Psychiatryand PsychiatricEpidemiology Social
Soc Psychiatry Psychiatr Epidemiol (1992)27:211-219 9 Springer-Verlag 1992
Changes in the prevalence of symptoms of depression and depression across Greece M. G. Madianos and C. N. Stefanis
Department of PsychiatryMedical School, University of Athens, Greece Accepted: April 14, 1992 Summary. This paper reports on the regional prevalence of symptoms of depression and clinical depression (current major depressive episodes) in Greece in the years 1978 and 1984. Prevalence rates were estimated from two extensive, nationwide cross-sectional home surveys on psychosocial issues and health, carried out in four geographical areas: the Greater Athens area, the Greater Thessaloniki area, the rest of the urban areas and rural areas. The methodology used, the sampling procedure and the screening instrument (The Center for Epidemiologic Studies-Depression Scale) were the same in both surveys. Within the 6-year period a substantial increase in the prevalence of symptoms of depression in all geographic areas was observed, with the Athenian respondents expressing a higher number of symptoms of depression than their counterparts from the other areas. The prevalence of current major depressive episodes, according to specific criteria matched with criteria from the DSM III R, was increased in t984 in Athens and in the rural areas only. We suggest that economic instability between 1978 and 1984 probably contributed to the changes in the rates of depressive disorders.
A series of epidemiologic community surveys have focused on the detection of the incidence and prevalence of depressive disorders during the last two decades (Hagnell et al. 1982; Lehtinen and Vaissanen 1981; Hafner 1985; Gastpar 1986; Murphy 1986). The majority of these surveys report on the prevalence rates of depressed mood in terms of the presence and persistence of dysphoric-dysthymic symptoms recorded on self-report rating scales (Blumenthal and Dielman 1975; Radloff 1977; Weissman and Klerman 1977; Weissman et al. 1977; Wing and Bebbington 1982; Myers and Weissman 1980). The findings of these surveys suggest that a significant proportion of the general population suffer from symptoms of depression. A number of investigators have questioned to what extent the social and cultural environment and socioeconomic factors contribute to psychopathologic processes and especially to depressive disorders in Western societies (Dohrenwend and Dohrenwend 1974; Craig and Van
Natta 1979; Bell et al. 1981; Dooley et al. 1981; Catalano and Dooley 1977; Sartorius 1986; Murphy 1986). During recent decades Greece has undergone profound socioeconomic and cultural changes through industrialization, urbanization and migration (Tsaoussis 1976). Traditional roles and family ties have been weakening and the extended family has been gradually transformed into the nuclear type (Madianos and Madianou, in press). Greek society, suffering serious economic changes reflected in a number of indicators, e, g. inflation and unemployment, has become more diversified and complex giving rise to psychosocial-stress-related disorders mostly prevalent in urban areas (Safilios-Rothschild 1972; Madianos 1983; Madianos et al. 1985; Mavreas et al. t986). Based on the assumption that socioeconomic change over recent years may have influenced the manifestation of depressive disorders in the Greek population, this paper attempts a comparison between the prevalence rates of symptoms of depression and the clinical depression rates, as estimated by two extensive, nationwide, cross-sectional home surveys, carried out in 1978 and 1984. The 1978 survey was the first Greek sociopsychiatric study using population survey methods. The second (1984) nationwide studywas a replication of the first one. Both surveys focussed on certain psychosocial issues and mental health and were conducted with the same methodology in large probability samples (Madianos and Zarnari 1983; Madianos et al. 1987). Due to the structure and the demands of the first nationwide survey and the lack of appropriate instrument ation at that time in Greece, a clinical diagnostic interview was excluded from the methodology. Symptoms of depression were detected by the use of the CES-D scale (Radloff 1977) supplemented by items on suicidal behaviour. There are many reports on the clinical significance of the use of the Center Epidemiologic Studies-Depression Scale (CES-D) in community studies, and on the definition of caseness (specific affective disorders), and also on where to place the threshold level that determines which symptoms are synonymous with the clinical criteria of the illness (Weissman et al. 1977; Weissman and Myers 1978; Boyd et al. 1982; Wing and Bebbington 1982). It is reported that the CES-D scale is more sensitive to the current diagnosis of
212 depressive illness and that there is a modest relationship between the self-reported symptoms of depression and the diagnosis of depression (Roberts and Vernon 1983; Boyd et al. 1982). The detection of major depressive episodes in the general population is made possible by the application of a specific methodology. The purpose of this paper was to answer the following questions: 9 Were there any differences in the prevalence rates of symptoms of depression between 1979 and 1984? 9 Were there any differences in the regional distribution of the prevalence rates of depressive symptomatology? 9 What were the predictors of depressed mood? 9 Were there any differences in the prevalence of major depressive episodes during the 6-year period (19781984) with respect to sex and regionalization? Method
interviewers, the majority of whom participated in both surveys. The duration of the interview ranged from 60 to 90 min. The overall response rate for the total sample in the 1978 survey was 92% and in the 1984 survey, 96.5 %. Thirty percent of the interviews were checked, mainly by a second visit or telephone contact.
