European Journal of Psychological Assessment

European Journal of Psychological Assessment The Beck Hopelessness Scale: Specific Factors of Method Effects? Marianna Szabó, Veronika Mészáros, Judit Sallay, Gyöngyi Ajtay, Viktor Boross, Àgnes Udvardy-Mészáros, Gabriella Vizin, and Dóra Perczel-Forintos Online First Publication, February 27, 2015. http://dx.doi.org/10.1027/1015-5759/a000240

CITATION Szabó, M., Mészáros, V., Sallay, J., Ajtay, G., Boross, V., Udvardy-Mészáros, À., Vizin, G., & Perczel-Forintos, D. (2015, February 27). The Beck Hopelessness Scale: Specific Factors of Method Effects?. European Journal of Psychological Assessment. Advance online publication. http://dx.doi.org/10.1027/1015-5759/a000240

Original Article

The Beck Hopelessness Scale Specific Factors of Method Effects? Marianna Szabó,1 Veronika Mészáros,2 Judit Sallay,2 Gyöngyi Ajtay,2 Viktor Boross,2 Àgnes Udvardy-Mészáros,2 Gabriella Vizin,2 and Dóra Perczel-Forintos2 1

School of Psychology, The University of Sydney, NSW, Australia, Department of Clinical Psychology, Semmelweis University, Budapest, Hungary

2

Abstract. The aim of the present study was to examine the construct and cross-cultural validity of the Beck Hopelessness Scale (BHS; Beck, Weissman, Lester, & Trexler, 1974). Beck et al. applied exploratory Principal Components Analysis and argued that the scale measured three specific components (affective, motivational, and cognitive). Subsequent studies identified one, two, three, or more factors, highlighting a lack of clarity regarding the scale’s construct validity. In a large clinical sample, we tested the original three-factor model and explored alternative models using both confirmatory and exploratory factor analytical techniques appropriate for analyzing binary data. In doing so, we investigated whether method variance needs to be taken into account in understanding the structure of the BHS. Our findings supported a bifactor model that explicitly included method effects. We concluded that the BHS measures a single underlying construct of hopelessness, and that an incorporation of method effects consolidates previous findings where positively and negatively worded items loaded on separate factors. Our study further contributes to establishing the cross-cultural validity of this instrument by showing that BHS scores differentiate between depressed, anxious, and nonclinical groups in a Hungarian population. Keywords: Beck Hopelessness Scale, method effect, validity, factor analysis

The concept of hopelessness was first introduced by Aaron Beck (1963), who observed that depressed individuals share a set of negative expectations concerning the self and the future. Subsequent research has shown that hopelessness is a powerful predictor of suicidal behaviors, and is associated with depression and a range of other clinical conditions (Beck, Brown, Berchick, Stewart, & Steer, 1990; Beck, Riskind, Brown, & Steer, 1988). Assessing and understanding hopelessness in the context of various forms of psychopathology is crucial, therefore. The most widely used measure of hopelessness is the Beck Hopelessness Scale (BHS; Beck, Weissman, Lester, & Trexler, 1974), a 20-item self-report instrument containing both negative (e.g., I don’t expect to get what I really want) and positive (e.g., I look forward to the future with hope and enthusiasm) statements about the future. An initial Principal Components Analysis (PCA) by the authors suggested the presence of three independent underlying dimensions. ‘‘Feelings about the Future’’ was thought to indicate affective associations, such as a lack of enthusiasm or faith, ‘‘Loss of Motivation’’ was concerned with giving up trying, and ‘‘Future Expectations’’ was thought to reflect a cognitive component of hopelessness, concerning beliefs about a dark and uncertain future. Since its first publication, the BHS has become a widely used instrument in research and clinical practice in Englishspeaking cultures, and the validity of its translated versions is beginning to be established in non-English-speaking cultures as well (Dozois & Covin, 2004). In countries with Ó 2015 Hogrefe Publishing

