Measurement invariance of the Beck Depression Inventory-Second Edition across race with South African university students

Abstract

Measurement invariance of the Beck Depression Inventory-Second Edition across race (Blacks and Whites) was examined in a sample of university students, from two universities from diverse geographical areas of South Africa (N = 870). Multigroup confirmatory factor analysis (i.e., means and covariance structures) was used to test the factorial invariance of the hierarchical four-factor structure, composed of three first-order factors (i.e., Negative Attitude, Performance Difficulty, and Somatic Complaints) and one second-order general factor (Depression) found with South African students. Evidence of measurement invariance was established at the level of configural, metric, and scalar invariance. However, there were some evidence of differential additive response style across race—with two non-invariant intercepts (Item 5 and 14) being identified. Results also revealed significant latent mean differences favoring Black students on the Performance Difficulty and Somatic Complaints factors but not on the Negative Attitude factor. Findings suggest that the Beck Depression Inventory-Second Edition provides an assessment of the severity of depressive symptoms that is equivalent across race in university students.

Keywords

BDI-II depression latent mean differences measurement invariance racial differences

The Beck Depression Inventory-Second Edition (BDI-II) is a commonly used measure of depressive mood by both clinicians and researchers in South Africa (see Kagee, Nel, & Saal, 2014; Pillay & Sargent, 1999; Steele & Edwards, 2008; Stellenberg & Abrahams, 2015; Vawda, 2014; Westaway & Wolmarans, 1992). In research, the instrument is mostly used in studies on inter-group variances in the processes and levels of depression or exploring variables associated with depressive symptoms (e.g., Hambrick et al., 2010). While research on comparative processes relates to differences in intra-group predictors of depression, studies of comparative levels focus on variances in between-group means. Inherent in both types of studies is the assumption that the BDI-II measures equivalent constructs in exactly the same manner across the concerned groups, that is, the scale is measurement invariant. If the instrument does not operate invariantly across groups, any conclusions on between-group differences may be artifactual and misleading (Byrne, Stewart, Kennard, & Lee, 2007; Cheung & Rensvold, 2002).

While there is some literature on the reliability and validity of the BDI-II in South Africa (e.g., Kagee et al., 2014; Steele & Edwards, 2008), little is known about its measurement invariance (MI). MI (i.e., the degree to which items or subtests have equivalent meaning across groups of examinees) is one method used to gather score validity evidence and to appraise construct-irrelevant variance (e.g., group affiliation). The testee’s score should not depend on construct-irrelevant variance. The likelihood of consequential mistakes is high, if conclusions are made for individuals on a test in the absence of MI (Cheung & Rensvold, 2002). Observed score differences can either represent true group mean differences or variances in the relation between the construct and the observed score that is not equivalent across groups (Cheung & Rensvold, 2002).

The few studies that have evaluated MI of the BDI-II across race or ethnicity have mainly displayed/found two patterns of results (i.e., full invariance and partial invariance). For example, Byrne et al. (2007) tested factorial invariance of the BDI-II across Hong Kong and American adolescents and found evidence of MI across the two cultures. Similar results were also reported with Chinese and European North American university students (Byrne & Stewart, 2006; Dere et al., 2015; Whisman, Judd, Whiteford, & Gelhorn, 2013). Whisman et al. (2013) found evidence for MI in the context of the hierarchical four-factor structure of the BDI-II between Whites and racial (Blacks and Asians) or ethnic (Latinos) minority groups. Hambrick et al. (2010) found little evidence for differential item functioning on BDI-II items between White and African American undergraduate students using item response theory (IRT). Lin (2012) also confirmed factorial invariance of the BDI among three racial/ethnic groups (i.e., non-Hispanic Blacks, non-Hispanic Whites, and Hispanic cardiac patients) of post-myocardial infarction patients.

Azocar, Arean, Miranda, and Munoz (2001) analyzed the BDI items for bias between Spanish and English speaking patients and their results supported MI of the BDI model, although with evidence of differential additive response style (ARS). Another study on the BDI in a sample of low-income non-Hispanic Blacks supported MI of their model (Grothe et al., 2005). Their findings also showed that non-Hispanic Blacks were less likely to endorse the item reflecting suicidal thoughts than non-Hispanic Whites. Similarly, there is some inconsistent evidence of mean differences on the BDI-II across groups defined by race or ethnicity (e.g., Hambrick et al., 2010). However, Carmody (2005) investigated item bias in the endorsement of symptoms on the BDI-II among White, Asian, and Latino Americans. In his study, White American students scored higher on three items (BDI-II Items 11, 14, and 17) than did Hispanic and Asian American students.

Hooper, Qu, Crusto, and Huffman (2012) using IRT and confirmatory factor analysis (CFA) examined scalar invariance with a sample of college students in the United States. Results from the analyses indicated that the items endorsed by Black American and White American college students were slightly different. In total, 23% of the items on the BDI-II functioned differently based on at least one comparison method (i.e., CFA or IRT). More specifically, for these race-related comparisons, symptom endorsement varied on five BDI-II items: Items 7, 8, 14, 15, and 21. Therefore, 5 items functioned differently, and 16 of the items functioned similarly in these racial group comparisons.

