Sage Journals: Discover world-class research

Abstract

Objective: There is conflicting evidence as to whether depression prevalence is increasing or is stable. Although birth cohort analysis studies show increasing prevalence, longitudinal studies do not. To date there are no published Australian studies providing long-term estimates of depression prevalence. The aim of the present study was to examine the increasing depression prevalence hypothesis in an Australian context.

Method: Data from the 1998 and 2004 South Australian Health Omnibus Surveys were re-analysed. In each survey there were >3000 participants who were interviewed. The data were weighted to achieve representativeness prior to analysis. Depression status was classified by the Primary Care Evaluation of Mental Disorders scale (the PRIME – MD) into major, other and no depressions. Both univariate and multivariable analyses were used to examine depression trends over time and to adjust the data for significant background variables.

Results: There was no significant increase in the cases of patients classified with major depression between 1998 and 2004; there was, however, a significant decrease in other depression. The overall unadjusted prevalence of major depression was 7.4% (6.8% in 1998 and 8.0% in 2004) and for other depression it was 9.5% (10.6% in 1998 and 8.4% in 2004). The highest levels of depression, both major and other, were observed among females aged 15–29 years, and the lowest levels among those aged ≥50 years. The strongest predictor of depression was poor overall health status.

Conclusion: It is time for the conventional wisdom that depression is increasing to be reconsidered, and for a more realistic perspective, developed from evidence-based analyses, to be accepted.

Keywords

depression prevalence PRIME-MD

There is a widespread belief that the prevalence of depression is increasing. The sources of this belief can be found in the literature reporting increases in depression prevalence by birth cohort [1–6]; increasing numbers of patients being diagnosed with depression or presenting for primary health care treatment [7–9]; high prevalence rates among older adults [10], [11]; burden of disease studies showing that depression is one of the leading causes if not the leading cause of disability [4], [9], [12–14]; and studies showing increased and increasing excess costs associated with depression [15], [16], including disability, self-harm, ischaemic heart disease and depression-related suicide [17], [18]. Based on extrapolating 1990 prevalence and treatment outcome data, it is predicted that by the year 2020 major depression will be the second cause (behind ischaemic heart disease) of the global disease burden [9], partly because of changing incidence and prevalence rates and expected better treatment outcomes for other conditions [13], [19]. Collectively, these studies from many different perspectives have contributed to the belief that depression is on the increase.

The hypothesis of increasing depression prevalence primarily rests on birth cohort analyses of cross-sectional data. These analyses show that those born between World War II and about 1965, have higher probabilities of suffering depression symptoms and that the first occurrence of these symptoms is at younger ages when compared with older birth cohorts [1–7]. These findings, however, have not been fully supported by other recent research. Mattisson et al. showed that among re-interviewed participants in the Swedish Lundby study, the annual age-standardized incidence rate of depression for male (female) was 3.3 (5.5)/1000 in the period 1947–1972 compared with 2.8 (5.1)/1000 in 1972–1999 [20]. Based on three cross-sectional samples over a 40 year timeframe, the Canadian Stirling County study reported that current depression prevalence remained constant at approximately 5% and that minor variations over time were probably a function of changes in the language used to describe depression [21].

Importantly, and again contrary to popular belief, there is little evidence to support the stereotyping of old age as a time of depression. In fact, the literature shows the opposite: that it is younger adults with the highest prevalence rates, particularly females [7], [21–32]. This extensive literature makes it important to examine depression prevalences in detail in order to enhance its understanding.

At present there is no Australian study reporting on changes in depression prevalence over time using any of the study designs reported above the analysis of cross-sectional data, longitudinal re-interviewing or repeated cross-sectional surveys. The present study reports on the prevalence of depression over a 6-year period using a repeated cross-sectional survey design that overcomes some of the limitations of both cross-sectional birth cohort analyses and longitudinal re-interviewing.

Methods

Participants

The paper draws upon the 1998 and 2004 South Australian Health Omnibus Surveys (SAHOS). A full description of the SAHOS methodology can be found in the Wilson et al. study [33]. Briefly, the SAHOS is an annual user-pays population-based survey for health organizations. The surveys analysed in this study consisted of metropolitan and rural samples. The metropolitan samples were based on the collectors’ districts used by the Australian Bureau of Statistics in the 1996 and 2001 censuses, respectively. Sampling in both surveys was based on probability of selection proportional to size, based on every fourth household. The rural samples were selected from towns with a population of 1000 or more in the 1996 or 2001 censuses, using the same procedure as for the metropolitan area. This sample frame comprised approximately 80% of the South Australian rural population.

One interview was conducted per household and, when more than one person over the age of 15 resided in that household, the respondent was the person who last had a birthday. Interviews were conducted by trained and experienced interviewers. For reliability purposes re-interviews for selected questions were conducted on a random 5% (1998) and 10% (2004) of each interviewer's interviews.

For the 1998 SAHOS, of the 4400 subjects initially sampled, 4289 were eligible. Of these, 408 were non-contactable, 658 refused, and 213 were absent, too ill or did not speak English; 3010 interviews were conducted; a 70% response rate. For the 2004 SAHOS, of 4700 initial cases, 532 were non-contac table or inaccessible, 202 were absent, did not speak English or were too ill, and 945 refused. The number of participants was 3015, a within-scope response of 72%. The South Australian Health Omnibus Advisory Committee independently vetted all questions in the survey and gave its ethical approval. Informed consent consistent with the Australian National Health and Medical Research Council requirements was obtained.

