A Measure of Depression in a Modern Asian Community: Singapore

The construct validity of two depression measures, Zung's Self-rating Depression Scale (SDS) and the Asian Adolescents Depression Scale (AADS), was investigated. Three studies were conducted using two samples collected in two stages, and the results were used to construct the Asian Depression Scale (ADS). Participants responded to the SDS and AADS in random order of presentation during stage 1; two months later, validation variables were collected. Study 1 found that the SDS is a reliable and valid measure of depression for Singaporean Chinese, but it does not cover the interpersonal dimension found in the AADS. Study 2 combined the two measures and found six factors. One of these factors, negative social self, which was a unique Asian depressive symptom cluster, consisted only of AADS items, while the affective manifestation and psychosomatic symptoms factor primarily consisted of items from the SDS. Study 3 selected high-loading items from the identified factors to construct the ADS, which showed excellent internal reliability, and good convergent and discriminant validity. Incremental predictive validity found on criterion data collected in stage 2, supported the nonspuriousness of the Asian Depression Scale.

Epidemiological studies of depression in Asians are rare and have used a variety of instruments, most of which were developed in non-Asian populations [17,18]. These studies have reported low rates of depression in the Asia Pacific region [17]. Recently, however, observers have noted a rapid increase in reported cases of depression in Asian populations [18,19], especially in regions such as Singapore, where there has been rapid industrial and economic development. The lack of consensus over diagnostic criteria, the lack of culturally appropriate norms [20], and the controversy over the cultural validity of the assessment tools used [6,18] make it difficult to accurately determine the prevalence of depression in Asian regions. Culturally appropriate assessment tools for identifying depression in non-Western regions are required [17,21]. This study constructs a comprehensive and valid measure for assessing depression in Singapore, a predominantly Chinese (75% of general population), ethnic Asian community.

Expressions of Depressive Symptoms in Asia
Earlier studies of depressive symptoms in Asian populations, especially the Chinese, have found that the Chinese tend 2 Depression Research and Treatment to de-emphasize affective symptoms; instead, they present somatized complaints (e.g., [3,5,22,23]). Using Zung's [24] Self-rating Depression Scale (SDS), Chang [16] found that Asian Americans showed a high salience of psychosomatic symptoms compared to other ethnic American groups. East Asian students and American students of Asian descent [10,20,25] exhibited depressive symptoms similar to other students, but with lower affective manifestations.
We have explored the depressive symptomatology of Asians living in Asian communities. We used in-depth interviews, followed by focus group discussions with children and adolescents. The reported depressive symptoms were factor analyzed to identify the symptom groups and to construct an Asian Adolescents Depression Scale (AADS) [26] and Asian Children's Depression Scale [27]. A common finding across both age groups was a socially-oriented dimension that contained negative feelings about self-other relationships. We think this newly identified symptom of depressionnegative social Self-reflects the culture's emphasis on the interdependent self-construal in regulating one's emotions [28,29], which is an emic expression of emotions in collectivist cultures [30].
The above review suggested the following features in the depressive symptoms of Asian populations: a relative deemphasis of affective manifestations, and a strong emphasis on the somatic and interpersonal symptoms. The AADS is high on interpersonal symptoms, while SDS has a considerable portion of somatized symptoms, with considerable overlap between them. Three empirical studies were conducted to develop a comprehensive measure of depression by amalgamating these two measures.
The current study aims to establish a valid and comprehensive instrument of depression to be used mainly for screening and survey of community level screening. We have chosen the AADS and the SDS for which extensive empirical data have been collected to support each instrument's internal reliability and easy interpretability for use in Asian community samples across a wide age range. We have also considered World Health Organization's International Statistical Classification of Diseases and Related Health Problems (ICD-10) [31]. The ten-item ICD-10 taps 10 general symptoms that are considered common expressions across different cultures that are covered in Zung's SDS but does not contain the variety and diversity of each symptom category as the SDS. Beck's Depression Index (BDI) is a theory-based instrument of 21-item scale used popularly mainly in North America. The BDI has met with mixed reviews in terms of the need for multiple periods of assessments [32] and is seldom used in Asian communities [33]. We therefore chose the AADS and SDS rather than the ICD-10 and BDI.