Given the limited research on the MI of the BDI-II, more studies are necessary to explore this issue further, particularly in the African context. Accordingly, the purpose of the current study was to examine the MI (i.e., configural, metric, and scalar invariance) and latent mean differences of the BDI-II across race in a South African university sample.

Method

Participants

A purposive sample of 870 students (M_age = 21.70 years, standard deviation [SD] = 13.51, 70% female) was selected from both the University of Limpopo (46%) and the University of Pretoria (54%). The inclusion criteria were as follows: (1) undergraduate university students, (2) command of English, and (3) predominantly White university and predominantly Black university (4) race. Besides accessibility, sampling from the institutions ensured that we had a heterogeneous sample (i.e., race and socio-economic status [SES]) which also approximated the ones used in previous studies on the BDI-II (Beck, Steer, & Brown, 1996; Byrne, Stewart, & Lee, 2004). The overwhelmingly female (72%) sample, while a concern, is not surprising given the sampling strategy used and the undergraduate enrollment rates of the two South African universities. For instance, the Department of Higher Education and Training’s (2011) Statistics on Post-School Education and Training in South Africa reports that female students comprise 54% of all students enrolled in contact programs at the University of Limpopo and 60% of all students enrolled in contact programs at the University of Pretoria. This part of the sample characteristics is given a detailed treatment in the Methodology section. The data consisted of undergraduate students from faculties of Humanities, Natural Sciences, Statistics, Law, and Management Sciences. In total, 67% of the participants identified themselves as ‘Black’, while 33% said they were ‘White’. The mean score (11.45; SD = 7.74) of the sample on the total BDI-II is outside the symptomatic range of the BDI-II suggested by Beck et al. (1996).

Instruments

BDI-II

The BDI-II is a 21-item self-report questionnaire (Beck et al., 1996) used to measure the severity of depressive symptoms in adolescents and adults. The respondent receives a score of 0–3 for each item; the total raw score is the sum of the 21 items, with the highest possible score of 63. The Cronbach’s alpha of the measure was .84 in this current study. Its correlations with clinical scales of depressive symptoms from measures such as the Minnesota Multiphasic Personality Inventory (MMPI) are normally in the range of r = .60 to .76 (Conoley, 1992).

Procedure

Participants were recruited from undergraduate classes at both the University of Limpopo and the University of Pretoria. The aim of the research was explained and instructions on completing the questionnaire were also given. After obtaining written consent from the students, the instrument was administered to them in English and in group format outside of their normal university hours. The instrument was self-administered, while seating arrangements ensured that confidentiality was preserved. The students were not compensated for taking part in the study, and they were debriefed at the end of the data collection session.

Ethical considerations

Ethical approval for the study was granted by the research and ethics committees of the University of Limpopo and University of Pretoria, respectively. All students consented (orally and in written form) to participation in the research. Taking part in the study was voluntary, while confidentiality and anonymity were assured.

Data analysis

The structural equation modeling (SEM) analysis was divided into two stages using the EQS 6.1 program (Bentler, 2007). Factorial invariance was based on multigroup confirmatory factor analysis (MCFA) (i.e., means and covariance structures [MACSs]) (Whisman et al., 2013; Wu, 2009, 2010). The robust maximum likelihood estimation method (Yuan–Bentler scaled method; Yuan & Bentler, 2000) was justified and employed since the BDI-II data represented a non-normal distribution (i.e., item-level skewness > 1.5 and item-level kurtosis > 2.5; Mardia multivariate kurtosis = 184.21; c.r. = 89.69). Given that the data comprise of responses from university students (i.e., a non-clinical population), evidence of both skewness and kurtosis is indeed not unexpected and is consistent with other BDI studies of community samples in general and adolescents in particular (e.g., Byrne, Baron, Larsson, & Melin, 1995; Byrne et al., 2004; Koenig, Isaacs, & Schwartz, 1994; Roberts, Lewinsohn, & Seeley, 1991). Hence, this aspect of the data should be of little concern.

However, given these distributional characteristics, what is imperative is that analyses be based on the correct statistics, that is, they should be based on algorithms designed to take this non-normality into account. The scaled chi-square and “robust” standard errors using maximum likelihood (ML) estimation is a method suggested by Satorra and Bentler (1988, 1994). This method has proven to be a good general approach for addressing nonnormality (Curran, West, & Finch, 1996; Hu, Bentler, & Kano, 1992), as opposed to the alternative of bootstrapping (for correcting standard errors), which is still indeterminate as to how well it performs under various conditions (e.g., specific bootstrapping approach and sample sizes) (Nevitt & Hancock, 2001). Adjustments to the chi-square, relative fit indices, and standard errors are based on weight matrix derived from an estimate of multivariate kurtosis. EQS 6.1 provides the option of using robust statistics, instead of the regular statistics for addressing this multivariate non-normality problem. Moreover, the robust maximum likelihood estimation method is a common method for addressing non-normal distribution in the BDI-II literature (see Byrne et al., 2007; Whisman et al., 2013; Wu, 2010; Wu & Huang, 2014 for examples).