extremely high suicide rates, such as Hungary, the introduction and valid use of the BHS has crucial clinical implications (Perczel-Forintos, Sallai, & Rózsa, 2010). In general, testing the construct validity of this instrument and its three proposed underlying dimensions is both theoretically and clinically important. If hopelessness in fact contains three components, these might differentially predict depression, suicidal behavior, or other forms of psychopathology. For example, it has been shown that the three components have differential associations with the number of physical symptoms and with a desire for hastened death among patients with AIDS (Rosenfeld, Gibson, Kramer, & Breitbart, 2004). Nevertheless, studies that examined the factor structure of the BHS yielded inconsistent results. While some researchers (Davidson, Tripp, Fabrigar, & Davidson, 2008; Dyce, 1996; Hill, Gallagher, Thompson, & Ishida, 1988; Iliceto & Fino, 2014; Rosenfeld et al., 2004; Steer, Iguchi, & Platt, 1994) supported a three-factor model resembling that offered by Beck et al. (1974), others suggested alternative solutions containing between one and five factors (e.g., Aish & Wasserman, 2001; Nissim et al., 2009; Pompili, Tatarelli, Rogers, & Lester, 2007; Steed, 2001; Steer, Beck, & Brown, 1997; Tanaka et al., 1998; Young, Halper, Clark, Scheftner, & Fawcett, 1992). In addition, several studies found that the 20 items do not converge in an interpretable structure, and recommended the use of either 18 (Tanaka et al., 1998), 16 (Steed, 2001), 15 (Davidson et al., 2008), 14 (Steer et al., 1994), or 4 (Aish European Journal of Psychological Assessment 2015 DOI: 10.1027/1015-5759/a000240

2

M. Szabó et al.: The Beck Hopelessness Scale

& Wasserman, 2001) items, with a large variability in the placement of these items on the suggested factors. A possible reason for the heterogeneity of the findings may be that some studies tested the structure of the BHS in psychologically healthy volunteers, while others used clinical samples. It has been shown that the BHS has relatively low reliability in nonclinical samples (see Dozois & Covin, 2004), and some researchers argued that studies involving such samples have limited utility in helping us understand hopelessness in individuals who exhibit various forms of psychopathology (Pompili et al., 2007; Rosenfeld et al., 2004). Nevertheless, the results of studies involving psychiatric patients also varied. Researchers who supported a three-factor solution disagreed in the placement of some of the 20 items on the factors (Dyce, 1996; Hill et al., 1988), or included less than 20 items in their solutions (Steer et al., 1994). Others concluded that a two-factor (Kao et al., 2012; Steer et al., 1997), or a one-factor (Aish & Wasserman, 2001; Young et al., 1992) solution fit the data best. Therefore, further research involving psychiatric samples is needed to confirm the structure of the hopelessness construct. Another reason for the reported inconsistencies may be that even among those studies that included comparable samples, a variety of analytical techniques have been employed. Importantly, items on the BHS have a binary response format that is not appropriate for traditional factor analysis. Although some researchers sought to overcome this problem by converting the item format into Likert-scales (Steed, 2001; Iliceto & Fino, 2014) this strategy alters the meaning of participants’ responses and reduces comparability with other studies. Alternatively, advanced factor analytic methods that are able to handle categorical data may be used, such as mean- and variance-adjusted weighted least squares parameter estimates (WLSMV), and the weighted root mean square residual (WRMR) as an index of model fit (Muthén & Muthén, 1998–2010; Yu, 2002). So far only one clinically relevant study utilized such methods. Aish and Wasserman (2001) tested several models in a sample of Swedish suicide attempters, and concluded that a four-item one-factor solution had excellent fit. While this solution may be statistically defensible, a four-item scale may have limited utility beyond providing a quick screen for potential suicidality. Among the large variety of solutions offered so far, two consistent findings emerged. In studies proposing more than one factor, the first factor tended to account for the largest amount of variance, prompting some researchers to speculate whether the results would in fact be best interpreted as supporting a one-factor solution (e.g., Dyce, 1996), and some reviewers to conclude that the BHS is arguably a unidimensional measure of hopelessness (Dozois & Covin, 2004). However, it has also been observed that in studies reporting more than one factor the majority of positively worded items tended to load on one factor, while the second and third factors tended to contain most of the negatively worded items (Beck et al., 1974; Dyce, 1996; Hill et al., 1988). While it can be argued that the factors represent ‘‘hopefulness’’ and ‘‘hopelessness,’’ respectively (Steed, 2001), it is also possible that this pattern indicates a European Journal of Psychological Assessment 2015

methodological artifact. The BHS may in fact reflect a unitary construct of hopelessness, as well as method effects resulting from item wording. Although this possibility may help to reconcile the reported pattern of item loadings in previous studies, no research has yet tested a model that explicitly included method effects. The aim of the present study, therefore, was to examine the construct validity of the BHS in a large Hungarian clinical sample. We first tested the original three-factor model using Confirmatory Factor Analysis (CFA). We followed this analysis with an exploratory approach to allow any alternative item loadings to empirically emerge. Further, we sought to explain previous observations concerning the separation of positively and negatively worded items. To that aim, we used CFA to test a two-factor model involving the distinct factors of ‘‘hopelessness’’ versus ‘‘hopefulness,’’ and a bifactor model involving a unitary ‘‘hopelessness’’ factor and two method factors indicating positive versus negative item wording. We compared these models with a one-factor model involving a unitary construct of hopelessness that did not involve method effects. Following the factor analyses, we sought to document the convergent and discriminant validity of the Hungarian BHS by assessing its relationship with depression and anxiety, as well as its ability to differentiate among various diagnostic groups.