Questionnaires The interviews in both surveys were conducted using a structured questionnaire, aimed at obtaining data on physical and mental health, suicidal behaviour and related psychosocial issues (i. e. help-seeking patterns, drug and alcohol use, and family medical history). Sociodemographic data were also obtained. The mental health status of the respondent was assessed by the use of the CES-D and the Langner scales (Radloff 1977; Langner 1962), standardized for the Greek population (Madianos et al. 1983; Madianos 1984).
Design Both surveys were cross-sectional. The first (study A) was carried out by the National Center for Social Research over a 6-month period in 1978 and the second (study B) by the Department of Psychiatry of Athens University over a 4-month period in 1984 (Madianos et al. 1987). Both samples were four stage systematic; the first comprised 4083 persons aged 18-64 years, and the second comprised 4292 respondents aged 12-17 years and 18-64 years. For purposes of comparability the group aged 12-17 years in the second survey was excluded, so that both survey samples included adults only. Both samples were drawn randomly from four geographic areas including Greater Athens, Greater Thessaloniki, rest of urban areas (15 cities) and semi-urban-rural areas (40 communities). The total population covered in both surveys was 9130000. The samples represented the whole of mainland Greece with the exception of the Aegean and Ionian islands (4.5 % of the total population), which were not covered by the sample for technical reasons. The sampling procedure was four stage systematic, including at the first and second stages, the random selection of the town location and the household block and at the third and fourth stages, the random selection of a household from the listings of the existing households in the block and finally, the selection of the individual respondent to be interviewed at random using the Kish selection grid (Kish 1965). The sociodemographic characteristics of both survey sample respondents are presented in Table 1. The data presented include the percentages of the actual numbers of persons surveyed by sex, age groups, years of schooling, socioeconomic status, marital status, employment, and place of residence. The slightly higher number of females compared to males in both sample corresponds to the general population ratios. Regarding the other variables, no differences were observed except for the unemployment rates. In the 1978 sample only 1.2% were unemployed, while the proportion of unemployed respondents in 1984 reached 4.6 %. A pilot study preceded the main surveys in order to test the validity of the research instruments. The respondents in both surveys were interviewed in their homes by trained
Measures of depressive disorders The prevalence of depressive symptomatology was assessed by the use of the CES-D scale developed by the Center for Epidemiological Studies, National Institute of
Table 1. Sociodemographiccharacteristics of respondents: studies
A and B nationwidesamples Sex Males Females Age (years) 19-29 3044 45454 Years of schooling 0-6 7-11 12 13 + SES I II III IV V Marital status Single Married Widowed Divorced Employmentstatus Unemployed Geography Greater Athens Thessaloniki Rest of urban areas Semi-uban-rural SES Socioeconomicstatus
1978 n:4083 [%] 44.6 55.4
1984 n :3706 [%] 44.8 55.2
23.0 33.7 43.2
25.0 35.0 40.0
57.4 10.8 20.5 8.2
51.7 16.5 19.2 12.6
9.3 26.7 45.1 15.6 7.3
6.3 25.2 44.6 15.0 8.9
16.3 77.4 4.4 1.9
20.8 74.5 2.8 1.9
1.2
4.6
34.5 9.1 18.8 37.7
35.2 7.8 20.3 36.7
213 Mental Health (USA). The scale consists of 20 items selected from previously developed scales (Comstock and Helsing 1976; Radloff 1977). These items represent the major symptoms in the clinical syndrome of depression. In our surveys we modified the scale for personal interview and we employed a 4-week-long criterion of symptom duration. The four possible responses to each scale item ranged from "rarely or none of the time" to "most or all of the time". Four items were scored in reverse. The range of possible scores was from 0 to 60. According to Weissman et al. (197"7) the scale has been proven to be valid for screening symptoms of depression in psychiatric populations. However, Boyd et al. (1982) have reported that there are discrepancies between the symptoms of depression and the diagnosis of major depression. The reason for these discrepancies originates in the structure of the CES-D scale items, some of which do not form a criterion symptom of major depression; other items are missing (e. g. questions on suicidal behaviour). A considerable number of the CES-D items, however, correspond to specific criteria in the Research Diagnostic Criteria of major depressive episodes (Boyd and Weissman 1983). The CES-D scale has been validated in clinical groups in terms of its internal consislLency and concurent validity (Madianos et al. 1983). A cut-off score of 16 and over provided the best estimates for sensitivity (81%) and specificity (78 %) when the