The present study evaluated the psychometric properties of the Dari version of the DASS-21 among Afghans. To date, the DASS-21 and DASS-42 have been translated into 57 different national languages, including Dari, Pashto, Persian, Arabic, Urdu, Hindi, Indonesian, Malaysian, Chinese, Thai, and Turkish (Psychology Foundation of Australia, 2023). Due to the fact that studies have been carried out in many different cultures, variations regarding the instrument’s internal structure and psychometric properties have been observed.
Various models have been observed in the literature, encompassing a range of structures such as a three-factor model, a second-order model comprising three factors (anxiety, depression, and stress), a tripartite model (consisting of anhedonia, physiological hyperarousal, and a general negative factor), a bifactor model (where items primarily load on a general factor of distress alongside specific factors of stress, depression, and anxiety), a two-factor model (comprising depression and anxiety/stress), a four-factor model (including a general factor of distress, alongside depression, anxiety, and stress factors), and a single-factor model of distress (Apóstolo et al., 2012; Lee, 2019; Nanthakumar et al., 2017). The findings lend support to the three-factor model (anxiety, depression, and stress) of the DASS-21, as corroborated by previous studies (e.g., Apóstolo et al., 2012).
The Cronbach’s alpha for the total DASS-21 was 0.94, and for depression, anxiety and stress subscales. The alpha values were .860, .832, .883 respectively. These findings indicated that the Dari DASS-21 demonstrates very good internal consistency. The findings also indicated a robust and consistent level of test-retest reliability for all subscales, with an ICC of 0.916, 0.870, and 0.784 for anxiety, depression, and stress, respectively. The ICC for the total DASS-21 was 0.924. These findings demonstrate that the Dari version of DASS-21 is a reliable psychometric instrument with excellent internal consistency. In the original validation investigation, the Cronbach’s alpha coefficients were recorded as 0.91 for depression, 0.81 for anxiety, and 0.89 for stress (Lovibond & Lovibond, 1995). The findings in the present study concur with those of the original study.
The overall Cronbach alpha concurs with (and is even better in some instances) than other studies from India (0.85 among adults aged between 18 and 25 years) (Sharma et al., 2020), Bangladesh (0.98 among university students) (Mostafa Alim et al., 2017), China (0.89 among school teachers) (Cao et al., 2023), Turkey (0.82 among non-clinical population) (Yıldırım et al., 2018), Vietnam (0.761 among adults) (Le et al., 2017), Malaysia (0.87 among students, teachers, and non-teaching staff of a university) (Thiyagarajan et al., 2022), Indonesia (0.91 among the general population) (Ifdil et al., 2022), and Greece (0.97 among the adult population) (Lyrakos et al., 2011).
In the present study, and as expected, significant positive associations were found between total score on the DASS-21 and total scores on both the CES-D-20 and the GHQ-28. There were also significant positive associations between the DASS-21-depression subscale scores and scores on both the CES-D-20 and GHQ-28 depression subscales. Similarly, scores on the DASS-21-anxiety subscale were strongly positively associated with scores on both the CES-D-20 and GHQ-28 anxiety subscale. Finally, scores on the DASS-21 stress subscale had positive associations with CES-D-20 and the GHQ-28 anxiety subscale. Combined, all of these findings demonstrate robust convergent validity of the DASS-21 with other scales and subscales assessing the same conditions.
The findings also concur with a 2017 Afghan study conducted with 1310 non-clinical randomly selected participants in Herat city using the DASS-42. In that study, the Cronbach’s alpha values were 0.888 for depression, 0.866 for stress, and 0.833 for anxiety (Shayan et al., 2021). The marginal discrepancies in outcomes could potentially be attributed to the research methodology, which entailed extended time intervals between the initial and subsequent administrations of the DASS-21, and/or the utilization of varied iterations of the DASS (Lee et al., 2019).
In the present study, all subscales and the overall DASS-21 exhibited favorable fit indices, supporting the proposed three-factor model structure. These results concur with studies conducted among clinical and non-clinical Canadian adults (Antony et al., 1998), non-clinical UK adults (Henry & Crawford, 2005), depressed Canadian adults (Clara et al., 2001), non-clinical US adults (Sinclair et al., 2012), and non-clinical Iranian adults (Asghari et al., 2008). However, test-retest reliability of the DASS-21 has not been observed in studies conducted in Australia (Gomez et al., 2014; Ownsworth et al., 2008,), Italy (Bottesi et al., 2015), and China (Wang et al., 2016). This occurrence could potentially be attributed to a methodological issue arising from prolonged intervals between the initial and subsequent administrations of the DASS-21 within the research protocol (Lee et al., 2019).
Although the present study used established psychometric scales, and had a relatively large sample, there were some limitations. First, the data were collected from non-clinical samples. Further studies should be conducted to replicate the present study with a clinical population. Second, information concerning the educational background of participants was not collected from participants in the present study. Different levels of education as well as different age groups may lead to differences in comprehension of the questions in the scale. Another limitation of this study was that the sample is not representative of all Afghans. The sample only contained participants from a few chosen Dari speaker provinces. Finally, although scale validations have to necessarily use self-report data, such data can still produce socially desirable responses and can still be fallible to memory recall.