Method Participants The factor analyses involved data from 905 clinic-referred individuals (72.33% women, M age = 35.97, SD = 13.52). A separate sample of 100 psychologically healthy participants (65% women, M age = 30.93, SD = 8.87 years) served as a comparison group to test the discriminant validity of the BHS; data from these participants were not included in the factor analyses. The clinic-referred sample (N = 905) was recruited from all successive intakes at two outpatient clinical psychology services during 2002– 2011. Primary diagnoses were established according to ICD-10 (WHO, 2013). Of the total group, 362 participants were diagnosed with Mood Disorders. A further 440 participants were diagnosed with Neurotic, Stress-related, and Somatoform disorders, for example, phobic or other anxiety disorders (n = 156), mixed-anxiety depression (n = 103), or adjustment disorders (n = 113). Additionally, 103 individuals received diagnoses for a variety of other clinical problems, including personality disorders (n = 43), eating disorders (n = 17), or schizophrenia (n = 10).

Materials The Beck Hopelessness Scale (Beck et al., 1974) consists of 20 items. Respondents use a true/false format to indicate whether they agree with each statement, referring to the Ó 2015 Hogrefe Publishing


previous week. The BHS was translated into Hungarian by the last author. This was followed by back-translation by a native English speaker who also speaks fluent Hungarian. Differences between the two translations were discussed and corrected. Cronbach’s alpha for the Hungarian version was .91 (Perczel-Forintos et al., 2010). The Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961) is a 21-item selfreport inventory for measuring the severity of depression. Participants rate each item on a 4-point scale (0 = not true at all, 3 = very much true). Reliability and validity indices for the BDI have been reported by Beck et al. (1988). We used the Hungarian version of the BDI (PerczelForintos, Ajtay, Barna, Kiss, & Komlósi, 2012), its internal consistency in the current sample was 0.92. The Beck Anxiety Inventory (BAI; Beck, Epstein, Brown, & Steer, 1988) consists of 21 items designed to measure symptoms of anxiety. Participants indicate the extent to which they experienced these symptoms during the past week, using a 4-point response format (0 = not at all, 3 = severely). We used the Hungarian version of the BAI (Perczel-Forintos et al., 2012). Internal consistency in the current sample was 0.71.

Procedure The study protocol was approved by the Institutional Review Board. Participants completed the questionnaires while waiting for their first interview with a clinical psychologist, as a routine part of initial assessment at the participating clinics. Diagnostic information was obtained at intake interview by intern clinical psychologists under the supervision of the last author. Diagnoses were established according to ICD-10 (WHO, 2013).

Results Factor Analyses First, we used CFA to evaluate the original three-factor structure of the BHS in the total clinical sample. Items were

3

set to load on the Affective, Cognitive, and Motivational factors as originally specified by Beck et al. (1974). We followed this analysis by testing a less restricted, exploratory three-factor model where all items were allowed to freely load on any factor. This analysis was expected to reveal whether the 20 items would converge in a three-factor structure that would approximate the original model proposed by Beck et al. Next, we used CFA to test further hypothetical solutions. For this series of analyses, the sample was randomly split into two subgroups; Subgroup 2 was used to cross-validate the results obtained in Subgroup 1. A model including two distinct content-related factors denoting ‘‘hopelessness’’ versus ‘‘hopefulness’’ was tested first, followed by a one-factor solution reflecting a single underlying hopelessness construct. Finally, we tested a bifactor model where one content-related factor (hopelessness) and two method-related factors (positive versus negative wording) were specified. For each factor analysis, we used Mplus 7.2 and employed WLSMV parameter estimates (Muthén & Muthén, 1998–2010). Goodness of fit was evaluated using chi-square, RMSEA, CFI, TLI, and the WRMR, an index recommended as an important addition to model fit assessment when binary variables are used (Yu, 2002). Acceptable model fit was defined by RMSEA < .05, CFI > .96, TLI > .95, and WRMR < 1.00 (Yu, 2002).

