- Email: [email protected]
Testing comparison models of DASS-12 and its reliability among adolescents in Malaysia
Testing comparison models of DASS-12 and its reliability among adolescents in Malaysia Zubaidah Jamil Osman, Firdaus Mukhtar, Hairul Anuar Hashim, Latiffah Abdul Latiff, Sherina Mohd Sidik, Hamidin Awang, Normala Ibrahim, Hejar Abdul Rahman, Siti Irma Fadhilah Ismail, Faisal Ibrahim, Esra Tajik, Norlijah Othman PII: DOI: Reference:
S0010-440X(14)00098-4 doi: 10.1016/j.comppsych.2014.04.011 YCOMP 51291
To appear in:
Comprehensive Psychiatry
Received date: Revised date: Accepted date:
1 November 2013 13 April 2014 22 April 2014
Please cite this article as: Osman Zubaidah Jamil, Mukhtar Firdaus, Hashim Hairul Anuar, Latiff Latiffah Abdul, Sidik Sherina Mohd, Awang Hamidin, Ibrahim Normala, Rahman Hejar Abdul, Ismail Siti Irma Fadhilah, Ibrahim Faisal, Tajik Esra, Othman Norlijah, Testing comparison models of DASS-12 and its reliability among adolescents in Malaysia, Comprehensive Psychiatry (2014), doi: 10.1016/j.comppsych.2014.04.011
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
SC
RI P
T
ACCEPTED MANUSCRIPT
MA NU
Testing comparison models of DASS-12 and its reliability among adolescents in Malaysia
Zubaidah Jamil Osman1, Firdaus Mukhtar1, Hairul Anuar Hashim2, Latiffah Abdul Latiff1,
ED
Sherina Mohd Sidik1, Hamidin Awang1, Normala Ibrahim1, Hejar Abdul Rahman1,
Faculty of Medicine and Health Sciences, Universiti Putra Malaysia, Malaysia Sport Science Unit, School of Medicine, Universiti Sains Malaysia, Malaysia
AC
2
CE
1
PT
Siti Irma Fadhilah Ismail1, Faisal Ibrahim1, Esra Tajik1 & Norlijah Othman1
Corresponding author:-
ACCEPTED MANUSCRIPT Dr. Firdaus Mukhtar, Department of Psychiatry, Universiti Putra Malaysia, 43400 Serdang,
AC
CE
PT
ED
MA NU
SC
RI P
T
Malaysia (Email: [email protected]; Tel: 03 8947 2543; Fax: 03 8941 4629).
ACCEPTED MANUSCRIPT Abstract Objective: The 21-item Depression, Anxiety and Stress Scale (DASS-21) is frequently used
T
in non-clinical research to measure mental health factors among adults. However, previous
RI P
studies have concluded that the 21 items are not stable for utilization among the adolescent population. Thus, the aim of this study is to examine the structure of the factors and to report
SC
on the reliability of the refined version of the DASS that consists of 12 items. Method: 2850 students (aged 13 to 17 years old) from three major ethnic in Malaysia
MA NU
completed the DASS-21. The study was conducted at 10 randomly selected secondary schools in the northern state of Peninsular Malaysia. The study population comprised secondary school students (Forms 1, 2 & 4) from the selected schools.
ED
Results: Based on the results of the EFA stage, 12 items were included in a final CFA to test the fit of the model. Using Maximum Likelihood procedures to estimate the model, the
PT
selected fit indices indicated a close model fit (χ² =132.94, df = 57, p = .000; CFI = .96; RMR = .02; RMSEA = .04). Moreover, significant loadings of all the unstandardized regression
CE
weights implied an acceptable convergent validity. Besides the convergent validity of the
AC
item, a discriminant validity of the subscales was also evident from the moderate latent factor inter-correlations, which ranged from .62 to .75. The subscale reliability was further estimated using Cronbach’s alpha and the adequate reliability of the subscales was obtained (Total= 76; Depression =.68; Anxiety = .53; Stress = .52). Conclusion: The new version of the 12-item DASS for adolescents in Malaysia (DASS-12) is reliable and has a stable factor structure, and thus it is a useful instrument for distinguishing between depression, anxiety and stress.
Keywords: DASS, depression, anxiety, adolescence, factor analysis, validation
ACCEPTED MANUSCRIPT 1. Introduction In Malaysia, three years ago National Health and Morbidity Survey [1] reported that
T
the overall prevalence of mental health problems among children between five to 16 years
RI P
using the Reporting Questionnaire for Children (RQC) was 20%. The age group of below 16 had the highest prevalence of 22.2% compared to the age group of 10-14 years (20.6%) and
SC
5-9 years (19.1%), respectively. The prevalence is worrying and highlights an on-going need for the continued strengthening and upgrading of Child and Adolescent Mental Health
MA NU
Services in Malaysia. Improved detection and recognition of early difficulties as well as more comprehensive early intervention programs, particularly among at-risk population groups are clearly needed. This survey has highlighted the extent of general mental health issues
ED
currently existing in the community, reflecting the unmet needs of the population nationwide. In order to improve detection and early referral of children and adolescents with
PT
mental health problems, it is suggested that a reliable tool be used for fast screening at community level [1]. The early identification of emotional distress, such as stress and
CE
anxiety, may be an important step in preventing the risk of the development of clinically
AC
significant psychological disorders. Importantly, the accurate identification of these negative effects relies on a psychometrically sound measurement instrument. Twenty years ago, Clark and Watson [2] highlighted that although anxiety and depression are phenomenologically distinct, it has been proven to be very difficult to distinguish between these constructs by empirical means, either by using clinical ratings or self-report measures. In measuring these negative effects, the Depression, Anxiety and Stress Scale-21 (DASS-21) [3] has been reported to be quite successful in terms of its stable factor structure and its psychometric properties, especially among the adult population, both in research and clinical services, in most types of languages. The Depression Scale assesses dysphoric mood states, including self-depreciation, lack of interest/involvement,
ACCEPTED MANUSCRIPT hopelessness, and anhedonia. The Anxiety Scale assesses arousal states, including autonomic arousal, muscular tension, and the anxious effect. Finally, the Stress Scale is reported to be
T
fairly consistent with the stressor and general tension [3].
RI P
However, a recent study by Hashim and colleagues [4] argued that the DASS-21 is stable or easily utilized in order to understand the younger population, especially adolescents.
SC
Hashim and colleagues concluded that a one-dimensional structure is consistently obtained and also suggested that it is important to revise the items to be utilised according to standard
MA NU
colloquial language, especially if this measure is to be used in a multi-ethnic population. An assessment of the psychometric properties of this measure among young respondents is critical given the fact that the youth population may not be able to report their experiences
ED
using items that have been designed for the adult population. Given the inconsistencies in relation to the factorial structure of the DASS-21among the younger population and the
PT
potential benefits that can be obtained from developmental research into the DASS-21, it is critically important to evaluate the psychometric properties of DASS-21 in younger
CE
respondents [4, 5].
AC
The original principal component analysis of the DASS items revealed that a stable 3factor solution of depression, anxiety and stress was the optimal fit [3]. Subsequent research has replicated the 3-factor solution in the adolescent population. At this stage, it has been clearly documented that the use of the DASS-21 is well-supported for the adult population, but it does not apply much to children and adolescents. Several studies have argued that the experience of depression and anxiety is expressed or experienced differently, and what more given the language discrepancy between English and other linguistic or cultural backgrounds [5]. When a self-report scale like the DASS-21 is used with children and adolescents, it is not measuring the three discriminate emotional dimensions as it does in adults. Rather, it
ACCEPTED MANUSCRIPT appears to be measuring a single distress dimension and it is more closely related to depression, to anxiety , or to general distress, or even that it is possible to draw distinctions
T
between these states. Herman and collegues [6] stated that the DASS-21 is a useful index of
symptoms of anxiety or depression as a core feature.
