Factor analysis of Zung Scale scores in a Japanese general population

Authors


Dr Fuminori Chida, Department of Neuropsychiatry, School of Medicine, Iwate Medical University, 19-1 Uchimaru, Morioka, Iwate 020-8505, Japan. Email: chida-psy@umin.ac.jp

Abstract

Abstract  The purpose of the present paper was to investigate the distribution of Zung Self-rating Depression Scale (SDS) scores in a general population and its factor structure. Questionnaires on SDS items were sent to 7136 randomly selected residents aged 20–79 years who lived in districts in Japan with high rates of suicide. Valid responses were received from 5547 residents (response rate: 77.7%). Factor analysis of the SDS scores was conducted. The SDS scores of the male subjects were significantly lower than those of the female subjects in all age groups. A reverse-J-shaped relationship was found between age groups and mean SDS scores for the male and female subjects. The highest mean score was in the age group of 20–39 years, and the lowest mean score was in the age group of 60–69 years for the male and female subjects. In factor analysis, two factors consisting of 12 items were extracted, and 10 of those 12 items covered six Diagnostic and Statistical Manual of Mental Disorders (4th edn; DSM-IV) criteria describing psychological disturbances of depression. The distribution of SDS scores differed depending on the age group. Major components of SDS in the subjects covered the DSM-IV criteria for psychological disturbances of depression.

INTRODUCTION

The rate of mortality from suicide in Japan is relatively high compared to that of other developed countries.1 It increased steeply to 25.4 per 100 000 (31 755 deaths) in 1998 and has not decreased since then. The steep rise in the rate of mortality from suicide in 1998 was caused by an increase in suicide among middle-aged men.2 Depression is thought to be one of the major causes of suicide; the risk of suicide among depressive patients is higher than that among other psychiatric patients or mentally healthy people.3 Extant psychological autopsy studies have also revealed that most suicide victims had suffered from mental disorders such as depression.4 Programs aimed at prevention of suicide have thus focused on depression.5,6 We focused on depression as one of the major causes of suicide in a community intervention trial in Japan.

The Zung Self-rating Depression Scale (SDS) and other simple self-reported scales for depression have been developed to identify people suffering from depression.7–9 The SDS has been used to assess depressive states in general populations.10–12 However, the usefulness of the SDS in representative samples has not been clarified.

In the present study we investigated the distribution of the SDS and its factor structure and examined the usefulness of the SDS for detecting depression in a general population.

METHODS

Study population

The present study was carried out as a part of the Suicide Prevention in North Rias Trial (SUNRIT), which aimed at improving people's knowledge of and attitude toward depression, and at reducing the rate of and mortality from suicide. The SUNRIT was conducted in Iwate Prefecture, a prefecture in northern Japan with a high rate of suicide.

Four of the six municipalities in Kuji District (Kuji, Yamagata, Ohno, and Taneichi) served as an intervention area (population: 69 000). Miyako District, located next to Kuji District, also participated in the trial, three of the six municipalities in the district (Iwaizumi, Taro, and Niisato) serving as a control areas (population: 21 000). Both districts are along the Pacific and have a low population density (Kuji District: 65.5 people/km2, Miyako District: 15.9 people/km2) and a high ratio of elderly people (population aged ≥65 years: Kuji District, 20.7%; Miyako District, 28.3%). The SUNRIT started in January 2002 and will continue until 2008.

We randomly selected subjects aged 20–79 years from lists of residents offered by the local governments. Residents unable to respond to our questionnaire, such as those who lived outside the municipality, those who were institutionalized, or those who were unable to answer due to severe illness, were excluded from the study.

Baseline survey

The baseline survey for this trial was conducted during February 2002 and June 2002. A questionnaire with a letter explaining the objective of the survey was sent to each subject. Items of the questionnaire consisted of (i) sociodemographic information (job, family members, martial status, education, financial situation, community participation etc.); (ii) lifestyles (smoking, drinking, diet, exercise etc.); (iii) mental health (knowledge of, views on, and attitudes toward mental disorders and suicide etc.); and (iv) the Japanese version of the SDS.13

Questionnaires were mailed to 7136 subjects. Participants were asked to complete the questionnaires if they agreed to participate in the study and to send them back to us. Written informed consent for participation in the study was obtained from all subjects. The respondents received an incentive, which was equivalent to ¥1000. Letters asking for response were sent twice to those who did not respond. Those who sent back incomplete answers were contacted and asked to complete the questionnaire. We had 5676 responses (a response rate of 79.5%).

The mean age of respondents was significantly higher than that of non-respondents (52.4 years vs 47.1 years, P < 0.001). Data from questionnaires in which there were no responses to more than two SDS items (questionnaires from 134 subjects) were not used for analysis. The ratio of subjects who did not respond to one item or two items was lowest in the youngest group both in male and female subjects (male, 0.9%; female, 1.7%; age group: 20–39 years). The ratio tended to rise with increasing age (male, 6.2%; female, 7.8%; age group: 70–79 years).

