Psychometric properties of the Urdu version of the Hospital Anxiety and Depression Scale (HADS) among pregnant women in Abbottabad, Pakistan

Study design and setting

This was a methodological study to examine the psychometric soundness of HADS/UV. A hospital-based cross-sectional design was conducted among pregnant women attending the gynaecology outpatient departments (OPDs) for their routine antenatal check-ups from April 2019 till August 2019 at Ayub Teaching Hospital, Abbottabad, Khyber Pakhtunkhwa province, Pakistan. Ayub Medical and Teaching Institution Abbottabad is a 1000-bed tertiary care teaching hospital and the largest healthcare facility of Northern Pakistan and provides major healthcare facilities to the general population. The obstetrics and gynaecology department is well equipped with advanced diagnostic and therapeutic facilities, along with trained staff.15

During the study period, 220 women attending routine antenatal care at Ayub Teaching Hospital were approached and encouraged to participate in the study. Women who consented to participate in the study were interviewed using a structured questionnaire comprising sociodemographic, family structure and relationship variables and maternal characteristics followed by the HADS.

Sample and sampling

A convenience sampling of 200 pregnant women finally completed the interview, while 20 women discontinued the interview and had missing data. Inclusion criteria were pregnancy in any trimester; willingness to participate; and not being diagnosed with psychological disorders before the commencement of the study, including anxiety and depression. A systematic presentation of the study procedure is shown in figure 1.

Request permission

Study tool

Test–retest reliability

A pilot test was conducted to ensure the items’ validity, sequence of questions, complete understanding and average time taken by participants for the interview. We conducted the pilot study before the start of the original survey, the questionnaire was tested–retested on 20 pregnant women attending the gynaecology OPDs for their routine antenatal check-ups, and these data were not included in the final analysis. Participants were requested to complete the translated HADS, along with the social demographic questionnaire, and the same questionnaire was provided to them again after a time interval of 2 weeks. For analysis, we expected the Cronbach alpha equal or greater than 0.70. Test–retest reliability signifies score stability over a time. Internal consistency was verified by examining the Cronbach coefficient alpha of >0.65 for HADS-A, HADS-D and the total HADS.

Reliability analysis

We conducted a reliability analysis of the HADS total, HADS-A and HADS-D subscales to ensure internal consistency. A Cronbach alpha reliability coefficient was considered satisfactory if the value was equal or greater than 0.65.19

Construct validity

Exploratory factor analysis (EFA) was examined on the 14 items of the HADS by means of survey among 200 pregnant women who attended routine antenatal care. Suitability of data and sampling adequacy was examined through assessment of Kaiser Meyer Olkin (KMO) > 0.6 and Bartlett test for sphericity. EFA was performed with varimax rotation and items loaded ≥ 0.3 were included.

Content Validity Index

Content validity was checked to ascertain whether the questionnaire's content is appropriate and the relevance of study purpose. A convenient sample of 10 health experts received the HADS/UV and were asked to rate the relevancy of items to its subscale using the 4-point Likert scale: 1 = not relevant, 2 = somewhat relevant, 3 = relevant and 4 = highly relevant. An item that received a score of 1 or 2 was considered ‘content invalid’, while a rating of 3 or 4 was considered ‘content valid’. Eight raters responded and Item-Level Content Validity Index (I-CVI) and Scale-Level Content Validity Index (S-CVI) were calculated accordingly using the average approach. CVI was calculated as the proportion of items that had ratings of 3 or 4 by the rating experts.

Data collection procedure

Informed consent was taken by the researcher before the start of the interview from pregnant women who visited the gynaecological OPDs. Participants were interviewed by gynaecological and obstetrical physicians. Two days of interviewing skills training were given to data collectors due to the sensitive nature of questions asked during the interview. The interview was conducted in a separated room. The average time taken by the each participant was 20–25 min. Written informed consent was provided by each of the 200 participants who agreed to take part.