scale was tested against the DSM III R criteria for current major depressive episodes (APA 1987). Based upon the limitation of our studies, in which a diagnostic interview classification was not used, we estimated the prevalence of the DSM III R current major depressive episodes by the selection of 15 items from the CES-D scale directly corresponding to the D S M I I I R criterion A symptoms, namely (1) depressed mood, (2) loss of interest or pleasure, (3) loss of weight/poor appetite, (4) insomnia, (5) psychomotor agitation or retardation, (6) low energy, (7) low selfesteem, and (8) poor concentration. Three other items were added regarding the 9th symptom of criterion A "suicidal ideation or suicide attempts". The duration criterion was met by asking the respondent if these symptoms were present for at least a 2-week period during the last month. Furthermore, we counted only those items that were rated "most or all of the time". Any respondent who reported the existence of a serious physical illness was excluded (criterion B for Current Major Depression Syndrome). It was not possible to meet the other two criteria (symptoms of delusions and hallucinations or history suggestive of schizophrenia) due to the limitations of the structure of the interview. Table 2 shows the relationship between the DSM III R criteria for current major depressive episodes and the criteria used in our studies. An interater reliability study on the agreement for current mai or depressive episode identification in a sub sample of respondents by two independent raters was carried out. The degree of agreement between the two raters, measured by the estimation of the Kappa (K) coefficient (Cohen 1960; Fleiss t 973; Shrout et al. 1987), was found to be high in both surveys: 0.78 and 0.81 respectively. The significance of K was determined by the estimation ofz and in both surveys was found to be highly significant at P < 0.0001.
Statistical methods
The prevalence of symptoms of depression in both surveys is presented in the form of cross tabulations, the CESD scale providing categorical scores. The statistical significance of the differences between 1978 and the 1984 prevalence rates was assessed by means of statistic z for the proportions with a specified characteristic in two independent samples (Fleiss, 1973). We also examined whether significant differences existed between the CES-D scale mean scores of the two studies by the application of Student's t-test. Finally, a stepwise multiple regression analysis was applied for the detection of the possible effects of all independent variables on the CES-D scale scores in both surveys (Cohen and Cohen 1975). Only main effects between the independent variables were included in the analysis (i. e. an additive model was used). The effect was measured by each regression coefficient and was, therefore, independent of the effects of the other variables. Finally, only those variables corresponding to a two-tailed significance level of 0.05 were included. The statistical analysis was performed by the use of the Statistical Package for Social Sciences, x version (SPSS-X 1983). Results
The estimated 1-month prevalence of persons expressing high numbers of symptoms of depression among the two nationwide samples by sex is shown in Table 3. Statistically significant differences were observed between the 1978 and 1984 studies in the prevalence rates of "cases" of depression in both male and female respondents by using as our criterion a score of greater than 16 on the CES-D scale. Table 4 presents the regional variation of the estimated 1-month prevalence of persons with a CES-D score of 16 or higher among the two nationwide samples. In all geographic areas there were statistically significant differences between the two studies in the 1-month prevalence rates of depressed respondents. In the second study, conducted 6 years later, a significant increase in the regional distribution of the prevalence rates of depressed persons was observed. Another finding was that the Athenian respondents in both surveys expressed higher numbers of symptoms of depression. The average CES-D scale scores of male and female respondents in both studies are presented in Table 5. Here, also, significant differences were observed in the number of reported symptoms of depression (average CES-D scores) between the 1978 and 1984 studies. The average CES-D scores of respondents drawn from the four geographic sample areas were found to be significantly higher in the 1984 study than in the 1978 study (Table 6). Table 7 presents an outline of the results of the multiple regression analysis with the CES-D scale scores as dependent variables, performed in both surveys. In the 1978 study the effect of the independent variables selected stepwise was found to be highly significant and explains 28 % of the total variance of the dependent variable. However the proportion of variance of the CES-D scale score explained by 11 independent variables was found to be slightly higher (33 %) in the second study.