Measures

Demographics

Data were collected for gender, age, birth country, relationship status, education attainment, workforce participation, and overall health status.

Due to sparse data (i.e. where the numbers in cells would violate the assumptions for stable analyses), demographic variables were recoded for use in the multivariate analysis. The classifications were: age (15–30/31–45/46–60/ ≥ 61 years), birth country (Australia/Other (where other refers to those born in the UK/Ireland, North-West Europe, South-East Europe, Asia, and Other), and education (primary or high or trade/certificate or diploma/degree).

Depression assessment

Depression assessment was undertaken using the mood module of the Primary Care Evaluation of Mental Disorders (PRIME-MD) instrument, which has demonstrated reliability and validity in primary health-care settings [34], [35]. It comprises of 16 questions related to symptoms of depression experienced nearly every day for the last 2 two weeks (categorized ‘yes’/'no’), and was embedded in the broader SAHOS. Scoring to determine depression caseness was undertaken at the data analysis stage using the PRIME-MD method for detection of DSM-IV diagnoses [36]. The PRIME-MD allows the identification of individual mood disorders, but for the purposes of these analyses respondents were classified as having major depression, other depressions or no depression. The DSM-IV definition of major depression is a disorder characterized by a depressed mood or loss of interest or pleasure nearly every day for the last 2 two weeks with at least five characteristic symptoms (change in appetite or weight, sleep, and psychomotor activity; decreased energy; feelings of worthlessness or guilt; difficulty thinking, concentrating, or making decisions; or recurrent thoughts of death or suicidal ideation, plans or attempts). Other depression includes dysthymia, other depression and partial remission of major depression, determined on the basis of having symptoms of depression that were not sufficient to fulfil the criteria for major depression. This classification is consistent with the reported trend effect in morbidity and impact on quality of life for the three groups, those with no depression showing the least morbidity and impact, those with major depression the highest and those with other depressions being intermediate [37].

Data analysis

Data were double-entered and verified. Missing data were collected by follow-up telephone interview. Data were then weighted to benchmarks derived from the 1996 and 2001 censuses and 2002 Estimated Resident Population [38] to provide estimates representative of the South Australian population in terms of age, gender and region (metropolitan, non-metropolitan), probability of household selection and household size.

Categorical data were analysed using X² test and continuous variables with analysis of variance. Logistic regression was used to examine adjusted models of depression.

All analyses were carried out using SPSS version 14 (SPSS, Chicago, IL, USA).

Results

Table 1 provides participant details. As shown, there was no significant difference between the 1998 and 2004 surveys by gender, country of birth, relationship status or health status. The 2004 cohort was older (mean 1998 age = 44 years, SD = 19 years) compared with 2004 mean age = 45 years (SD = 19 years), F = 8.68, df = 1,6023, p < 0.01), and was more highly educated and was less likely to be homemaking and more likely to be retired.

Table 1.

Demographic characteristics

	SAHOS survey
	1988	2004	Statistics
Gender
Female	1546	1535
Male	1464	1480	X²=0.12, df = 1, p = 0.73
Country of birth
Australia	2266	2244
UK/Ireland	381	367
Other	363	404	X²=2.56, df = 2, p = 0.28
Age group (years)
15–29	797	727
30–49	1147	1082
50–69	696	800
≥ 70	370	406	X²=14.01, df = 3, p < 0.01
Relationship status
Partnered (married/de facto)	1851	1865
Separated/Divorced	221	257
Widowed	187	168
Never married	748	725	X²=4.13, df = 3, p = 0.25
Household size
1 person	441	441
2 persons	1600	1642
3 persons	570	537
4+ persons	339	395	X²=1.54, df = 3, p = 0.67
Education attainment
Primary	663	556
High	1019	979
Trade	373	380
Certificate/Diploma	599	681
Degree	356	419	X²=20.63, df = 4, p < 0.01
Labourforce participation
Full-time employment	1145	1169
Part-time employment	463	509
Unemployed	105	64
Home duties	515	331
Student	270	285
Retired	460	562
Other	53	93	X²=73.94, df = 6, p < 0.01
Health status
Excellent	663	594
Very good	1088	1086
Good	777	820
Fair	373	377
Poor	109	138	X²=8.37, df = 4, p = 0.08

SAHOS, South Australian Health Omnibus Surveys.

Table 2 shows a univariate analysis of the proportions classified with depression for each of the two cohorts, by age group and gender. As shown in the table, there were no significant differences by gender by age group between 1998 and 2004. The largest differences in proportions were among females aged 15–29 and 30–49 years, both cohorts of which had a non-significant 4% increase in the proportion classified with major depression between 1998 and 2004. The only statistically significant finding reported in Table 2 is a decline in the proportion of those classified with other depression; this finding, however, was not confirmed in the multivariable analysis presented in Table 4.

Table 2.

