of .80, and a moderate correlation of .5. This study protocol was approved by the institutional review board of the hospital. All participants were informed of the study's purpose, the potential risks, their right to withdraw from the study at any time, and the process, and the participants signed the consent form prior to entering the study.
Measures
Personal Information Questionnaire
This questionnaire gathered data about gender, age, occupation, and family disease history.
Chinese Version of the Mini International Neuropsychological Interview
The Mini International Neuropsychological In- terview (MINI) is a structured interview scale developed based on the DSM-IV. This scale can be used to diagnose 17 diseases, including manic episodes, hypomanic episodes, and major depres- sion. This structured interview takes about 20 minutes and is very effective for epidemiological investigation. When the original scale was pub- lished abroad, it was regarded as simple, easy to use, and precise. The validity of the MINI was assessed by comparing its diagnostic concordance with that of the Structured Clinical Interview for DSM-IV Diagnoses and that of the Composite International Diagnostic Interview for the International Statistical Classification of Disease and Related Health Problems (ICD-10). Generally, the sensitivity of this scale is at least .70, and the specificity is more than .85 (Adewuya, 2006). The MINI was translated into Chinese and found to have a sensitivity of .79, specificity of .94, and overall accuracy rate of .88 (Lin, Chen, Liu, & Chen, 2002).
Procedure
The MDQ-C was examined through three phases: Phase 1, translation and backtranslation with establishment of face and content validity; Phase 2, examination for internal consistency, construct validity, sensitivity, and specificity; and phase 3, contrast group comparisons.
Translation of the MDQ
The MDQ-C was translated following Brislin's (1970) steps for translation:
1. Find competent bilingual translators who are familiar with the content.
2. After practice, one translator was asked to translate from the source language (SL, i.e., English) to the target language (TL, i.e., Chinese). Another translator was asked to blindly backtranslate from the TL into the SL.
3. Several raters examined the original SL, the TL, and backtranslation versions for meaning errors.
4. When the meaning errors were nil, the target version was pretested on monolingual TL populations, and revisions were made in both the SL and TL versions if the pretest showed problems of comprehension.
Upon completion of the first draft, we invited 15 patients to fill out the MDQ-C and asked them to provide comments regarding the content, meaning, and expression of the scale. Then, we revised the scale, established content validity, and invited five experts in the field to test face validity. Finally, the terminology and content of the scale were revised based on the experts' opinions. In addition, the scale itself could be revised if necessary according to the experts' opinions.
Examination of the Psychometric Properties of the MDQ-C
A total of 170 patients with mood disorders were recruited and completed the personal infor- mation questionnaire and the MDQ-C. The participants were evaluated using the MINI and DSM-IV-TR as criterion standard for inclusion in the study by on-site psychiatrists. The examination of reliability included tests for internal consistency and test–retest reliability. Internal consistency of the MDQ-C was assessed by determining the coefficient alpha, a measure of the degree to which scale items measure a homogeneous construct or characteristic. Test–Retest reliability over a 2- week interval was determined by correlating the total scale scores on the MDQ-C at the initial and subsequent administrations. Because of the elapsed time between patients' visits to the hospital, we assessed the test–retest reliability of the MDQ-C in 54 patients who revisited the hospital within 2 weeks. Construct validity was established by factor analysis and contrast group comparison. We adopted the principal component analysis (PCA) with varimax rotation to extract mutual factors.
56
LIN ET AL
The contrast group was used to compare persons with and without bipolar disorder matched by age and sex. A difference analysis using the test of homogeneity of proportions was performed to distinguish the two groups of people by character- istics that differed significantly (170 bipolar disorders and 88 subjects without psychiatric illness from local communities).
Receiver operating characteristic (ROC) curves were used to calculate bipolar sensitivity, specific- ity, positive predictive value, overall accuracy rate, and the best cutoff point. These properties were compared with the MINI diagnosis. The area under the curve (AUC) is a comprehensive means of comparing different tests or different scoring procedures for one test. The AUC can be used to estimate the accuracy of a diagnostic test.
Data were analyzed using the SPSS package for Windows 15.0. Basic demographics were analyzed through descriptive analyses. Pearson's correlation test was used for test–retest reliability, and Cronbach's alpha coefficient was used to evaluate internal consistency. To examine validity, a content validity index (CVI) greater than 80% was regarded as a standard for testing expert validity (higher scores mean greater relevance, less duplication, and greater appropriateness). Student's t test was used to see if there were differences between groups in the contrast group comparison. The MDQ-C was analyzed using PCA with iteration, followed by a varimax rotation and Kaiser normalization for factor analysis, to analyze the structure of the first part of the questionnaire. The ROC analysis was further used to determine the best cutoff point to distinguish patients with or without mood disorders.
RESULTS Demographic Characteristics
Of 176 subjects who met the recruitment criteria, 4 (2.27%) did not complete the questionnaires, and 2 (1.14%) failed to complete the screening process. Therefore, 170 subjects remained for analysis. The average age of the study participants was 38.94 ± 13.77 years, and 97 (57.06%) were female. In addition, 88.22% of them had a high school diploma. On the basis of the MINI diagnostic criteria, 67 (39.41%) of the participants were identified as having bipolar I disorder (BD I), 25 (14.71%) as having bipolar II disorder (BD II), and
3 (1.76%) as having bipolar disorder NOS, and 75 (44.12%) had major depression.
Reliability
On the basis of the 15 items of the questionnaire, the Cronbach's alpha, representing internal consis- tency reliability, was .82 (n = 170). The item-to- total correlation coefficient for each item ranged between .40 and .68 (P b .01). The test–retest reliability was evaluated based on 54 subjects who retook the MDQ-C at 2-week intervals, and the correlation coefficient was .76 (P b .01).
Validity
Content Validity Index
The CVI was used to quantify the extent of agreement between the experts. The advisory committee included two bilingual language experts, two psychiatrists, and one psychiatric nurse who assessed content validity. Given the 15 items in the scale, five experts reached agreement on 12 of them, and the overall CVI of this scale was .80. The CVI indicated that 80% of the screening items were viewed as extremely appropriate by the experts, and the content of the scale is acceptable, so it can be applied clinically.
Contrasted Group Comparison
The researcher recruited 88 subjects without psychiatric illness from local communities who matched the study patients by age and sex. These subjects also filled out the MDQ-C. Using Stu- dent's t test to analyze the difference between people with and without psychiatric illness, we found that patients with psychiatric disorders had an average MDQ-C score of 7.91 (SD = 2.68), and the average score among the comparison group was 1.40 (SD = 1.55). The results of the MDQ reached a statistically significant difference between groups (t = 19.07, P b .05). This finding indicated that the scale possessed good discrimination capability.
Factor Analysis
To understand the factor construction of the MDQ-C, we first performed the Bartlett's test of sphericity and Kaiser–Meyer–Olkin measure of sampling adequacy (KMO) to examine the first part of the questionnaire (13 items). The results showed that the KMO was .82 (N.80), and the result of the