Abstract
Probably only those involved with a major survey understand the extensive commitment required for such a survey to be prepared, run, analysed and written up. Congratulations are in order to all who played a role in the Australian 2007 National Survey of Mental Health and Wellbeing (NSMHWB), particularly to the Advisory Group for the survey, the Australian Bureau of Statistics, who actually ran the survey and performed the field work, and the researchers who have put together the suite of papers for this issue of the Australian and New Zealand Journal of Psychiatry. It is good to see, even at this early stage in the analysis, that policy implications have been drawn, where possible.
This editorial sets out to compare the findings from the Australian 2007 NSMHWB and the New Zealand Mental Health Survey (NZMHS), which was carried out in late 2003 and 2004. Both have resulted in an issue of the Australian and New Zealand Journal of Psychiatry: the October 2006 issue for the NZMHS [1–11] and this issue for the 2007 NZMHWB [12–17], which allows readers to obtain an overview of the main findings all at once instead of having to track papers across multiple journals as they are published. Of course there will be many more papers from each survey but to have the core descriptive papers presented together in one issue is a bonus not available to many of the other countries that have also participated in the World Mental Health Survey Initiative.
Before comparing the findings it is necessary to set the context by describing the similarities and differences between the two surveys. The major features of the surveys were the same: they were both cross-sectional surveys with national stratified multistage probability samples, they both interviewed participants face to face using a computer-assisted personal interview, and both used the English-language version of the Composite International Diagnostic Interview (CIDI 3.0), a fully structured interview that can be administered by trained lay interviewers, making it possible to obtain large samples. The common features make these two surveys comparable with each other and with other surveys in the World Mental Health Consortium (www.hcp.med.harvard.edu/wmh/).
There are, however, some important differences between the 2007 Australian NSMHWB and the New Zealand NZMHS. Although the CIDI 3.0 was used in both countries, different approaches were taken to meet the common goal of having an interview with an average completion time of 90 min, the maximum average length thought to be acceptable to potential participants in the absence of remuneration. Certain diagnostic and demographic sections were omitted from both surveys but only the New Zealand survey retained the Part 1/Part 2 structure whereby a subset of participants completed some sections. This enabled more sections to be included in New Zealand but at the expense of the sample size for those included in Part 2. In Australia all participants entered or at least were screened for all sections included in the interview.
The second major difference is that diagnoses reported so far from the NZMHS are DSM-IV diagnoses [18], whereas those from the 2007 NSMHWB are ICD-10 diagnoses [19], as was also the situation for reports from the 1997 NSMHWB. The CIDI 3.0 can yield either so that it is possible to have DSM-IV diagnoses from both surveys or ICD-10 diagnoses but that has not been done yet, and may not be done. In as much as these diagnostic systems differ, results from the 2007 NSMHWB and the NZMHS may differ even if prevalence, comorbidity, and help seeking do not differ between the two countries.
The third major difference is a non-trivial difference in response rate: 60.0% in Australia in 2007 (although 78% in 1997) [15], and 70.2% in New Zealand [20]. (The NZMHS response rate is usually quoted as 73.3% but that counted as ineligible those unable to participate because of difficulties with English or illness or other disability, whereas the 2007 NSMHWB did not count these as ineligible.) Under assumptions of comparable characteristics related to response rates, differences in response rates can be crudely accounted for when comparing prevalences. If a reported prevalence in Australia is less than that in New Zealand it is possible to calculate the prevalence that would have had to occur in the 10.2% not responding in Australia for the two countries to have similar prevalence, had response rates been equal. Similarly it can be calculated that if the observed Australian prevalence is more than 1.17 times that for New Zealand then Australia would have a higher prevalence even if response rates were equal (to see this, set the prevalence to zero in the 10.2% not participating). These simple calculations provide some guidelines for looking at differences in results.
In spite of these differences between the surveys it is clear that both countries have around 40% of the adult population who have already experienced a mental disorder at some time in their lives, with around 20% experiencing a disorder in the past 12 months. Mental disorders are often comorbid with each other and with chronic physical conditions, mental disorders are disabling and yet there is often no contact with treatment services for current disorder. In this Australia and New Zealand are also very like the USA [21, 22]. Furthermore, although reported prevalences are lower in some other World Mental Health Consortium sites [23, 24], the pattern of comorbidity and lack of treatment is common to all, although to varying extents.
Prevalence comparisons
In this section 95% confidence intervals are not quoted. Readers should bear in mind that in these two large surveys the 95% confidence intervals are less than ±1% for overall prevalences [6, 10, 15].
The prevalence of any disorder prior to interview, the lifetime prevalence, was 45.5% in Australia (2007 NSMHWB) and 39.5% in New Zealand (NZMHS), with 12 month prevalences of 20.0% and 20.7%, respectively. The prevalences, particularly the 12 month prevalences, are remarkably similar but they are not quite comparable because of inclusion of more disorders in the NZ interview and the use of ICD-10 in Australia and DSM-IV in New Zealand.