Testing the Original Three-Factor Model We first tested a three-factor model with item placements as specified by Beck et al. (1974) in the total clinical sample (N = 905). The model fit indices presented in the first row of Table 1 show that although model fit was adequate according to some of the indices, the RMSEA and the WRMR indicated inadequate fit. Standardized factor correlations were 0.848 between the Motivational and Affective factors, 0.977 between the Motivational and Cognitive factors, and 0.925 between the Cognitive and Affective factors, indicating poor differentiation. All items loaded significantly on their allocated factors, but item 10 had a very low loading of 0.156. Modification Indices suggested 12 instances of potential misspecification of item placements.

Table 1. Model fit indices for all CFA models tested in the clinical sample Total sample (N = 905) Three-factor 20-item Subsample 1 (N = 442) One-factor 20-item Two-factor 20-item Bifactor 20-item Subsample 2 (N = 463) One-factor 18-item Two-factor 18-item Bifactor 18-item

v2

RMSEA

652.62

.057

N/A 367.914 319.337 382.256 259.533 186.912

p RMSEA .05

CFI

TLI

WRMR

.008

.965

.960

1.498

.052 .051

.347 .443

.971 .975

.967 .969

1.125 0.987

.063 .045 .036

.002 .845 .995

.954 .981 .989

.948 .978 .986

1.295 1.041 0.810

Note. One-factor 18-item and two-factor 18-item = BHS items 10 and 13 excluded. Ó 2015 Hogrefe Publishing

European Journal of Psychological Assessment 2015

4


Table 2. Oblimin rotated loadings and factor correlations obtained from EFA of a three-factor model in the total clinical sample (n = 905) Original item numbers and abbreviated content 16. Never get what I want 14. Things won’t work out 20. No use in really trying 9. Just don’t get the breaks 17. Very unlikely to get real satisfaction 2. Might as well give up 11. Ahead of me is unpleasantness 12. Don’t expect to get what I really want. 7. Future seems dark 15. Have great faith in the future 1. Look forward to the future with hope 10. Experiences prepared well for future 6. Expect to succeed 8. Expect to get more good things 19. Look forward to more good times 13. Expect to be happier than now 18. Future seems vague and uncertain 3. Helped knowing can’t stay that way 5. Have enough time to accomplish things 4. Can’t imagine life in 10 years Factor correlations Factor 1 Factor 2 Factor 3

Factor 1

Factor 2

Factor 3 0.250

0.912 0.872 0.871 0.777 0.766 0.737 0.663 0.616 0.562

0.353

0.205 0.213 0.382 0.822 0.735 0.653 0.603 0.483 0.473 0.234 0.468 0.408 0.393 0.234

1.000 0.660 0.031

1.000 0.094

0.423 0.319 0.325 0.349 0.306 0.233

0.572 0.493 0.412 0.271

Original factor Motivational Cognitive Motivational Motivational Motivational Motivational Motivational Motivational Cognitive Affective Affective Cognitive Affective Cognitive Affective Affective Cognitive Motivational Affective Cognitive

1.000

Note. Loadings < 0.20 are omitted for clarity. Factor loadings >.40 are in bold.

Noting the large number of misspecifications indicated, we proceeded to explore a model where a correlated three-factor structure was specified but items were allowed to freely load on any factor. This less restricted three-factor model fit the data well (v2 = 304.183, df = 133, p < .0001; RMSEA = .038, CFI = .988, TLI = .982), but the pattern of item loadings differed substantially from the original pattern. As Table 2 shows, the first factor was defined primarily by the original ‘‘Motivational’’ items, while the second factor included a mixture of ‘‘Affective’’ and ‘‘Cognitive’’ items, with item 10 double loading on the first and second factor. The third factor contained only three items, and was primarily defined by item 13.