214 Table 2. The DSM III R criteria for current major depressive syndrome compared with the criteria used in the present studies D S M III R criteria
CES-D Items
Duration: A t least 2 weeks in the last month Frequency
1. Depressed mood most of the daya
1.1 1 could not shake the blues 1.21 felt depressed 1.3 1 felt sad
"Most of the time" "Most of the time" "Most of the time"
2. Loss of interest or pleasure a
2.1 1 was happy u 2.21 felt hopeful about the future u 2.31 enjoyed life b
"Most of the time" "Most of the time" "Most of the time"
3. Loss of weight, poor apetite
3.1 My appetite was poor c
"Most of the time"
4. Insomnia
4.1 Restless sleep
"Most of the time"
5. Psychomotor agitation or retardation
5.1 1 felt fearful 5.21 felt that everything I did was an effort 5.31 talked less than usual
"Most of the time" "Most of the time" "Most of the time"
6. Low energy-fatigue
6.1 1 could not "get going"
"Most of the time"
7. Low self-esteem
7.11 felt that I was just as good as other people u 7.2 My life had been a failure
"Most of the time" "Most of the time"
8. Poor concentration
8.1 1 had trouble keeping my mind on what I was doing New items added
"Most of the time"
9. Suicidal ideation or suicide attempt
9.1 1 thought of taking my life 9.21 have tried to kill myself 9.31 really wanted to die
"Most of the time" Yes Yes
B No organic etiology
B Exclusion of any respondent reporting a somatic disease
A A t least five of the following:
C Absence of delusions or hallucinations D No symptoms or history suggestive of schizophrenia A t least one of the two symptoms b Reversed score c T h e r e was an additional question on loss of weight > 10 kg
Table 3. Estimated 1-month prevalence of symptoms of depression" in two nationwide samples by sex: Study A n = 4083, study B n = 3706 Sex
Study A
Study B
n
%
Males Females
175 527
9.6 23.3
255 770
Total
702
17.0
1024
% Change
z
Significance (P)
15.4 37.6
+ 5.8 + 14.3
3.60 10.20
16 score on the CES-D scale
Table 4. Regional distribution of the estimated 1-month prevalence of symptoms of depression" among the two nation-wide samples: Study A n = 4083, Study B n - 3706 Areas
Study A n
Study B
% Change
z
Significance ( P )
%
n
298
21.1
403
30.8
+ 9.7
5.74
< 0.0001
56
15.1
81
27.9
+ 12.8
8.50
< 0.0001
3. Rest of urban areas
119
15.5
174
23.2
+ 7.7
8.60
< 0.0001
4. Semi-urban-rural
259
16.8
348
25.6
+ 8.8
5.73
< 0.0001
1. Greater Athens 2. Thessaloniki
Base f o r %
1.1411 1.1307 2.366 2. 290 3.766 3. 749 4.1540 4.1360 a A > 1 6 s c o r e o n t h e CES-Dscale
%
215 Table 5. Average CES-D scores of male and female respondents: study A n = 4083, study B n = 3706
Study A
Study B
Sex
n
2
S.E.
n
~
S.E.