RI P
the severity of a mixed affective and/or oppositional/disruptive behavioral disorder which has
SC
In a study conducted by Szabó and Lovibond [7] involving 7- to 14- year old children and adolescents, a 2-factor structure incorporating anxiety and a combined
MA NU
depression-stress item best fitted the data. Similarly, in another study involving 11- to 15year old adolescents conducted by Duffy and collegues [8] the findings did not support the original 3-factor structure (depression-anxiety-stress) of DASS-21. Instead, the researchers
ED
only observed a reasonable model fit when the items of the DASS-21 were grouped into 2 factors, Physiological Hyperarousal and General Negative Effect, while allowing some
PT
error terms to co-vary. On the basis of the results, it was speculated that young people might not yet have developed the ability to differentiate between depression, anxiety, and stress.
CE
Instead, adolescents are likely to report their experience as a more general, negative
AC
experience and physiological arousal. The inconsistent factor structure of the DASS-21 in the younger population is also evident in two studies that are more recent. In one of those studies, Szabó [9] revealed that the DASS-21 was best used as a measure of anxiety and depression instead of the 3-factor structure in 11- to 15-year old children and adolescents. Contrary to Szabó [9], Tully and colleagues [10] observed that the DASS-21 was best seen as a measure of depression, physiological arousal, and general negative effect. Consistent across these studies was the view of the researchers that emotional differentiation is still developing in younger respondents and they may not be able to fully appreciate the differentiation between depression, anxiety, and stress as reflected in the DASS-21 items [4]. Furthermore, Szabó [9] contended that the DASS contained several expressions and words that might not be familiar
ACCEPTED MANUSCRIPT to adolescents. Thus, this accounted for the failure to obtain a clear factor structure when using the DASS-21 among younger respondents.
T
1.1 Aims of study
RI P
The aims of this study are (a) to test the goodness fit of the current data against previous models of the DASS; (b) to conduct an exploratory and confirmatory analysis and to
SC
report on its reliability in a sample of adolescents in Malaysia. It was hypothesized that the DASS-21 would exhibit similar factor structures as in the original version and exhibit
AC
CE
PT
ED
MA NU
acceptable reliability indices.
ACCEPTED MANUSCRIPT 2. Material and Methods 2.1 Study design and sampling
T
The participants were 2980 secondary school students (boys =46.9% and girls = 53.1%)
RI P
ranging in age from between 13 to 17 years (13 years old = 31.9%, 14 years old 31.5%, 15 years old 2%, 16 years old 33.1% and 17 years 1.5%, respectively). The majority of them
SC
were Malays (53.9%), and the rest were Chinese, Indians and other races (32.2, 11.8 and 2.1 per cent, respectively). The Muslims were the most represented in this study (54.3%), while
MA NU
the rest comprised Buddhists (28.4%), Hindus (10.8%), Christians (5.7%) and other religions (0.8%). In terms of levels of study, the percentages were quite balanced at 35.3%, 30.3% and 34.4% of the participants being in Form One, Form Two and Form Four, respectively.
ED
The study was conducted at 10 randomly selected secondary schools in the Pasir Gudang district. The study population comprised secondary school students (Forms 1, 2 & 4)
PT
from the selected schools. A multistage cluster sampling method was used in the study and a total number of 3000 students were selected based on the estimated sample size (Lwanga &
CE
Lemeshow, 1991). The inclusion criteria included all Forms 1, 2 and 4 students who
AC
consented to the study. Students who consented to participate in the study and were permitted to do so by their parents were given the questionnaire to be completed. The administration of the questionnaire took place in a classroom setting and was supervised by the second and third authors. The participants spent an average of 10 minutes to complete the questionnaire.
2.2 Research instrument Depression, Anxiety and Stress Scale-21. The DASS-21 [3] is a measure of three distinct negative effects: depression, stress and anxiety. The respondents were to indicate the extent to which they experienced each of the symptoms on a 4-point Likert scale ranging from 0 (Did not apply to me at all) to 3 (Applied to me very much or most of the time). The Malay
ACCEPTED MANUSCRIPT translated version of the DASS-21 was used in the present study. In an adult population, the
T
initial assessment of the translated version revealed adequate validity and reliability indices.
RI P
2.3 Ethical approval
The Ethical Committee of the University and the Ministry of Education, Malaysia granted the
MA NU
their parents prior to the data collection.
SC
permission to conduct the study. Written consent was also obtained from the respondents and
2.4 Data Analysis
Four statistical procedures were utilized. Descriptive statistics were used for the data screening. An exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) were
ED
performed to establish the factorial validity of the measure. In addition, Cronbach’s alpha
PT
coefficients were computed to evaluate the reliability of the subscales. For EFA, the principal axis factoring with orthogonal rotation (Varimax) was used and the
CE
main criterion for the initial number of factors to be extracted was eigenvalues greater than 1.
AC
For the CFA, multiple fit indices were used. The selected indices were the chi-square statistic (χ ²), the root mean square residuals (RMR)[11],the goodness fit index (GFI), the comparative fit index (CFI)[11], and the root mean square error of approximation (RMSEA)[12]. A good model fit is indicated by values of .90 or higher for the GFI and CFI. RMR values of less than .05 reflect a close fit, while values of .1 or lower indicate a reasonable fit for the RMR [11]. For the RMSEA, values of .05 or lower indicate a close fit while values less than .08 indicate an acceptable fit [12]. Additionally, the expected cross validation index (ECVI), and the Parsimony-Adjusted CFI (PCFI) were also evaluated. For ECVI, models with smaller values indicate the best potential of replication in samples of equivalent size and precision of the ECVI is presented in confidence interval. Lastly, PCFI
ACCEPTED MANUSCRIPT takes into account the complexity of the model and values above 0.70 indicates good fit, with
AC
CE
PT
ED
MA NU
SC
RI P
T
higher values indicating better fit.
ACCEPTED MANUSCRIPT 3.Results Prior to conducting the primary analyses, the total sample was randomly divided into two
T
separate samples; a validation sample (n = 1488) and a calibration sample (n = 1474). Each
RI P
sample was checked for accuracy, missing values, and multivariate outliers. The number of missing values was minimal (<5%) for all the remaining cases, and the mean substitution was
multivariate outliers using a cut-off of .001.
SC
therefore employed, where necessary. The Mahalanobis distance was used to identify the
MA NU
For the validation sample, 131 cases were identified as outliers and removed from further analysis, leaving a total of 1357 cases for subsequent analysis. For the calibration sample, 147 cases were identified as outliers and removed from the dataset. The distributions
ED
of both samples were further assessed using skewness and kurtosis statistics. An inspection of the skewness and kurtosis indices indicated a departure from normality, with most measures
PT
exhibiting some positive skewness. However, no variable transformations were performed.. Skewness and kurtosis statistics of each individual item for both calibration and validation
AC
CE
samples are presented in Table 2.