Study 1: Internal Structure and Construct
Validity of Self-Rating Depression Scale  [24]. This 20-item self-rating depression scale was developed by Zung [24] and has been widely used for both clinical and community samples. It has been used in Asian communities and was found to show acceptable reliability. A for the current study has been found to be around .80.

The Asian Adolescent Depression Scale.
For convergent and discriminant validity: measures of positive affects (PA) [34] and measures of negative affect (NA) [34] and depression measured by the abbreviated version of Asian Adolescent Depression Scale (AADS) [26]. The AADS consists of 24 items and has shown high internal reliability. For the current sample the internal reliability was found to be α = .945.

Test Anxiety
Scale. The 20-item test anxiety scale has been widely used to assess anxiety related to being evaluated. For the present sample, the internal reliability was found to be .912. [36]. The Hope Scale constructed by Snyder et al. [36] is a 12-item measure and is divided into the hope agency (6 items) and hope pathway (6 items) subscales. For the present sample the internal reliability has been found to be .70 for the hope subscale and .65 for the pathway subscale, relatively low but acceptable. The Life Orientation (LOT): The Life Orientation Scale [37]. The life orientation test (LOT) was developed to assess individual differences in generalized optimism versus pessimism. Using factor analysis, we trimmed off the items with insignificant loading. The remaining 9-item LOT showed a high internal reliability of .87 for the present sample.

3.3.
Procedures. Data were collected in two stages. In the first stage, data were collected on the SDS, the AADS, Depression Research and Treatment 3 and demographic variables of the exploration sample. In the second stage, data were collected from the validation sample on a set of validation variables, including positive and negative affects such as optimism, hope, test anxiety, life satisfaction, and Asian subjective wellbeing.

Exploratory Factor
Analysis. Data analysis was guided by our objective to identify underlying dimensions that are meaningful and parsimonious. To achieve this objective, we employed a series of exploratory factor analysis and conceptual grouping of items into meaningful categories to construct a hypothesized factor structural model to be subject to confirmatory factor analysis.
First, a principal components analysis with Promax rotation was conducted on the items in the Self-rating Depression Scale. We chose Promax rotation for two reasons: (1) we hypothesized that the factors were correlated, however, (2) we did not have a apriori hypothesis as to the magnitude of factor correlation, therefore we did not use Oblimin rotation but chose Promax rotation as suggested by Gorsuch [38]. The scree plot suggested a three-factor solution, with eigenvalues and variance accounted for values being 5.80 and 29.00% for factor one; 1.84 and 9.21% for factor two; 1.40 and 7.01% for factor three, respectively.
Factor two was labeled "loss of self-efficacy" and included items such as "I find it easy to make decisions (reversed coded)" and "I find it easy to do the things I used to do (reversed coded)." Factor three was labeled "physical symptoms" and included items such as "I eat as much as I used to (reverse coded)" and "I have trouble with constipation." Additional factor analysis was conducted on Factor one, which initially included conceptually mixed items. Two conceptually coherent factors emerged: loss of life directions and affective manifestations.
Finally, by reviewing factor loadings and sorting items into meaningful categories, four factors were identified for the SDS: loss of life direction (existentialist factor) (four items), loss of self-efficacy (five items), affective manifestations (two items), and physical symptoms (nine items). Table 1 presents the EFA results, where LLD refers to loss of life directions, LSE refers to loss of self-esteem, AM refers to affective manifestations and PS refers to physical symptoms.

Internal Reliability.
Internal reliability of the SDS was .801 and .803 for the exploratory sample and the validation sample, respectively. Both were acceptable, but were considerably lower than the internal reliability of .945 of AADS in the validation sample.