Characteristics of the study participants

The demographic, obstetric and gynaecological characteristics of participating women are presented in table 1. A total of 200 pregnant women were included, and the mean age was 26.7 (5.0) years. Half of the women (102 (51%)) were residents in urban areas; 115 (57%) had a history of at most two gravida; and 79 (39.5%) had a history of at least two deliveries. History of previous abortions was reported in 31 women (15.5%). One-third completed up to elementary school and 34 (17%) completed university level, while 55 (27.5%) could not have formal education. Twenty-eight (14%) and 11 women (5.5%) were reported to have hypertension and diabetes, respectively. Most of the respondents belonged to joint family systems (159 (79.5%)).

Assessment of validity

Regarding face validity, reviewers were satisfied with questionnaire appearance. Moreover, questions were clear and easy to understand; however, we made some changes for questions 3 and 4 after the feedback received from the reviewers and made it simpler. For content validity, the I-CVI and S-CVI scores were calculated accordingly. The I-CVI scores of anxiety and depression items ranged between 0.75 and 1.0 and between 0.88 and 1.0, respectively. The S-CVI scores of anxiety and depression items were 0.947 and 0.948, respectively. For HADS, the CVI was 0.947 (table 2).

Analysis of reliability involved assessment of the Cronbach alpha coefficient. The value was adequate (> 0.7), and after removal of each item, the Cronbach alpha did not reveal any major deviation in the overall internal consistency of the HADS. Majority of the corrected item–total correction analysis was above 0.4; however, the lowest value was in the statement ‘I can enjoy a good book or radio or TV programme’, and the value was 0.15 (table 3).

Internal consistency reliability analysis

The anxiety and depression subscale had a Cronbach alpha equal to 0.82 and 0.64, respectively. However, the Cronbach alpha value for whole scale was good (α=0.84). All anxiety items showed a significant correlation with the anxiety subscale (range 0.56–0.77, p<0.001). Similarly, depression items showed a significant correlation with the depression subscale (range 0.39–0.70, p<0.001). Furthermore, items of HADS also showed a significant correlation with the full scale (range 0.27–0.75, p<0.001) (table 4). The two subscales correlated significantly with each other (Pearson’s coefficient r=0.894, p<0.001). In turn, each subscale showed a significant good correlation with the full scale (r=0.930 and r=0.894, p<0.001) for the anxiety and depression subscales, respectively.

Exploratory factor analysis

Prior to factor extraction, sampling adequacy was checked to ensure suitability for factor analysis. The Kaiser-Meyer-Olkin (KMO) and the Bartlett test of sphericity (BTS) showed adequate sampling. The KMO yielded an index of 0.847 and high significant BTS (χ²=800.327, df=91, p<0.001). The number of factors were assessed using eigenvalues greater than 1, which suggested two-factor solution and explained 42.75% of the total response variance. The two-factor solution is the most frequently reported one in literature. The first and second factors explained 34.97% and 7.78% of the variability of data, respectively (table 5).

Main findings

In this study, the HADS was validated on a sample of Pakistani pregnant women living in the district of Abbottabad. Our findings are of clinical importance in terms of validation of a screening tool for anxiety and depression. The HADS/UV proved to be acceptable for pregnant women.

The language of the original English questionnaire was highly metaphorical, and some items had to be carefully translated to adapt to the Urdu language of the Pakistani population. Translation of the HADS items faced some problems, and some items were retranslated at least twice, especially item 8 D4, item 9 A5 and item 11 A6, because of ambiguity of words. Ambiguity was also noticed with the Polish20 and Malayalam versions.14 In our opinion, the translation was sufficient; however, further scrutiny of the meaning is suggested to match with the Urdu language and local context.

The original research of HADS tested its psychometric soundness on a group of hospitalised patients who suffered from different diseases. The scale is proven to be valid and reliable for other groups of patients as well: for oncology,21 22 HIV-infected patients,23 hospitalised elderly,24 gynaecology patients,12 14 patients with coronary artery diseases,25 patients in the emergency department26 and among patients in primary healthcare.27 In turn, our study contributes to the body of knowledge about irregularities of psychometric findings of HADS noticed in the literature28 when it is applied on patients with chronic diseases to others who suffer from physiological changes during pregnancy.