3.1 Exploratory Factor Analysis using the validation sample Following Tabachnick and Fidell’s [13] suggestion, the data was examined for its factorability. The correlation matrices revealed a substantial number of correlations greater than .33, suggesting the factorability of the data set. Furthermore, acceptable values of the Kaiser-Meyer-Olkin value (.93), and a significant value (p < .001) of Barlett’s Test of Sphericity also suggested that the data were suitable for factor analysis. Moreover, the Tolerance (0.647-0.872) and VIF (1.20-1.64) values suggested that multicollinearity and singularity were not a threat in this dataset.
ACCEPTED MANUSCRIPT In the subsequent analysis, the principal axis factoring with orthogonal rotation (Varimax) was used. The main criterion for the initial number of factors to be extracted was
T
eigenvalues greater than 1. On the basis of this criterion, three factors were initially extracted,
RI P
accounting for 39% of the total variance. The initial eigenvalues and their respective variances are presented in Table 2. Additionally, the item communalities and the initial item
SC
loadings are presented in Table 3.
For the item retention, an arbitrary criterion of a factor loading greater than 0.32 on
MA NU
one factor was employed with no other factor loadings greater than 0.32. (13) Items which (a) failed to load on any factor, (b) loaded on more than one factor, and (c) cross loaded on different factors were removed from the analysis one by one.
ED
These analyses led to the deletion of a total of 9 items. Although item 2 failed to meet the specified criteria, it was decided to maintain it throughout the analysis on the basis of its
PT
theoretical significance. Moreover, although item 20 met the item loading criteria for item retention, it was decided to remove this item on the grounds of item homogeneity. The final
CE
item loadings are presented in Table 4.
AC
3.2 Confirmatory Factor Analysis using calibration sample Based on the results from the EFA stage, 12 items were included in a final CFA to test the fit of the model. Using Maximum Likelihood procedures to estimate the model, the selected fit indices indicated a close model fit (χ² =132.94, df = 57, p = .000; GFI = 0.98; CFI = 0.96; PCFI = 0.74; RMR = .02; RMSEA = .04; ECVI = 0.14) 3.3 Internal consistency and its validity Moreover, the significant loadings of all the unstandardized regression weights implied an acceptable convergent validity (see Figure 1). Besides the convergent validity of the item, the discriminant validity of the subscales was also evident from the moderate latent factor inter-correlations, which ranged from .62 to .75. The subscale reliability was further
ACCEPTED MANUSCRIPT estimated using Cronbach’s alpha and an adequate reliability of the subscales was obtained
AC
CE
PT
ED
MA NU
SC
RI P
T
(Depression =.68; Anxiety = .53; Stress = .52).
ACCEPTED MANUSCRIPT 4. Discussion This study aims to further test the factorial validity of the DASS-21. The results
T
suggested that the DASS-21 did not indicate a goodness fit for the data. Therefore, an
RI P
exploratory factor analysis and further confirmatory factor analysis were conducted in order to confirm its factor structure.
SC
The items that have been deleted are items number 5, 8, 9, 11, 12,13,15,16 and 20. The items that remain in the depression domain are items number 3, 10, 17 and 21, while for
MA NU
anxiety the items that remain are 2, 4, 7, and 19. In addition, for the stress domain, the remaining items are 1, 6, 14 and 18. Contrary to previous studies that attempted to obtain the best solution (i.e. factor structures that best describe their respective data), this study aims to
ED
find the simplest solution to the model (i.e., depression, anxiety, and stress) by removing items that are complex and exhibit evidence of overlapping in their content. Indeed, this
PT
problem can be observed from prior studies that have shown an extremely high latent factor inter correlation. Furthermore, there is also evidence that previous studies employed
AC
variances.
CE
covariates between some pairs of items, which indicate that those items share unknown
The initial exploratory analysis reinforced the notion that the DASS items can indeed be grouped into three different factors. However, the findings also reinforced the idea that some items may be complex. A closer examination of each item revealed that the anxiety subscale tends to cluster into two main themes, cognitive and somatic. This is parallel with recent conceptualizations of anxiety which suggest that anxiety can be manifested in cognitive as well as somatic forms. While the somatic-related item forms a distinct group, the cognitive-related items load on either with the stress or depression items. For the depression subscales, it was observed that the items that were loading together with the stress items were two items with the word
ACCEPTED MANUSCRIPT semangat and one item that had two underlying emotions – murung (depressed) and sedih (sad). While the cause of the double loading on items 5 and 13 is more difficult to explain, it
T
could have been because of the keyword used in the translation of the two items. Indeed, both
RI P
items used the word ‘semangat’, which in the Malay language reflects something that is spirited, exuberant, or with vigour. While the word ‘initiative’ (item 5) may be better
SC
translated as ‘dorongan’, the translation nevertheless conveys its meaning, though not literally.
MA NU
The accurate identification of a term is important among adolescents for several reasons, of which two are most important. Firstly, it has been implied that a term could actually precede the elevation, recurrence, and exacerbation of severe emotional disorders.
ED
Adolescents who report higher levels of stress are four times more likely to exhibit depressive symptoms. Indeed, an elevation in the rates of depressed moods associated with stress
PT
represents a substantial risk of developing clinically significant depressive disorders and impaired functioning. Therefore, an accurate assessment may eventually assist in a proper
CE
referral and faster intervention or arrangements for rehabilitation of the student and the
AC
family. This could also actually help the rapport, performance and achievement of the students’ respective schools. While depression and anxiety have clear cognitive and somatic manifestations and are clearly defined in the literature, stress has a more general effect and has a wide variety of symptoms. One stress item that failed to load tended to be lengthy in its translation and may cause confusion, while the remaining two loaded with depression items. The present study is important for at least two reasons. First, although the psychometric properties of the Malaysian adapted DASS-21 have been established in adult populations, the first attempt to investigate the goodness fit of the DASS with shorter items (i.e. 12) is also unique and essential in younger respondents. Second, a psychometrically sound instrument contributes to
ACCEPTED MANUSCRIPT the accurate identification of elevated rates of stress and depressed moods, which may curb the risk of developing clinically significant depressive and anxiety disorders in adolescents.
T
Therefore with these 12 items, the DASS could hasten the process of screening at the
RI P
community level in particular and also assist teachers at the school level to identify students who are at risk.
SC
There are a few limitations in this study. For instance, the age range among the sample population is between 13 and 16 years only due to a ruling by the Ministry that
MA NU
research should not be conducted on students who will be undergoing comprehensive examinations at the ages of 15 and 17 years. Besides that, the ethnic and gender distributions are also unequal and are dominated by samples of Malays and females. Despite its
ED
limitations, this is the first study that has attempted to test the shortest version of the DASS while still maintaining its factorial distribution through a series of exploratory and
PT
confirmatory factor analyses, and which found its reliability to be good. The items finalized in this analysis may be used to create scores on the prevalence of depression, anxiety, and
CE
stress. This version has significantly reduced the symptoms associated with the three negative
AC
effects and thus may not represent the extent of each negative effect. Further assessments in the future can be conducted on concurrent validity, test-retest reliability, specificity and sensitivity and also to establish its scoring interpretation for the 12item DASS. However, it would be better to make changes to the mentioned items to make them more acceptable to all the ethnic groups in Malaysia prior to conducting research to further evaluate their psychometric properties. 4.1 Conclusion To sum up, this study suggests that the 12-item DASS can be utilised and has been considered to have stable factorial validity, adequate reliability and validity measures to be
ACCEPTED MANUSCRIPT used among adolescents in clinical or research assessments and treatment outcome measures avenues.
