data on the Beck Depression Inventory – Data were obtained from 15,233 college stuSecond Edition (BDI-II) in college students. Method: dents drawn from 17 universities in the United States, weighted to match the gender and race/ethnicity of enrollment in degree-granting institutions. Results: Descriptive statistics, point prevalence of individuals exceeding cutoff scores, and mean differences by gender and race/ethnicity were provided. Because the distribution of BDI-II scores was not normal, percentile ranks for raw scores were provided for the total sample and separately by gender and race/ethnicity for the total sample and by race/ethnicity for men and women. Normative data were used to calculate the Reliable Change Index on the BDI-II for college students. Conclusion: Because the distribution of BDI-II scores demonstrated significant skewness and non-normal kurtosis, percentile ranks are important to consider in C 2015 Wiley Periodicals, interpreting scores on the measure, in addition to descriptive statistics.  Inc. J. Clin. Psychol. 71:898–907, 2015. Keywords: Beck Depression Inventory – Second Edition; BDI-II; norms; normative data; Reliable Change Index

In the United States, it is estimated that nearly one out of every 12 adults (8.3% of the population) will meet criteria for a major depressive episode in the prior 12 months (Kessler et al., 2010). Worldwide, depression has been identified as the leading cause of disability (Ustun, AyusoMateos, Chatterji, Mathers, & Murray, 2004) and the second leading cause of burden (Ferrari et al., 2013). Approximately 40% of 18- to 24-year-olds in the United States are enrolled in college (National Center for Education Statistics, 2014). Therefore, understanding depression in college students is important for understanding depression in young adults. Existing evidence suggests that depression is widespread among college students. For example, the American College Health Association (ACHA)–National College Health Assessment II survey found that 35% of female and 28% of male participants report that they “felt so depressed it was difficult to function” some time within the last 12 months, and 14% of female and 8% of male participants report that they were diagnosed or treated by a professional for depression within the last 12 months (ACHA, 2014). Furthermore, many depressed college students do not receive treatment. For example, a recent study found that 85% of students with moderately severe to severe levels of depressive symptoms were not receiving any form of psychiatric treatment (Garlow et al., 2008). Given the widespread prevalence of depression, and further given its significant personal and societal impact, psychometrically sound measures of depressive symptoms are necessary for the assessment of the presence and severity of depression. One of the most widely accepted measures for assessing depressive symptom severity is the Beck Depression Inventory – Second Edition (BDI-II; Beck, Steer, & Brown, 1996), which was developed to replace the Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961) and its successor, the BDIIA (Beck, Rush, Shaw, & Emery, 1979; Beck & Steer, 1987). This research was supported by a grant from the Brain & Behavior Research Foundation. We thank the researchers who generously provided us with the raw data from their published studies to include in the current analysis. Please address correspondence to: Mark A. Whisman, University of Colorado Boulder, Department of Psychology and Neuroscience, 345 UCB, Boulder, CO 80309-0345. E-mail: [email protected] JOURNAL OF CLINICAL PSYCHOLOGY, Vol. 71(9), 898–907 (2015) Published online in Wiley Online Library (wileyonlinelibrary.com/journal/jclp).

 C 2015 Wiley Periodicals, Inc. DOI: 10.1002/jclp.22188

Normative Data on the BDI-II in College Students

899

To be meaningful, tests such as the BDI-II must have an empirical frame of reference. Generally, a test score is interpreted as indicating a person’s standing relative to other people in some group; norms provide a basis for such comparisons. Normative data on most symptom measures such as the BDI-II are generally presented in terms of means and standard deviations (e.g., Kendall & Sheldrick, 2000). To evaluate the score of an individual or the mean score of a group to these normative data, raw scores are often converted to standard scores (i.e., z or T scores, which are expressed in terms of the mean and standard deviation). This procedure allows one to compare not only the score of the person or the group relative to the normative group but also the standard scores on one measure to the standard score on another measure (e.g., to compare a standard score on a measure of depressive symptoms to a standard score on a measure of anxiety symptoms). Standard scores, in turn, can be used to compute percentile ranks, which provide an index of how common or uncommon a person’s score is in the normative population. Percentile ranks can be used to evaluate the percentage of people who score at or below a particular score, which allow clinicians or researchers to quantify the abnormality or rarity of an individual’s score and, therefore, have the advantage of being easy to understand. There is, however, a major limitation of using percentiles based on standardized scores for symptom measures of psychopathology such as the BDI-II. The underlying assumption when using z scores (or T scores) is that the distribution of scores follows the standard normal distribution, such that there is a fixed relationship between z scores and percentiles. However, measures of psychopathology symptoms are unlikely to be normally distributed. Most individuals are likely to report few symptoms of depression, for example, resulting in a distribution with positive skewness. To the extent that measured constructs such as depressive symptoms are not normally distributed, the percentile ranks based on z or T scores will not be accurate, which can contribute to inaccurate interpretations of scores (Crawford, Garthwaite, & Slick, 2009). For example, Gawrysiak, Nicholas, and Hopko (2009) reported the percentile corresponding to the mean BDI-II in a treatment-seeking sample of college students, based on the mean and standard deviation of a nonclinical student sample, which may not be accurate if the distribution of BDI-II scores in college students is not normal. The present study was conducted to (a) provide normative data on the BDI-II in college students, including providing percentile ranks, based on the assumption that scores on the BDIII would not be normally distributed; and (b) to illustrate potential uses of the normative data by computing the Reliable Change index (RC; Jacobson, Follette, & Revenstorf, 1984; Jacobson & Truax, 1991), a means for establishing that change between two time points (e.g., pre- to posttreatment) is statistically reliable.

