CanJPsychiatry 2014;59(11):591–596
Original Research
Changing Perceptions of Mental Health in Canada Scott B Patten, MD, PhD1; Jeanne V A Williams, MSc2; Dina H Lavorato, MSc2; Kirsten M Fiest, PhD3; Andrew G M Bulloch, PhD4; JianLi Wang, PhD5 1
Professor, Department of Community Health Sciences, University of Calgary, Calgary, Alberta; Member, Mathison Centre for Mental Health Research and Education, Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta. Correspondence: Department of Community Health Sciences, 3rd Floor TRW Building, University of Calgary, 3280 Hospital Drive NW, Calgary, AB T2N 4Z6; [email protected].
2
Research Associate, Department of Community Health Sciences, University of Calgary, Calgary, Alberta.
3
Postdoctoral Fellow, Department of Clinical Neurosciences, University of Calgary, Calgary, Alberta.
4
Professor, Department of Community Health Sciences, University of Calgary, Calgary, Alberta; Interim Director, Mathison Centre for Mental Health Research and Education, Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta.
5
Associate Professor, Department of Psychiatry, University of Calgary, Calgary, Alberta; Member, Mathison Centre for Mental Health Research and Education, Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta.
Key Words: perceived mental health, general health, epidemiologic studies, distress, cross-sectional studies, secular trends Received February 2014, revised, and accepted May 2014.
Objective: Epidemiologic studies typically assess mental health using diagnostic measures or symptom severity measures. However, perceptions are also important. The objective of our study was to evaluate trends in perceived mental health in Canada during the past 20 years using data collected in a series of surveys. Method: Perceived mental health status, the stressfulness of most days, and perceived general health, have been repeatedly measured in national surveys. In our study, the resulting frequencies and 95% confidence intervals were calculated. Distress was also assessed in the same surveys with the Kessler 6 Psychological Distress Scale, and analyzed using mean scores and frequencies based on cut-points. Data synthesis used forest plots. Time trends were assessed using random effects meta-regression models. Results: No detectable changes in distress were found. Similarly, self-rated general health remained stable. However, over time, Canadians became slightly more likely to report that their mental health was merely fair or poor. Conversely, they have been progressively less likely to perceive that their lives are quite a bit or extremely stressful. Conclusion: While these observations are ecological, the 2 trends may be related: distressing emotional experiences may increasingly be interpreted as evidence of a disturbance of mental health rather than a reaction to stressful circumstances. These changing perceptions should not be misinterpreted as an epidemic of poor mental health. WWW
Changer les perceptions de la maladie mentale au Canada Objectif : Les études épidémiologiques évaluent habituellement la santé mentale à l’aide de mesures diagnostiques ou de la gravité des symptômes. Cependant, les perceptions sont également importantes. L’objectif de notre étude était d’évaluer les tendances de la santé mentale perçue au Canada au cours des 20 dernières années, au moyen de données recueillies dans une série d’enquêtes. Méthode : L’état perçu de santé mentale, le caractère stressant de la plupart des journées, et la santé générale perçue ont été mesurés à répétition dans les enquêtes nationales. Dans notre étude, les fréquences qui en ont résulté et les intervalles de confiance à 95 % ont été calculés. La détresse a aussi été évaluée dans ces mêmes enquêtes avec l’Échelle de détresse psychologique de Kessler-6, et analysée à l’aide des scores moyens et des fréquences en fonction des seuils d’inclusion. La synthèse des données utilisait des graphiques en forêt. Les tendances chronologiques ont été évaluées à l’aide de modèles de méta-régression d’effets aléatoires. Résultats : Aucun changement détectable de détresse n’a été constaté. De même, la santé générale auto-déclarée est demeurée stable. Toutefois, avec le temps, les Canadiens ont été légèrement plus susceptibles de percevoir que leur vie est assez ou extrêmement stressante. www.TheCJP.ca
The Canadian Journal of Psychiatry, Vol 59, No 11, November 2014 W 591
Original Research
Conclusion : Bien que ces observations soient écologiques, les 2 tendances peuvent être reliées : les expériences de détresse émotionnelle peuvent être interprétées de plus en plus comme une preuve de perturbation de la santé mentale plutôt que comme une réaction à des circonstances stressantes. Ces perceptions changeantes ne doivent pas être interprétées à tort comme une épidémie de mauvaise santé mentale.
A
ccording to the World Health Organization, mental health is characterized by a “state of well-being in which an individual realizes his or her own abilities, (and) can cope with the normal stresses of life.”1 This definition illustrates Wakefield’s2 assertion that mental health and illness exist on a boundary between biological facts and social values. Social values and scientific knowledge change over time and with them the boundaries of normality are likely to shift.3,4 Apart from a single study,5 there has been little examination of whether perceptions of mental health, stress, and distress are occurring in Canada. In our study, we make use of recently available data and employ contemporary tools for data synthesis to examine trends in perceived mental health, stress, and distress.
Methods
The NPHS and CCHS are national health surveys targeting the Canadian household population. The NPHS was a longitudinal study that collected data between 1994 and 2010. Cross-sectional data files are available for the 1994, 1996, and 1998 data collection cycles.6 The CCHS7 program collected data initially in 2000 and thereafter every 2 years until 2007 at which time an annual data collection cycle was initiated. Two-year data files have remained available, however, and were used in our analysis. These 2-year files are labelled using the year in which the data collection was initiated, for example, the data from 2007 and 2008 are labelled 2007. All of these studies can support inference to the Canadian household population when the data are analyzed using appropriate sampling weights and bootstrap variance estimation procedures. An item assessing perceived mental health has been included in several of these surveys: “In general, would you say your mental health is: excellent?, very good?, good?, fair?, or poor?” The frequency of fair or poor mental health status was defined a priori as the primary measure of perceived mental health in our study. At the level of face validity, these categories reflect a perceived problem with mental health. Other distinctions, for example, those between excellent and very good, are more difficult to interpret and Abbreviations CCHS
Canadian Community Health Survey
CCHS 1.2
Canadian Community Health Survey: mental health and well-being
K6
Kessler 6 Psychological Distress Scale
NPHS
National Population Health Survey
592 W La Revue canadienne de psychiatrie, vol 59, no 11, novembre 2014
Clinical Implications •
Canadians increasingly perceive their mental health as fair or poor.