T
Acknowledgement
RI P
The authors thank participants for their contribution to the study. This research is funded by Universiti Putra Malaysia’s RU grant to Proffesor Latiffah Abdul Latiff and research
SC
team.There is no financial conflict of interest in this study. References
MA NU
[1] National Health Morbidity Survey. Mental Health Status for Children and Adolescents. Institute of Public Health Malaysia.2011.
[2] Clark LA , Watson D . Theoretical and empirical issues in differentiating depression
ED
from anxiety. In J. Becker & A.Kleinman (Eds). Advances in Mood Disorders. Vol.I. Psychological aspects of depression. Hillsdale, NJ: Erbaum.1990.
PT
[3] Lovibond SH , Lovibond PF . Manual for the Depression Anxiety Stress Scales, Sydney: Psychology Foundation. 1995.
CE
[4] Hashim, HA, Golok F, Ali R . Factorial validity and internal consistency of
AC
Malaysian Adapted Depression Anxiety Stress Scale-21 in an adolescent sample, Int Jour Collabor Res Internal Med Pub Health,2011 3, 29-39 [5] Patrick J, Dyck M, Bramston P. Depression Anxiety Stress Scale: is it valid for children and adolescents?, J Clin Psychol. 2010, 66, 996-1007. [6] Herman K, Ostrander R, Walkup J, Sylva S, March J. Emperically derived subtypes of adolescent depression: Latent profile analysis of co-occuring symptoms in the treatment of adolescents with depression study, J Consult Clin Psych, 2007,75, 716-728. [7] Szabo M , Lovibond PF . Anxiety, depression and tension/stress in children, J Psychopathol Beh, 2000, 28 (3): 195-205.
ACCEPTED MANUSCRIPT [8] Duffy CJ, Cunningham EG, Moore SM . The factor structure of mood states in an early adolescents sample. J Adol, 2005, 28, 677-680.
T
[9] Szabo, M .The short version of the Depression Anxiety Stress Scales (DASS-21):
RI P
Factor Structure in a young adolescent sample. J of Adol, 2010, 33 (1):1-8
[10] Tully PJ, Zajac IT, Venning AJ . The structure of anxiety and depression in a
SC
normative sample of younger and older Australian adolescents. J Abn Child Psych,2001, 37: 717-726
MA NU
[11] Bentler PM . Comparative fit indexes in structural models. Psychol Bull.1990, 107:238246
[12] Browne MW , Cudeck R. Alternate ways of assessing model fit In Bollen K.,
ED
Long J.S., eds. Testing structural equation models. Newbury Park, CA: Sage: 1993: 132-162
PT
[13] Tabachnick BG , Fidell, LS . Using multivariate Statistics. Fourth Edition Boston,
AC
CE
Massachusettes: Allyn and Bacon, 2001.
ACCEPTED MANUSCRIPT Table 1. Descriptive statistics of individual items for both calibration and validation samples
T
SC
0.81 0.95 0.70 0.52 1.03 0.77 0.63 1.31 0.83 0.66 0.58 0.83 0.63 1.10 0.70 0.63 0.58 1.14 0.67 0.60 0.44
Validation sample Std. Skewness Kurtosis Deviation Statistic Std. Statistic Std. Error Error 0.75 0.76 0.07 0.43 0.13 0.07 0.13 0.89 0.65 -0.34 0.07 0.13 0.83 0.95 0.06 0.07 0.13 0.80 1.50 1.47 0.07 0.13 0.91 0.52 -0.58 0.07 0.13 0.77 0.83 0.38 0.07 0.13 0.83 1.22 0.76 0.07 0.13 0.96 0.29 -0.84 0.07 0.13 0.89 0.84 -0.16 0.07 0.13 0.84 1.09 0.35 0.07 0.13 0.74 1.12 0.72 0.07 0.13 0.94 0.90 -0.17 0.07 0.13 0.79 1.15 0.75 0.07 0.13 0.95 0.50 -0.69 0.07 0.13 0.79 0.96 0.31 0.07 0.13 0.78 1.16 0.86 0.07 0.13 0.80 1.29 0.96 0.07 0.13 0.98 0.49 -0.78 0.07 0.13 0.85 1.17 0.65 0.07 0.13 0.82 1.27 0.87 0.07 0.13 0.78 1.77 2.30
RI P
Mean
MA NU
ED
PT
0.86 0.98 0.76 0.57 1.09 0.78 0.67 1.33 0.88 0.71 0.64 0.86 0.67 1.14 0.71 0.67 0.61 1.16 0.67 0.65 0.51
CE
Item 1 Item 2 Item 3 Item 4 Item 5 Item 6 Item 7 Item 8 Item 9 Item 10 Item 11 Item 12 Item 13 Item 14 Item 15 Item 16 Item 17 Item 18 Item 19 Item 20 Item 21
AC
Mean
Calibration sample Std. Skewness Kurtosis Deviation Statistic Std. Statistic Std. Error Error 0.06 0.79 0.80 0.38 0.12 0.06 0.12 0.90 0.64 -0.38 0.06 0.12 0.88 0.93 -0.01 0.06 0.12 0.86 1.42 1.03 0.06 0.12 0.94 0.52 -0.62 0.06 0.12 0.82 0.90 0.32 0.06 0.12 0.85 1.14 0.53 0.06 0.12 1.00 0.27 -0.97 0.06 0.12 0.95 0.82 -0.34 0.06 0.12 0.89 1.11 0.34 0.06 0.12 0.81 1.18 0.80 0.06 0.12 0.95 0.86 -0.29 0.06 0.12 0.84 1.16 0.68 0.06 0.12 0.97 0.47 -0.75 0.06 0.12 0.83 1.05 0.47 0.06 0.12 0.81 1.13 0.70 0.06 0.12 0.86 1.31 0.80 0.06 0.12 0.99 0.49 -0.79 0.06 0.12 0.87 1.17 0.53 0.06 0.12 0.88 1.26 0.72 0.12 0.84 1.59 0.06 1.50
ACCEPTED MANUSCRIPT
Table 2. Initial eigenvalues and variances of each factor
Cumulative
Variance
%
5.812
27.678
27.678
2
1.364
6.495
34.173
3
1.099
5.233
39.406
AC
CE
PT
ED
MA NU
1
RI P
Total
% of
T
Initial Eigenvalues
SC
Factor
ACCEPTED MANUSCRIPT Table 3. items communalities and initial factor loading Factor Loadings
Communalities Extraction
2
3
.168
.179
.314
DASS 6
.230
.281
.456
DASS 8
.136
.161
DASS 11
.391
.422
.428
.388
DASS 12
.353
.364
.417
.357
DASS 14
.206
.244
.428
DASS 18
.147
.178
.342
DASS 3
.192
.208
DASS 5
.277
.280
DASS 10
.294
.345
DASS 13
.347
.350
DASS 16
.311
.375
DASS 17
.320
.387
DASS 21
.352
.454
DASS 2
.128
.130
DASS 4
.243
DASS 7
.183
DASS 9
.240
DASS 15
.302
DASS 19
.242
MA NU .368
.347
.543 .423
.310
.533 .590 .651
ED .281
.234 .310
.410
PT
.241
CE .272
AC
DASS 20
.394
SC
DASS 1
.429
.322
.515
.325
.379
.405 .288
T
1
RI P
Initial
.337 .605
.340
.333
ACCEPTED MANUSCRIPT
Table 4: Final item loadings
3 .353
DASS 6
.475
DASS 14
.423
DASS 18
.377 .365
DASS 10
.548
DASS 17
.589
DASS 21
.644
DASS 2
PT
.250
AC
DASS 19
.471
CE
DASS 4 DASS 7
ED
DASS 3
MA NU
DASS 1
RI P
2
SC
1
T
Factor Loading
.425 .575
ACCEPTED MANUSCRIPT FIGURE 1. Standardized regression weight for factor loading and latent factor
AC
CE
PT
ED
MA NU
SC
RI P
T
intercorrelations.