The overall prevalence of anxiety disorders would be expected to differ because the NZMHS interview included specific phobia, the most common anxiety disorder (10.8% lifetime, 7.3% 12 month) but the 2007 NSMHWB interview did not. Therefore the apparent equality of the overall prevalences for anxiety disorders may be misleading: lifetime prevalences are reported as 26.3% for Australia and 24.9% for New Zealand, 12 month prevalences are 14.4% and 14.8%. Use of the individual disorders assessed in both countries provides a fairer comparison [25]. The largest differences between Australia and New Zealand occur for post-traumatic stress disorder (PTSD) and panic disorder, with prevalences in Australia approximately double those in New Zealand.
For PTSD the lifetime prevalences are 12.2% in Australia and 6.0% in New Zealand, with 6.4% and 3.0% for 12 month prevalences. Even if there was no PTSD in the extra 10.2% who did not respond in Australia, prevalences would still be 10.4% and 5.5%. Results for the USA were very similar to those for New Zealand (6.8% and 3.5%) [21, 22]. Before investigating further why Australia has so much PTSD, by looking at the types of trauma reported, the first step should be to calculate prevalences using the same diagnoses. Inspection of the criteria shows that DSM-IV requires a reaction of intense fear, helplessness or horror and more persistence of symptoms than ICD-10, so these criteria may account for the apparent differences in prevalence.
Similarly, for panic disorder DSM-IV requires 1 month of concern about recurrence of attacks or consequences or a change in behaviour but ICD-10 does not, so diagnostic differences may account for the differences in the prevalences for panic disorder: lifetime prevalences of 5.2% in Australia and 2.7% in New Zealand, 12 month prevalences of 2.6% and 1.7%. Differences in obsessive–compulsive disorder (OCD) prevalences probably reflect revisions to the interview between the two surveys, so that participants were not screened out of the OCD section so early. Agoraphobia is reported in the 2007 NSMHWB whereas the NZMHS reported agoraphobia without panic. Prevalences for social phobia and generalized anxiety disorder were close in both countries (well within sampling error).
The reported prevalence of affective disorders is lower for Australia: lifetime prevalences are 15.9% for Australia and 20.2% for New Zealand; for 12 month prevalence the percentages are 4.1% and 6.7%, respectively. Again cautions about possible DSM-IV/ICD-10 diagnoses are in order. In Australia ICD-10 depression has lifetime prevalence of 11.6% and 12 month prevalence of 4.1%, whereas the prevalences for New Zealand are 16.0% and 5.7%. Bipolar disorder also appears less common in Australia but this is because the New Zealand prevalences [6, 10] include subthreshold cases (hypomania but not meeting criteria for a major depressive episode). If only bipolar I and II are included the New Zealand prevalences are perhaps a little lower: lifetime 2.9% for Australia, 1.7% for New Zealand, 12 month 1.8% and 1.0%.
Are Australians more likely to be affected by alcohol or drugs? Here comparisons are confounded by both diagnostic differences and by differences in the interviews. Early versions of the CIDI 3.0, including those used in the USA [21, 22] and in New Zealand, routed past the dependence section participants who did not report any symptom of abuse ever in their lifetime. This routing occurred in the alcohol section and in the drug section. After criticism [26] this routing was removed. The likely effect of the skip past dependence was investigated and found to be trivial for cocaine [27] and cannabis [28] but to reduce prevalence for alcohol [29]. Note that the 34% reduction mistakenly claimed by Grant et al. [30] arose from calculating the percent of those with 12 month dependence who did not report 12 month abuse when the CIDI 3.0 skip was for lifetime abuse. The consequence of the skip is that dependence on alcohol will have been underestimated in New Zealand, as discussed previously [31]. The second interview difference is that in the standard CIDI 3.0, as used in New Zealand, participants were screened into the abuse questions only if they responded positively to either of three abuse type questions about alcohol or drugs. In contrast, in Australia those who had 12 drinks of alcohol in the past 12 months or who had used a type of drug more than five times were asked abuse questions. I think that the screening is weaker in the Australian interview so I would expect more to enter the abuse section and perhaps more to be diagnosed with abuse. The Australian interview prohibits investigation of successful abstinence after previous dependence; the New Zealand interview may particularly underestimate dependence for women. The reported lifetime prevalence of alcohol abuse for Australia is 18.9% whereas that for New Zealand is 13.2%. Overall, the reported prevalences of any substance disorder are 24.7% in Australia and 12.3% in New Zealand, with 12 month prevalences of 5.1% and 3.5%.
These considerations point out the problems of comparisons from tables of results without paying attention to the diagnoses used and whether or not there were changes to the interview. Even surveys that are as alike as the 2007 NSMHWB and the NZMHS need to be compared with great care, and if only published results are used the conclusions may be that it is not clear if there are real differences. Comparisons between other countries within the World Mental Health Surveys Initiative have all so far been made using DSM-IV diagnoses, which has greatly simplified comparisons [23, 24, 32].