The hypothesized hierarchical four-factor structure of the BDI-II composed of three first-order factors (i.e., Negative Attitude [NA], Performance Difficulty [PD], and Somatic Complaints [S]) and one second-order general factor (Depression), established in university students is presented in Figure 1. This CFA model follows from a rigorous validation of the BDI-II across Black and White South African non-clinical students (see Makhubela & Mashegoane, 2015) and is consistent with past research on obtained factor structures in American and Chinese college student samples (e.g., Byrne et al., 2007). The model was used to determine the baseline model (i.e., separately for the race groups and the total sample). This second-order model is supported by theory and evidence of substantial mean correlations (NA, PD = .76; NA, S = .78; PD, S = .83) across groups (Makhubela & Mashegoane, 2015). As such, this model serves as a more parsimonious structure of the BDI-II for non-clinical students. Preliminary models (across race groups) converged on the same hierarchical factor structure obtained in previous studies (e.g., Byrne et al., 2007; Makhubela & Mashegoane, 2015; Whisman et al., 2013). After this baseline model was shown to be consistent with the data, the analyses then proceeded to test the equivalence of this model across race groups, using a series of ordered steps (see Byrne, 2006; Chen, Sousa, & West, 2005). Results of MI and MACS analysis were reported based on the following multiple criteria: the Yuan–Bentler scaled Chi-square test (Y–Bχ²) < 1.5 (χ²/degrees of freedom [df]) < 1.5 (Yuan & Bentler, 2000), comparative fit index (CFI) ⩾ .95, standardized root mean-square residual (SRMR) ⩽ .05, and root mean-square error of approximation (RMSEA) ⩽ .06 and its 90% confidence interval (CI) (Hu & Bentler, 1999).

Figure 1.

Hypothesized hierarchical second-order model.

Given this baseline model, the tenability of invariance across race groups was examined by fitting a series of increasingly restrictive models. The tenability of each restriction was evaluated based on the differences in fits between the restricted model and the less-restricted model (Δ*CFI & Δ*RMSEA) (Chen, 2007; Cheung & Rensvold, 2002). Given the support for MI, the MACSs were separately estimated to test latent mean differences for each latent construct by means of setting the latent mean value to zero for Blacks and freely estimating for Whites. The differences between the latent means were evaluated on the basis of the z statistic (Wald’s significance test) (Aiken, Stein, & Bentler, 1994; Hancock, 2001). Latent mean differences are preferable than observed mean scores when evaluating race differences on depression. The latter confounds factor and indicator variance, while the former is calculated by controlling for measurement error (Cheung & Rensvold, 2000). The extent of confounding may have significant effects on observed mean differences.

Results

Measurement invariance

In Table 1, tests of the hypothesized BDI-II structure, the configural model, revealed a good fit to the data for both Blacks and Whites (Y–Bχ²_[369] = 448.66; *CFI = .96; SRMR = .04; *RMSEA = .02, with 90% CI = .01–.02). All factor loadings were viable and statistically significant. This indicates that Blacks and Whites have an equivalent basic conceptualization of depression. This model serves as the baseline against which all remaining models are compared in the process of determining evidence of invariance. Furthermore, all MCFA results for MI across race, show that all the four nested models represented a good fit to the data (*CFIs = .95–.96, SRMRs = .04–.06, *RMSEAs = .02–.04). Model 2 (metric invariance), in which all lower-order factor loadings were equally constrained, also represented a good fit to the data with a resulting Δ*CFI and Δ*RMSEA values of 0. The satisfaction of metric invariance implies that all items of BDI-II have equal salience for Blacks and Whites.

Table 1.

Goodness-of-fit statistics for tests for invariance of BDI-II hierarchical structure for Blacks and Whites.

Model and constraints	Y–B χ²	df	*CFI	SRMR	*AIC	*RMSEA	90% *RMSEA CI	Model comparison	Δ*CFI	Δ*RMSEA
1. Configural invariance	448.66	369	.96	.04	289.34	.02	.01, .03
2. Lower and higher-order factor loadings invariant	469.59	379	.96	.05	288.41	.02	.02, .03	2 vs. 1	0	0
3. Intercepts invariant	633.77	392	.95	.05	150.23	.04	.03, .04	3 vs. 2	−.01	.02
4. Latent factor means invariant	597.98	389	.95	.06	180.02	.04	.03, .04	4 vs. 3	0	0

df: degrees of freedom; Y–Bχ²: Yuan–Bentler scaled chi-square test; *CFI: robust comparative fit index; *90% RMSEA CI: robust root mean square error of approximation and its 90% confidence interval; SRMR: standardized root mean-square residual; *AIC: robust Akaike information criterion; Δ*CFI: comparative fit index difference value; Δ*RMSEA: robust root mean square error of approximation difference value.