Depression prevalence, by gender, age group and year, univariate analysis

	Age (years)	1998			2004
		No symptoms (%)	Other (%)	Major (%)	No symptoms (%)	Other (%)	Major (%)
Female	15–29	77.1	13.0	9.9	74.6	11.0	14.4
	30–49	81.7	11.0	7.2	80.7	8.5	10.7
	50–69	81.8	9.7	8.5	85.4	6.4	8.1
	70 +	82.8	10.9	6.3	85.6	8.9	5.5

Male	15–29	86.8	10.4	2.7	90.6	7.0	2.4
	30–49	82.0	9.9	8.1	84.4	8.3	7.2
	50–69	85.8	9.6	4.7	84.3	8.6	7.1
	70 +	86.0	10.0	4.0	85.3	9.4	5.3

All	15–29	81.9	11.7	6.4	82.8	8.9	8.3
	30–49	81.9	10.5	7.7	82.5	8.4	9.1
	50–69	83.6	9.8	6.6	85.0	7.4	7.6
	70 +	84.3	10.3	5.4	85.4	9.1	5.4
Overall		82.6	10.6∗	6.8	83.6	8.4∗	8.0

Statistics = χ². There are no significant differences by gender by age group between 1998 and 2004.

∗χ²=10.76, df = 2, p < 0.01.

Examination of the relationship between depression classification and taking medication for depression showed that, among those with no symptoms of depression, in the 1998 cohort depression medication was taken by 1.6%, rising to 3.1% in 2004 (odds ratio (OR) = 1.98, 95% confidence interval (CI) = 1.32–2.97).

It is possible that the results in Table 2 were confounded by the differences in the samples reported in Table 1, and depression medication status. Table 3 shows a multivariable logistic regression model predicting major depression status, where the estimates over time are adjusted by differences in the cohorts.

Table 3.

Predicting major depression

	Reference group	Comparator	OR	95%CI
Gender	Male	Female	1.53	1.19–1.96
Age group (years)	15–29s	30–49	1.18	0.85–1.65
		50–69	0.57	0.38–0.86
		≥70	0.39	0.22–0.68
Education attainment	Primary/High/Trade	Certificate/Diploma	0.82	0.62–1.09
		Degree	0.54	0.35–0.83
Relationship status	Partnered (Married/De facto)	Separated/Divorced	2.19	1.59–3.03
		Widowed	0.87	0.51–1.49
		Never married	1.87	1.34–2.60
Labourforce status	Working/Student	Homemaker	2.02	1.47–2.79
		Unemployed/Other	2.54	1.80–3.58
		Retired	1.21	0.78–1.86
Health status	Excellent	Very good	1.96	1.25–3.07
		Good	3.32	2.13–5.17
		Fair	8.84	5.60–13.94
		Poor	21.85	12.98–36.79
Depression medication	Not taking	Taking	5.08	3.73–6.94
Study cohort	1998	2004	1.12	0.90–1.39

CI, confidence interval; OR, odds ratio.

Those with other depressions were coded as no depression. Country of birth was excluded from the model because this was non-significant.

Hosmer and Lemeshow χ²=5.22, df = 8, p = 0.73.

One variable excluded from the multivariable analyses in Tables 3–4 was household size. When we examined household size by depression we found a significant association. Those living alone were significantly more likely to be classified with both minor (excluding major depression cases) and major depression (excluding minor depression cases) than were those living with others (OR_minor=1.52, 95%CI = 1.21–1.92; OR_major=2.18, 95%CI = 1.72–2.76). However, the association between relationship status and household size showed that 97% of those who were living alone were not in a relationship (χ²=3,436.40, df = 9, p < 0.0001). Furthermore, of those living in a household with others and who were aged 15–19 years 53% had never partnered, suggesting that they were still living in the family home, most likely with their parents and/or siblings. We interpreted these findings to suggest that living alone or living with others was a function of a person's relationship and family status. We therefore excluded household size from the multivariate analyses because it would have been a confounder.

The significant predictors of depression, when compared with the reference groups, were being female (53% more likely to be classified with depression), being separated/divorced (2.2 times more likely) or having never married (87% more likely), being a homemaker (twice as likely), being unemployed (2.6-fold more likely), having very good overall health (twice as likely), good overall health (3.3 times more likely), fair overall health (8.8 times more likely) or poor overall health (21.9 times more likely).

The protective predictors were being older (those aged 50–59 years and those aged ≥70 years were 43% and 61% less likely to be classified with depression), and having completed a university degree (46% less likely).

An important finding was the higher proportion of females classified with depression when compared with males (Table 3). When examined by age group, however, in both the 1998 and 2004 samples this effect was confined to those aged 15–29 years (χ² ₁₉₉₈=19.81 and χ² ₂₀₀₄=40.24, both df = 2, p < 0.001; no other comparisons statistically significant).

As was expected, those taking depression medication were five times more likely to be classified with depression. To examine whether depression medication may have confounded the model (those without symptoms and who were taking depression medication may have unduly influenced the classification of depression), the model was re-run excluding depression medication. No significant changes to the model were observed, suggesting that the increase in depression medication between 1998 and 2004 played no significant part in ameliorating the prevalence of depression. The OR in the model excluding depression medication for year (2004 when compared with 1998) was 1.20 (95%CI = 0.97–1.48).

In short, there was no significant effect over time after adjusting for background variables. This adjusted multivariable finding is consistent with the univariate finding in Table 2.

The same analysis was conducted for other depressions, excluding those with major depression (Table 4). The significant predictors of other depression were, when compared with the reference group, being separated or divorced (69% more likely to be depressed), being unemployed/other (2.2 times more likely), having very good health (79% more likely), good health (2.4 times more likely), fair health (5.36 times more likely), poor health (11 times more likely) and taking depression medication (4.07 times more likely). The protective predictors were being older (aged 50–69 years and ≥70 years, both 43% less likely to be classified with other depression).

Table 4.