t
df
P
Males Females
1820 2263
5.94 10.12
0.18 0.22
1660 2046
10.35 15.02
0.30 0.34
23.96 12.07
3479 4308
< 0.005 < 0.005
Total
4083
8.30
0.14
3706
13.24
0.16
22.46
7787
< 0.005
Table 6. Average CES-D scores across the country: study A n = 4083, study B n = 3706
Study A Areas 1. Greater Athens
~ 9.05
S.E. 0.56
n 1307
2 13.55
S.E. 0.42
r 9.37
df 2716
P < 0.005
366
6.46
0.43
290
11.34
0.52
5.30
3372
< 0.005
766 1540
6.84 7.63
0.12 0.19
749 1360
11.59 12.01
0.39 0.40
9.70 9.75
1513 2899
Soc Psychiatry Psychiatr Epidemiol (1992)27:211-219 9 Springer-Verlag 1992
Changes in the prevalence of symptoms of depression and depression across Greece M. G. Madianos and C. N. Stefanis
Department of PsychiatryMedical School, University of Athens, Greece Accepted: April 14, 1992 Summary. This paper reports on the regional prevalence of symptoms of depression and clinical depression (current major depressive episodes) in Greece in the years 1978 and 1984. Prevalence rates were estimated from two extensive, nationwide cross-sectional home surveys on psychosocial issues and health, carried out in four geographical areas: the Greater Athens area, the Greater Thessaloniki area, the rest of the urban areas and rural areas. The methodology used, the sampling procedure and the screening instrument (The Center for Epidemiologic Studies-Depression Scale) were the same in both surveys. Within the 6-year period a substantial increase in the prevalence of symptoms of depression in all geographic areas was observed, with the Athenian respondents expressing a higher number of symptoms of depression than their counterparts from the other areas. The prevalence of current major depressive episodes, according to specific criteria matched with criteria from the DSM III R, was increased in t984 in Athens and in the rural areas only. We suggest that economic instability between 1978 and 1984 probably contributed to the changes in the rates of depressive disorders.
A series of epidemiologic community surveys have focused on the detection of the incidence and prevalence of depressive disorders during the last two decades (Hagnell et al. 1982; Lehtinen and Vaissanen 1981; Hafner 1985; Gastpar 1986; Murphy 1986). The majority of these surveys report on the prevalence rates of depressed mood in terms of the presence and persistence of dysphoric-dysthymic symptoms recorded on self-report rating scales (Blumenthal and Dielman 1975; Radloff 1977; Weissman and Klerman 1977; Weissman et al. 1977; Wing and Bebbington 1982; Myers and Weissman 1980). The findings of these surveys suggest that a significant proportion of the general population suffer from symptoms of depression. A number of investigators have questioned to what extent the social and cultural environment and socioeconomic factors contribute to psychopathologic processes and especially to depressive disorders in Western societies (Dohrenwend and Dohrenwend 1974; Craig and Van
Natta 1979; Bell et al. 1981; Dooley et al. 1981; Catalano and Dooley 1977; Sartorius 1986; Murphy 1986). During recent decades Greece has undergone profound socioeconomic and cultural changes through industrialization, urbanization and migration (Tsaoussis 1976). Traditional roles and family ties have been weakening and the extended family has been gradually transformed into the nuclear type (Madianos and Madianou, in press). Greek society, suffering serious economic changes reflected in a number of indicators, e, g. inflation and unemployment, has become more diversified and complex giving rise to psychosocial-stress-related disorders mostly prevalent in urban areas (Safilios-Rothschild 1972; Madianos 1983; Madianos et al. 1985; Mavreas et al. t986). Based on the assumption that socioeconomic change over recent years may have influenced the manifestation of depressive disorders in the Greek population, this paper attempts a comparison between the prevalence rates of symptoms of depression and the clinical depression rates, as estimated by two extensive, nationwide, cross-sectional home surveys, carried out in 1978 and 1984. The 1978 survey was the first Greek sociopsychiatric study using population survey methods. The second (1984) nationwide studywas a replication of the first one. Both surveys focussed on certain psychosocial issues and mental health and were conducted with the same methodology in large probability samples (Madianos and Zarnari 1983; Madianos et al. 1987). Due to the structure and the demands of the first nationwide survey and the lack of appropriate instrument ation at that time in Greece, a clinical diagnostic interview was excluded from the methodology. Symptoms of depression were detected by the use of the CES-D scale (Radloff 1977) supplemented by items on suicidal behaviour. There are many reports on the clinical significance of the use of the Center Epidemiologic Studies-Depression Scale (CES-D) in community studies, and on the definition of caseness (specific affective disorders), and also on where to place the threshold level that determines which symptoms are synonymous with the clinical criteria of the illness (Weissman et al. 1977; Weissman and Myers 1978; Boyd et al. 1982; Wing and Bebbington 1982). It is reported that the CES-D scale is more sensitive to the current diagnosis of
212 depressive illness and that there is a modest relationship between the self-reported symptoms of depression and the diagnosis of depression (Roberts and Vernon 1983; Boyd et al. 1982). The detection of major depressive episodes in the general population is made possible by the application of a specific methodology. The purpose of this paper was to answer the following questions: 9 Were there any differences in the prevalence rates of symptoms of depression between 1979 and 1984? 9 Were there any differences in the regional distribution of the prevalence rates of depressive symptomatology? 9 What were the predictors of depressed mood? 9 Were there any differences in the prevalence of major depressive episodes during the 6-year period (19781984) with respect to sex and regionalization? Method
interviewers, the majority of whom participated in both surveys. The duration of the interview ranged from 60 to 90 min. The overall response rate for the total sample in the 1978 survey was 92% and in the 1984 survey, 96.5 %. Thirty percent of the interviews were checked, mainly by a second visit or telephone contact.