Model 3 (scalar invariance) additionally constrains the intercepts to be equal between groups, thus forcing equality of the variances/covariance matrices between the races. Once again, results yielded a good fit and Δ*CFI value of .01. Likewise, for the final remaining model tested, the multigroup model retains the same good fit and the Δ*CFI value never exceeds .01. However, scalar invariance was achieved by removing two constraints of intercepts (Item 5 and 14) (see Table 2). The findings of two non-invariant intercepts implied that there is differential ARS bias (Cheung & Rensvold, 2000) for the BDI-II across race groups. That is, Whites systematically endorse higher item responses in Item 14, while Blacks do the same for Item 5. Nonetheless, consistent with Cheung and Rensvold’s (2002) and Chen’s (2007) recommendations, we consider this model to exhibit evidence of invariance across the two race groups.

Table 2.

The intercepts of non-invariant items in Model 3 across race.

	Estimated values
	Factor	Blacks	Whites
Item 5 (guilt)	S	.51	.47
Item 14 (worthlessness)	NA	.57	.67

S: Somatic Complaints; NA: Negative Attitude.

Latent mean difference

As seen in Table 3, there were no significant latent mean differences on the NA factor (z = −.23, p > .05, d = .06) between the race groups. However, latent mean differences on the PD factor (z = 4.16, p < .05, d = .41) and the S factor (z = 3.74, p < .05, d = .66) were significant, with Blacks endorsing higher scores (see Cohen, 1988).

Table 3.

Differences between Blacks and Whites on latent constructs.

Latent mean analyses
Dimension	Model 3 z^a	d ^b
NA	−.23	.06
PD	4.16*	.41
S	3.74*	.66

NA: Negative Attitude; PD: Performance Difficulty; S: Somatic Complaints.

z = Wald significance test.

d = effect sizes associated with the latent mean difference was calculated using Hancock’s (2001) guidelines.

p < .05.

Discussion

As far as the researcher is aware, this study is the first to establish MI of the BDI-II across race groups in South Africa. Evidence of MI in the context of the hierarchical four-factor structure of the BDI-II across race was found. MI was established at the level of configural, metric, and scalar invariance. Specifically, across models in which there was increasingly restricted parameterization on the variance/covariance matrices of the indicators, there was consistent evidence that the hierarchical four-factor structure provided robust fit with the data. Furthermore, the Δ*CFI and Δ*RMSEA values for comparisons between models 1 (configural model), 2, 3, and 4 were all negligible.

Scalar invariance was established with two non-invariant item intercepts for race (Items 5: guilt and 14: worthlessness), providing empirical evidence of construct validity for the BDI-II for purposes of making race latent mean comparisons. These findings of two non-invariant intercepts across race implied that there is differential ARS bias (Cheung & Rensvold, 2000) for the BDI-II across these groups. That is, Blacks systematically endorse a relatively higher item response in Item 5, whereas Whites systematically endorse a higher item response in Item 14. The presence of differential ARS of the BDI-II in the current study confirms Wu’s (2010) findings. Moreover, these findings are consistent with the established empirical evidence in South Africa, which shows that, while certain core symptoms of depression are encountered across the different cultures, it is also true that depressive symptomatology often has unique manifestations in some of these cultures as a result of social and cultural differences (Mosotho, Louw, Calitz, & Esterhuyse, 2008). According to Hassim and Wagner (2013) culture exerts pathogenic, psychoselective, psychoplastic, patho-elaborating, psychofacilitating, and psychoreactive effects to the experience of psychopathology. As such, culture modifies and modulates the expression and the structured manifestation of depressive symptoms in South African race groups (see Schlebusch, 2005).

This pattern of results is corroborated by the majority of sub-Saharan African studies that have established that depression in Blacks is predominantly characterized by feelings of guilt and somatic presentations (Bhuti & Mohamud, 2003; Hollifield, Katon, Spain, & Pule, 1990; Patel, Abas, Broadhead, Todd, & Reeler, 2001; Reynolds & Swartz, 1993; Rogers, 1992; Swartz, 1998; Tomlinson, Swartz, Kruger, & Gureje, 2007). Steele and Edwards (2008) report that in South Africa, Xhosa-speaking respondents are more troubled by guilt than any other symptom of depression. Similarly, four studies conducted among Sesotho-speaking depression patients in the Free State province also reported that participants expressed feelings of guilt more than any other depressive symptoms (Jordaan & Joubert, 1994; Mosotho et al., 2008; Rossouw, 1998). In their interpretative review of selected literature, Tomlinson et al. (2007) reported that guilt as a symptom of depression has been a source of cultural variation in Africa.

Although the presence of racial differences in depressive symptoms is well-known, the direction and explanations for the differences are complex and at times unclear (Nolen-Hoeksema, Larson, & Grayson, 1999). An array of social, cultural, and personality reasons for greater vulnerability of certain races to specific depressive symptoms have been offered. For instance, the level of Westernization has often been associated with variations in the manifestation of depression. People from Western cultures are said to psychologize their depression (i.e., emotional and cognitive report of distress), whereas people from non-Western cultures are said to somatize their depression (i.e., report distress in the form of bodily complaints and physiological symptoms) (Canino, Rubio-Stipec, Canino, & Escobar, 1992).