Predicting other depressions

	Reference group	Comparator	OR	95%CI
Gender	Male	Female	1.09	0.89–1.33
Age group (years)	15–29	30–49	0.91	0.68–1.22
		50–69	0.57	0.40–0.81
		≥70	0.57	0.36–0.92
Education attainment	Primary/High/Trade	Certificate/Diploma	0.95	0.75–1.20
		Degree	0.99	0.74–1.33
Relationship status	Partnered (Married/De facto)	Separated/Divorced	1.69	1.24–2.31
		Widowed	1.02	0.66–1.58
		Never married	1.30	0.98–1.74
Labourforce status	Working/Student	Homemaker	1.28	0.95–1.74
		Unemployed/Other	2.21	1.57–3.12
		Retired	1.00	0.70–1.43
Health status	Excellent	Very good	1.79	1.31–2.43
		Good	2.40	1.75–3.27
		Fair	5.36	3.79–7.57
		Poor	10.69	6.78–16.86
Depression medication	Not taking	Taking	4.07	2.85–5.82
Study cohort	1998	2004	0.74	0.62–0.89

CI, confidence interval; OR, odds ratio.

Those with major depression were excluded from the model. Country of birth was excluded from the model because this was non-significant.

Hosmer and Lemeshow χ²=8.86, df = 8, p = 0.35.

Regarding the survey cohort, those surveyed in 2004, when compared with the 1998 cohort, were significantly less likely to suffer other depression, after adjustment for the other variables in the model. To examine the impact of depression medication on this finding, the model was re-run without this variable. No significant differences were reported for any of the variables in the model. The OR in the model excluding depression medication for year 2004 when compared with 1998 was 0.78 (95%CI = 0.65–0.93), compared with 0.74 for the full model in Table 4. A separate logistic regression analysis examined the relationship between current antidepression medication, depression classification and cohort year. The results showed a 90% increase in current antidepression medication (OR_adj=1.91, 95%CI = 1.29–2.82) between 1998 and 2004, but no interaction effect between depression classification and cohort year. These two results suggest that the significantly increased consumption of current antidepression medication had no statistically significant effect on other depression classification between 1988 and 2004.

Discussion

This study examined depression prevalence over a 6-year period in South Australia. The two samples were drawn in the identical manner and were very similar in size, the instruments used and data collection methods.

The limitations of the study are that it was repeated cross-sectional rather than longitudinal, that the PRIME-MD may have misclassified cases, and that the time interval between 1998 and 2004 was too short to observe increases in depression prevalences. The repeated cross-sectional nature of the study, however, has the advantage that it allows for changes in the population to be reflected in the samples. In this case, the data were adjusted for potential differences through the use of multivariable analyses.

Regarding possible misclassification by the PRIME-MD, this would have affected both samples similarly. The PRIME-MD has demonstrated reliability and validity in primary health-care settings: Boyer et al. reported that when compared with physician diagnosis the sensitivity was 1.00 and the specificity 0.91 [34], [35]. Previous population-based prevalence studies have used other assessments including the Composite International Diagnostic Interview (CIDI) [27], [39]; psychiatric interviews [20], [40], the Hopkins Symptom Checklist-25 (HSCL-25) [5], the Research Diagnostic Criteria (RCD) or the DSM-III criteria [4] and the Diagnostic Interview Schedule (DIS) [7], [40]. In short, a wide variety of measures and/or interviews have been used in population prevalence studies. This diversity of measurement is consistent with Kendler and Gardner's contention that the classification of major depression is but a convenient cut-off along a continuum of symptoms [41]. Because there has been no direct head-to-head validation of these different measures (although most are based on either the DSM-III or the ICD-10) it is possible that different prevalence estimates are a function of measurement differences.

That the time interval may have been too short to detect increases in depression prevalence is unlikely given that significant decreases in other depression were observed (Table 2). It is acknowledged, however, that birth cohort studies showing increases in depression have typically compared retrospective rates over 30-year periods [1–7]. This observation raises the issue of study power to detect differences between the two cohorts. Based on sample size calculation for logistic regression [42], the study sample size would have been sufficient to statistically detect a difference of 16% compared with the observed difference of 12% in major depression.

Subject to these caveats, the findings showed that there was no significant increase in either major or other depression in South Australia between 1998 and 2004, as reported in our previous two papers [15], [16]. For major depression, this finding held for both the univariate analyses by gender and age group presented in Table 2 and the multivariable analyses presented in Table 3. The adjusted observed increase of 12% (Table 3) between 1998 and 2004 in major depression classification could have occurred by chance, subject to the caveat of study power. For other depression, on both the univariate and multivariable analyses (Tables 2 and 4) there was a statistically significant decrease in classification by 26% between 1998 and 2004. To our knowledge this is the first study reporting such an outcome, other than the longitudinal Swedish Lundby study, which reported lower incidence rates as participants aged. The authors attributed this to possible forgetfulness of mild episodes [20]. These findings are also consistent with the Canadian Stirling County Study, which collected data in three cross-sectional surveys conducted over a 40-year timeframe. Even acknowledging that there were variations between surveys, the researchers concluded that over the 40-years the rates of depression were stable, although there were gender and age re-distributions [21]. The results are also consistent with Sacker and Wiggin's finding that psychological distress remained stable in the UK between 1981 and 2000 [43].