S0010-440X(14)00098-4 doi: 10.1016/j.comppsych.2014.04.011 YCOMP 51291
To appear in:
Comprehensive Psychiatry
Received date: Revised date: Accepted date:
1 November 2013 13 April 2014 22 April 2014
Please cite this article as: Osman Zubaidah Jamil, Mukhtar Firdaus, Hashim Hairul Anuar, Latiff Latiffah Abdul, Sidik Sherina Mohd, Awang Hamidin, Ibrahim Normala, Rahman Hejar Abdul, Ismail Siti Irma Fadhilah, Ibrahim Faisal, Tajik Esra, Othman Norlijah, Testing comparison models of DASS-12 and its reliability among adolescents in Malaysia, Comprehensive Psychiatry (2014), doi: 10.1016/j.comppsych.2014.04.011
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
SC
RI P
T
ACCEPTED MANUSCRIPT
MA NU
Testing comparison models of DASS-12 and its reliability among adolescents in Malaysia
Zubaidah Jamil Osman1, Firdaus Mukhtar1, Hairul Anuar Hashim2, Latiffah Abdul Latiff1,
ED
Sherina Mohd Sidik1, Hamidin Awang1, Normala Ibrahim1, Hejar Abdul Rahman1,
Faculty of Medicine and Health Sciences, Universiti Putra Malaysia, Malaysia Sport Science Unit, School of Medicine, Universiti Sains Malaysia, Malaysia
AC
2
CE
1
PT
Siti Irma Fadhilah Ismail1, Faisal Ibrahim1, Esra Tajik1 & Norlijah Othman1
Corresponding author:-
ACCEPTED MANUSCRIPT Dr. Firdaus Mukhtar, Department of Psychiatry, Universiti Putra Malaysia, 43400 Serdang,
AC
CE
PT
ED
MA NU
SC
RI P
T
Malaysia (Email: [email protected]; Tel: 03 8947 2543; Fax: 03 8941 4629).
ACCEPTED MANUSCRIPT Abstract Objective: The 21-item Depression, Anxiety and Stress Scale (DASS-21) is frequently used
T
in non-clinical research to measure mental health factors among adults. However, previous
RI P
studies have concluded that the 21 items are not stable for utilization among the adolescent population. Thus, the aim of this study is to examine the structure of the factors and to report
SC
on the reliability of the refined version of the DASS that consists of 12 items. Method: 2850 students (aged 13 to 17 years old) from three major ethnic in Malaysia
MA NU
completed the DASS-21. The study was conducted at 10 randomly selected secondary schools in the northern state of Peninsular Malaysia. The study population comprised secondary school students (Forms 1, 2 & 4) from the selected schools.
ED
Results: Based on the results of the EFA stage, 12 items were included in a final CFA to test the fit of the model. Using Maximum Likelihood procedures to estimate the model, the
PT
selected fit indices indicated a close model fit (χ² =132.94, df = 57, p = .000; CFI = .96; RMR = .02; RMSEA = .04). Moreover, significant loadings of all the unstandardized regression
CE
weights implied an acceptable convergent validity. Besides the convergent validity of the
AC
item, a discriminant validity of the subscales was also evident from the moderate latent factor inter-correlations, which ranged from .62 to .75. The subscale reliability was further estimated using Cronbach’s alpha and the adequate reliability of the subscales was obtained (Total= 76; Depression =.68; Anxiety = .53; Stress = .52). Conclusion: The new version of the 12-item DASS for adolescents in Malaysia (DASS-12) is reliable and has a stable factor structure, and thus it is a useful instrument for distinguishing between depression, anxiety and stress.
Keywords: DASS, depression, anxiety, adolescence, factor analysis, validation
ACCEPTED MANUSCRIPT 1. Introduction In Malaysia, three years ago National Health and Morbidity Survey [1] reported that
T
the overall prevalence of mental health problems among children between five to 16 years
RI P
using the Reporting Questionnaire for Children (RQC) was 20%. The age group of below 16 had the highest prevalence of 22.2% compared to the age group of 10-14 years (20.6%) and
SC
5-9 years (19.1%), respectively. The prevalence is worrying and highlights an on-going need for the continued strengthening and upgrading of Child and Adolescent Mental Health
MA NU
Services in Malaysia. Improved detection and recognition of early difficulties as well as more comprehensive early intervention programs, particularly among at-risk population groups are clearly needed. This survey has highlighted the extent of general mental health issues
ED
currently existing in the community, reflecting the unmet needs of the population nationwide. In order to improve detection and early referral of children and adolescents with
PT
mental health problems, it is suggested that a reliable tool be used for fast screening at community level [1]. The early identification of emotional distress, such as stress and
CE
anxiety, may be an important step in preventing the risk of the development of clinically
AC
significant psychological disorders. Importantly, the accurate identification of these negative effects relies on a psychometrically sound measurement instrument. Twenty years ago, Clark and Watson [2] highlighted that although anxiety and depression are phenomenologically distinct, it has been proven to be very difficult to distinguish between these constructs by empirical means, either by using clinical ratings or self-report measures. In measuring these negative effects, the Depression, Anxiety and Stress Scale-21 (DASS-21) [3] has been reported to be quite successful in terms of its stable factor structure and its psychometric properties, especially among the adult population, both in research and clinical services, in most types of languages. The Depression Scale assesses dysphoric mood states, including self-depreciation, lack of interest/involvement,
ACCEPTED MANUSCRIPT hopelessness, and anhedonia. The Anxiety Scale assesses arousal states, including autonomic arousal, muscular tension, and the anxious effect. Finally, the Stress Scale is reported to be
T
fairly consistent with the stressor and general tension [3].
RI P
However, a recent study by Hashim and colleagues [4] argued that the DASS-21 is stable or easily utilized in order to understand the younger population, especially adolescents.
SC
Hashim and colleagues concluded that a one-dimensional structure is consistently obtained and also suggested that it is important to revise the items to be utilised according to standard
MA NU
colloquial language, especially if this measure is to be used in a multi-ethnic population. An assessment of the psychometric properties of this measure among young respondents is critical given the fact that the youth population may not be able to report their experiences
ED
using items that have been designed for the adult population. Given the inconsistencies in relation to the factorial structure of the DASS-21among the younger population and the
PT
potential benefits that can be obtained from developmental research into the DASS-21, it is critically important to evaluate the psychometric properties of DASS-21 in younger
CE
respondents [4, 5].
AC
The original principal component analysis of the DASS items revealed that a stable 3factor solution of depression, anxiety and stress was the optimal fit [3]. Subsequent research has replicated the 3-factor solution in the adolescent population. At this stage, it has been clearly documented that the use of the DASS-21 is well-supported for the adult population, but it does not apply much to children and adolescents. Several studies have argued that the experience of depression and anxiety is expressed or experienced differently, and what more given the language discrepancy between English and other linguistic or cultural backgrounds [5]. When a self-report scale like the DASS-21 is used with children and adolescents, it is not measuring the three discriminate emotional dimensions as it does in adults. Rather, it
ACCEPTED MANUSCRIPT appears to be measuring a single distress dimension and it is more closely related to depression, to anxiety , or to general distress, or even that it is possible to draw distinctions
T
between these states. Herman and collegues [6] stated that the DASS-21 is a useful index of
symptoms of anxiety or depression as a core feature.
