Psychometric Properties of the Arabic DASS-21 Scale

The current study aims to investigate the psychometric properties of the Arabic DASS-21 questionnaire. A large-scale cross-sectional study of the Arabic adaptation of the DASS-21 scores. The participants (n= 1235) came from a variety of backgrounds. Overall, the reliability coefficient, Cronbach’s alpha, for the gross DASS-21 score, was 0.94 (95% confidence interval CI between 0.93 and 0.94), indicative of excellent internal consistency. The three-factor model fitted acceptably for the full 21-item dataset. Nonetheless, some items were noted to load onto different factors to their original structure. Our study gives evidence of the excellent reliability, internal consistency, and convergent validity of the Arabic adaptation of DASS-21. We also confirm the acceptability of the current three-factor composition for the DASS-21 in its Arabic version. We may conclude that, in its current format, Arabic DASS-21 can be used in many public and secondary health settings to detect depression, anxiety, and stress symptoms.


Background:
The DASS-21 is an established self-report survey that is composed of a total of 21 questions.The respondent assigns a rating to each question, ranging from 0 (indicating no applicability) to 3 (indicating high applicability).The 21 items are further subdivided into three subcategories: 7 items for depression assessment, 7 items for anxiety assessment, and 7 items for stress assessment.
Many studies assessed the psychometric qualities of the original DASS-21 in its English language format over the last three decades [Brown et al., 1997].The factor structure was reported to be consistent and validity measures were found to be quite good [Lee, 2019].
Many studies attempted to investigate the psychometric qualities of the DASS-21 in its various formats.
The DASS-21 was first proposed by Lovibond and Lovibond [1995].They recruited over 700 participants to examine the psychometric qualities of the Depression Anxiety Stress Scales in contrast to the well-known Beck Depression Inventory (BDI) and the Beck Anxiety Inventory (BAI).The calculated relationships of the BDI and BAI for the DASS-21 were 0.81 and 0.74, respectively.They reported the factor structure to be 'satisfactory' with DASS being superior to BDI in terms of brevity.
Lee [2016] formulated a modified one-factor CFA model of DASS-21.He surveyed a group of 255 students.The latent factor was related to psychological distress, with high-value factor loadings of 0.70.He went further to examine the convergent, discriminant, and nomological validity of this saturated modified one-factor model of DASS-21.Additionally, he measured the convergent validity to be 0.87 with a high correlation between the three subscales of depression, anxiety, and distress that ranged between 0.5 and 0.75.
A Turkish version of DASS-21 [Hekimoglu et al., 2012] was found to have excellent discriminant and stable psychometric properties.
A recent Saudi-based study [Mahfouz et al., 2020] utilized the DASS-21 in investigating the stress levels among Jazan University students and their sleep quality.However, the researchers did not report the psychometric traits of the survey as per the outcomes of their data.It was not clear if they even used an Arabic version of the questionnaire.
Egyptian researchers [Ali et al., 2017] attempted to validate an Arabized version of the DASS-21 among 149 patients with substance use disorders.Although they did not carry out a full confirmatory factor analysis, they focused their analysis on the α coefficient [which was acceptable, but not excellent].They calculated coefficients of reproducibility and scalability and found them even less acceptable.
They discovered that a few of the items were highly challenging and had weak discrimination.They thought that these things were 'problematic' and needed more investigation.
Very little, if any, is known about Arabic DASS-21 in terms of advanced psychometric properties.The primary objective of the current investigation was to examine psychometric properties of the Arabic DASS-21 questionnaire.We also aimed at evaluating the factor structure of the Arabic DASS-21 questionnaire, examining its internal unity, and its convergent, discriminant, and nomological validity, in a large-scale sample of Saudi Arabia citizens.

Method for the Validation:
Demographic and DASS-21 data were collected through a predesigned questionnaire mounted onto an online portal that was distributed to potential respondents by sharing the unique link.The dataset in total was automatically transferred into the Microsoft Excel system.The advanced statistical methods to exami ne the psychometric properties included confirmatory factor analysis.We probed different factor structures including one-factor, two-factor, three-factor, and four-factor structures and compared the models in terms of absolute fit indices such as the chi-squared test, root mean square error of approximation (RMSEA), comparative fit index (CFI), goodness-of-fit index (GFI), root mean square residual (RMR), and standardized root mean square residual (SRMR).Cronbach's alpha and hierarchical omega will be used to examine the unity of the total DASS-21 scale and its three subscales.