Method Participants Participants included undergraduate students from 17 universities across the United States. We used data from multiple universities to enhance the generalizability of the results, including generalizability to racial and ethnic minority students. To identify studies that included BDI-II data on minority students, we first conducted a PsycINFO literature search for articles published using the following terms: (a) “BDI-II” or “Beck Depression Inventory – Second Edition”; (b) “student” or “undergraduate”; and (c) “minority” or “African American” or “Black” or “Hispanic” or “Latino” or “Asian.” In addition, we reasoned that studies that included large samples of undergraduates would be more likely to yield larger numbers of minority students, and because factor analytic studies would tend to be based on relatively large samples, we conducted a separate search using the same search terms for the BDI-II and the keywords “factor analysis.” Because of the potential problems with translated versions of the BDI-II, we limited our search to English language version of the measure. Finally, because we intended to weight the data to reflect the gender and ethnic distribution of college students nationwide, we limited the search to college samples within the United States. The search identified 33 studies. We then contacted the authors of these studies and requested

900

Journal of Clinical Psychology, September 2015

the raw data on the BDI-II, as well as data on gender, race/ethnicity, and age (if available). Of the 33 requests, 18 authors provided complete data from their study; an additional two authors provided raw data on the BDI-II but did not have data on race/ethnicity. Thus, the overall response rate was 61%, and we obtained usable data from 55% of the studies. This response rate was better than some other requests for data for reanalysis (e.g., Wicherts, Borsboom, Kats, & Molenaar, 2006). We also collected additional data on the BDI-II at the University of Colorado Boulder. Therefore, the study is based on combined data from 18 published studies and one unpublished sample, which represents data from 17 universities from diverse geographical regions in the United States. Informed consent was obtained from participants in all samples, and the Institutional Review Board at the University of Colorado Boulder approved the use of deidentified data. Information regarding the studies providing data, including the university from which the data were obtained, initial sample size, and sample description, can be found in Table 1. We excluded participants who did not provide information on gender (n = 17) or race/ethnicity (n = 30), or who had out of bounds scores on individual items (n = 5), leaving an initial sample of 15,349 participants. We also excluded 116 participants who were missing data on more than two (i.e., >10%) of the items on the BDI-II, which included 104 people who were missing data on all items; item mean substitution was used for participants who were missing one (n = 215) or two (n = 136) items. The final sample comprised data from 15,233 college students. Because data were drawn from 17 universities in the United States, these data should be more representative than data obtained at only one site. However, there may still be biases in the probability that particular subgroups of individuals in the reference population (i.e., all undergraduate students in the United States) were selected, which can lead to a sample having different characteristics to the population it aims to represent. Consequently, we used information provided by the United States Department of Education, National Center for Education Statistics (i.e., the Higher Education General Information Survey; Kim, 2011) to compute sample weights to weight the data to match the gender and race/ethnicity of the total fall enrollment in degree-granting institutions in the United States for 2007. We used 2007 enrollment data because a new category–Two or More Races–was added to the racial/ethnic classifications for enrollment in 2008, which affects the aggregate for each of the other racial/ethnic categories. The Higher Education General Information Survey percentages of students enrolled in degreegranting institutions in the United States by gender and race/ethnicity were divided by the percentages of participants in the pooled sample to create sample weights; values 1.0 indicate that a particular subgroup was underrepresented. Sample weights were applied to (i.e., multiplied by) participants’ responses so that responses of people who share gender and race/ethnicity characteristics with groups who were underrepresented in the sample were given more weight than the responses of people who shared gender and race/ethnicity characteristics with overrepresented groups. Demographic characteristics of the final pooled sample, including the actual number and actual and weighted percentage of participants by gender and race/ethnicity are presented in Table 2. The mean age of the weighted sample was 19.32 years (standard deviation [SD] = 4.87), although data on age were available from only 13,912 participants.