•
These changes may reflect improving mental health literacy.
•
Changing beliefs about the cause of psychiatric symptoms may be altering the ways in which people describe and respond to symptoms of distress.
Limitations •
Data for our study were derived from self-report, making it impossible to disentangle changes in etiologic attribution from actual health changes.
•
The perceptions examined in our study were assessed using single items rather than detailed measurement instruments.
•
Regional differences were not examined.
are perhaps of less relevance to psychiatry and psychiatric epidemiology. For descriptive purposes, the frequencies of each response option were also examined. Perceived stress was evaluated using an item with the following wording: “Thinking about the amount of stress in your life, would you say that most days are: not at all stressful?, not very stressful?, a bit stressful?, quite a bit stressful? or extremely stressful?” We estimate the proportion reporting quite a bit or extremely stressful, again for reasons of face validity, in the primary analysis. The uncategorized item responses were also tabulated. For comparative purposes, a similar item for general health status was included in the analysis, and was treated similarly in the analysis. Notably, the survey participants were not provided with definitions of any of these concepts. As such, the responses reflect their perceptions of the targeted issues, including their ways of defining what is meant by them. The surveys incorporated a widely used nonspecific distress measure called the K6.8 The frequency of elevated distress (score of 13 or more) on the K6 and mean K6 scores were estimated, the former being suggestive of severe distress likely to be indicative of a disorder.8 Estimates were summarized using forest plots. We examined heterogeneity using the I2 statistic, which is the percentage of variance that can be explained by heterogeneity. Metaregression was used to quantify change over time. Time was represented as years since first assessment of an outcome variable so that the intercept term in these regression models represented a 1994 baseline and the slope term represented change per subsequent year. In addition to time, www.LaRCP.ca
Changing Perceptions of Mental Health in Canada
Figure 1 Prevalence of fair or poor mental health in Canada, 2002 to 2012a Prevalence of fair or poor mental health, %
we explored the impact of including indicator variables for survey group (NPHS, CCHS general health, and CCHS mental health surveys). As recommended by Higgins and Thompson9 for protection against type 1 error in metaregression, permutation tests (with n = 1000 Monte Carlo trials) were used to assess P values in the meta-regression models. The analyses used Stata’s10 metan command and were conducted in the Prairie Regional Data Centre in Calgary. The University of Calgary Ethics Review Board waived ethical approval in view of the nature of the study.
Results
Table 1 summarizes the number of available observations on each of the outcome variables in each survey cycle. The analyses reported are based on between 354 045 (distress) and 843 636 (perceived general health) respondents.
We conducted a similar analysis looking at the prevalence of fair or poor perceived general health, available from the same 7 surveys and also for the 2000 CCHS. The frequency of reporting fair or poor general health was generally higher than that for mental health, but with a smaller range of variation and the frequency did not change over time. All of the estimates fell between 11.1% in 2002 and 2012 and 12.0% in 2000. In a meta-regression model the effect of the survey group was not significant (β for general health surveys = 0.002; t = 1.14, df = 4, P = 0.32) nor was the effect of time (β = –0.00001; t = 0.09, df = 4, P = 0.93). The proportion of the population reporting that most days of their life were quite a bit or extremely stressful could be estimated from 8 surveys starting with the 2000 CCHS. The observed pattern was that of diminishing prevalence (Figure 2). In the meta-regression analysis there was no significant effect of the survey group (β for general health surveys with mental health surveys as a baseline category = www.TheCJP.ca
8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0
Year a
The dashed lines are fitted values from a meta-regression model.
Figure 2 Prevalence of quite a bit or extremely stressful lives in Canada, 2000 to 2012a 29.0 Prevalence of high stress, %
Perceived mental health was included in 7 studies starting in 2002 and most recently in 2012. The lowest prevalence of fair or poor mental health was in the general health 2003 CCHS survey (4.7%, 95% CI 4.5% to 4.8%). The 2 highest estimates were from the 2 mental health surveys (the 2002 and 2012 CCHS survey, which focused on mental health). The 2002 estimate was 6.9% (95% CI 6.3% to 7.2%) and the 2012 estimate was 7.8% (95% CI 7.2% to 8.3%). Figure 1 presents the pattern of change over time. In the meta-regression analysis, the effect of the survey group (mental health or general health) was significant with the beta coefficient for general (compared with mental) health surveys being –0.023 (t = 12.42, df = 4, P < 0.001) and that for time was 0.001 (t = 5.51, df = 4, P = 0.005). The latter parameter indicates a 0.1% increase in the prevalence of fair or poor mental health with each passing year, leading to an about 1% increase over the span of these studies. Inclusion of these 2 covariates in a meta-regression model decreased the I2 from 97.6% in an overall pooling to 44.3%. Application of the permutation test did not meaningfully alter the observed P values for the survey group (P < 0.001) or time (P = 0.004).
9.0
27.0 25.0 23.0 21.0 19.0 17.0 15.0
Year of survey a
The dashed lines are fitted values from a meta-regression model. High stress refers to self-reported stress levels of quite a bit or extremely stressful.