A more psychological report of depression has also been found among more Westernized groups of non-Western societies (Kwang-Iel, Dongen, & Dae-Ho, 1999). Somatization of depressive symptoms has been observed in various non-Western cultures, such as in South Africa (Ngcobo & Pillay, 2008), Nigeria (Abiodun, 1995), China (Kwang-Iel et al., 1999), and United Arab Emirates (Hamdi, Amin, & Abou-Saleh, 1997). Therefore, that the endorsement of guilt (Item 5) should be higher for Blacks than Whites in the current study is inconsistent with other depression research (i.e., socialization literature) in reporting the tendency of Blacks to somatize than psychologize. However, this finding is consistent with research reporting no significant demographic differences between people who tend to somatize and those who psychologize (Blazer, Landerman, Hays, Simonsick, & Saunders, 1998; Razali & Hasanah, 1999).

A possible explanation for this direction in difference can be the fact that South Africa is more Westernized/acculturated, and therefore, we expected that evaluations of self by Blacks, as measured by the BDI-II, would necessarily reflect the worldview espoused by the society. This finding, of higher guilt in Blacks, is also contradictory to the implication of external locus-of-control beliefs in Blacks in the explanation of depression symptoms across cultures (see April, Dharani, & Peters, 2012; Lambert, Herman, Bynum, & Ialongo, 2009; Sanchez, Lambert, & Ialongo, 2011). An external locus-of-control in Blacks is attributed to the fact that Blacks have limited opportunities as an economic minority group in South Africa (Hamad, Fernald, Karlan, & Zinman, 2008). So, theoretically, Blacks are expected to report more feelings of being punished and less self-blame (Bhagwanjee, Parekh, Paruk, Petersen, & Subedar, 1998; Hamad et al., 2008; Pillay & Sargent, 2003; Rumble, Swartz, Parry, & Zwarenstein, 1996; Strebel, Msomi, & Stacey, 1999).

However, the stronger endorsement of Item 14 (worthlessness) by Whites than by Blacks is consonant with the literature, in that Whites typically psychologize their depression. This may also be attributed to the fact that people in individualistic cultures may be more in tune with their private emotional states, while those socialized in collectivistic cultures are more responsive to the promotion of the well-being of their in-group and thus emotions are used more discreetly (Canino et al., 1992). Accordingly, emotional states are a more immediate and significant source for the self-efficacy appraisals of individuals raised in idiocentric systems than in allocentric cultures.

The evidence of scalar invariance also implies that clinicians and researchers can use the BDI-II with greater confidence given the generalizability of the instrument’s properties between Black and White university students in South Africa. Contrary to most previous studies investigating race differences on overall depression (e.g., Hankin et al., 1998; Schuch, Roest, Nolen, Penninx, & de Jonge, 2014), the present study examined group differences at specific factor levels. Results revealed significant latent mean differences that favored Blacks in terms of PD and S factors. The results on differential race endorsements of items are explainable if we consider that people from non-Western cultures, especially Black South Africans, tend to somatize their depression (i.e., report affect in the form of bodily complaints and physiological symptoms) (Canino et al., 1992; Halbreich et al., 2007; Reynolds & Swartz, 1993; Rochat, Tomlinson, Barnighausen, Newell, & Stein, 2011; Rogers, 1992; Swartz, 1998; Tomlinson et al., 2007).

The present study found two non-invariant intercepts across race (Item 5 and 14) in the BDI-II with negligible effects on latent mean differences. Although scalar invariance properties of the BDI-II across race groups are established, practitioners are urged to interpret race differences corresponding to these non-invariant item intercepts with restraint. These intercepts failed to display equivalent measurement across race groups, suggesting that these items overestimated the corresponding factor for one group. In that regard, the item depicting guilt overestimated its respective factors for Blacks, whereas that depicting worthlessness overestimated its factor for Whites.

Conclusion

Since factorial invariance was obtained for analyses constraining factor structure and loadings, these results suggest that it is also appropriate to compare correlates of depressive symptoms across groups (Whisman et al., 2013). These findings corroborate past research on the MI of the BDI-II within student populations across race (Byrne, Baron, Larsson, & Melin, 1996; Byrne et al., 2007; Campbell, Roberti, Maynard, & Emmanuel, 2009; Hooper et al., 2012; Whisman et al., 2013; Wu, 2010; Wu & Huang, 2014). It can therefore be concluded that the BDI-II appears to measure depressive mood equivalently across Black and White student respondents in South Africa. However, the reported findings are not without limitations. Given that the sample was overwhelmingly female (72%) and also restricted to university students, this precludes the possibility to generalize the results to other populations.

Footnotes

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Abiodun

O. A.

(1995). A study of somatization among primary care patients in Nigeria. International Journal of Mental Health, 23, 81–89.

Aiken

L. S.

Stein

J. A.

Bentler

P. M.

(1994). Structural equation analyses of clinical subpopulation differences and comparative treatment outcomes: Characterizing the daily lives of drug addicts. Journal of Consulting and Clinical Psychology, 62, 488–499.

April

K. A.

Dharani

Peters

(2012). Impact of locus of control expectancy on level of well-being. Review of European Studies, 4, 124–137.

Azocar

Arean

Miranda

Munoz

R. F.

(2001). Differential item functioning in a Spanish translation of the Beck Depression Inventory. Journal of Clinical Psychology, 57, 355–365.