With regard to content validity, the CVI of HADS items revealed high and excellent agreement among experts (k≥0.88), indicating good content validity, except for item 11 A6 (k=0.72), which is below the recommended value (k>0.75).29 Therefore, the question should be reconsidered to make it more understandable and culturally suitable. The S-CVI of the HADS subscales, anxiety and depression, was above the recommended value (0.90).30 Similarly, Zimmermann et al 29 stated that an S-CVI score above 0.83 is sufficient and satisfactory, indicating no ambiguity in understanding the questionnaire. The satisfactory S-CVI indicates cultural relevance.

The HADS/UV of both scales is comparable to previous studies12 31 and showed satisfactory internal consistency. The Cronbach alpha of HADS scales was good (α=0.84) and above recommended value and met Klines' criteria.20 The Cronbach alpha coefficient of the HADS subscales was good for anxiety (α=0.82) but below the recommended value for depression (α=0.64). By and large, the HADS-A subscale showed better psychometric properties than the HADS-D.13 Our study has reported good reliability scores; hence, we recommend that HADS/UV should be used in the clinical setting for screening of anxiety and depression in Pakistan.

Test–retest reliability was also excellent and intraclass correlation was 0.93, which was similar to that of previous studies32 and better than that of the Hyland et al study.21 The stability of the test–retest score over 2 weeks and high alpha coefficient support the strength of the HADS for use among pregnant women. Correlation between two factors, HADS-A and HADS-D, was excellent (0.667), similar to the Greek,33 Iranian,22 Arabic26 and Pashto versions.9 The correlation was strong between HADS of both scales and the subscale scores.21 22 26

The EFA showed unequal distribution of items among the two factors, especially for factor 2. It was bidimensionally similar to the original English published by Zigmond and Snaith34 Majority of ex-studies documented the two-factor model,9 12 13 22 23 26 33 and some reported three factors.20 25 Three exceptions were found, which reported a unidimensional model.23 In the EFA, most items of the HADS-A were loaded in factor 1 except for item 9 A5. A5 was also an unstable item and problematic in Iranian22 and Arabic26 translations. In contrast, item 14 (D7) was loaded less than 0.3 in factor 2 (depression); item ‘D7’ load remained <0.3 in factor 1 (anxiety). Items ‘A4’, ‘D3’ and ‘A7’ shared their loadings between factor 1 and 2. So, we would recommend that it is better to rephrase this item 14 (D7) in future studies as it asks about enjoying a good book, radio or TV programme, and it gave us mixed response due to the different medium of entertainment sources.

In turn, the extracted factors’ correlation matrix showed no high correlation, typically between 0.03 and 0.545, in the initial factor analysis, and very few items were highly correlated with others (>0.50). Items 10 'D5' and item 14 'D7' correlated low (>0.2) with most items. This irregularity of items' distribution and low correlation were observed in the Karimova and Martin31 study and suggests an unsatisfactory model and could be attributed to the impact of biological changes occurring during pregnancy on the perception of emotional components of the HADS. The explained variance of the HADS/UV is 42.75%, which is quite similar to Waqas et al 12 and a bit far from the Watrowski and Rohde20 finding in the same population.

Limitation

This study has certain limitations; first, convergent and criterion validity were not done. Perhaps further validation approaches might be useful to test external validity of the HADS/UV. Second, the study sample was not representative of all three pregnancy trimesters.

Implications

The study revealed applicability of the HADS/UV as a screening tool to determine and measure depression and anxiety among pregnant women. Screening and early detection of psychological distresses, including anxiety and depression, will be very useful to overcome the burden associated with anxiety and depression. Healthcare workers can use this tool to predict early symptoms of anxiety and depression. Further validation studies in specific pregnancy trimesters are required.