Questionnaires The interviews in both surveys were conducted using a structured questionnaire, aimed at obtaining data on physical and mental health, suicidal behaviour and related psychosocial issues (i. e. help-seeking patterns, drug and alcohol use, and family medical history). Sociodemographic data were also obtained. The mental health status of the respondent was assessed by the use of the CES-D and the Langner scales (Radloff 1977; Langner 1962), standardized for the Greek population (Madianos et al. 1983; Madianos 1984).
Design Both surveys were cross-sectional. The first (study A) was carried out by the National Center for Social Research over a 6-month period in 1978 and the second (study B) by the Department of Psychiatry of Athens University over a 4-month period in 1984 (Madianos et al. 1987). Both samples were four stage systematic; the first comprised 4083 persons aged 18-64 years, and the second comprised 4292 respondents aged 12-17 years and 18-64 years. For purposes of comparability the group aged 12-17 years in the second survey was excluded, so that both survey samples included adults only. Both samples were drawn randomly from four geographic areas including Greater Athens, Greater Thessaloniki, rest of urban areas (15 cities) and semi-urban-rural areas (40 communities). The total population covered in both surveys was 9130000. The samples represented the whole of mainland Greece with the exception of the Aegean and Ionian islands (4.5 % of the total population), which were not covered by the sample for technical reasons. The sampling procedure was four stage systematic, including at the first and second stages, the random selection of the town location and the household block and at the third and fourth stages, the random selection of a household from the listings of the existing households in the block and finally, the selection of the individual respondent to be interviewed at random using the Kish selection grid (Kish 1965). The sociodemographic characteristics of both survey sample respondents are presented in Table 1. The data presented include the percentages of the actual numbers of persons surveyed by sex, age groups, years of schooling, socioeconomic status, marital status, employment, and place of residence. The slightly higher number of females compared to males in both sample corresponds to the general population ratios. Regarding the other variables, no differences were observed except for the unemployment rates. In the 1978 sample only 1.2% were unemployed, while the proportion of unemployed respondents in 1984 reached 4.6 %. A pilot study preceded the main surveys in order to test the validity of the research instruments. The respondents in both surveys were interviewed in their homes by trained
Measures of depressive disorders The prevalence of depressive symptomatology was assessed by the use of the CES-D scale developed by the Center for Epidemiological Studies, National Institute of
Table 1. Sociodemographiccharacteristics of respondents: studies
A and B nationwidesamples Sex Males Females Age (years) 19-29 3044 45454 Years of schooling 0-6 7-11 12 13 + SES I II III IV V Marital status Single Married Widowed Divorced Employmentstatus Unemployed Geography Greater Athens Thessaloniki Rest of urban areas Semi-uban-rural SES Socioeconomicstatus
1978 n:4083 [%] 44.6 55.4
1984 n :3706 [%] 44.8 55.2
23.0 33.7 43.2
25.0 35.0 40.0
57.4 10.8 20.5 8.2
51.7 16.5 19.2 12.6
9.3 26.7 45.1 15.6 7.3
6.3 25.2 44.6 15.0 8.9
16.3 77.4 4.4 1.9
20.8 74.5 2.8 1.9
1.2
4.6
34.5 9.1 18.8 37.7
35.2 7.8 20.3 36.7
213 Mental Health (USA). The scale consists of 20 items selected from previously developed scales (Comstock and Helsing 1976; Radloff 1977). These items represent the major symptoms in the clinical syndrome of depression. In our surveys we modified the scale for personal interview and we employed a 4-week-long criterion of symptom duration. The four possible responses to each scale item ranged from "rarely or none of the time" to "most or all of the time". Four items were scored in reverse. The range of possible scores was from 0 to 60. According to Weissman et al. (197"7) the scale has been proven to be valid for screening symptoms of depression in psychiatric populations. However, Boyd et al. (1982) have reported that there are discrepancies between the symptoms of depression and the diagnosis of major depression. The reason for these discrepancies originates in the structure of the CES-D scale items, some of which do not form a criterion symptom of major depression; other items are missing (e. g. questions on suicidal behaviour). A considerable number of the CES-D items, however, correspond to specific criteria in the Research Diagnostic Criteria of major depressive episodes (Boyd and Weissman 1983). The CES-D scale has been validated in clinical groups in terms of its internal consislLency and concurent validity (Madianos et al. 1983). A cut-off score of 16 and over provided the best estimates for sensitivity (81%) and specificity (78 %) when the