Participants and Method.
Participants in both subsamples were administered the SDS and AADS in random order. The items in the Self-rating Depression Scale and the Asian Adolescent Scale were combined for analyses, resulting in a total of 40 items. A series of EFA and CFA were conducted to test for the best fit for the underlying factors and the internal structure.

Exploratory Factor Analyses.
A principal components analysis with Promax rotation was conducted on the 40 items. The scree plot suggested a three factor solution, with eigenvalues and variance accounted for values of 13.15 and 32.88% for factor one (16 items), 2.29 and 5.72% for factor two (15 items), and 2.15 and 5.37% for factor three (nine items), respectively. Factor three was conceptually coherent, with items such as "I get tried for no reason," "my heart beats faster than normal," and "I have trouble sleeping at night." This factor was labeled "physical symptoms." factor one and factor two contained conceptually mixed items, so further factor analyses were conducted.
After analysis of the 15 items in Factor One, two facets were identified: "Loss of Life Direction (existentialist factor)," which included items such as "my life is pretty full (reversed coded)" and "I feel hopeful about the future (reversed coded)," and "negative self and relationship," which included items such as "I do not like going out with friends or meeting people" and "I feel that others would be better off if I were dead." The remaining items consisted of symptoms of affective manifestations and were combined with factor three.
Further analysis of factor two yielded additional facets. The first and third facetswith items such as "I take a long time to decide on things" and "my mind is as clear as it used to be (reversed coded)"-were grouped into one factor, that was labeled "loss of cognitive efficacy." The second facet included items such as "I feel that have no energy to do things most of the time" and "I do not feel like doing anything," and was labeled "loss of motivation." In the final analysis, six factors were identified: loss of life direction (existentialist factor) (five items), negative self and relationship esteem (five items), loss of cognitive efficacy (eight items), loss of motivation (seven items), affective manifestations (five items) and physical symptoms (ten items). This six-factor solution formed the hypothesized factor model to be tested by confirmatory factor analysis. Table 2 presents the results of EFA on the combined item pool.
As mentioned earlier, we aimed at identifying the best fit internal structure that was meaningful and parsimonious. We hypothesized that some of the factors identified may be empirically grouped into higher order factors. Table 3 presents the correlations, reliabilities, and descriptive statistics of the six factors for subsample one and subsample two. The pattern of correlations between the six factors suggests that the six factors could be regrouped into three higher order factors. Loss of life direction and negative social self were highly correlated with each other in the two samples (r 1 = .68, P 1 < .05 and r 2 = .74, P 2 < .05, resp.), hence they were grouped into one higher order factor, labeled "loss of life direction". Loss of cognitive efficacy and loss of motivation were also highly correlated with each other in both samples (r 1 = .72, P 1 < .05 and r 2 = .78, P 2 < .05, resp.) and were grouped into a second higher order factor, labeled "loss of self efficacy". In subsample two, physical symptoms had a higher correlation with affective manifestations (r 2 = .61, P 2 < .05), than with the other factors, hence, these two factors were grouped into a third higher order factor, labeled "affective and physical symptoms."

Correlations between Factors.
Confirmatory factor analyses were then conducted to test and confirm the conceptual appropriateness of the threefactor structure of depression.
Subsample One. Using the item mean of each factor as a manifested indicator, with two indicators for each of the three latent factors, a reasonably good fit was found for the higher order structure (χ 2 (6) = 22.70, P < .05, GFI = .96, AGFI = .85, SRMR = .04).
Subsample Two. Using the item mean of each factor as an indicator, with two indicators for each of the three latent factors, a good fit was found for the higher order structure (χ 2 (6) = 6.83, P = ns, GFI = .99, AGFI = .95, SRMR = .02).