In summary I am agnostic but sceptical that Australia has much higher prevalence of PTSD and panic disorder than New Zealand but I think that there may be somewhat less depression and possibly more substance disorder.
Comorbidity and disability
Given the qualification made about diagnostic differences, it is not appropriate to discuss in detail the specific percentages with comorbidity between disorder groups. Both surveys show that comorbidity is common both between individual mental disorders and between mental disorder classes [7–9, 16]. Furthermore there is comorbidity between physical and mental disorders even though mental disorders are more common in the young and physical disorders in the old. Disorder and disorder comorbidity are associated with disability.
These papers describe the observed patterns and point out the implications for services because of this comorbidity but, as expected in the first set of descriptive papers, they do not model these relationships between mental disorders to look for underlying factors or classes to account for the observed patterns. Slade and Watson, applying confirmatory factor analysis to data from the 1997 NSMHWB, replicated earlier findings of three major factors of distress, fear and externalizing [33]. A different approach was taken by Kessler et al., who used latent class analysis to cope with non-linear relationships, and were able to see not only the major groupings found by Slade and others but also more specific relationships such as the co-occurrence of mania/hypomania and depression, which characterizes bipolar disorder [22].
Health service use
The 1997 and 2007 surveys in Australia asked about perceived need for mental health care [14], providing information that complements data on actual use of health services and inferences about required care based on disorder prevalences. No data on perceived need was collected in New Zealand so no comparisons can be made.
Burgess et al. report on the service use for mental health problems in the previous 12 months [12]. Overall 11.9% (11.0–12.8%) of Australians made treatment contact compared with 13.4% (12.7–14.2%) in New Zealand [5]. Although comparisons conditional on disorder are complicated by the diagnosis issues previously discussed, it is worth noting that the percentage making treatment contact is slightly lower in Australia both for any disorder (34.9% in Australia, 38.9% in New Zealand) and for no disorder (6.1% vs 7.2%).
The Australian interview more cleanly distinguished between community and hospital care than the New Zealand interview, which measured hospitalizations but did not record where contact with individual types of practitioners took place. Also New Zealand has not reported data specifically about psychologists, whereas the 2007 NSMHWB has been able to see the increase in consultation with psychologists since changes in their health system that permit such consultation through Medicare.
Suicidal thoughts and behaviours
Comparisons of the prevalences for suicidal ideation, plan and attempt are straightforward because the same questions were asked and there are no diagnostic differences to deal with. Consequently it seems safe to present a table of comparisons that can stand alone, without major qualifications being made in the text.
CI, confidence interval; NSMHWB, National Survey of Mental Health and Wellbeing; NZMHS, New Zealand Mental Health Survey.
Thoughts on ways of carrying out, analysing and reporting surveys
National surveys require considerable investment of money and time. In Australia and New Zealand funding for such surveys has come from central government (the Australian Government Department of Health and Aging and the New Zealand Ministry of Health, with some supplementary funding in New Zealand from the Health Research Council and the Alcohol Advisory Council). A field provider is required with national coverage. In Australia the Australian Bureau of Statistics (ABS) set up the survey with an Advisory Board, carried out the field work and holds the original dataset. In New Zealand the contract was won by a national research team, and a private provider (the National Research Bureau) with previous experience of national surveys was the field provider. Funders require reports. In Australia the initial report came from the ABS in the standard form of a government report with many tables and explanatory notes but no interpretation [25]. In New Zealand the research team had exclusive access to the data until mid 2007 and could carry out whatever analyses they wanted, although they discussed plans with the Ministry of Health Mental Health Directorate. The report [36], nonetheless, was required to be policy neutral, under an agreement whereby Statistics New Zealand allowed other Ministries to carry out surveys, subject to quality requirements.
The consequence has been that the New Zealand research team carried out their own analyses with full access to the data whereas the Australian researchers have so far been restricted to tables produced by the Australian Bureau of Statistics (ABS). The ABS has not yet released confidentialized (i.e. de-identified) unit record files (curfs), as the Australian researchers have lamented. Furthermore, even when the curfs are available it is possible that the ABS will have suppressed some of the fine detail required by the researchers to answer important questions, such as relative ages of onset for different disorders. In contrast, the New Zealand team had to restrain themselves from policy comment in their report to the Ministry of Health (although they did place the findings in the context of the literature), with the consequence that when writing the papers almost simultaneously with the report they did not focus on policy implications.
I hope that the Australian researchers will soon be able to access curfs for the more detailed analyses they are keen to do, and that funding will be forthcoming for this purpose. I also hope that this dataset will become available for use within the World Mental Health Surveys Initiative so that Australian data can join that cross-national collaboration.