Beck

A. T.

Steer

R. A.

Brown

G. K.

(1996). Beck Depression Inventory manual (2nd ed.). San Antonio, TX: Psychological Corp.

Bentler

P. M.

(2007). EQS 6.1: Structural equations program manual. Encino, CA: Multivariate Software.

Bhagwanjee

Parekh

Paruk

Petersen

Subedar

(1998). Prevalence of minor psychiatric disorders in adult African rural community in South Africa. Psychological Medicine, 28, 1137–1147.

Bhuti

Mohamud

(2003). Cultural adaptation of mental health measures: Improving the quality of clinical practice and research. British Journal of Psychiatry, 183, 185–186.

Blazer

D. G.

Landerman

L. R.

Hays

J. C.

Simonsick

E. M.

Saunders

W. B.

(1998). Symptoms of depression among community-dwelling elderly African Americans and White older adults. Psychological Medicine, 28, 1311–1320.

10.

Byrne

B. M.

(2006). Structural equation modeling with EQS: Basic concepts, applications, and programming (2nd ed.). Mahwah, NJ: Lawrence Erlbaum.

11.

Byrne

B. M.

Baron

Larsson

Melin

(1995). The Beck Depression Inventory: Testing and cross-validating a second-order factorial structure for Swedish nonclinical adolescents. Behaviour Research and Therapy, 33, 345–356.

12.

Byrne

B. M.

Baron

Larsson

Melin

(1996). Measuring depression for Swedish nonclinical adolescents: Factorial validity and equivalence of the Beck Depression Inventory across gender. Scandinavian Journal of Psychology, 37, 37–45.

13.

Byrne

B. M.

Stewart

S. M.

(2006). The MACS approach to testing for multigroup invariance of a second-order structure: A walk through the process. Structural Equation Modeling, 13, 287–321.

14.

Byrne

B. M.

Stewart

S. M.

Lee

P. W. H.

(2004). Validating the Beck Depression Inventory–II for Hong Kong community adolescents. International Journal of Testing, 4, 199–216.

15.

Byrne

B. M.

Stewart

S. W.

Kennard

B. D.

Lee

P. W. H.

(2007). The Beck Depression Inventory II: Testing for measurement equivalence and factor mean differences across Hong Kong and American adolescents. International Journal of Testing, 7, 293–309.

16.

Campbell

M. H.

Roberti

J. W.

Maynard

Emmanuel

M. K.

(2009). Factorial structure of depressive symptoms in Anglophone Caribbean university students: Psychometric properties of the Beck Depression Inventory-II. InterAmerican Journal of Psychology, 43, 577–585.

17.

Canino

I. A.

Rubio-Stipec

Canino

Escobar

J. I.

(1992). Functional somatic symptoms: A cross-ethnic comparison. American Journal of Orthopsychiatry, 62, 605–612.

18.

Carmody

D. P.

(2005). Psychometric characteristics of the Beck Depression Inventory-II with college students of diverse ethnicity. International Journal of Psychiatry in Clinical Practice, 9, 22–28.

19.

Chen

F. F.

(2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling, 14, 464–504.

20.

Chen

F. F.

Sousa

K. H.

West

S. G.

(2005). Testing measurement invariance of second-order factor models. Structural Equation Modeling, 12, 471–492.

21.

Cheung

G. W.

Rensvold

R. B.

(2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9, 233–255.

22.

Cheung

G. W.

Rensvold

R. B.

(2000). Assessing extreme and acquiescence response sets in cross-cultural research using structural equation modeling. Journal of Cross-Cultural Psychology, 31, 187–212.

23.

Cohen

(1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum.

24.

Conoley

C. W.

(1992). Review of the Beck Depression Inventory [Revised edition]. In Kramer

J. J.

Conoley

J. C.

(Eds.), Eleventh mental measurements yearbook (pp. 78–79). Lincoln, NE: Buros Institute of Mental Measurements.

25.

Curran

P. J.

West

S. G.

Finch

J. F.

(1996). The robustness of test statistics to nonnormality and specification error in confirmatory factor analysis. Psychological Methods, 1, 16–29.

26.

Department of Higher Education and Training. (2011). Statistics on post-school education and training in South Africa. Pretoria, South Africa: Author.

27.

Dere

Watters

C. A.

S. C.

Bagby

R. M.

Ryder

A. G.

Harkness

K. L.

(2015). Cross-cultural examination of measurement invariance of the Beck Depression Inventory–II. Psychological Assessment, 27, 68–81. doi:10.1037/pas0000026

28.

Grothe

K. B.

Dutton

G. R.

Jones

G. N.

Bodenlos

Ancona

Brantley

P. J.

(2005). Validation of the Beck Depression Inventory-II in a low-income African American sample of medical outpatients. Psychological Assessment, 17, 110–114.

29.

Halbreich

Renato

D. A.

Calil

Douki

Gaszner

Jadresic

. . . Trivedi

J. K.

(2007). Culturally-sensitive complaints of depressions and anxieties in women. Journal of Affective Disorders, 102, 159–176.