The most likely cause of the reduction in other depression observed in the present study is the increased consumption of current antidepression medication (a 90% increase between 1998 and 2004), reflecting the hypothesis that these would ameliorate depression symptoms and that the main beneficiaries would be those with other depression. The sub-analysis of depression classification by current antidepression medication use reported here, however, did not support this hypothesis. Between 1998 and 2004 antidepressant medication rose 97% among those without symptoms, compared with a 73% increase among those classified with other depressions and a 63% increase among those with major depression. These findings raise the distinct possibility that the observed increase in antidepression medication was largely due to overprescription. The implication is that increased consumption of antidepression medication played no significant role in influencing depression prevalence at the community level – indeed it is consistent with Mulder's contention concerning the ‘medicalisation of unhappiness’ [44] and other recent findings concerning the lack of efficacy of antidepression medication [45]. Our findings are also consistent with those from an Italian study reporting no decline in hospital admission for depression despite increases in antidepressant medication consumption [46].

As shown in Table 2, the overall unadjusted prevalence of major depression was 7.4% (6.8% in 1998 and 8.0% in 2004), and for other depression it was 9.5% (10.6% in 1998 and 8.4% in 2004). These findings are consistent with, but slightly higher than, other Australian-based population studies into depression prevalence. The 1997 Australian National Survey of Mental Health and Wellbeing of Adults, using DSM-IV criteria and the CIDI, reported a 12 month prevalence of 5.1%, with 3.4% for males and 6.8% for females [27], [39]. Andrews et al. reported a 12 month prevalence for major depression of 6.7% and a 1 month prevalence of 3.3% [27]. Additionally, 1.3% of the population suffered from dysthymia (ICD-10 criteria) over 12 months and 1.1% over the last month. Re-analysis of the same data provided Henderson et al. with 12 month prevalences of 5.8% overall, and 7.4% and 4.2% females and males, respectively [29]. Although there are several possible explanations for these differences, including sampling methodology, response bias or response artefacts, or inclusion of ‘brief depression’ [47], Goldney et al. noted that they may be due to the CIDI exclusion criteria, which were designed to exclude those with symptoms due to medications, alcohol or drug use, a physical illness or injury, those who had not consulted a clinician, who considered their condition did not interfere a ‘lot’ with their everyday life and those who had taken medication for their symptoms [48]. None of these exclusions applied in the present study.

Consistent with other reports and reviews [1], [3], [6], [21–25], [30], [32], [39], [47], [49], females were significantly more likely to be classified as suffering depression when compared with males (Table 3). However, sub-analyses of this finding by age group showed that this was confined to those aged 15–29 years; there were no significant differences between males and females in any other age groups for either the 1998 or 2004 surveys. It should be noted that few previous studies reporting gender differences have broken these down by age groups. Myers et al. showed similar results in the US Epidemiologic Catchment Area study, although the peak years for females were between 25–44 years [22]; Kessler et al. reported results consistent with those of this study [23]. In their reporting of the results from the US National Comorbidity Survey they showed that the highest prevalence rate was among those aged 15–24 years although they did not break this down by gender. The only recent study reporting this appears to be the Stirling County study, which reported that the highest prevalence was among younger females [22]. Previously both Klerman, and Jorm, in review articles, had reported higher prevalences among younger females [7], [24]. Sacker and Wiggins reported the same finding for general psychological distress [43]. Three possible explanations are that: (i) this is due to differences in social roles [24]; (ii) young men are more likely to forget depression episodes and women to remember them [50]; or (iii) due to changes in mental health literacy younger females may be more likely to recognize depression symptoms earlier and seek treatment for these [15], [51]. All three possible explanations would be consistent the Wilhelm and Parker's observation that there may be a need for different classification definitions in epidemiological studies [50]. This finding is therefore important, suggesting that public health interventions may be most effective where they are age and gender cohort targeted.

That there was no significant increase in depression prevalence among older adults (Table 2) is consistent with the literature disputing the stereotyping of ageing [25], [26]. This may be partly due to the use of a standard diagnostic instrument, the PRIME-MD, which attempts to classify cases based on DSM-IV clinical criteria. This finding is consistent with Klerman's observation that depression increases in the elderly are associated with instruments that assess symptoms (older people report more symptoms) but not when assessments rely upon clinical classifications [7].

The statistically significant lower rates of both major depression and other depression among those aged ≥50 years may reflect that mental health (other than cognitive impairment, such as dementia) seems to improve over the lifespan [10], [28], [31], [52–56], perhaps due to a decrease in negative affect [57], leading to discounting of negative feelings that may be more important to younger adults. Certainly the findings from this present study are consistent with other reports showing that older adults have lower prevalences when compared with younger adults [22], [27–32], although this is not confirmed in all studies [58].

The finding of lower probabilities of depression classification among older adults has implications for the increasing depression hypothesis. Examination of the data by birth cohort by age of onset, as has been widely done in studies showing increases in depression [2], [3], [6], [7], [59], was not possible because age of first depression onset was not collected. The implication, however, is that discounting of negative experiences could partly explain birth cohort findings, particularly given that age of onset studies generally rely upon memory for age of first onset – a method that is notoriously inaccurate because it ignores memory effects that systematically distort recall of health problems [20], [47], [60–62], particularly where there is a long time between assessments, temporary exacerbation or remission of health conditions [63], and Hawthorne or placebo effects [64], [65].

Another possible explanation of differences between studies may relate to continued study participation in longitudinal studies, where as few as 50% of the original participants may be available for follow-up analyses [20]. The repeated cross-sectional design of this study overcomes this as an explanation of the study findings.