For questionnaires in which there were no responses to one item or two items, scores of the missing items were substituted by mean scores in the same sex and age groups. We determined whether this procedure distorted the mean score by comparing with the mean score in subjects who responded to all 20 items and found no significant difference between the mean scores (data not shown). Eventually, 5547 SDS scores, equivalent to a response rate of 77.7%, were used for analysis. The response rates were different among age groups. The youngest age group had the lowest response rate and older age groups tended to have higher response rates.

Statistical analysis

The subjects were divided into five age groups: 20–39, 40–49, 50–59, 60–69, and 70–79 years. Analysis of variance was performed to compare the SDS scores between sexes and between age groups. The SDS scores were divided into four levels according to the classification by Barrett et al.: no or insignificant symptomatology, 20–39 points; mild depressive state, 40–47 points; moderate depressive state, 48–55 points; and severe depressive state, ≥56 points.14 Factor analysis was conducted on the SDS scores using the Promax rotation algorithm with various criteria for the number of factors, such as Kaiser's criterion (eigenvalues >1.0), a scree test, and the interpretability of resulting factor structures. The cut-off point of the initial eigenvalue was set at 1.0, and the primary criterion for item inclusion was set at least at 0.40 (absolute value). Factor analysis was also performed on the SDS scores for each sex group and each age group. spss for Windows Ver. 11.0. (SPSS, Chicago, IL, USA) was used for all statistical analysis.

Ethics

The present study was approved by the Ethics Committee of Iwate Medical University.

RESULTS

Descriptive findings

The mean age of the male subjects was significantly lower than that of the female subjects (51.9 and 52.8 years, respectively, P = 0.034). The mean SDS score for all subjects was 39.3 points (Table 1). The mean SDS score of male subjects (38.2 points) was significantly lower than that for female subjects (40.3 points; P < 0.001). The mean scores for male subjects were significantly lower than those for female subjects in all age groups (P < 0.01).

Table 1. . Zung self-rating depression scale scores (mean ± SD)
 AllAllMenWomen
20–39 years40–49 years50–59 years60–69 years70–79 yearsPAll20–39 years40–49 years50–59 years60–69 years70–79 yearsP
  • SDS, Zung Self-Rating Depresion Scale.

  • †*

    P < 0.05;

  • †**

    P < 0.01 by t-test for sex.

  • anova for age group.

  • ††χ2 test for age group.

  • *

    P < 0.05;

  • **

    P < 0.01 by Tulcey's method.

Participants (n)55472602579539514568402 2945665515559733473 
Response rate (%)77.775.464.173.577.483.485.4 79.973.580.183.783.180.7 
Age (Mean ± SD)52.4 ± 16.0*51.9 ± 15.9      52.8 ± 16.1      
SDS score (Mean ± SD)39.3 ± 7.8**38.2 ± 7.639.1 ± 7.638.7 ± 7.638.3 ± 7.537.1 ± 7.238.0 ± 7.9< 0.00140.3 ± 7.941.2 ± 8.240.3 ± 7.739.8 ± 7.639.5 ± 7.740.8 ± 8.4< 0.001
High or moderate
SDS score (%)
13.710.613.111.510.77.210.70.025††16.320.615.315.013.018.20.002††

In the male subjects, a reverse-J-shaped relationship was observed between the mean scores and age groups. The highest mean score was in the age group of 20–39 years. The lowest mean score was in the age group of 60–69 years, and the mean score in this age group was significantly lower than those in the age groups of 20–39 years and 40–49 years (P < 0.01). Similar patterns were observed in the female subjects.

The highest mean score was in the age group of 20–39 years, and the lowest mean score was in the age group of 60–69 years, which was significantly lower than that in the age group of 20–39 years (P < 0.01).

High or moderate SDS scores (i.e. severe or moderate depressive states) were found in 13.7% of the subjects. The percentage of male subjects with high or moderate SDS scores (10.6%) was significantly lower than the percentage of female subjects with high or moderate SDS scores (16.3%; P < 0.01). The age group with the lowest percentage of subjects with high or moderate SDS scores was the age group of 60–69 years for both male and female subjects.