30.

Hamad

Fernald

L. C. H.

Karlan

D. S.

Zinman

(2008). Social and economic correlates of depressive symptoms and perceived stress in South African adults. Journal of Epidemiology & Community Health, 62, 538–544.

31.

Hambrick

J. P.

Rodebaugh

T. L.

Balsis

Woods

C. M.

Mendez

J. L.

Heimberg

R. G.

(2010). Cross-ethnic measurment equivalence of measures of depression, social anxiety, and worry. Assessment, 17, 155–171.

32.

Hamdi

Amin

Abou-Saleh

M. T.

(1997). Performance of the Hamilton Depression Rating Scale in depressed patients in the United Arab Emirates. Acta Psychiatrica Scandinavica, 96, 416–423.

33.

Hancock

G. R.

(2001). Effect size, power, and sample size determination for structured means modeling and MIMIC approaches to between-groups hypothesis testing of means on a single latent construct. Psychometrika, 66, 373–388.

34.

Hankin

B. L.

Abramson

L. Y.

Moffitt

T. E.

Silva

P. A.

McGee

Angell

K. E.

(1998). Development of depression from preadolescence to young adulthood: Emerging gender differences in a 10-year longitudinal study. Journal of Abnormal Psychology, 107, 128–140.

35.

Hassim

Wagner

(2013). Considering the cultural context in psychopathology formulations. South African Journal of Psychiatry, 19, 4–10.

36.

Hollifield

Katon

Spain

Pule

(1990). Anxiety and depression in a village in Lesotho, Africa: A comparison with the United States. British Journal of Psychiatry, 156, 343–350.

37.

Hooper

L. M.

Crusto

Huffman

L. E.

(2012). Scalar equivalence self-rated depressive symptomatology as measured by the Beck Depression Inventory-II: Do racial and gender differences in college students exist? Psychology, 3, 762–774.

38.

L.-T.

Bentler

P. M.

(1999). Cut-off criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.

39.

L.-T.

Bentler

P. M.

Kano

(1992). Can test statistics in covariance structure analysis be trusted? Psychological Bulletin, 112, 351–362.

40.

Jordaan

H. L.

Joubert

(1994). The prevalence of depression among black patients in a primary care practice. Geneeskunde: The Medical Journal, 7, 27–34.

41.

Kagee

Nel

Saal

(2014). Factor structure of the Beck Depression Inventory-II among South Africans receiving antiretroviral therapy. AIDS Care, 26, 257–262. doi:10.1080/09540121.2013.802278

42.

Koenig

L. J.

Isaacs

A. M.

Schwartz

J. A. J.

(1994). Sex differences in adolescent depression and loneliness: Why are boys lonelier if girls are more depressed? Journal of Research in Personality, 28, 27–43.

43.

Kwang-Iel

Dongen

Dae-Ho

(1999). Depressive symptoms in Koreans, Korean-Chinese and Chinese: A transcultural study. Transcultural Psychiatry, 36, 303–316.

44.

Lambert

Herman

K. C.

Bynum

Ialongo

(2009). Perceptions of racism and depressive symptoms in African American adolescents: The role of perceived academic and social control. Journal of Youth and Adolescence, 38, 519–531.

45.

Lin

H. C.

(2012). Racial/ethnic differences in depressive symptoms and treatment effect among patients with myocardial infarction from the enhancing recovery in coronary heart disease (ENRICHD) trial (Unpublished doctoral dissertation). University of Miami, Coral Gables, FL.

46.

Makhubela

M. S.

Mashegoane

(2015). Validation of the Beck Depression Inventory-II in South Africa: Factorial validity and longitudinal measurement invariance in university students. South African Journal of Psychology. Advance online publication. doi:10.1177/0081246315611016.

47.

Mosotho

N. L.

Louw

D. A.

Calitz

F. J. W.

Esterhuyse

K. G. F.

(2008). Depression among Sesotho speakers in Mangaung, South Africa. African Journal of Psychiatry, 11, 35–43.

48.

Nevitt

Hancock

G. R.

(2001). Performance of bootstrapping approaches to model test statistics and parameter standard error estimation in structural equation modeling. Structural Equation Modeling, 8, 353–377.

49.

Ngcobo

Pillay

B. J.

(2008). Depression in African women presenting for psychological services at a general hospital. African Journal of Psychiatry, 11, 133–137.

50.

Nolen-Hoeksema

Larson

Grayson

(1999). Explaining the gender difference in depressive symptoms. Journal of Personality and Social Psychology, 77, 1061–1072.

51.

Patel

Abas

Broadhead

Todd

Reeler

(2001). Depression in developing countries: Lessons from Zimbabwe. British Medical Journal, 322, 482–484.

52.

Pillay

Sargent

(1999). Relationship of age and education with anxiety, depression, and hopelessness in a South African community sample. Perceptual and Motor Skills, 89, 881-884.

53.

Pillay

A. L.

Sargent

(2003). Descriptive profile of sex and psychiatric diagnosis among rural and peri-urban clinic attenders in South Africa. Psychological Reports, 92, 595–598.

54.