The key predictor of depression in the multivariable analyses presented in Tables 3 and 4 was overall health status, suggesting a direct relationship between health and depression. This finding is consistent with previous reports showing that prior physical health, disability and social support play major roles in the onset of depressive symptoms [66–68] and are strongly associated with current depression status [37], [69–72]. The other significant predictors (labourforce participation, education attainment and relationship status in Tables 3 and 4) are also consistent with the literature [32], [73], suggesting a complex web of social factors co-existing with, caused by or causing depression. We note that the univariate relationship between living alone and depression is also consistent with the literature [72], although the findings suggest it is confounded with relationship status. These findings suggest that mental health programmes aimed at treating depression need to go beyond simple mental health care if such treatments are to have population-wide impacts. It is possible that these relationships partly explain the very limited impacts reported above of antidepression medication: if the causes and associations of depression symptoms are not addressed, medication by itself may have limited effect.

Conclusion

This is the first Australian study to examine the increasing depression hypothesis. Although doubts about this hypothesis first surfaced over 20 years ago, the hypothesis still has widespread currency and has been used to predict the burden of disease in the future.

Using a repeated cross-sectional study design over a 6-year period, with more than 3000 participants at each wave, where the data were weighted to be representative of the South Australian population, no statistically significant increases in either major or other depressions were observed regardless of whether univariate or multivariable analyses were used. Indeed, the only over-time statistically significant finding over time showed a reduction in the proportion of cases classified with other depression.

The other significant findings were that current major and other depressions decreased with aging and that the cohort with highest levels of major depression were females aged 15–29 years.

The results of this study, whilst in need of replication over a longer period of time, would suggest that it is time for the conventional wisdom that depression is increasing to be reconsidered, and for a more realistic perspective, developed from evidence-based analyses, to be accepted.

Footnotes

Acknowledgements

This study was support by unrestricted grants from Ely Lily Australia, Janssen-Cilag, Lundbeck Australia, Organon Australia, Sanofi-Synthelabo, Wyeth Australia, Ramsay Health Care Australia and University of Melbourne. None of these organizations participated in the design of the study, the analysis of the data or in the preparation of the manuscript.

References

1. Bland

. Epidemiology of affective disorders: a review. Can J Psychiatry 1997; 42: 367–377.

2. Cross National Collaborative Group . The changing rate of major depression: cross-national comparisons. JAMA 1992; 268: 3098–4105.

3. Joyce

Oakley-Browne

Wells

Bushnell

Hornblow

. Birth cohort trends in major depression: increasing rates and earlier onset in New Zealand. J Affect Disord 1990; 18: 83–89.

4. lerman

Weissmann

. Increasing rates of depression. JAMA 1989; 261: 2229–2235.

5. Sandanger

Nygard

Ingebrigtsen

Sorensen

Dalgard

. Prevalence, incidence and age at onset of psychiatric disorders in Norway. Soc Psychiatry Psychiatr Epidemiol 1999; 34: 570–579.

6. Weissman

Bland

Joyce

Newman

Wells

Wittchen

. Sex differences in rates of depression: cross-national perspectives. J Affect Disord 1993; 29: 77–84.

7. Kerman

. The current age of youthful melancholia. Evidence for increase in depression among adolescents and young adults. Br J Psychiatry 1988; 152: 4–14.

8. Olfson

Shea

Feder

. Prevalence of anxiety, depression, and substance use disorders in an urban general medicine practice. Arch Fam Med 2000; 9: 876–883.

9. WHO . The world health report: Mental health: new understanding, new hope. Office of Publications, World Health Organization, Geneva 2001; 2001.

10.

10. Alonso

Angermeyer

Bernert

. Prevalence of mental disorders in Europe: results from the European Study of the Epidemiology of Mental Disorders (ESEMeD) project. Acta Psychiatr Scand Suppl 2004; 109(420)21–27.

11.

11. Kennedy

Kelman

Thomas

Wisniewski

Metz

Bijur

. Hierarchy of characteristics associated with depressive symptoms in an urban elderly sample. Am J Psychiatry 1989; 146: 220–225.

12.

12. Mathers

Vos

Stevenson

. The burden of disease and injury in Australia. Australian Institute of Health and Welfare, Canberra 1999.

13.

13. Murray

CJL

, Lopez

. The global burden of disease: a comprehensive assessment of mortality and disability from diseases, injuries, and risk factors in 1990 and projected to 2020. Cambridge, MA: Harvard School of Public Health on behalf of the World Health Organization and the World Bank, 1996.

14.

14. Ustun

Ayuso-Mateos

Chatterji

Mathers

Murray

. Global burden of depressive disorders in the year 2000. Br J Psychiatry 2004; 184: 386–392.

15.

15. Goldney

Fisher

Dal Grande

Taylor

Hawthorne

. Have education and publicity about depression made a difference? A comparison of prevalence, service use and excess costs in South Australia: 1998 and 2004. Aust N Z J Psychiatry 2007; 41: 38–53.

16.

16. Hawthorne

Cheok

Goldney

Fisher

. The excess cost of depression in South Australia: a population-based study. Aust N Z J Psychiatry 2003; 37: 362–373.

17.

17. Australian Institute of Health and Welfare. Heart, stroke and vascular diseases, Australian facts, Cardiovascular Disease Series N10. Canberra: Australian Institute of Health and Welfare and the Heart Foundation of Australia, 1999.