RI P
the severity of a mixed affective and/or oppositional/disruptive behavioral disorder which has
SC
In a study conducted by Szabó and Lovibond [7] involving 7- to 14- year old children and adolescents, a 2-factor structure incorporating anxiety and a combined
MA NU
depression-stress item best fitted the data. Similarly, in another study involving 11- to 15year old adolescents conducted by Duffy and collegues [8] the findings did not support the original 3-factor structure (depression-anxiety-stress) of DASS-21. Instead, the researchers
ED
only observed a reasonable model fit when the items of the DASS-21 were grouped into 2 factors, Physiological Hyperarousal and General Negative Effect, while allowing some
PT
error terms to co-vary. On the basis of the results, it was speculated that young people might not yet have developed the ability to differentiate between depression, anxiety, and stress.
CE
Instead, adolescents are likely to report their experience as a more general, negative
AC
experience and physiological arousal. The inconsistent factor structure of the DASS-21 in the younger population is also evident in two studies that are more recent. In one of those studies, Szabó [9] revealed that the DASS-21 was best used as a measure of anxiety and depression instead of the 3-factor structure in 11- to 15-year old children and adolescents. Contrary to Szabó [9], Tully and colleagues [10] observed that the DASS-21 was best seen as a measure of depression, physiological arousal, and general negative effect. Consistent across these studies was the view of the researchers that emotional differentiation is still developing in younger respondents and they may not be able to fully appreciate the differentiation between depression, anxiety, and stress as reflected in the DASS-21 items [4]. Furthermore, Szabó [9] contended that the DASS contained several expressions and words that might not be familiar
ACCEPTED MANUSCRIPT to adolescents. Thus, this accounted for the failure to obtain a clear factor structure when using the DASS-21 among younger respondents.
T
1.1 Aims of study
RI P
The aims of this study are (a) to test the goodness fit of the current data against previous models of the DASS; (b) to conduct an exploratory and confirmatory analysis and to
SC
report on its reliability in a sample of adolescents in Malaysia. It was hypothesized that the DASS-21 would exhibit similar factor structures as in the original version and exhibit
AC
CE
PT
ED
MA NU
acceptable reliability indices.
ACCEPTED MANUSCRIPT 2. Material and Methods 2.1 Study design and sampling
T
The participants were 2980 secondary school students (boys =46.9% and girls = 53.1%)
RI P
ranging in age from between 13 to 17 years (13 years old = 31.9%, 14 years old 31.5%, 15 years old 2%, 16 years old 33.1% and 17 years 1.5%, respectively). The majority of them
SC
were Malays (53.9%), and the rest were Chinese, Indians and other races (32.2, 11.8 and 2.1 per cent, respectively). The Muslims were the most represented in this study (54.3%), while
MA NU
the rest comprised Buddhists (28.4%), Hindus (10.8%), Christians (5.7%) and other religions (0.8%). In terms of levels of study, the percentages were quite balanced at 35.3%, 30.3% and 34.4% of the participants being in Form One, Form Two and Form Four, respectively.
ED
The study was conducted at 10 randomly selected secondary schools in the Pasir Gudang district. The study population comprised secondary school students (Forms 1, 2 & 4)
PT
from the selected schools. A multistage cluster sampling method was used in the study and a total number of 3000 students were selected based on the estimated sample size (Lwanga &
CE
Lemeshow, 1991). The inclusion criteria included all Forms 1, 2 and 4 students who
AC
consented to the study. Students who consented to participate in the study and were permitted to do so by their parents were given the questionnaire to be completed. The administration of the questionnaire took place in a classroom setting and was supervised by the second and third authors. The participants spent an average of 10 minutes to complete the questionnaire.
2.2 Research instrument Depression, Anxiety and Stress Scale-21. The DASS-21 [3] is a measure of three distinct negative effects: depression, stress and anxiety. The respondents were to indicate the extent to which they experienced each of the symptoms on a 4-point Likert scale ranging from 0 (Did not apply to me at all) to 3 (Applied to me very much or most of the time). The Malay
ACCEPTED MANUSCRIPT translated version of the DASS-21 was used in the present study. In an adult population, the
T
initial assessment of the translated version revealed adequate validity and reliability indices.
RI P
2.3 Ethical approval
The Ethical Committee of the University and the Ministry of Education, Malaysia granted the
MA NU
their parents prior to the data collection.
SC
permission to conduct the study. Written consent was also obtained from the respondents and
2.4 Data Analysis
Four statistical procedures were utilized. Descriptive statistics were used for the data screening. An exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) were
ED
performed to establish the factorial validity of the measure. In addition, Cronbach’s alpha
PT
coefficients were computed to evaluate the reliability of the subscales. For EFA, the principal axis factoring with orthogonal rotation (Varimax) was used and the
CE
main criterion for the initial number of factors to be extracted was eigenvalues greater than 1.
AC
For the CFA, multiple fit indices were used. The selected indices were the chi-square statistic (χ ²), the root mean square residuals (RMR)[11],the goodness fit index (GFI), the comparative fit index (CFI)[11], and the root mean square error of approximation (RMSEA)[12]. A good model fit is indicated by values of .90 or higher for the GFI and CFI. RMR values of less than .05 reflect a close fit, while values of .1 or lower indicate a reasonable fit for the RMR [11]. For the RMSEA, values of .05 or lower indicate a close fit while values less than .08 indicate an acceptable fit [12]. Additionally, the expected cross validation index (ECVI), and the Parsimony-Adjusted CFI (PCFI) were also evaluated. For ECVI, models with smaller values indicate the best potential of replication in samples of equivalent size and precision of the ECVI is presented in confidence interval. Lastly, PCFI
ACCEPTED MANUSCRIPT takes into account the complexity of the model and values above 0.70 indicates good fit, with
AC
CE
PT
ED
MA NU
SC
RI P
T
higher values indicating better fit.
ACCEPTED MANUSCRIPT 3.Results Prior to conducting the primary analyses, the total sample was randomly divided into two
T
separate samples; a validation sample (n = 1488) and a calibration sample (n = 1474). Each
RI P
sample was checked for accuracy, missing values, and multivariate outliers. The number of missing values was minimal (<5%) for all the remaining cases, and the mean substitution was
multivariate outliers using a cut-off of .001.
SC
therefore employed, where necessary. The Mahalanobis distance was used to identify the
MA NU
For the validation sample, 131 cases were identified as outliers and removed from further analysis, leaving a total of 1357 cases for subsequent analysis. For the calibration sample, 147 cases were identified as outliers and removed from the dataset. The distributions
ED
of both samples were further assessed using skewness and kurtosis statistics. An inspection of the skewness and kurtosis indices indicated a departure from normality, with most measures
PT
exhibiting some positive skewness. However, no variable transformations were performed.. Skewness and kurtosis statistics of each individual item for both calibration and validation
AC
CE
samples are presented in Table 2.
3.1 Exploratory Factor Analysis using the validation sample Following Tabachnick and Fidell’s [13] suggestion, the data was examined for its factorability. The correlation matrices revealed a substantial number of correlations greater than .33, suggesting the factorability of the data set. Furthermore, acceptable values of the Kaiser-Meyer-Olkin value (.93), and a significant value (p < .001) of Barlett’s Test of Sphericity also suggested that the data were suitable for factor analysis. Moreover, the Tolerance (0.647-0.872) and VIF (1.20-1.64) values suggested that multicollinearity and singularity were not a threat in this dataset.