0.15; t = 2.28, df = 5, P = 0.07), but the effect of time was significant (β = –0.003; t = 4.32, df = 5, P = 0.008). There was considerable unexplained heterogeneity in the model. The overall I2 was 97.7% (τ2 = 0.0002). With adjustment for time the I2 diminished to 92.5% (τ2 = 0.0001). The P value for time deriving from the permutation test was less than 0.001. An important question is whether these changes in perceived mental health have occurred along with actual changes in experiences of mental distress. We initially examined the prevalence of elevated distress using the K6 cut-point of 13 or more. There were 11 available estimates of this parameter between 1994 and 2012, ranging between a low of 1.6% in the 2009 CCHS to a high of 3.0% in the 2000 CCHS. Metaregression with the study group variables identified no The Canadian Journal of Psychiatry, Vol 59, No 11, November 2014 W 593
a
Original Research
Table 1 Sample size availability from the National Population Health Survey (NPHS) and the Canadian Community Health Survey (CCHS) cycles Perceived mental health
Perceived general health
Perceived stress
Distress
NPHS 1994
n/a
n/a
n/a
16 694
NPHS 1996
n/a
n/a
n/a
70 032
NPHS 1998
n/a
n/a
n/a
14 723
CCHS 2000
n/a
131 476
118 105
n/a
CCHS 2002a
36 960
36 973
36 969
36 840
CCHS 2003
135 442
135 422
127 481
12 416
CCHS 2005
130 555
132 782
126 112
63 040
CCHS 2007
128 751
131 748
125 620
28 042
CCSH 2009
122 174
124 725
124 390
41 096
CCHS 2011
122 571
125 405
124 821
46 219
25 088
25 105
25 086
24 943
Survey
CCHS 2012 a
a
Mental health and well-being surveys6,7
n/a = not applicable
significant effects, either for the study group (all P > 0.44) or time (β = –0.00008; t = 0.36, df = 9, P = 0.73). The pooled estimate from random effects meta-analysis was 2.1% (95% CI 1.9% to 2.3%).
disclose symptoms, potentially offsetting expected gains in mental health status owing to improved services. Such an effect is also a possible interpretation of the Canadian data. Unchanging levels of distress may also represent a lack of success in addressing mental health problems in the population, or there may have been a lack of power to detect favourable changes. The latter possibility seems unlikely as small changes in other variables over time did achieve significance.
The analysis of mean distress ratings included 11 available estimates. Considerable heterogeneity was observed in the estimated means (I2 > 99%), with the estimated mean distress ratings varying from a low of 1.9 in 2003 to 3.5 in 1994. Time was not significant in meta-regression models, irrespective of whether these included a survey group (β = –0.03; t = 0.72, df =7, P = 0.49) or not (β = –0.02, t = 1.1; df = 9, P = 0.32). In the absence of a survey group or time effect, the pooled mean may be regarded as the best estimate of mean distress. This pooled mean was 2.9 (95% CI 2.7 to 3.1). The moment-based estimate of betweenstudy variance (τ2) was 0.12.
The idea that perceptions of mental health and illness are fluid rather than static over time is supported by population surveys conducted in Germany, 20 years apart.12 In Angermeyer et al,12 perceptions of etiology, effectiveness of different treatment modalities, attitudes, and desired social distance often changed, but in ways that were not consistent across diagnostic categories.
To examine the broader set of responses to the perceived health and stress items, the response frequencies were also tabulated in their original categories. The frequencies of self-reported mental health and perceived stress are depicted in Figure 3 and Figure 4 using data from the first and last general health surveys assessing these variables. The figures indicate that most of the population, as expected, do not report poor mental health, nor do they report high levels of stress. For comparison, online eFigure 5 presents response frequencies for general health. There is no increased frequency of fair or poor general health in the 2011 data as compared with the 2000 CCHS 1.1.
Our study confirms a preliminary observation5 of a trend toward more negative perceptions of mental health and diminished perceptions of stress. This change may mean that Canadians’ mental health really is deteriorating and (or) that the level of stress in their lives is actually diminishing. However, another interpretation is that distressing experiences, which appear not to be changing, are increasingly being attributed to mental health issues as opposed to stressful events. Of course, this conclusion is speculative as the data are ecological: the data do not identify changes occurring within individual people, only trends in independent samples over time.
Discussion
These results are consistent with Australian data in showing no change in distress over time in that country,11 and add to these results by including more time points and greater methodological consistency of the data sources. The authors of the Australian work speculated that changing mental health literacy may have created a greater willingness to 594 W La Revue canadienne de psychiatrie, vol 59, no 11, novembre 2014
The target population for the most recent CCHS consists of about 30 million Canadians. This provides some indication of the magnitude of the effects observed here. The 1% change in perceptions that mental health is fair or poor reflects about 300 000 more Canadians placing themselves in this category than would have been expected had the baseline frequencies persisted. The 4% decrease in the proportion reporting high stress would translate into more www.LaRCP.ca
Changing Perceptions of Mental Health in Canada
Figure 3 Self-reported mental health, 2003 and 2012a 45 40
Population , %
35 30 25 20 15 10 5 0 Excellent
Very good
2003 a
Good
Fair
Poor
2011
The first and last available surveys recording this information.