18.

18. Lonnqvist

. Psychiatric aspects of suicidal behaviour: depression. The international handbook of suicide and attempted suicide, Hawton

van Heeringen

. John Wiley & Sons, Chichester 2000.

19.

19. Murray

Lopez

. Alternative projections of mortality and disability by cause 1990–2020: Global Burden of Disease Study. Lancet 1997; 349: 1498–1504.

20.

20. Mattisson

Bogren

Nettelbladt

Munk-Jorgensen

Bhugra

. First incidence depression in the Lundby Study: a comparison of the two time periods 1947–1972 and 1972–1997. J Affect Disord 2005; 87: 151–160.

21.

21. Murphy

Laird

Monson

Sobol

Leighton

. A 40-year perspective on the prevalence of depression: the Stirling County Study. Arch Gen Psychiatry 2000; 57: 209–215.

22.

22. Myers

Weissman

Tischler

. Six-month prevalence of psychiatric disorders in three communities 1980 to 1982. Arch Gen Psychiatry 1984; 41: 959–967.

23.

23. Kessler

McGonagle

Zhao

. Lifetime and 12-month prevalence of DSM-III-R psychiatric disorders in the United States. Results from the National Comorbidity Survey. Arch Gen Psychiatry 1994; 51: 8–19.

24.

24. Jorm

. Sex and age differences in depression: a quantitative synthesis of published research. Aust N Z J Psychiatry 1987; 21: 46–53.

25.

25. Copeland

Beekman

Dewey

. Cross-cultural comparison of depressive symptoms in Europe does not support stereotypes of ageing. Br J Psychiatry 1999; 174: 322–329.

26.

26. Copeland

Beekman

Dewey

. Depression in Europe. Geographical distribution among older people. Br J Psychiatry 1999; 174: 312–321.

27.

27. Andrews

Henderson

Hall

. Prevalence, comorbidity, disability and service utilisation. Overview of the Australian National Mental Health Survey. Br J Psychiatry 2001; 178: 145–153.

28.

28. Henderson

Jorm

Korten

Jacomb

Christensen

Rodgers

. Symptoms of depression and anxiety during adult life: evidence for a decline in prevalence with age. Psychol Med 1998; 28: 1321–1328.

29.

29. Henderson

Andrews

Hall

. Australia's mental health: an overview of the general population survey. Aust N Z J Psychiatry 2000; 34: 197–205.

30.

30. Oakley-Browne

Joyce

Wells

Bushnell

Hornblow

. Christchurch Psychiatric Epidemiology Study, part II: Six month and other period prevalences of specific psychiatric disorders. Aust N Z J Psychiatry 1989; 23: 327–340.

31.

31. Regier

Boyd

Burke

Jr ,. One-month prevalence of mental disorders in the United States. Based on five epidemiologic catchment area sites. Arch Gen Psychiatry 1988; 45: 977–986.

32.

32. Riddle

Smith

. Millennium cohort: the 2001–2003 baseline prevalence of mental disorders in the U.S. military. J Clin Epidemiol 2007; 60: 192–201.

33.

33. Wilson

Wakefield

Taylor

. The South Australian Health Omnibus Survey. Health Promotion J Aust 1992; 2(3)47–49.

34.

34. Boyer

Bisserbe

Weiller

. How efficient is a screener? A comparison of the PRIME-MD patient questionnaire with the SDDS–PC screen. Int J Methods Psychiatr Res 1997; 7: 27–32.

35.

35. Spitzer

Williams

Kroenke

. Utility of a new procedure for diagnosing mental disorders in primary care. The PRIME-MD 1000 study. JAMA 1994; 272: 1749–1756.

36.

36. American Psychiatric Association . Diagnostic and statistical manual of mental disorders, fourth edition. American Psychiatric Association, Washington, DC 1994.

37.

37. Goldney

Fisher

Wilson

Cheok

. Major depression and its associated morbidity and quality of life in a random, representative Australian community sample. Aust N Z J Psychiatry 2000; 34: 1022–1029.

38.

38. Australian Bureau of Statistics. Census of population and housing: selected social and housing characteristics, Australia, Report No.: 2015.0. Canberra: Australian Bureau of Statistics, 2002.

39.

39. Australian Bureau of Statistics. Mental health and wellbeing: profile of adults in Australia 1997, ABS Catalogue N. 4326.0. Canberra: Australian Bureau of Statistics, Commonwealth of Australia, 1998.

40.

40. Murphy

Horton

Laird

Monson

Sobol

Leighton

. Anxiety and depression: a 40-year perspective on relationships regarding prevalence, distribution, and comorbidity. Acta Psychiatr Scand 2004; 109: 355–375.

41.

41. Kendler

Gardner

Jr . Boundaries of major depression: an evaluation of DSM-IV criteria. Am J Psychiatry 1998; 155: 172–177.

42.

42. Hsieh

Bloch

Larsen

. A simple method of sample size calculation for linear and logistic regression. Stat Med 1998; 17: 1623–1634.

43.

43. Sacker

Wiggins

. Age-period-cohort effects on inequalities in psychological distress, 1981–2000. Psychol Med 2002; 32: 977–990.

44.

44. Mulder

. An epidemic of depression or the medicalisation of unhappiness. N Z Fam Physician 2005; 32: 161–163.

45.

45. Nemeroff

Entsuah

Benattia

Demitrack

Sloan

Thase

. Comprehensive analysis of remission (COMPARE) with venlafaxine versus SSRIs. Biol Psychol 2008; 63: 424–434.

46.

46. Guaiana

Andretta

Corbari

. Antidepressant drug consumption and public health indicators in Italy, 1955 to 2000. J Clin Psychiatry 2005; 66: 750–755.

47.

47. Angst

. Epidemiology of depression. Psychopharmacology 1992; 106 (Suppl): S71–S74.

48.

48. Goldney

Hawthorne

Fisher

. Is the Australian National Survey of Mental Health a reliable guide for health planners? A methodological note on the prevalence of depression. Aust N Z J Psychiatry 2004; 38: 635–638.

49.

49. Spaner

Bland

Newman

. Major depressive disorder. Acta Psychiatr Scand Suppl 1994; 376: 7–15.

50.

50. Wilhelm

Parker

. Sex differences in lifetime depression rates: fact or artefact?. Psychol Med 1994; 24: 97–111.

51.

51. Goldney

Fisher

Dal Grande

Taylor

. Changes in mental health literacy about depression: South Australia, 1998 to 2004. Med J Aust 2005; 183: 134–137.

52.

52. Australian Bureau of Statistics . National Health Survey: SF-36 population norms, Australia. Australian Bureau of Statistics, Canberra 1997.

53.

53. Christensen

Jorm

Mackinnon

. Age differences in depression and anxiety symptoms: a structural equation modelling analysis of data from a general population sample. Psychol Med 1999; 29: 325–339.

54.

54. Hawthorne

Osborne

Sansoni

Taylor

. The SF36 Version 2: critical analysis of population weights, scoring algorithms and population norms. Qual Life Res 2007; 16: 661–673.

55.

55. Hybels

Blazer

. Epidemiology of late-life mental disorders. Clin Geriatr Med 2003; 19: 663–696.

56.

56. Trollor

Anderson

Sachdev

Brodaty

Andrews

. Prevalence of mental disorders in the elderly: the Australian National Mental Health and Well-Being Survey. Am J Geriatr Psychiatry 2007; 15: 455–466.

57.

57. Charles

Reynolds

Gatz

. Age-related differences and change in positive and negative affect over 23 years. J Pers Soc Psychol 2001; 80: 136–151.

58.

58. Andrade

Caraveo-Anduaga

Berglund

. The epidemiology of major depressive episodes: results from the International Consortium of Psychiatric Epidemiology (ICPE) Surveys. Int J Methods Psychiatr Res 2003; 12: 3–21.

59.

59. Weissman

Leaf

Tischler

. ffective disorders in five United States communities. Psychol Med 1988; 18: 141–153.

60.

60. Giuffra

Risch

. Diminished recall and the cohort effect of major depression: a simulation study. Psychol Med 1994; 24: 375–383.

61.

61. Foddy

. Constructing questions for interviews and questionnaires: theory and practice in social research. Cambridge University Press, Cambridge 1993.

62.

62. Parker

. Are the lifetime prevalence estimates in the ECA study accurate?. Psychol Med 1987; 17: 275–282.

63.

63. Norman

Stratford

Regehr

. Methodological problems in the retrospective computation of responsiveness to change: the lesson of Cronbach. J Clin Epidemiol 1997; 50: 869–879.

64.

64. Macebo

Farre

Banos

. Placebo effect and placebos: what are we talking about? Some conceptual and historical considerations. Eur J Clin Pharmacol 2003; 59: 337–342.

65.

65. Hrobjartsson

Gotzsche

. Is the placebo powerless? An analysis of clinical trials comparing placebo with no treatment. N Engl J Med 2001; 344: 1594–1602.

66.

66. Cole

Dendukuri

. Risk factors for depression among elderly community subjects: a systematic review and meta-analysis. Am J Psychiatry 2003; 160: 1147–1156.

67.

67. Kaplan

Roberts

Camacho

Coyne

. Psychosocial predictors of depression. Prospective evidence from the human population laboratory studies. Am J Epidemiol 1987; 125: 206–220.

68.

68. Phifer

Murrell

. Etiologic factors in the onset of depressive symptoms in older adults. J Abnorm Psychol 1986; 95: 282–291.

69.

69. Braam

Prince

Beekman

. Physical health and depressive symptoms in older Europeans. Results from EURODEP. Br J Psychiatry 2005; 187: 35–42.

70.

70. Cacioppo

Hughes

Waite

Hawkley

Thisted

. Loneliness as a specific risk factor for depressive symptoms: cross-sectional and longitudinal analyses. Psychol Aging 2006; 21: 140–151.

71.

71. Lenze

Rogers

Martire

. The association of late-life depression and anxiety with physical disability: a review of the literature and prospectus for future research. Am J Geriatr Psychiatry 2001; 9: 113–135.

72.

72. Prince

Harwood

Blizard

Thomas

Mann

. Impairment, disability and handicap as risk factors for depression in old age. The Gospel Oak Project V. Psychol Med 1997; 27: 311–321.

73.

73. Hawthorne

. Perceived social isolation in a community sample: its prevalence and correlates with aspects of peoples’ lives. Soc Psychiatry Psychiatr Epidemiol 2008; 43: 140–150.

Depression Prevalence: Is it Really Increasing?

Abstract

Keywords

Methods

Participants

Measures

Demographics

Depression assessment

Data analysis

Results

Discussion

Conclusion

Footnotes

Acknowledgements

References