Alternative Models For the next set of analyses, the clinical sample was split into two random groups. Initial CFAs were conducted in subsample 1 (n = 442), followed by cross-validating the obtained models in subsample 2 (n = 463). There were no significant differences between the two subsamples regarding background demographic variables. Subsample 1 We first tested a two-factor model where all positively worded items were set to load on Factor 1 (‘‘Hopefulness’’) European Journal of Psychological Assessment 2015

and all negatively worded items on Factor 2 (‘‘Hopelessness’’). Fit indices presented in Table 1 show that although the TLI and CFI suggested adequate fit, RMSEA and WRMR did not meet recommended thresholds. Item loadings for the two-factor model are shown in Table 3. Although all items loaded significantly on their allocated factors, item 10 had a very low loading. The only Modification Index above 10 for this model suggested that item 10 load on both factors. Items 13 and 5 also had relatively low loadings, although item 5 approached acceptable threshold. Finally, a high factor correlation indicated little differentiation between the two factors. Nevertheless, a one-factor model where all items were set to load on a single ‘‘hopelessness’’ factor failed to converge in this subsample. Next, we tested a bifactor model where all items were set to load on a single content-related factor (‘‘hopelessness’’), all positively worded items were set to load on Method Factor 1, and all negatively worded items on Method Factor 2. The fit indices for this model are also presented in Table 1, and item loadings are given in Table 4. Most items loaded highly on the content factor, and less strongly on the method factors. Consistent with the other models, item 10 failed to load significantly on the content factor, and item 13 exhibited the second lowest loading. The only Modification Index over 10 suggested that item 13 also load on the second method factor (tapping into both positive and negative method-related effects). Ó 2015 Hogrefe Publishing


5

Table 3. Standardized factor loadings from CFA testing a two-factor model in subgroup 1 (n = 442) Original item numbers and abbreviated content

Factor 1

6. Expect to succeed 1. Look forward to the future with hope 19. Look forward to more good times 15. Great faith in the future. 8. Expect to get more good things 3. Helped knowing can’t stay that way 5. Enough time to accomplish things 13. Expect to be happier than now. 10. Experiences prepared well for future. 7. Future seems dark 20. No use in really trying 11. Ahead of me is unpleasantness 12. Don’t expect to get what I really want 16. Never get what I want 9. Just don’t get the breaks 18. Future seems vague and uncertain 2. Might as well give up 14. Things won’t work out 17. Very unlikely to get real satisfaction 4. Can’t imagine life in 10 years Factor correlations Factor 1 Factor 2

Factor 2

0.889 0.828 0.803 0.798 0.698 0.524 0.492 0.427 0.218 0.903 0.893 0.845 0.815 0.800 0.798 0.790 0.777 0.761 0.706 0.592 1.000 0.861

1.000

Note. Relatively low factor loadings are in italics.

Table 4. Standardized factor loadings from CFA of orthogonal bifactor model specifying one content-related factor and two method-related factors in subgroup 1 (n = 442) Original item numbers and abbreviated content 6. 19. 1. 15. 8. 3. 5. 13. 10. 7. 11. 18. 20. 12. 9. 2. 14. 16. 17. 4.

Factor 1 (content)

Factor 2 (method)

0.826 0.744 0.739 0.692 0.635 0.468 0.438 0.392 0.131 0.908 0.817 0.809 0.799 0.788 0.781 0.719 0.680 0.660 0.659 0.598

0.214 0.229 0.397 0.541 0.276 0.262 0.247 0.128 0.458

Expect to succeed Look forward to more good times Look forward to the future with hope Great faith in the future. Expect to get more good things Helped knowing can’t stay that way Enough time to accomplish things Expect to be happier than now Experiences prepared well for future Future seems dark Ahead of me is unpleasantness Future seems vague and uncertain No use in really trying Don’t expect to get what I really want. Just don’t get the breaks Might as well give up Things won’t work out Never get what I want. Very unlikely to get real satisfaction Can’t imagine life in 10 years.

Factor 3 (method)

0.071 0.212 0.029 0.455 0.206 0.160 0.330 0.420 0.616 0.280 0.021

Note. Relatively low factor loadings on the content factor are in italics.

Cross-Validating: Subsample 2 Each of the models tested so far indicated that item 10 did not contribute substantially to the definition of the Ó 2015 Hogrefe Publishing

underlying construct, and item 13 also had low or inconsistent loadings. Therefore, we proceeded to cross-validate these models in subsample 2 with the omission of items 10 and 13. A series of CFAs in subsample 2 (n = 463) European Journal of Psychological Assessment 2015

6


Table 5. Descriptive statistics for the 18-item BHS in a nonclinical sample and in separate diagnostic groups Nonclinical Phobic and anxiety disorders Mixed-anxiety depression Mood disorders

n

Mean

SD

a

100 148 93 340

2.03 5.82 8.85 10.41

2.11 4.34 4.47 4.92

.678 .869 .856 .884

tested the model fit of the one-factor, two-factor, and bifactor solutions described above, including 18 items only. As shown in Table 1, although the fit indices improved for each model, only the 18-item bifactor model had excellent fit according to all indices.