Figure 4 Proportions reporting their lives to be quite a bit or extremely stressful, 2000 to 2012a 45 40 35
Population, %
30 25 20 15 10 5 0
Not at all stressful
Not very stressful
2000 a
Quite a bit stressful
Extremely stressful
2011
The first and last available surveys recording this information.
than 1 million fewer people in the high stress category than would have been expected had the baseline frequencies persisted. In our analysis, we observed a tendency for more negative ratings of mental health to occur in surveys targeting mental health than in general health surveys. It is plausible that the observed difference relates to the placement of the item in the survey questionnaires. In the 2 mental health and wellbeing surveys, this item was included in the screening section for the Composite International Diagnostic Interview,13 such that respondents may have been primed by a series of preceding mental health questions. Another possibility is that simply knowing that a survey is about mental health may have caused respondents to alter their perceptions of their own mental health. In the general health iterations of www.TheCJP.ca
A bit stressful
the CCHS, the perceived mental health item was included in a general health module that was administered near the start of the interview. Surprisingly, even greater heterogeneity was observed with distress ratings. This heterogeneity may reflect an impact of social conditions, such as economic or employment security, which may change in a nonlinear fashion over time. In general, the extent of heterogeneity observed was large, despite the shared items and methodological similarities of the various surveys. This suggests a need for caution when comparing results from different surveys and highlights a need for replication of such results. There are evidently unknown determinants of differences in these estimates, which could produce misleading impressions of temporal trends. The Canadian Journal of Psychiatry, Vol 59, No 11, November 2014 W 595
Original Research
An advantage of pooling at the study level, as was done in our study, rather than at the individual level (as in a pooled analysis), is that the estimates can be appropriately weighted and standard errors can be accurately estimated using survey-specific techniques, such as the use of sampling weights and bootstrap variance estimation. An implication of this approach is that the sampling weights employed in each survey support inference to the household population in the year of each survey. As such, changes in the demographic structure of the population over time could influence the temporal trends observed. A methodological concern with the meta-analysis of frequency estimates is the possible need for variance stabilizing transformations, as meta-analytic weights are inversely proportional to the variance of estimated proportions, whereas the variance is expected to be smaller for smaller proportions. As smaller proportions may therefore be given greater weight, bias could occur. Appropriate transformations require specification of a numerator and denominator for each proportion and are therefore inconsistent with recommended populationweighting and variance estimation procedures. We have explored the impact of variance stabilizing transformations (such as the double arcsine transformation) in unweighted data using the metafor package in the software R. In instances where we have applied these approaches, they have led to very similar results to the more standard methods employed in the reported analyses. The main limitation of our study pertains to the limitations of the underlying data sources. As the included studies were large-scale health surveys, they tend to include brief items and scales rather than more detailed research interviews and schedules. Perceived mental health and perceived stress were measured only with single items of unknown validity (apart from face validity). Also, the K6 is a fairly rudimentary 6-item screening scale and is by no means a comprehensive assessment of mental health status.
Acknowledgements
Dr Patten is a Senior Health Scholar with Alberta Innovates, Health Solutions (AIHS). At the time of this work, Kirsten Fiest was supported by a PhD Studentship from AIHS.
596 W La Revue canadienne de psychiatrie, vol 59, no 11, novembre 2014
The estimates reported in our paper used data collected by Statistics Canada, but the analysis and results are the sole responsibility of the authors and do not reflect the views of Statistics Canada. This work was supported by an operating grant from the Canadian Institutes of Health Research (MOP-130415).
References
1. World Health Organization (WHO). Fact sheet 220. Mental health: strengthening our response [Internet]. Geneva (CH): WHO; 2014 [cited 2014 May 21]. Available from: http://www.who.int/mediacentre/factsheets/fs220/en/. 2. Wakefield JC. The concept of mental disorder. On the boundary between biological facts and social values. Am Psychol. 1992;47:373–388. 3. Wakefield JC, First MB. Clarifying the boundary between normality and disorder: a fundamental conceptual challenge for psychiatry. Can J Psychiatry. 2013;58(11):603–605. 4. Wakefield JC, First MB. The importance and limits of harm in identifying mental disorder. Can J Psychiatry. 2013;58(11):618–621. 5. Simpson KR, Meadows GN, Frances AJ, et al. Is mental health in the Canadian population changing over time? Can J Psychiatry. 2012;57(5):324–331. 6. Statistics Canada. National Population Health Survey—household component—cross-sectional (NPHS). Ottawa (ON): Statistics Canada; 2007 [cited 2013 Dec 30]. Available from: http://www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSurvey& SDDS=3236&lang=en&db=imdb&adm=8&dis=2. 7. Statistics Canada. Canadian Community Health Survey—annual component [Internet]. Ottawa (ON): Statistics Canada; 2013 [cited 2013 Dec 30]. Available from: http://www23.statcan.gc.ca/ imdb/p2SV.pl?Function=getSurvey&SDDS=3226&lang=en& db=imdb&adm=8&dis=2. 8. Kessler RC, Green JG, Gruber MJ, et al. Screening for serious mental illness in the general population with the K6 screening scale: results from the WHO World Mental Health (WMH) survey initiative. Int J Methods Psychiatr Res. 2010;19(Suppl 1):4–22. 9. Higgins JP, Thompson SG. Controlling the risk of spurious findings from meta-regression. Stat Med. 2004;23(11):1663–1682. 10. StataCorp. Stata user’s guide: version 12.1. College Station (TX): Stata Corporation; 2012. 11. Jorm AF, Reavley NJ. Changes in psychological distress in Australian adults between 1995 and 2011. Aust N Z J Psychiatry. 2012;46(4):352–356. 12. Angermeyer MC, Matschinger H, Schomerus G. Attitudes of the German public to restrictions on persons with mental illness in 1993 and 2011. Epidemiol Psychiatr Sci. 2014;23(3):263–270. 13. Kessler RC, Ustun TB. The World Mental Health (WMH) Survey Initiative Version of the World Health Organization (WHO) Composite International Diagnostic Interview (CIDI). Int J Methods Psychiatr Res. 2004;13:83–121.
www.LaRCP.ca
Original Research
Changing Perceptions of Mental Health in Canada Scott B Patten, MD, PhD1; Jeanne V A Williams, MSc2; Dina H Lavorato, MSc2; Kirsten M Fiest, PhD3; Andrew G M Bulloch, PhD4; JianLi Wang, PhD5 1
Professor, Department of Community Health Sciences, University of Calgary, Calgary, Alberta; Member, Mathison Centre for Mental Health Research and Education, Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta. Correspondence: Department of Community Health Sciences, 3rd Floor TRW Building, University of Calgary, 3280 Hospital Drive NW, Calgary, AB T2N 4Z6; [email protected].