Discussion.
Of the six factors that emerged from the joint item pool, the negative social self factor, which consists mainly of AADS items, appears to be a specifically Asian depressive manifestation. The loss of self-efficacy factor, which includes loss of cognitive efficacy and loss of motivation symptoms, corresponds with young Asian's worries about achievement and failure. The SDS's factor of affective and somatic symptoms confirms the earlier findings on Asian My heart beats faster than normal 0.53 S 10 I get tried for no reason 0.65 S 13 I am restless and can't keep still 0.11 S 15 I am more irritable than usual 0.53 A 7 I am more bad tempered than before 0.55 Note. "S" refers to items in the Self-rating Depression Scale; "A" refers to items in the Asian Adolescent Depression Scale. LLD: loss of life direction; NSS: negative social self; LCE: loss of cognitive efficacy; LOM: loss of motivation; AM: affective manifestations; PS: physical symptoms. depressive complaints. The integrated symptomatology of AADS and SDS is a more comprehensive reflection of the contemporary Asian experience of depression [18] than either AADS or SDS alone.

Study 3: An Asian Depression Measure
Using the integrated symptomatology of depression derived in Study 2, we constructed the Asian Depression Scale (ADS) and assessed its internal reliability, construct validity, and incremental validity.

Participants and Method.
Data from the validation sample were used to construct and validate the ADS. Materials used were as described above.

Scale Construction.
We examined the individual item loadings in all three factors and chose the higher loading items and deleted redundant items. The resultant scale maintained the comprehensive factor structure of the integrated symptomatology.

Data Analysis.
We explored the internal reliability, the convergent and predictive validity of the newly constructed ADS. We conducted exploratory factor analysis followed by confirmatory factor analysis (EFA and CFA) for the ADS. Finally, using validating data collected at stage 2, we tested whether the incremental validity of the ADS was a spurious measure over SDS or AADS. Table 4 presents the items and factors of the Asian Depression Scale. Using the 24-item ADS, the three-factor structure yielded a reasonable fit (χ 2 (249) = 403.84, P < .05, GFI = .83, AGFI = .79, SRMR = .06), lending empirical support to the validity of loss of meaning of life, the loss of self-efficacy, and affectivesomatic symptoms as dimensions constituting the ADS (see Table 3).

Internal Reliability.
The ADS had excellent psychometric properties. Cronbach's alpha was .964.

Concurrent Validity.
The intercorrelation between the ADS with the AADS were r = .980 and P < .000, and SDS, r = .861 and P < .000; both significantly high, suggesting that the ADS measures the same underlying construct as AADS and SDS.

Discriminant and Convergent
Validity. The ADS was negatively correlated with positive affects and cognitions. It My heart feels heavy 1 I often feel like crying 24 My heart beats faster than normal 10 I get tried for no reason 12 I am more bad tempered than before Note: Numbers refer to item numbering in the Asian Depression Scale.

5.
3. Discussion. The ADS demonstrates excellent internal reliability and is highly correlated with the ADS and the SDS. Furthermore, the ADS correlates with both positive and negative affects, subjective wellbeing and hope, and life satisfaction in the theoretically correct directions. This pattern of correlations supports the conclusion that the ADS is a measure of depression. Most importantly, the ADS demonstrates significant incremental validity over the AADS and SDS in predicting positive and negative outcomes, suggesting that it is not a spurious measure of depression for the Asian population.

General Discussion
The amalgamated items from the SDS and the AADS have excellent internal coherence and culturally meaningful factors, suggesting that the combined list provides a more accurate assessment of depression for Singaporean adults than either measure alone.
The ADS yields incremental validity on the positive and negative criteria over its source measures, suggesting that it is not a spurious measure of depression over the AADS or the SDS.
These empirical studies were conducted with Singaporean Chinese samples; however, this should not greatly limit the application of the ADS. The social dimension identified here reflects a shared Asian cultural emphasis on the interdependence of self-construal. The ADS is more suitable for other collectivist Asian cultures than the standardized Western measures.
Cross-cultural comparisons with non-Asian populations should also be conducted to test the boundary of cultural limitations of the instrument. Furthermore, as the ADS was developed on community samples, studies should be conducted to test the utility of the scale in clinical settings.