Reliability and Validity of the 18-Item Hungarian BHS We proceeded to calculate total scores for the 18-item version of the BHS to explore its reliability and validity. The Mean BHS score was 8.13 (SD = 5.13, N = 844), and internal consistency estimate (Cronbach’s alpha) was .898 in the total clinical sample. To compare with traditional Cronbach’s alpha values, we calculated model-based reliability indices (Brunner, Nagy, & Wilhelm, 2012). Coefficient omega, indicating the proportion of variance attributable to a blend of the global hopelessness factor and the specific method factors was .789. Coefficient omega hierarchical, indicating the proportion of variance in hopelessness scores that is attributable to the content factor only, was .620. In the clinical sample as a whole, the 18-item BHS scores had a correlation of 0.69 with the BDI ( p < .001, N = 766) and 0.36 with the BAI ( p < .001, N = 841). Finally, we compared mean BHS scores among groups of individuals diagnosed with mood disorders, mixed-anxiety depression, phobic and other anxiety disorders, and psychiatrically healthy individuals. Descriptive statistics are shown in Table 5. The significant overall ANOVA, F(3, 677) = 107.25, p < .001, and linear trend, F(1, 677) = 303.59, p < .001, indicated that BHS scores increased gradually among these groups, with the nonclinical group obtaining the lowest and the depressed group the highest scores. Internal consistency estimates were similar among the clinical groups, but appeared relatively lower in the nonclinical group.

Discussion The present study explored the factor structure and convergent and discriminant validity of the Beck Hopelessness Scale (Beck et al., 1974) in a large Hungarian clinical sample. We used CFA to test the original three-factor model hypothesized to reflect ‘‘Feelings about the Future,’’ ‘‘Loss European Journal of Psychological Assessment 2015

of Motivation,’’ and ‘‘Future Expectations,’’ followed by exploring a less restricted three-factor model where all items were allowed to freely load on any factor. The CFA indicated inadequate model fit and a large number of misspecifications. The exploratory model failed to replicate the original pattern of item loadings. Only a few items loaded on their expected factors, and the third factor was defined by a very small number of items. The majority of negatively worded items loaded on the first factor, and the majority of positively worded items loaded on the second. We proceeded to test further hypotheses using CFA in two randomly established subsamples of clinic-referred individuals. We tested a two-factor, a one-factor, and a bifactor model in the first subsample. Each of the models tested in this series of analyses suggested that items 10 and 13 contributed little to the definition of the underlying construct. Therefore, we proceeded to cross-validate these models in the second subsample with the inclusion of 18 items only. These analyses indicated that the 18-item bifactor model provided the best fit for the data. We concluded that the BHS reflects a unitary construct of hopelessness, with method effects resulting from item wording. Our findings considering the factor structure of the BHS have significant theoretical and practical implications. The importance of modeling method effects to improve our understanding of various constructs has been increasingly emphasized in the literature. Our findings are consistent with previous studies where such effects have been found to underlie the structure of other well-known instruments, for example, the Rosenberg Self-Esteem Scale (Tomas & Oliver, 1999). An incorporation of method effects in the structure of the BHS consolidates previous findings, where either one-factor solutions were reported (Young et al., 1992), or positively and negatively worded items loaded on separate factors (e.g., Hill et al., 1988; Rosenfeld, et al., 2004). In addition to providing an explanation for previous results and a clearer delineation of hopelessness as a unitary construct, modeling both content and method-related factors allowed a more precise assessment of the reliability of the BHS in this study. Our results show that Cronbach’s alpha overestimates the reliability of the BHS, and provide a more precise assessment of the amount of variance accounted for by the underlying construct itself. The results also add to a growing body of evidence questioning the three-factor structure of the BHS proposed by Beck et al. (1974). Considering the limited support for this structure, researchers and clinicians need to exercise caution when using subscale scores derived from the three putative components. Although differential associations of the components with other measures, for example, with a desire for hastened death have previously been reported, the majority of correlations obtained in that study were in fact similar for all three components (Rosenfeld et al., 2004). A more recent study also reported that the three components were associated with depression, aggression, and various personality traits to similar degrees (Iliceto & Fino, 2014). Therefore, a single score assessing the unitary construct of hopelessness would be more advisable to use both in research and clinical practice. In the present study, Ó 2015 Hogrefe Publishing


total scores derived from the 18-item Hungarian language BHS had a stronger relationship with depression than with anxiety, and were able to discriminate between individuals diagnosed with depression, mixed-anxiety depression, phobic and other anxiety disorders, and those with no diagnosed disorders. These results are consistent with those of previous findings (Beck, Riskind, et al., 1988), and support the validity and potential utility of the BHS as a measure of a unitary hopelessness construct in a Hungarian clinical population. Similar to previous studies (e.g., Aish & Wasserman, 2001; Steer et al., 1994), our results also indicated that not all 20 items of the BHS contribute substantially to the definition of hopelessness. Each model we tested indicated that items 10 and 13 were weakly or inconsistently associated with the underlying construct. Item 10 (My experiences prepared me well for future) had the lowest loading in the original PCA reported by Beck et al. (1974) and has been identified as a weak item by others (e.g., Steer et al., 1994). It appears that this statement may be understood as having either an optimistic or a pessimistic tone, and may therefore be interpreted inconsistently by the respondents. Indeed, in our clinical experience, respondents often ask for clarification for this item when completing the BHS. The second item we selected for deletion, item 13 (When I think about the future, I expect to be happier than I am now), has not previously been identified as a weak item. Therefore, it is possible that this item may have been affected by the translation process. For example, the Hungarian word used to translate the English expression ‘‘I expect . . . (to be happier)’’ can also be understood as ‘‘I hope . . . (to be happier),’’ lending a less clearly positive meaning to this item. Considering the nature and aims of our study, it cannot be ruled out that other language or cultural differences affected our results. While the results make both theoretical and statistical sense and are largely consistent with previous data, future studies in other cultures are needed to establish whether the BHS indeed reflects a unitary construct of hopelessness with method effects, rather than two dimensions of ‘‘hopelessness’’ versus ‘‘hopefulness,’’ or the three commonly used dimensions initially offered by Beck et al. (1974). Replication of the inclusion and exclusion of particular items is also necessary in varied cultural and language groups, and in both clinic-referred and psychiatrically healthy individuals. For example, because we know that hopelessness is associated with wide range of psychopathologies, we included a heterogeneous psychiatric sample to study this phenomenon. Further studies may now explore the same question in more homogenous groups, in particular in depressed and suicidal individuals. Although total BHS scores differentiated between various diagnostic groups and a nonclinical sample in our study, it also needs to be acknowledged that internal consistency appeared to be substantially lower in the nonclinical sample, compared to the clinic-referred groups. This finding supports the contention that the construct or measurement of hopelessness needs to be considered separately in clinical versus nonclinical samples, and calls for future research to examine possible differences in the experience of hopelessness in Ó 2015 Hogrefe Publishing

7

psychiatrically healthy and in clinic-referred individuals (Dozois & Covin, 2004; Rosenfeld et al., 2004). Finally, because it was not possible to establish a standardized environment for data collection, uncontrollable environmental factors in the data collection process may have influenced our results and further underline the need for replication. In the meantime, our study is the first to use modern factor analytic methods in a large clinical sample to reconcile previously inconsistent results concerning the BHS. Our data have shown that hopelessness is a unitary construct, and that method effects associated with positive and negative item wording need to be acknowledged when using this instrument in research and clinical practice.