DASS-21 Scale Architecture:
The DASS-21 comprises three subscales: the depression scale, anxiety scale, and distress scale.Each subscale is composed of seven items.Subscores for depression, anxiety, and stress are obtained by adding up the scores for the respective items.
The depression scale of the DASS-21 encompasses items that assess divergent aspects such as anhedonia (item 3: "I was unable to feel any good emotions at all"), inertia (item 5: "It was tough for me to muster the motivation to take action"), hopelessness (item 10: "It seemed like there was nothing to look forward to"), dysphoria (item 13: "I had a depressed and blue mood"), lack of interest/ involvement (item 16: "I could never get excited about anything"), self-deprecation (item 17: "I considered myself to be a person of little value"), and devaluation of life (item 21: "I believed that life had no purpose").
On the other hand, the anxiety scale of the DASS-21 evaluates autonomic arousal (item 2: "I noticed that my mouth was dry") and (item 19: "When I wasn't physically active, I could feel my heart beating faster or skipping a beat, for example), " skeletal muscle effects (item 4: "I had a hard time breathing" (e.g., excessively rapid breathing, breathlessness in the absence of physical exertion)") and (item 7: "I experienced trembling (e.g., in the hands)"), situational anxiety (item 9: "I was concerned that there would be circumstances when I'd freak out and seem foolish"), and subjective experience of anxious affect (item 15: "I felt on the verge of panic") and (item 20: "I experienced fear for no apparent reason").
The stress scale within the DASS-21 is specifically designed to identify chronic nonspecific arousal.It assesses diversified aspects such as difficulty in relaxing (item 12: "I had a hard time unwinding") and (item 1: "My ability to relax was challenging"), nervous arousal (item 8: "I felt that my anxious energy was being expended heavily"), being easily upset/ agitated (item 11: "I could feel my temper rising"), irritability/ overreactivity (item 18: "I felt that I was quite sensitive") and (item 6: "I often reacted excessively to circumstances"), and impatience (item 14: "Anything that prevented me from completing my task infuriated me").

Results:
Out of the total participants (n = 1,235), 481 individuals (38.9%) were identified as males, while 754 individuals (61.1%) were females.The majority of participants, specifically 653 individuals (52.9%), were under the age of 25.Table 1 shows a detailed display of the basic demographic factors of the participants.See also Supplementary File 1 for visualization of the DASS-21 scores and subscored broken-up according to different sociodemographic characteristics.We utilized the results of our test to estimate the population-related normative values in accordance with Arabized DASS-21 score, as illustrated in Figure 1 and Table 2.
The normative mean DASS-21 stress subscore in our sample was 7.3 (SD = 5.0), ranging between 0 and 21; the median DASS-21 stress subscore was 7.  Overall, the reliability coefficient, Cronbach's alpha, for the total DASS-21 score was 0.94 (95% confidence interval CI between 0.93 and 0.94), indicative of excellent internal consistency.All its items were of comparable impact on the overall reliability estimate.Further information is available in Supplementary File 2.
The reliability coefficient for the depression subscale of the Arabic DASS-21 was 0.87 (95% CI between 0.86 and 0.88), signifying good internal consistency.None of its items has a significant impact on the overall reliability estimate.
Similarly, the Cronbach's alpha reliability coefficient for the anxiety subscale was 0.84 (95% CI between 0.82 and 0.85), reflecting good internal consistency.All items were comparable in terms of effect on estimates for reliability coefficients.
In addition, the Cronbach's alpha reliability coefficient for the stress subscale was 0.86 (95% CI between 0.85 and 0.87), implying good internal consistency.None of its items has a significan impact on the overall reliability estimate.