scale was tested against the DSM III R criteria for current major depressive episodes (APA 1987). Based upon the limitation of our studies, in which a diagnostic interview classification was not used, we estimated the prevalence of the DSM III R current major depressive episodes by the selection of 15 items from the CES-D scale directly corresponding to the D S M I I I R criterion A symptoms, namely (1) depressed mood, (2) loss of interest or pleasure, (3) loss of weight/poor appetite, (4) insomnia, (5) psychomotor agitation or retardation, (6) low energy, (7) low selfesteem, and (8) poor concentration. Three other items were added regarding the 9th symptom of criterion A "suicidal ideation or suicide attempts". The duration criterion was met by asking the respondent if these symptoms were present for at least a 2-week period during the last month. Furthermore, we counted only those items that were rated "most or all of the time". Any respondent who reported the existence of a serious physical illness was excluded (criterion B for Current Major Depression Syndrome). It was not possible to meet the other two criteria (symptoms of delusions and hallucinations or history suggestive of schizophrenia) due to the limitations of the structure of the interview. Table 2 shows the relationship between the DSM III R criteria for current major depressive episodes and the criteria used in our studies. An interater reliability study on the agreement for current mai or depressive episode identification in a sub sample of respondents by two independent raters was carried out. The degree of agreement between the two raters, measured by the estimation of the Kappa (K) coefficient (Cohen 1960; Fleiss t 973; Shrout et al. 1987), was found to be high in both surveys: 0.78 and 0.81 respectively. The significance of K was determined by the estimation ofz and in both surveys was found to be highly significant at P < 0.0001.
Statistical methods
The prevalence of symptoms of depression in both surveys is presented in the form of cross tabulations, the CESD scale providing categorical scores. The statistical significance of the differences between 1978 and the 1984 prevalence rates was assessed by means of statistic z for the proportions with a specified characteristic in two independent samples (Fleiss, 1973). We also examined whether significant differences existed between the CES-D scale mean scores of the two studies by the application of Student's t-test. Finally, a stepwise multiple regression analysis was applied for the detection of the possible effects of all independent variables on the CES-D scale scores in both surveys (Cohen and Cohen 1975). Only main effects between the independent variables were included in the analysis (i. e. an additive model was used). The effect was measured by each regression coefficient and was, therefore, independent of the effects of the other variables. Finally, only those variables corresponding to a two-tailed significance level of 0.05 were included. The statistical analysis was performed by the use of the Statistical Package for Social Sciences, x version (SPSS-X 1983). Results
The estimated 1-month prevalence of persons expressing high numbers of symptoms of depression among the two nationwide samples by sex is shown in Table 3. Statistically significant differences were observed between the 1978 and 1984 studies in the prevalence rates of "cases" of depression in both male and female respondents by using as our criterion a score of greater than 16 on the CES-D scale. Table 4 presents the regional variation of the estimated 1-month prevalence of persons with a CES-D score of 16 or higher among the two nationwide samples. In all geographic areas there were statistically significant differences between the two studies in the 1-month prevalence rates of depressed respondents. In the second study, conducted 6 years later, a significant increase in the regional distribution of the prevalence rates of depressed persons was observed. Another finding was that the Athenian respondents in both surveys expressed higher numbers of symptoms of depression. The average CES-D scale scores of male and female respondents in both studies are presented in Table 5. Here, also, significant differences were observed in the number of reported symptoms of depression (average CES-D scores) between the 1978 and 1984 studies. The average CES-D scores of respondents drawn from the four geographic sample areas were found to be significantly higher in the 1984 study than in the 1978 study (Table 6). Table 7 presents an outline of the results of the multiple regression analysis with the CES-D scale scores as dependent variables, performed in both surveys. In the 1978 study the effect of the independent variables selected stepwise was found to be highly significant and explains 28 % of the total variance of the dependent variable. However the proportion of variance of the CES-D scale score explained by 11 independent variables was found to be slightly higher (33 %) in the second study.