References Aish, A. M., & Wasserman, D. (2001). Does Beck’s Hopelessness Scale really measure several components? Psychological Medicine, 31, 367–372. Beck, A. T. (1963). Thinking and depression. Archives of General Psychiatry, 9, 326–333. Beck, A. T., Brown, G., Berchick, R. J., Stewart, B. L., & Steer, R. A. (1990). Relationship between hopelessness and ultimate suicide: A replication with psychiatric outpatients. American Journal of Psychiatry, 147, 190–195. Beck, A. T., Epstein, N., Brown, G., & Steer, R. A. (1988). An inventory for measuring clinical anxiety: Psychometric properties. Journal of Consulting and Clinical Psychology, 56, 893–897. Beck, A. T., Riskind, J. H., Brown, G., & Steer, R. A. (1988). Levels of hopelessness in DSM-III disorders: A partial test of content specificity in depression. Cognitive Therapy and Research, 12, 459–469. Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry, 4, 561–571. Beck, A. T., Weissman, A., Lester, D., & Trexler, L. (1974). The measurement of pessimism: The hopelessness scale. Journal of Consulting and Clinical Psychology, 42, 861–865. Brunner, M., Nagy, G., & Wilhelm, O. (2012). A tutorial on hierarchically structured constructs. Journal of Personality, 80, 796–846. Davidson, M. A., Tripp, D. A., Fabrigar, L. R., & Davidson, P. R. (2008). Chronic pain assessment: A seven-factor model. Pain Research & Management, 13, 299–308. Dozois, D., & Covin, R. (2004). The Beck Depression Inventory-II (BDI-II), the Beck Hopelessness Scale (BHS) and Beck Scale for Suicidal Ideation (BSS). In M. Hersen (Ed.), Comprehensive handbook of psychological assessment (Vol. 2, Ch. 5, pp. 50–85). Hoboken, NJ: Wiley. Dyce, J. A. (1996). Factor structure of the Beck Hopelessness Scale. Journal of Clinical Psychology, 52, 555–558. Hill, R. D., Gallagher, D., Thompson, L. W., & Ishida, T. (1988). Hopelessness as a measure of suicidal intent in the depressed elderly. Psychology and Aging, 3, 230–232. Iliceto, P., & Fino, E. (2014). Beck Hopelessness Scale (BHS): A second-order confirmatory factor analysis. European Journal of Psychological Assessment. Advance online publication. http://dx.doi.org/10.1027/1015-5759/a000201 Kao, Y. C., Liu, Y. P., & Lu, C. W. (2012). Beck Hopelessness Scale: Exploring its dimensionality in patients with schizophrenia. Psychiatric Quarterly, 83, 241–255. Muthén, L. K., & Muthén, B. O. (1998-2010). Mplus User’s Guide (Sixth Edition). Los Angeles, CA: Muthén & Muthén. European Journal of Psychological Assessment 2015

8


Nissim, R., Flora, D. B., Cribbie, R. A., Zimmerman, C., Gagliese, L., & Rodin, G. (2009). Factor structure of the Beck Hopelessness Scale in individuals with advanced cancer. Psychooncology, 19, 255–263. Perczel-Forintos, D., Ajtay, G. Y., Barna, C. S., Kiss, Z. S., & Komlósi, S. (2012). Kérdo}ívek, becslo}skálák a klinikai pszichológiában [Questionnaires and Assessment Methods in Clinical Psychology]. Budapest: Semmelweis Kiadó. Perczel-Forintos, D., Sallai, J., & Rózsa, S. (2010). Adaptation of the Beck Hopelessness Scale in Hungary. Psychological Topics, 19, 307–321. Pompili, M., Tatarelli, R., Rogers, J. R., & Lester, D. (2007). The Hopelessness Scale: A factor analysis. Psychological Reports, 100, 375–378. Rosenfeld, B., Gibson, C., Kramer, M., & Breitbart, W. (2004). Hopelessness and terminal illness: The construct of hopelessness in patients with advanced AIDS. Palliative & Supportive Care, 2, 43–53. Steed, L. (2001). Further validity and reliability evidence for Beck Hopelessness Scale scores in a nonclinical sample. Educational and Psychological Measurement, 61, 303–316. Steer, R. A., Beck, A. T., & Brown, G. K. (1997). Factors of the Beck Hopelessness Scale: Fact or artifact? Multivariate Experimental Clinical Research, 11, 131–144. Steer, R. A., Iguchi, M. Y., & Platt, J. J. (1994). Hopelessness in IV drug-users not in treatment and seeking HIV testing and counseling. Drug and Alcohol Dependence, 34, 99–103. Tanaka, E., Sakamoto, S., Ono, Y., Fujihara, S., & Kitamura, T. (1998). Hopelessness in a community population: Factorial structure and psychosocial correlates. Journal of Social Psychology, 138, 581–590.

European Journal of Psychological Assessment 2015

Tomas, J. M., & Oliver, A. (1999). Rosenberg’s Self-Esteem Scale: Two factors or method effects. Structural Equation Modeling: A Multidisciplinary Journal, 6, 84–98. WHO. (2013, November 13). International classification of diseases. [web page]. Retrieved from http://www.who.int/ classifications/icd/en/ Young, M. A., Halper, I. S., Clark, D. C., Scheftner, W., & Fawcett, J. (1992). An item-response theory evaluation of the Beck Hopelessness Scale. Cognitive Therapy and Research, 16, 579–587. Yu, C.-Y. (2002). Evaluating cutoff criteria of model fit indices for latent variable models with binary and continuous outcomes. Unpublished doctoral dissertation. University of California, Los Angeles, CA.

Date of acceptance: August 22, 2014 Published online: February 27, 2015

Dóra Perczel-Forintos Department of Clinical Psychology Semmelweis University Tömõ street 25–29 1083 Budapest Hungary E-mail [email protected]

Ó 2015 Hogrefe Publishing