2
Research Associate, Department of Community Health Sciences, University of Calgary, Calgary, Alberta.
3
Postdoctoral Fellow, Department of Clinical Neurosciences, University of Calgary, Calgary, Alberta.
4
Professor, Department of Community Health Sciences, University of Calgary, Calgary, Alberta; Interim Director, Mathison Centre for Mental Health Research and Education, Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta.
5
Associate Professor, Department of Psychiatry, University of Calgary, Calgary, Alberta; Member, Mathison Centre for Mental Health Research and Education, Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta.
Key Words: perceived mental health, general health, epidemiologic studies, distress, cross-sectional studies, secular trends Received February 2014, revised, and accepted May 2014.
Objective: Epidemiologic studies typically assess mental health using diagnostic measures or symptom severity measures. However, perceptions are also important. The objective of our study was to evaluate trends in perceived mental health in Canada during the past 20 years using data collected in a series of surveys. Method: Perceived mental health status, the stressfulness of most days, and perceived general health, have been repeatedly measured in national surveys. In our study, the resulting frequencies and 95% confidence intervals were calculated. Distress was also assessed in the same surveys with the Kessler 6 Psychological Distress Scale, and analyzed using mean scores and frequencies based on cut-points. Data synthesis used forest plots. Time trends were assessed using random effects meta-regression models. Results: No detectable changes in distress were found. Similarly, self-rated general health remained stable. However, over time, Canadians became slightly more likely to report that their mental health was merely fair or poor. Conversely, they have been progressively less likely to perceive that their lives are quite a bit or extremely stressful. Conclusion: While these observations are ecological, the 2 trends may be related: distressing emotional experiences may increasingly be interpreted as evidence of a disturbance of mental health rather than a reaction to stressful circumstances. These changing perceptions should not be misinterpreted as an epidemic of poor mental health. WWW
Changer les perceptions de la maladie mentale au Canada Objectif : Les études épidémiologiques évaluent habituellement la santé mentale à l’aide de mesures diagnostiques ou de la gravité des symptômes. Cependant, les perceptions sont également importantes. L’objectif de notre étude était d’évaluer les tendances de la santé mentale perçue au Canada au cours des 20 dernières années, au moyen de données recueillies dans une série d’enquêtes. Méthode : L’état perçu de santé mentale, le caractère stressant de la plupart des journées, et la santé générale perçue ont été mesurés à répétition dans les enquêtes nationales. Dans notre étude, les fréquences qui en ont résulté et les intervalles de confiance à 95 % ont été calculés. La détresse a aussi été évaluée dans ces mêmes enquêtes avec l’Échelle de détresse psychologique de Kessler-6, et analysée à l’aide des scores moyens et des fréquences en fonction des seuils d’inclusion. La synthèse des données utilisait des graphiques en forêt. Les tendances chronologiques ont été évaluées à l’aide de modèles de méta-régression d’effets aléatoires. Résultats : Aucun changement détectable de détresse n’a été constaté. De même, la santé générale auto-déclarée est demeurée stable. Toutefois, avec le temps, les Canadiens ont été légèrement plus susceptibles de percevoir que leur vie est assez ou extrêmement stressante. www.TheCJP.ca
The Canadian Journal of Psychiatry, Vol 59, No 11, November 2014 W 591
Original Research
Conclusion : Bien que ces observations soient écologiques, les 2 tendances peuvent être reliées : les expériences de détresse émotionnelle peuvent être interprétées de plus en plus comme une preuve de perturbation de la santé mentale plutôt que comme une réaction à des circonstances stressantes. Ces perceptions changeantes ne doivent pas être interprétées à tort comme une épidémie de mauvaise santé mentale.
A
ccording to the World Health Organization, mental health is characterized by a “state of well-being in which an individual realizes his or her own abilities, (and) can cope with the normal stresses of life.”1 This definition illustrates Wakefield’s2 assertion that mental health and illness exist on a boundary between biological facts and social values. Social values and scientific knowledge change over time and with them the boundaries of normality are likely to shift.3,4 Apart from a single study,5 there has been little examination of whether perceptions of mental health, stress, and distress are occurring in Canada. In our study, we make use of recently available data and employ contemporary tools for data synthesis to examine trends in perceived mental health, stress, and distress.
Methods
The NPHS and CCHS are national health surveys targeting the Canadian household population. The NPHS was a longitudinal study that collected data between 1994 and 2010. Cross-sectional data files are available for the 1994, 1996, and 1998 data collection cycles.6 The CCHS7 program collected data initially in 2000 and thereafter every 2 years until 2007 at which time an annual data collection cycle was initiated. Two-year data files have remained available, however, and were used in our analysis. These 2-year files are labelled using the year in which the data collection was initiated, for example, the data from 2007 and 2008 are labelled 2007. All of these studies can support inference to the Canadian household population when the data are analyzed using appropriate sampling weights and bootstrap variance estimation procedures. An item assessing perceived mental health has been included in several of these surveys: “In general, would you say your mental health is: excellent?, very good?, good?, fair?, or poor?” The frequency of fair or poor mental health status was defined a priori as the primary measure of perceived mental health in our study. At the level of face validity, these categories reflect a perceived problem with mental health. Other distinctions, for example, those between excellent and very good, are more difficult to interpret and Abbreviations CCHS
Canadian Community Health Survey
CCHS 1.2
Canadian Community Health Survey: mental health and well-being
K6
Kessler 6 Psychological Distress Scale
NPHS
National Population Health Survey
592 W La Revue canadienne de psychiatrie, vol 59, no 11, novembre 2014
Clinical Implications •
Canadians increasingly perceive their mental health as fair or poor.
•
These changes may reflect improving mental health literacy.
•
Changing beliefs about the cause of psychiatric symptoms may be altering the ways in which people describe and respond to symptoms of distress.
Limitations •
Data for our study were derived from self-report, making it impossible to disentangle changes in etiologic attribution from actual health changes.
•
The perceptions examined in our study were assessed using single items rather than detailed measurement instruments.
•
Regional differences were not examined.
are perhaps of less relevance to psychiatry and psychiatric epidemiology. For descriptive purposes, the frequencies of each response option were also examined. Perceived stress was evaluated using an item with the following wording: “Thinking about the amount of stress in your life, would you say that most days are: not at all stressful?, not very stressful?, a bit stressful?, quite a bit stressful? or extremely stressful?” We estimate the proportion reporting quite a bit or extremely stressful, again for reasons of face validity, in the primary analysis. The uncategorized item responses were also tabulated. For comparative purposes, a similar item for general health status was included in the analysis, and was treated similarly in the analysis. Notably, the survey participants were not provided with definitions of any of these concepts. As such, the responses reflect their perceptions of the targeted issues, including their ways of defining what is meant by them. The surveys incorporated a widely used nonspecific distress measure called the K6.8 The frequency of elevated distress (score of 13 or more) on the K6 and mean K6 scores were estimated, the former being suggestive of severe distress likely to be indicative of a disorder.8 Estimates were summarized using forest plots. We examined heterogeneity using the I2 statistic, which is the percentage of variance that can be explained by heterogeneity. Metaregression was used to quantify change over time. Time was represented as years since first assessment of an outcome variable so that the intercept term in these regression models represented a 1994 baseline and the slope term represented change per subsequent year. In addition to time, www.LaRCP.ca
Changing Perceptions of Mental Health in Canada
Figure 1 Prevalence of fair or poor mental health in Canada, 2002 to 2012a Prevalence of fair or poor mental health, %
we explored the impact of including indicator variables for survey group (NPHS, CCHS general health, and CCHS mental health surveys). As recommended by Higgins and Thompson9 for protection against type 1 error in metaregression, permutation tests (with n = 1000 Monte Carlo trials) were used to assess P values in the meta-regression models. The analyses used Stata’s10 metan command and were conducted in the Prairie Regional Data Centre in Calgary. The University of Calgary Ethics Review Board waived ethical approval in view of the nature of the study.
Results
Table 1 summarizes the number of available observations on each of the outcome variables in each survey cycle. The analyses reported are based on between 354 045 (distress) and 843 636 (perceived general health) respondents.
We conducted a similar analysis looking at the prevalence of fair or poor perceived general health, available from the same 7 surveys and also for the 2000 CCHS. The frequency of reporting fair or poor general health was generally higher than that for mental health, but with a smaller range of variation and the frequency did not change over time. All of the estimates fell between 11.1% in 2002 and 2012 and 12.0% in 2000. In a meta-regression model the effect of the survey group was not significant (β for general health surveys = 0.002; t = 1.14, df = 4, P = 0.32) nor was the effect of time (β = –0.00001; t = 0.09, df = 4, P = 0.93). The proportion of the population reporting that most days of their life were quite a bit or extremely stressful could be estimated from 8 surveys starting with the 2000 CCHS. The observed pattern was that of diminishing prevalence (Figure 2). In the meta-regression analysis there was no significant effect of the survey group (β for general health surveys with mental health surveys as a baseline category = www.TheCJP.ca
8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0
Year a
The dashed lines are fitted values from a meta-regression model.
Figure 2 Prevalence of quite a bit or extremely stressful lives in Canada, 2000 to 2012a 29.0 Prevalence of high stress, %
Perceived mental health was included in 7 studies starting in 2002 and most recently in 2012. The lowest prevalence of fair or poor mental health was in the general health 2003 CCHS survey (4.7%, 95% CI 4.5% to 4.8%). The 2 highest estimates were from the 2 mental health surveys (the 2002 and 2012 CCHS survey, which focused on mental health). The 2002 estimate was 6.9% (95% CI 6.3% to 7.2%) and the 2012 estimate was 7.8% (95% CI 7.2% to 8.3%). Figure 1 presents the pattern of change over time. In the meta-regression analysis, the effect of the survey group (mental health or general health) was significant with the beta coefficient for general (compared with mental) health surveys being –0.023 (t = 12.42, df = 4, P < 0.001) and that for time was 0.001 (t = 5.51, df = 4, P = 0.005). The latter parameter indicates a 0.1% increase in the prevalence of fair or poor mental health with each passing year, leading to an about 1% increase over the span of these studies. Inclusion of these 2 covariates in a meta-regression model decreased the I2 from 97.6% in an overall pooling to 44.3%. Application of the permutation test did not meaningfully alter the observed P values for the survey group (P < 0.001) or time (P = 0.004).
9.0
27.0 25.0 23.0 21.0 19.0 17.0 15.0
Year of survey a
The dashed lines are fitted values from a meta-regression model. High stress refers to self-reported stress levels of quite a bit or extremely stressful.
0.15; t = 2.28, df = 5, P = 0.07), but the effect of time was significant (β = –0.003; t = 4.32, df = 5, P = 0.008). There was considerable unexplained heterogeneity in the model. The overall I2 was 97.7% (τ2 = 0.0002). With adjustment for time the I2 diminished to 92.5% (τ2 = 0.0001). The P value for time deriving from the permutation test was less than 0.001. An important question is whether these changes in perceived mental health have occurred along with actual changes in experiences of mental distress. We initially examined the prevalence of elevated distress using the K6 cut-point of 13 or more. There were 11 available estimates of this parameter between 1994 and 2012, ranging between a low of 1.6% in the 2009 CCHS to a high of 3.0% in the 2000 CCHS. Metaregression with the study group variables identified no The Canadian Journal of Psychiatry, Vol 59, No 11, November 2014 W 593
a
Original Research
Table 1 Sample size availability from the National Population Health Survey (NPHS) and the Canadian Community Health Survey (CCHS) cycles Perceived mental health
Perceived general health
Perceived stress
Distress
NPHS 1994
n/a
n/a
n/a
16 694
NPHS 1996
n/a
n/a
n/a
70 032
NPHS 1998
n/a
n/a
n/a
14 723
CCHS 2000
n/a
131 476
118 105
n/a
CCHS 2002a
36 960
36 973
36 969
36 840
CCHS 2003
135 442
135 422
127 481
12 416
CCHS 2005
130 555
132 782
126 112
63 040
CCHS 2007
128 751
131 748
125 620
28 042
CCSH 2009
122 174
124 725
124 390
41 096
CCHS 2011
122 571
125 405
124 821
46 219
25 088
25 105
25 086
24 943
Survey
CCHS 2012 a
a
Mental health and well-being surveys6,7
n/a = not applicable
significant effects, either for the study group (all P > 0.44) or time (β = –0.00008; t = 0.36, df = 9, P = 0.73). The pooled estimate from random effects meta-analysis was 2.1% (95% CI 1.9% to 2.3%).
disclose symptoms, potentially offsetting expected gains in mental health status owing to improved services. Such an effect is also a possible interpretation of the Canadian data. Unchanging levels of distress may also represent a lack of success in addressing mental health problems in the population, or there may have been a lack of power to detect favourable changes. The latter possibility seems unlikely as small changes in other variables over time did achieve significance.
The analysis of mean distress ratings included 11 available estimates. Considerable heterogeneity was observed in the estimated means (I2 > 99%), with the estimated mean distress ratings varying from a low of 1.9 in 2003 to 3.5 in 1994. Time was not significant in meta-regression models, irrespective of whether these included a survey group (β = –0.03; t = 0.72, df =7, P = 0.49) or not (β = –0.02, t = 1.1; df = 9, P = 0.32). In the absence of a survey group or time effect, the pooled mean may be regarded as the best estimate of mean distress. This pooled mean was 2.9 (95% CI 2.7 to 3.1). The moment-based estimate of betweenstudy variance (τ2) was 0.12.
The idea that perceptions of mental health and illness are fluid rather than static over time is supported by population surveys conducted in Germany, 20 years apart.12 In Angermeyer et al,12 perceptions of etiology, effectiveness of different treatment modalities, attitudes, and desired social distance often changed, but in ways that were not consistent across diagnostic categories.
To examine the broader set of responses to the perceived health and stress items, the response frequencies were also tabulated in their original categories. The frequencies of self-reported mental health and perceived stress are depicted in Figure 3 and Figure 4 using data from the first and last general health surveys assessing these variables. The figures indicate that most of the population, as expected, do not report poor mental health, nor do they report high levels of stress. For comparison, online eFigure 5 presents response frequencies for general health. There is no increased frequency of fair or poor general health in the 2011 data as compared with the 2000 CCHS 1.1.
Our study confirms a preliminary observation5 of a trend toward more negative perceptions of mental health and diminished perceptions of stress. This change may mean that Canadians’ mental health really is deteriorating and (or) that the level of stress in their lives is actually diminishing. However, another interpretation is that distressing experiences, which appear not to be changing, are increasingly being attributed to mental health issues as opposed to stressful events. Of course, this conclusion is speculative as the data are ecological: the data do not identify changes occurring within individual people, only trends in independent samples over time.
Discussion
These results are consistent with Australian data in showing no change in distress over time in that country,11 and add to these results by including more time points and greater methodological consistency of the data sources. The authors of the Australian work speculated that changing mental health literacy may have created a greater willingness to 594 W La Revue canadienne de psychiatrie, vol 59, no 11, novembre 2014
The target population for the most recent CCHS consists of about 30 million Canadians. This provides some indication of the magnitude of the effects observed here. The 1% change in perceptions that mental health is fair or poor reflects about 300 000 more Canadians placing themselves in this category than would have been expected had the baseline frequencies persisted. The 4% decrease in the proportion reporting high stress would translate into more www.LaRCP.ca
Changing Perceptions of Mental Health in Canada
Figure 3 Self-reported mental health, 2003 and 2012a 45 40
Population , %
35 30 25 20 15 10 5 0 Excellent
Very good
2003 a
Good
Fair
Poor
2011
The first and last available surveys recording this information.
Figure 4 Proportions reporting their lives to be quite a bit or extremely stressful, 2000 to 2012a 45 40 35
Population, %
30 25 20 15 10 5 0
Not at all stressful
Not very stressful
2000 a
Quite a bit stressful
Extremely stressful
2011
The first and last available surveys recording this information.
than 1 million fewer people in the high stress category than would have been expected had the baseline frequencies persisted. In our analysis, we observed a tendency for more negative ratings of mental health to occur in surveys targeting mental health than in general health surveys. It is plausible that the observed difference relates to the placement of the item in the survey questionnaires. In the 2 mental health and wellbeing surveys, this item was included in the screening section for the Composite International Diagnostic Interview,13 such that respondents may have been primed by a series of preceding mental health questions. Another possibility is that simply knowing that a survey is about mental health may have caused respondents to alter their perceptions of their own mental health. In the general health iterations of www.TheCJP.ca
A bit stressful
the CCHS, the perceived mental health item was included in a general health module that was administered near the start of the interview. Surprisingly, even greater heterogeneity was observed with distress ratings. This heterogeneity may reflect an impact of social conditions, such as economic or employment security, which may change in a nonlinear fashion over time. In general, the extent of heterogeneity observed was large, despite the shared items and methodological similarities of the various surveys. This suggests a need for caution when comparing results from different surveys and highlights a need for replication of such results. There are evidently unknown determinants of differences in these estimates, which could produce misleading impressions of temporal trends. The Canadian Journal of Psychiatry, Vol 59, No 11, November 2014 W 595
Original Research
An advantage of pooling at the study level, as was done in our study, rather than at the individual level (as in a pooled analysis), is that the estimates can be appropriately weighted and standard errors can be accurately estimated using survey-specific techniques, such as the use of sampling weights and bootstrap variance estimation. An implication of this approach is that the sampling weights employed in each survey support inference to the household population in the year of each survey. As such, changes in the demographic structure of the population over time could influence the temporal trends observed. A methodological concern with the meta-analysis of frequency estimates is the possible need for variance stabilizing transformations, as meta-analytic weights are inversely proportional to the variance of estimated proportions, whereas the variance is expected to be smaller for smaller proportions. As smaller proportions may therefore be given greater weight, bias could occur. Appropriate transformations require specification of a numerator and denominator for each proportion and are therefore inconsistent with recommended populationweighting and variance estimation procedures. We have explored the impact of variance stabilizing transformations (such as the double arcsine transformation) in unweighted data using the metafor package in the software R. In instances where we have applied these approaches, they have led to very similar results to the more standard methods employed in the reported analyses. The main limitation of our study pertains to the limitations of the underlying data sources. As the included studies were large-scale health surveys, they tend to include brief items and scales rather than more detailed research interviews and schedules. Perceived mental health and perceived stress were measured only with single items of unknown validity (apart from face validity). Also, the K6 is a fairly rudimentary 6-item screening scale and is by no means a comprehensive assessment of mental health status.
Acknowledgements
Dr Patten is a Senior Health Scholar with Alberta Innovates, Health Solutions (AIHS). At the time of this work, Kirsten Fiest was supported by a PhD Studentship from AIHS.
596 W La Revue canadienne de psychiatrie, vol 59, no 11, novembre 2014
The estimates reported in our paper used data collected by Statistics Canada, but the analysis and results are the sole responsibility of the authors and do not reflect the views of Statistics Canada. This work was supported by an operating grant from the Canadian Institutes of Health Research (MOP-130415).
References
1. World Health Organization (WHO). Fact sheet 220. Mental health: strengthening our response [Internet]. Geneva (CH): WHO; 2014 [cited 2014 May 21]. Available from: http://www.who.int/mediacentre/factsheets/fs220/en/. 2. Wakefield JC. The concept of mental disorder. On the boundary between biological facts and social values. Am Psychol. 1992;47:373–388. 3. Wakefield JC, First MB. Clarifying the boundary between normality and disorder: a fundamental conceptual challenge for psychiatry. Can J Psychiatry. 2013;58(11):603–605. 4. Wakefield JC, First MB. The importance and limits of harm in identifying mental disorder. Can J Psychiatry. 2013;58(11):618–621. 5. Simpson KR, Meadows GN, Frances AJ, et al. Is mental health in the Canadian population changing over time? Can J Psychiatry. 2012;57(5):324–331. 6. Statistics Canada. National Population Health Survey—household component—cross-sectional (NPHS). Ottawa (ON): Statistics Canada; 2007 [cited 2013 Dec 30]. Available from: http://www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSurvey& SDDS=3236&lang=en&db=imdb&adm=8&dis=2. 7. Statistics Canada. Canadian Community Health Survey—annual component [Internet]. Ottawa (ON): Statistics Canada; 2013 [cited 2013 Dec 30]. Available from: http://www23.statcan.gc.ca/ imdb/p2SV.pl?Function=getSurvey&SDDS=3226&lang=en& db=imdb&adm=8&dis=2. 8. Kessler RC, Green JG, Gruber MJ, et al. Screening for serious mental illness in the general population with the K6 screening scale: results from the WHO World Mental Health (WMH) survey initiative. Int J Methods Psychiatr Res. 2010;19(Suppl 1):4–22. 9. Higgins JP, Thompson SG. Controlling the risk of spurious findings from meta-regression. Stat Med. 2004;23(11):1663–1682. 10. StataCorp. Stata user’s guide: version 12.1. College Station (TX): Stata Corporation; 2012. 11. Jorm AF, Reavley NJ. Changes in psychological distress in Australian adults between 1995 and 2011. Aust N Z J Psychiatry. 2012;46(4):352–356. 12. Angermeyer MC, Matschinger H, Schomerus G. Attitudes of the German public to restrictions on persons with mental illness in 1993 and 2011. Epidemiol Psychiatr Sci. 2014;23(3):263–270. 13. Kessler RC, Ustun TB. The World Mental Health (WMH) Survey Initiative Version of the World Health Organization (WHO) Composite International Diagnostic Interview (CIDI). Int J Methods Psychiatr Res. 2004;13:83–121.
www.LaRCP.ca