Measure The BDI-II is a 21-item self-report questionnaire that is designed to measure the severity of depressive symptoms in adolescents and adults (Beck et al., 1996). The questions relate to symptoms of depression with respect to the “past two weeks, including today” and are rated on a 4-point scale ranging from 0 to 3. Items are summed to create a total score, ranging from 0 to 63, with higher scores reflecting greater severity. Prior research indicates that the measure is internally consistent, differentiates between depressed and nondepressed individuals, and correlates highly with other measures of depressive symptoms and depression-related constructs

Normative Data on the BDI-II in College Students

901

Table 1 Studies Providing Data Study

University

N

Description of sample Undergraduates at public university Undergraduates in psychology course at public university Undergraduate in psychology courses at public university Undergraduates at public university Freshman at public university Undergraduates in psychology courses at two public universities Undergraduates in psychology course at public university Undergraduate psychology students at public university Undergraduates at public university Undergraduates in psychology course at public university Undergraduates at public university Undergraduates at public university Undergraduates at public university Undergraduates in psychology course at public university Undergraduate psychology students at public university Undergraduates in psychology or education at two public universities Undergraduates in psychology course at public university Undergraduates at private university Undergraduates in psychology course at public university

Cheref et al. (2015)

Hunter College

690

Chesin & Jeglic (2012)

City University of New York

554

Cukrowicz et al. (2012)

Florida International University

976

Dvorak et al. (2013)

University of South Carolina

Geisner et al. (2012)

University of Washington

867

Harari et al. (2005)

University of Akron; University of South Carolina

317

Klibert et al. (2005)

University of South Alabama

475

Lamis et al. (2014)

University of South Carolina

1,200

Lamis & Malone (2011)

University of South Carolina

1,020

Lewandowski et al. (2006)

University of North Carolina at Greensboro

1,258

Norton (2005)

University of Houston

540

Okazaki (2002)

University of Illinois at Urbana-Champaign University of Houston

400

Sashidharan et al. (2012)

Southern Illinois University Edwardsville

278

Scarpa et al. (2002)

Virginia Polytechnic Institute and State University

518

Storch et al. (2004)

University of Florida; Louisiana State University

414

Whisman

University of Colorado Boulder

Whisman et al. (2000)

Yale University

576

Wiebe & Penley (2005)

University of Texas at El Paso

895

Philipp et al. (2008)

1,101

923

2,399

902

Journal of Clinical Psychology, September 2015

Table 2 Demographic Characteristics of the Pooled Sample and BDI-II Means and Standard Deviations by Demographic Group BDI-II score Sample characteristic Total White Black Asian Latino Other Men White Black Asian Latino Other Women White Black Asian Latino Other

N

% of total

Weighted %

Mean

SD

15,233 8,999 1,891 1,333 2,479 531 4,609 2,904 402 492 631 180 10,624 6,095 1,489 841 1,848 351

59.1 12.4 8.8 16.3 3.5 30.3 19.1 2.6 3.2 4.1 1.2 69.7 40.0 9.8 5.5 12.1 2.3

66.6 13.6 6.9 11.8 1.1 42.4 29.1 4.8 3.2 4.9 0.4 57.6 37.5 8.8 3.7 6.9 0.7

9.14 8.44 9.32 11.82 11.18 10.56 8.22 7.59 8.53 11.20 9.62 9.15 9.81 9.09 9.75 12.36 12.29 11.45

8.45 8.09 8.44 9.64 8.93 9.22 8.06 7.74 8.01 9.48 8.30 8.09 8.67 8.30 8.64 9.76 9.20 9.81

Note. BDI-II = Beck Depression Inventory – Second Edition; SD = standard deviation.

(for a review, see Dozois & Covin, 2004); the measure demonstrated good internal consistency in the pooled sample (α = .91). Recently, factorial invariance of the BDI-II has been found across gender and race/ethnicity in college students (Whisman, Judd, Whiteford, & Gelhorn, 2013). Kendall, Hollon, Beck, Hammen, and Ingram (1987) provided recommendations for BDI cutoff scores for classifying people as nondepressed (0–9), mildly dysphoric (10–15), and dysphoric (if people fail to meet diagnostic criteria for major depressive episode) or depressed (if they meet criteria; 16); Dozois, Dobson, and Ahnberg (1998) developed corresponding cutoff values for the BDI-II of 0) was 7.23 (SD = 5.47; range = 0–21). Using the cutoff scores provided by Dozois et al. (1998), we estimated that 74% of undergraduates were nondepressed (i.e., had BDI-II scores of

Normative Data on the Beck Depression Inventory--Second Edition (BDI-II) in College Students.

This study was conducted to provide normative data on the Beck Depression Inventory--Second Edition (BDI-II) in college students...
92KB Sizes 1 Downloads 12 Views