ACCEPTED MANUSCRIPT In the subsequent analysis, the principal axis factoring with orthogonal rotation (Varimax) was used. The main criterion for the initial number of factors to be extracted was
T
eigenvalues greater than 1. On the basis of this criterion, three factors were initially extracted,
RI P
accounting for 39% of the total variance. The initial eigenvalues and their respective variances are presented in Table 2. Additionally, the item communalities and the initial item
SC
loadings are presented in Table 3.
For the item retention, an arbitrary criterion of a factor loading greater than 0.32 on
MA NU
one factor was employed with no other factor loadings greater than 0.32. (13) Items which (a) failed to load on any factor, (b) loaded on more than one factor, and (c) cross loaded on different factors were removed from the analysis one by one.
ED
These analyses led to the deletion of a total of 9 items. Although item 2 failed to meet the specified criteria, it was decided to maintain it throughout the analysis on the basis of its
PT
theoretical significance. Moreover, although item 20 met the item loading criteria for item retention, it was decided to remove this item on the grounds of item homogeneity. The final
CE
item loadings are presented in Table 4.
AC
3.2 Confirmatory Factor Analysis using calibration sample Based on the results from the EFA stage, 12 items were included in a final CFA to test the fit of the model. Using Maximum Likelihood procedures to estimate the model, the selected fit indices indicated a close model fit (χ² =132.94, df = 57, p = .000; GFI = 0.98; CFI = 0.96; PCFI = 0.74; RMR = .02; RMSEA = .04; ECVI = 0.14) 3.3 Internal consistency and its validity Moreover, the significant loadings of all the unstandardized regression weights implied an acceptable convergent validity (see Figure 1). Besides the convergent validity of the item, the discriminant validity of the subscales was also evident from the moderate latent factor inter-correlations, which ranged from .62 to .75. The subscale reliability was further
ACCEPTED MANUSCRIPT estimated using Cronbach’s alpha and an adequate reliability of the subscales was obtained
AC
CE
PT
ED
MA NU
SC
RI P
T
(Depression =.68; Anxiety = .53; Stress = .52).
ACCEPTED MANUSCRIPT 4. Discussion This study aims to further test the factorial validity of the DASS-21. The results
T
suggested that the DASS-21 did not indicate a goodness fit for the data. Therefore, an
RI P
exploratory factor analysis and further confirmatory factor analysis were conducted in order to confirm its factor structure.
SC
The items that have been deleted are items number 5, 8, 9, 11, 12,13,15,16 and 20. The items that remain in the depression domain are items number 3, 10, 17 and 21, while for
MA NU
anxiety the items that remain are 2, 4, 7, and 19. In addition, for the stress domain, the remaining items are 1, 6, 14 and 18. Contrary to previous studies that attempted to obtain the best solution (i.e. factor structures that best describe their respective data), this study aims to
ED
find the simplest solution to the model (i.e., depression, anxiety, and stress) by removing items that are complex and exhibit evidence of overlapping in their content. Indeed, this
PT
problem can be observed from prior studies that have shown an extremely high latent factor inter correlation. Furthermore, there is also evidence that previous studies employed
AC
variances.
CE
covariates between some pairs of items, which indicate that those items share unknown
The initial exploratory analysis reinforced the notion that the DASS items can indeed be grouped into three different factors. However, the findings also reinforced the idea that some items may be complex. A closer examination of each item revealed that the anxiety subscale tends to cluster into two main themes, cognitive and somatic. This is parallel with recent conceptualizations of anxiety which suggest that anxiety can be manifested in cognitive as well as somatic forms. While the somatic-related item forms a distinct group, the cognitive-related items load on either with the stress or depression items. For the depression subscales, it was observed that the items that were loading together with the stress items were two items with the word
ACCEPTED MANUSCRIPT semangat and one item that had two underlying emotions – murung (depressed) and sedih (sad). While the cause of the double loading on items 5 and 13 is more difficult to explain, it
T
could have been because of the keyword used in the translation of the two items. Indeed, both
RI P
items used the word ‘semangat’, which in the Malay language reflects something that is spirited, exuberant, or with vigour. While the word ‘initiative’ (item 5) may be better
SC
translated as ‘dorongan’, the translation nevertheless conveys its meaning, though not literally.
MA NU
The accurate identification of a term is important among adolescents for several reasons, of which two are most important. Firstly, it has been implied that a term could actually precede the elevation, recurrence, and exacerbation of severe emotional disorders.
ED
Adolescents who report higher levels of stress are four times more likely to exhibit depressive symptoms. Indeed, an elevation in the rates of depressed moods associated with stress
PT
represents a substantial risk of developing clinically significant depressive disorders and impaired functioning. Therefore, an accurate assessment may eventually assist in a proper
CE
referral and faster intervention or arrangements for rehabilitation of the student and the
AC
family. This could also actually help the rapport, performance and achievement of the students’ respective schools. While depression and anxiety have clear cognitive and somatic manifestations and are clearly defined in the literature, stress has a more general effect and has a wide variety of symptoms. One stress item that failed to load tended to be lengthy in its translation and may cause confusion, while the remaining two loaded with depression items. The present study is important for at least two reasons. First, although the psychometric properties of the Malaysian adapted DASS-21 have been established in adult populations, the first attempt to investigate the goodness fit of the DASS with shorter items (i.e. 12) is also unique and essential in younger respondents. Second, a psychometrically sound instrument contributes to
ACCEPTED MANUSCRIPT the accurate identification of elevated rates of stress and depressed moods, which may curb the risk of developing clinically significant depressive and anxiety disorders in adolescents.
T
Therefore with these 12 items, the DASS could hasten the process of screening at the
RI P
community level in particular and also assist teachers at the school level to identify students who are at risk.
SC
There are a few limitations in this study. For instance, the age range among the sample population is between 13 and 16 years only due to a ruling by the Ministry that
MA NU
research should not be conducted on students who will be undergoing comprehensive examinations at the ages of 15 and 17 years. Besides that, the ethnic and gender distributions are also unequal and are dominated by samples of Malays and females. Despite its
ED
limitations, this is the first study that has attempted to test the shortest version of the DASS while still maintaining its factorial distribution through a series of exploratory and
PT
confirmatory factor analyses, and which found its reliability to be good. The items finalized in this analysis may be used to create scores on the prevalence of depression, anxiety, and
CE
stress. This version has significantly reduced the symptoms associated with the three negative
AC
effects and thus may not represent the extent of each negative effect. Further assessments in the future can be conducted on concurrent validity, test-retest reliability, specificity and sensitivity and also to establish its scoring interpretation for the 12item DASS. However, it would be better to make changes to the mentioned items to make them more acceptable to all the ethnic groups in Malaysia prior to conducting research to further evaluate their psychometric properties. 4.1 Conclusion To sum up, this study suggests that the 12-item DASS can be utilised and has been considered to have stable factorial validity, adequate reliability and validity measures to be
ACCEPTED MANUSCRIPT used among adolescents in clinical or research assessments and treatment outcome measures avenues.
T
Acknowledgement
RI P
The authors thank participants for their contribution to the study. This research is funded by Universiti Putra Malaysia’s RU grant to Proffesor Latiffah Abdul Latiff and research
SC
team.There is no financial conflict of interest in this study. References
MA NU
[1] National Health Morbidity Survey. Mental Health Status for Children and Adolescents. Institute of Public Health Malaysia.2011.