214 Table 2. The DSM III R criteria for current major depressive syndrome compared with the criteria used in the present studies D S M III R criteria
CES-D Items
Duration: A t least 2 weeks in the last month Frequency
1. Depressed mood most of the daya
1.1 1 could not shake the blues 1.21 felt depressed 1.3 1 felt sad
"Most of the time" "Most of the time" "Most of the time"
2. Loss of interest or pleasure a
2.1 1 was happy u 2.21 felt hopeful about the future u 2.31 enjoyed life b
"Most of the time" "Most of the time" "Most of the time"
3. Loss of weight, poor apetite
3.1 My appetite was poor c
"Most of the time"
4. Insomnia
4.1 Restless sleep
"Most of the time"
5. Psychomotor agitation or retardation
5.1 1 felt fearful 5.21 felt that everything I did was an effort 5.31 talked less than usual
"Most of the time" "Most of the time" "Most of the time"
6. Low energy-fatigue
6.1 1 could not "get going"
"Most of the time"
7. Low self-esteem
7.11 felt that I was just as good as other people u 7.2 My life had been a failure
"Most of the time" "Most of the time"
8. Poor concentration
8.1 1 had trouble keeping my mind on what I was doing New items added
"Most of the time"
9. Suicidal ideation or suicide attempt
9.1 1 thought of taking my life 9.21 have tried to kill myself 9.31 really wanted to die
"Most of the time" Yes Yes
B No organic etiology
B Exclusion of any respondent reporting a somatic disease
A A t least five of the following:
C Absence of delusions or hallucinations D No symptoms or history suggestive of schizophrenia A t least one of the two symptoms b Reversed score c T h e r e was an additional question on loss of weight > 10 kg
Table 3. Estimated 1-month prevalence of symptoms of depression" in two nationwide samples by sex: Study A n = 4083, study B n = 3706 Sex
Study A
Study B
n
%
Males Females
175 527
9.6 23.3
255 770
Total
702
17.0
1024
% Change
z
Significance (P)
15.4 37.6
+ 5.8 + 14.3
3.60 10.20
16 score on the CES-D scale
Table 4. Regional distribution of the estimated 1-month prevalence of symptoms of depression" among the two nation-wide samples: Study A n = 4083, Study B n - 3706 Areas
Study A n
Study B
% Change
z
Significance ( P )
%
n
298
21.1
403
30.8
+ 9.7
5.74
< 0.0001
56
15.1
81
27.9
+ 12.8
8.50
< 0.0001
3. Rest of urban areas
119
15.5
174
23.2
+ 7.7
8.60
< 0.0001
4. Semi-urban-rural
259
16.8
348
25.6
+ 8.8
5.73
< 0.0001
1. Greater Athens 2. Thessaloniki
Base f o r %
1.1411 1.1307 2.366 2. 290 3.766 3. 749 4.1540 4.1360 a A > 1 6 s c o r e o n t h e CES-Dscale
%
215 Table 5. Average CES-D scores of male and female respondents: study A n = 4083, study B n = 3706
Study A
Study B
Sex
n
2
S.E.
n
~
S.E.
t
df
P
Males Females
1820 2263
5.94 10.12
0.18 0.22
1660 2046
10.35 15.02
0.30 0.34
23.96 12.07
3479 4308
< 0.005 < 0.005
Total
4083
8.30
0.14
3706
13.24
0.16
22.46
7787
< 0.005
Table 6. Average CES-D scores across the country: study A n = 4083, study B n = 3706
Study A Areas 1. Greater Athens
~ 9.05
S.E. 0.56
n 1307
2 13.55
S.E. 0.42
r 9.37
df 2716
P < 0.005
366
6.46
0.43
290
11.34
0.52
5.30
3372
< 0.005
766 1540
6.84 7.63
0.12 0.19
749 1360
11.59 12.01
0.39 0.40
9.70 9.75
1513 2899