SDS scores for each item

High scores were found for ‘decreased libido’ (item 6, 2.87 points), ‘hopelessness’ (item 14, 2.73 points), ‘emptiness’ (item 18, 2.51 points), and ‘dissatisfaction’ (item 20, 2.66 points), and low scores were found for ‘weight loss’ (item 7, 1.33 points), ‘psychomotor retardation’ (item 12, 1.38 points), and ‘suicidal ideation’ (item 19, 1.32 points). Significantly lower scores were found in the male subjects than in the female subjects: for ‘depressed affect’ (item 1) the difference between male and female was 0.13 points; ‘crying spells’ (item 3), 0.26 points; ‘sleep disturbance’ (item 4), 0.14 points; ‘decreased libido’ (item 6), 0.61 points; ‘constipation’ (item 8), 0.39 points; ‘tachycardia’ (item 9), 0.10 points; ‘fatigue’ (item 10), 0.15 points; ‘confusion’ (item 11), 0.08 points; ‘psychomotor retardation’ (item 12), 0.08 points; ‘irritability’ (item 15), 0.11 points; and ‘indecisiveness’ (item 16), 0.10 points (P < 0.01). In contrast, for the scores for ‘emptiness’, the difference between male and female was 0.10 points and that for ‘dissatisfaction’ was 0.11 points; these were significantly higher in the male subjects than in the female subjects (P < 0.01). Scores for all items except ‘constipation’ (item 8) and ‘agitation’ (item 13) were significantly different in the age groups (P < 0.01).

Factor analysis

Factor analysis was performed on the SDS scores for all male subjects, all female subjects, all subjects, and subjects in each age group. Ten to 15 items were extracted by factor analysis for male subjects, female subjects, all subjects, and subjects in each age group. Three factors were given in all subjects, two factors in male subjects, and three factors in female subjects. Table 2 shows numbers of factors and items with loading values of at least 0.40. In the factor analysis for all subjects, 13 of the 20 items were extracted. For the male subjects, the 14 items were extracted. For the female subjects, 13 items were extracted. There was no significant difference between numbers of extracted items in the male and female groups or between numbers of extracted items in the age groups. Similar patterns were observed in each age group for male and female subjects.

Table 2. Factor analysis with Promax rotation of SDS by age group
No. factorsAll 3MenWomen
All
2
20–39
5
40–49
4
50–59
3
60–69
5
70–79
4
All
3
20–39
4
40–49
4
50–59
4
60–69
4
70–79
6
  • SDS, Zung Self-rating Depression Scale.

  • Factors that produced eigenvalues >2.0 were selected for further analysis.

  • ○, Primary criterion for item inclusion was a loading of at least 0.400 (absolute value).

 1. Depressed affect 
 2. Diurnal variation    
 3. Crying spells
 4. Sleep disturbance     
 5. Decreased appetite         
 6. Decreased libido            
 7. Weight loss            
 8. Constipation         
 9. Tachycardia   
10. Fatigue 
11. Confusion   
12. Psychomotor retardation      
13. Agitation
14. Hopelessness    
15. Irritability   
16. Indecisiveness          
17. Personal devaluation 
18. Emptiness 
19. Suicidal ideation       
20. Dissatisfaction

Factor analysis was re-performed on items with loading values of more than 0.40 in all male subjects, all female subjects, and all subjects. In all subjects, factor I consisted of item 1, item 2, item 3, item 4, item 9, item 10, item 11, item 13, and item 15, and factor II consisted of item 14, item 18, and item 20 (Table 3). The eigenvalues of factor I and factor II were 3.9 and 2.6, respectively. In all male subjects, the two-factor solution explained 50.0% of total variance. Results of factor analysis were similar. The eigenvalues of factor I and factor II were 4.0 and 2.7, respectively. The two-factor solution explained 47.9% of total variance. In all female subjects, the two-factor solution explained 50.8% of total variance. Results of factor analysis were similar. The eigenvalues of factor I and factor II were 3.9 and 2.7, respectively.

Table 3. Factor analysis with Promax rotation of the Zung Self-Rating Depression Scale§
FactorAllMenWomen
IIIIIIIII
  • Bold, primary criterion for item inclusion was a loading of no less than 0.4 (absolute value).

  • Excluded items were items 5, 6, 7, 8, 12, 16 and 17 for all subjects; items 5, 6, 7, 8, 16 and 19 for male subjects; and items 5, 6, 7, 8, 12, 14 and 16 for female subjects.

  • Factors that produced eigenvalues >1.0 were selected for further analysis.

  • §

    Loaded value for factor I and II from the factor analysis among all subjects and male/female subjects.