Confirmatory Factor Analysis:
We carried out a confirmatory factor analysis to guarantee that the three dimensions of depression, stress, and anxiety conform satisfactorily to the Arabic adaptation of the DASS-21.We sequentially fitted one-factor, twofactor, three-factor, and four-factor models to the dataset.The three-factor model fitted acceptably for the full 21-item dataset.Nonetheless, some items were noted to load onto different factors to their original structure.
Item 5: "It was tough for me to muster the motivation to take action" exhibited a stronger association with the stress factor compared to the supposed depression factor, although the difference was marginal.Additionally, item 9: "I was concerned that there would be circumstances when I'd freak out and seem foolish" demonstrated a significantly stronger association with the stress factor than the original anxiety factor.Furthermore, item 11: "I could feel my temper rising" showed a slightly stronger association with the depression factor instead of the expected stress factor.

Psychometric Properties of the Arabic…
Dr. Fatimah Alharbi, Dr. Mugtaba Osman Moreover, item 14: "Anything that prevented me from completing my task infuriated me" deviated from the stress factor and loaded significantly better onto the depression factor.However, a fuller 4-factor structure fitted better than models with fewer factors.An additional dimension (composed of item 1: "My ability to relax was challenging" and item 12: "I had a hard time unwinding") is required for the sample data to conform acceptably to the theoretical covariance structure.See Supplementary File 3 for a detailed account of the four-factor loading.We note that the meaning of those two items can be almost identical when translated into Arabic tongue.
We further attempted to perform the confirmatory factor analysis of a three-factor model after the removal of the two poorly performing items (item 1: "My ability to relax was challenging" and item 12: "I had a hard time unwinding").
Although not a better fit than the 4-factor model, this item-deleted threefactor model was noticeably superior to the full three-factor model in the entire dataset (see Table 3 and Table 5).
The comparative fit index (CFI) > 0.95, Tucker-Lewis index (TLI) > 0.95, and root mean square error of approximation (RMSEA) < 0.06 are chosen for a good fit [Hu et al., 1999].A good fit to the dataset is revealed with an RMSEA value < 0.05 and an acceptable fit is achieved when the RMSEA is < 0.08 [Browne and Cudeck ,1993].CFI, comparative fit index; TLI, Tucker-Lewis index; RMSEA, root mean square error of approximation; CI, confidence interval; df, degrees of freedom; SRMR, standardized root-mean-square residuals.Three-factor*: the original three-dimensional structure.

Discriminant Validity Analysis:
Discriminant validity is considered an index of the difference between the three underlying constructs (namely, depression, stress, and anxiety).
These constructs are inherently similar; therefore, a strong correlation is expected between them and, hence, low discriminatory power.We opted to use a multiply operationalized procedure method to probe the correlation matrix of z-transformed scores of each of the three constructs [Cole et al., 1981].The discriminant validity was poor for the three constructs.

Convergent Validity Analysis:
A measurement scale's convergent validity is established when there is a strong correlation between its discrete components.[Duckworth and Kern, 2011].We examined the correlation between each of the DASS-21 items and Our study gives evidence to the excellent reliability, internal consistency, and convergent validity for the Arabic adaptation of DASS-21.
We also confirm the acceptability of the current three-factor structure for the DASS-21 in its Arabic version.We may conclude that, in its current format, Arabic DASS-21 can be used in many public and secondary health settings to detect depressive, anxiety, and stress symptoms.
However, we have also identified room for improvement.
A couple of stress-related items seem to topple away from the overall DASS-21 construct, namely, item 1: "My ability to relax was challenging" and item 12: "I had a hard time unwinding.Also, a couple of items were detected to go better with the depressive construct rather than the stress construct in the Arabic DASS-21 adaptation, namely, item 11: "I could feel my temper rising" and item 14: "Anything that prevented me from completing my task infuriated me."Moreover, a couple of factors performed better with the stress construct rather than the depressive or the anxiety constructs, namely, item 5: "It was tough for me to muster the motivation to take action" and item 9: "I was concerned that there would be circumstances when I'd freak out and seem foolish." The three items should be reworded for future research, and focus groups should be organized to learn more about respondents' attitudes toward these particular questions.The current research has numerous strengths.The specimen size was large and exemplary of Saudi citizens to a great extent.However, the cross-sectional and online design would not elucidate bidirectional associations or selection bias, as computer and internet literacy is an important filter in selecting participants.

Figure 1 .
Figure 1.Normative values for the DASS-21 subscales among our Saudi subjects

Figure 2 :
Figure 2: Distribution of DASS-21 score by the adjusted effect of geographic region Conclusion: