Reliability and Validity of Depression Anxiety Stress Scale (DASS)-21 in Screening Common Mental Disorders Among Postpartum Women in Malawi

Background: Approximately one in five women who have recently gave birth suffer from some form of common mental disorder (CMD) particularly depression and/or anxiety.Most available CMD screening tools in most lowand middle income countries do not screen for more than one mental health problem. Having a screening tool that is free to use, short in assessment time and used to screen for more than one CMD is appealing in a resource constrained setting. This study aimed at assessing the reliability and validity of the Depression, Anxiety, and Stress Scale-21 items (DASS-21) for screeningCMDsexperienced by women who have recently given birth in Malawi. Method: We used the gold-standard design for scale validation of comparing the responses to the Chichewa translated and adapted DASS-21 instrument with an independently administered Structured Clinical Interview for DSM-IV (SCID). Internal reliability was reported using Cronbach’s alpha coefficient. The ability of DASS-21 and Edinburgh Postpartum Depression Scale (EPDS) to discriminate cases from non-cases was assessed by receiver operating characteristics (ROC) analysis and its curve was used to select cut-off points for DASS-21 and EPDS that maximise both sensitivity and specificity. Results: : The overall Cronbach’s alpha for DASS-21 scale was 0.74. The internal consistency for DASS-21 depression subscale, anxiety subscale and stress subscale were 0.66, 0.29 and 0.52 respectively. The area under the ROC curve was 0.76 (95% CI: 0.6086; 0.9102) for DASS-21 depression subscale (DASS-D) and 0.65 for DASS-21 anxiety subscale (DASS-A). At a cut-off point of one or more, the sensitivity and specificity for DASS-D was 69.2% and 75.5% whilst DASS-A was 52.9% and 75.5% respectively. Pearson correlation coefficient for the association between DASS-D and EPDS was r = 0.61, p<0.001. Conclusion: DASS-21 as an overall item has good internal reliability but none of its subscales (DASS-D, DASS-A, DASS-S) have satisfactory internal reliability.