 1. Depressed affect0.70−0.020.69−0.020.680.01
 2. Diurnal variation0.460.110.480.100.480.07
 3. Crying spells0.64−0.020.59−0.010.630.02
 4. Sleep disturbance0.490.020.500.040.470.00
 9. Tachycardia0.49−0.070.47−0.030.49−0.09
10. Fatigue0.59−0.060.58−0.050.60−0.09
11. Confusion0.680.090.700.060.700.06
12. Psychomotor retardation  0.510.07  
13. Agitation0.600.050.630.000.600.07
14. Hopelessness0.080.450.050.48  
15. Irritability0.660.030.67−0.020.650.05
17. Personal devaluation  0.030.500.010.49
18. Emptiness−0.090.91−0.090.94−0.110.97
19. Suicidal ideation0.310.13  0.310.18
20. Dissatisfaction0.040.650.070.580.130.57
Eigenvalues3.92.64.02.73.92.7
Contribution (%)30.020.028.619.330.020.8

DISCUSSION

Background and setting of the study

The number of suicides in Japan has been increasing since 1998, the rate of mortality from suicide now being the highest since World War II. Iwate Prefecture, where our survey areas were located, has had one of the highest rates of suicide in Japan during the past 20 years. Our survey areas, Kuji and Miyako, have a particularly high rate of suicide in Iwate. The SDS was given to residents in the areas with high rates of suicide. The SDS scores were statistically analyzed to examine the distribution in a general population and its factor structure. Then the results of factor analysis were compared with Diagnostic and Statistical Manual of Mental Disorders (4th edn; DSM-IV) criteria for depression in order to consider the validity of the SDS for detecting depression in a general population.

A limitation of the present study was that the surveys were conducted in the region with the highest rates of suicide in Japan, and the findings cannot be generalized to explain the tendency in the whole Japanese population.

Descriptive findings

There have been a few studies in which the distributions of SDS scores in age groups in Japan have been investigated.15–17 In recent studies in Japan and in other countries, however, mean scores of the SDS in all age groups were not compared.18,19 Subjects in recent studies were limited to groups of young, middle-aged or elderly subjects. Studies using subjects of relatively young age have shown that young subjects tend to have high scores and that scores become lower with advancing age.15,16 In contrast, studies using elderly subjects have shown that elderly subjects tend to have high scores.17–19 In the present study we obtained SDS scores for subjects with a wide age range, although the response rates were different among age groups. The difference of the SDS scores between male and female in the youngest age group with the lowest response rate was similar to those in other age groups. In addition, similar patterns were observed in each age group for male and female subjects in the factor analysis. Therefore, it is unlikely that the difference of the response rates among age groups distorted the distribution of the SDS scores, therefore we were able to compare scores among all age groups.

In both male and female groups, reverse-J-shaped relationships were found between mean SDS scores and age groups. The lowest mean score was in the age group of 60–69 years. Therefore, influence of sex and age should be taken into consideration when using the SDS as a screening test for detecting depression. For example, the cut-off point of SDS should be changed depending on the age group and sex.

Summary of factor analysis

In the factor analysis of SDS scores, two factors consisting of 12 SDS items were extracted. Although the SDS is different from the DSM-IV in the administration, in that the former is self-reported and the latter is based on a clinical interview, we compared the 12 items with the DSM-IV criteria for depression in order to examine the usefulness of SDS as a screening test. We referred to Zung's studies on the SDS.10,20 Zung compared SDS items with DSM-III criteria and suggested that 17 SDS items (excluding item 2, 8, and 9) covered the DSM-III criteria. In the present study, 10 out of the extracted 12 SDS items covered six DSM-IV criteria describing psychological disturbances of depression. ‘Diurnal variation’ (item 2) and ‘tachycardia’ (item 9) did not correspond to DSM-IV criteria. In contrast, three DSM-IV criteria (weight loss or weight gain, item 3; feeling of worthlessness or excessive or inappropriate guilt, item 7; and recurrent thought of death and suicide ideation, item 9) were not covered by the SDS items extracted by factor analysis. In other words, somatic items of the SDS tended to be excluded by factor analysis in the present study. This suggests that somatic items may be influenced by some factors other than depression.

Strategy for screening for depression

The results of the present study suggest that the SDS is not suitable for detecting depression in individuals who mainly show somatic symptoms. Such individuals tend to consult family physicians about their somatic disturbances. For prevention of suicide, family physicians are also expected to have skills to detect depression in patients manifesting somatic symptoms. There is a need to develop a strategy for screening for depression among residents using a combination of the SDS for detecting depression in individuals with psychological symptoms, and clinical interviews in a medical setting for detecting depression in individuals with somatic symptoms. For the development of such a strategy, further study is needed to determine whether residents screened out by the SDS suffer from major depression.

CONCLUSIONS

Major components of SDS items in subjects in the present study covered the DSM-IV criteria describing psychological disturbances of depression. Psychological factors should be focused on as part of the strategy to detect major depression among residents.

ACKNOWLEDGMENTS

The present study was supported in part by a Grant-in-Aid for Scientific Research from the Ministry of Health and Welfare and a Grant from Keiryokai Research Foundation No.924069. The authors wish to thank colleagues of the School of Medicine, Iwate Medical University for technical support.

Ancillary