[2] Clark LA , Watson D . Theoretical and empirical issues in differentiating depression
ED
from anxiety. In J. Becker & A.Kleinman (Eds). Advances in Mood Disorders. Vol.I. Psychological aspects of depression. Hillsdale, NJ: Erbaum.1990.
PT
[3] Lovibond SH , Lovibond PF . Manual for the Depression Anxiety Stress Scales, Sydney: Psychology Foundation. 1995.
CE
[4] Hashim, HA, Golok F, Ali R . Factorial validity and internal consistency of
AC
Malaysian Adapted Depression Anxiety Stress Scale-21 in an adolescent sample, Int Jour Collabor Res Internal Med Pub Health,2011 3, 29-39 [5] Patrick J, Dyck M, Bramston P. Depression Anxiety Stress Scale: is it valid for children and adolescents?, J Clin Psychol. 2010, 66, 996-1007. [6] Herman K, Ostrander R, Walkup J, Sylva S, March J. Emperically derived subtypes of adolescent depression: Latent profile analysis of co-occuring symptoms in the treatment of adolescents with depression study, J Consult Clin Psych, 2007,75, 716-728. [7] Szabo M , Lovibond PF . Anxiety, depression and tension/stress in children, J Psychopathol Beh, 2000, 28 (3): 195-205.
ACCEPTED MANUSCRIPT [8] Duffy CJ, Cunningham EG, Moore SM . The factor structure of mood states in an early adolescents sample. J Adol, 2005, 28, 677-680.
T
[9] Szabo, M .The short version of the Depression Anxiety Stress Scales (DASS-21):
RI P
Factor Structure in a young adolescent sample. J of Adol, 2010, 33 (1):1-8
[10] Tully PJ, Zajac IT, Venning AJ . The structure of anxiety and depression in a
SC
normative sample of younger and older Australian adolescents. J Abn Child Psych,2001, 37: 717-726
MA NU
[11] Bentler PM . Comparative fit indexes in structural models. Psychol Bull.1990, 107:238246
[12] Browne MW , Cudeck R. Alternate ways of assessing model fit In Bollen K.,
ED
Long J.S., eds. Testing structural equation models. Newbury Park, CA: Sage: 1993: 132-162
PT
[13] Tabachnick BG , Fidell, LS . Using multivariate Statistics. Fourth Edition Boston,
AC
CE
Massachusettes: Allyn and Bacon, 2001.
ACCEPTED MANUSCRIPT Table 1. Descriptive statistics of individual items for both calibration and validation samples
T
SC
0.81 0.95 0.70 0.52 1.03 0.77 0.63 1.31 0.83 0.66 0.58 0.83 0.63 1.10 0.70 0.63 0.58 1.14 0.67 0.60 0.44
Validation sample Std. Skewness Kurtosis Deviation Statistic Std. Statistic Std. Error Error 0.75 0.76 0.07 0.43 0.13 0.07 0.13 0.89 0.65 -0.34 0.07 0.13 0.83 0.95 0.06 0.07 0.13 0.80 1.50 1.47 0.07 0.13 0.91 0.52 -0.58 0.07 0.13 0.77 0.83 0.38 0.07 0.13 0.83 1.22 0.76 0.07 0.13 0.96 0.29 -0.84 0.07 0.13 0.89 0.84 -0.16 0.07 0.13 0.84 1.09 0.35 0.07 0.13 0.74 1.12 0.72 0.07 0.13 0.94 0.90 -0.17 0.07 0.13 0.79 1.15 0.75 0.07 0.13 0.95 0.50 -0.69 0.07 0.13 0.79 0.96 0.31 0.07 0.13 0.78 1.16 0.86 0.07 0.13 0.80 1.29 0.96 0.07 0.13 0.98 0.49 -0.78 0.07 0.13 0.85 1.17 0.65 0.07 0.13 0.82 1.27 0.87 0.07 0.13 0.78 1.77 2.30
RI P
Mean
MA NU
ED
PT
0.86 0.98 0.76 0.57 1.09 0.78 0.67 1.33 0.88 0.71 0.64 0.86 0.67 1.14 0.71 0.67 0.61 1.16 0.67 0.65 0.51
CE
Item 1 Item 2 Item 3 Item 4 Item 5 Item 6 Item 7 Item 8 Item 9 Item 10 Item 11 Item 12 Item 13 Item 14 Item 15 Item 16 Item 17 Item 18 Item 19 Item 20 Item 21
AC
Mean
Calibration sample Std. Skewness Kurtosis Deviation Statistic Std. Statistic Std. Error Error 0.06 0.79 0.80 0.38 0.12 0.06 0.12 0.90 0.64 -0.38 0.06 0.12 0.88 0.93 -0.01 0.06 0.12 0.86 1.42 1.03 0.06 0.12 0.94 0.52 -0.62 0.06 0.12 0.82 0.90 0.32 0.06 0.12 0.85 1.14 0.53 0.06 0.12 1.00 0.27 -0.97 0.06 0.12 0.95 0.82 -0.34 0.06 0.12 0.89 1.11 0.34 0.06 0.12 0.81 1.18 0.80 0.06 0.12 0.95 0.86 -0.29 0.06 0.12 0.84 1.16 0.68 0.06 0.12 0.97 0.47 -0.75 0.06 0.12 0.83 1.05 0.47 0.06 0.12 0.81 1.13 0.70 0.06 0.12 0.86 1.31 0.80 0.06 0.12 0.99 0.49 -0.79 0.06 0.12 0.87 1.17 0.53 0.06 0.12 0.88 1.26 0.72 0.12 0.84 1.59 0.06 1.50
ACCEPTED MANUSCRIPT
Table 2. Initial eigenvalues and variances of each factor
Cumulative
Variance
%
5.812
27.678
27.678
2
1.364
6.495
34.173
3
1.099
5.233
39.406
AC
CE
PT
ED
MA NU
1
RI P
Total
% of
T
Initial Eigenvalues
SC
Factor
ACCEPTED MANUSCRIPT Table 3. items communalities and initial factor loading Factor Loadings
Communalities Extraction
2
3
.168
.179
.314
DASS 6
.230
.281
.456
DASS 8
.136
.161
DASS 11
.391
.422
.428
.388
DASS 12
.353
.364
.417
.357
DASS 14
.206
.244
.428
DASS 18
.147
.178
.342
DASS 3
.192
.208
DASS 5
.277
.280
DASS 10
.294
.345
DASS 13
.347
.350
DASS 16
.311
.375
DASS 17
.320
.387
DASS 21
.352
.454
DASS 2
.128
.130
DASS 4
.243
DASS 7
.183
DASS 9
.240
DASS 15
.302
DASS 19
.242
MA NU .368
.347
.543 .423
.310
.533 .590 .651
ED .281
.234 .310
.410
PT
.241
CE .272
AC
DASS 20
.394
SC
DASS 1
.429
.322
.515
.325
.379
.405 .288
T
1
RI P
Initial
.337 .605
.340
.333
ACCEPTED MANUSCRIPT
Table 4: Final item loadings
3 .353
DASS 6
.475
DASS 14
.423
DASS 18
.377 .365
DASS 10
.548
DASS 17
.589
DASS 21
.644
DASS 2
PT
.250
AC
DASS 19
.471
CE
DASS 4 DASS 7
ED
DASS 3
MA NU
DASS 1
RI P
2
SC
1
T
Factor Loading
.425 .575
ACCEPTED MANUSCRIPT FIGURE 1. Standardized regression weight for factor loading and latent factor
AC
CE
PT
ED
MA NU
SC
RI P
T
intercorrelations.