Razali

S. M.

Hasanah

C. I.

(1999). Somatization of depression: Prevalence and characteristics features. International Medical Journal, 6, 27–31.

55.

Reynolds

Swartz

(1993). Professional constructions of ‘lay’ illness: ‘Nerves’ in a rural ‘coloured’ community in South Africa. Social Science & Medicine, 36, 657–663.

56.

Roberts

R. E.

Lewinsohn

P. M.

Seeley

J. R.

(1991). Screening for adolescent depression: A comparison of depression scales. Journal of the American Academy of Child and Adolescent Psychiatry, 30, 58–66.

57.

Rochat

T. J.

Tomlinson

Barnighausen

Newell

Stein

(2011). The prevalence and clinical presentation of antenatal depression in rural South Africa. Journal of Affective Disorders, 135, 362–373.

58.

Rogers

(1992). Exploratory models of illness amongst primary health care users in Mamre (Unpublished master’s thesis). University of Cape Town, South Africa.

59.

Rossouw

(1998). Die voorkoms van major depressive in Botshabelo: ‘n Epidemiologies-patogenetiese ondersoek [The prevalence of major depression in Botshabelo: An Epidemiological-pathogenical investigation] (Unpublished doctoral dissertation). University of the Free State, Bloemfontein, South Africa.

60.

Rumble

Swartz

Parry

Zwarenstein

(1996). Prevalence of psychiatric morbidity in the adult population of a rural South African village. Psychological Medicine, 26, 997–1007.

61.

Sanchez

Y. M.

Lambert

S. F.

Ialongo

N. S.

(2011). Adverse life events and depressive symptoms in African American youth: The role of control-related beliefs. Depression Research and Treatment. 2011, 1-13.

62.

Satorra

Bentler

P. M.

(1988). Scaling corrections for chi-square statistics in covariance structure analysis. In 1988 ASA Proceedings of the Business and Economic Statistics Section (pp. 308–313). Alexandria, VA: American Statistical Association. Retrieved from http://www.amstat.org/meetings/csp.cfm

63.

Satorra

Bentler

P. M.

(1994). Corrections to test statistics and standard errors in covariance structure analysis. In von Eye

Clogg

C. C.

(Eds.), Latent variables analysis: Applications for developmental research (pp. 399-419). Thousand Oaks, CA: Sage.

64.

Schlebusch

(2005). Suicidal behaviour in South Africa. Pietermaritzburg, South Africa: University of KwaZulu-Natal Press.

65.

Schuch

J. J.

Roest

A. M.

Nolen

W. A.

Penninx

B. W.

de Jonge

(2014). Gender differences in major depressive disorder: Results from the Netherlands study of depression and anxiety. Journal of Affective Disorders, 156, 156–163.

66.

Steele

G. I.

Edwards

D. J. A.

(2008). Development and validation of the Xhosa translations of the Beck Inventories: 2. Item analyses and internal consistency. Journal of Psychology in Africa, 18, 217–226.

67.

Stellenberg

E. L.

Abrahams

J. M.

(2015). Prevalence of and factors influencing postnatal depression in a rural community in South Africa. African Journal of Primary Health Care & Family Medicine, 7(1), 1–8.

68.

Strebel

Msomi

Stacey

(1999). A gender and racial epidemiological profile of public psychiatric hospitals in the Western Cape. South African Journal of Psychology, 29, 53–61.

69.

Swartz

(1998). Culture and mental health: A Southern African view. Cape Town, South Africa: Oxford University Press.

70.

Tomlinson

Swartz

Kruger

Gureje

(2007). Manifestations of affective disturbances in sub-Saharan Africa: Key themes. Journal of Affective Disorders, 102, 191–198.

71.

Vawda

N. B. M.

(2014). Depressive symptoms, burnout and the impact of events in non-professional volunteer counselors in Durban, South Africa. Journal of Psychiatry, 17, 512–513.

72.

Westaway

M. S.

Wolmarans

(1992). Depression and self-esteem: Rapid screening for Depression in Black, low literacy, hospitalized Tuberculosis patients. Social Science and Medicine, 35, 1311-1315.

73.

Whisman

M. A.

Judd

C. M.

Whiteford

N. T.

Gelhorn

H. L.

(2013). Measurement invariance of the Beck Depression Inventory-Second Edition (BDI-II) across gender, race, and ethnicity in college students. Assessment, 20, 419–428.

74.

P.-C.

(2009). Differential functioning of the Chinese version of Beck Depression Inventory-II in adolescent gender groups: Use of a multiple-group mean and covariance structure model. Social Indicators Research, 96(3), 535-550.

75.

P.-C.

(2010). Measurement invariance and latent mean differences of the Beck Depression Inventory II across gender groups. Journal of Psychoeducational Assessment, 28, 551–563.

76.

P.-C.

Huang

T.-W.

(2014). Gender-related invariance of the Beck Depression Inventory II for Taiwanese adolescent samples. Assessment, 21(2) 218–226.

77.

Yuan

K.-H.

Bentler

P. M.

(2000). Robust mean and covariance structure analysis through iteratively reweighted least squares. Psychometrika, 65, 43–58.