Synthesis Through Simulation ... - economic-analysis.ca · Conclusions: Simulation models can act as a platform for economic analyses and epidemiologic projections and can support

CanJPsychiatry 2012;57(12)765-771

Original Research

Synthesis Through Simulation: Insights on the Epidemiologyof Mood and Anxiety Disorders in Canada

Scott B Patten, MD, PhD ;̂ Elizabeth Lin, PhD ;̂ Patricia J Martens, PhD ;̂ David Stiff, PhD";Paul Smetanin, MQF, BEcon (DBA Candidate)^; Carol E Adair,' Professor, Departments of Community Health Sciences and Psychiatry, University of Calgary, Calgary, Alberta; Member, Hotchkiss Brain institute, Calgary,Alberta.Correspondence: Department of Community Health Sciences, University of Calgary, 3rd Floor, TRW Building, 3280 Hospital Drive NW, Calgary, AB T2N 4Z6;[email protected].

'Scientist, Provincial System Support Program, Health Systems Research and Consulting Unit, Centre for Addiction and Mental Health, Toronto, Ontario;Assistant Professor, Department of Psychiatry, University of Toronto, Toronto, Ontario.

'Professor and Director, Manitoba Centre for Health Policy, Department of Community Health Sciences, Faculty of Medicine, University of Manitoba,Winnipeg, Manitoba.

"Senior Manager, Research, RiskAnalytica, Toronto, Ontario.

^CEO, RiskAnalytica, Toronto, Ontario

^Adjunct Associate Professor, Departments of Psychiatry and Community Health Sciences, University of Caigary, Calgary, Alberta.

Key Words: data synthesis,simulation, mathematicalmodelling, mood disorders,depressive disorders, anxietydisorders

Received March 2012, revised,and accepted June 2012.

Objective: Prevalence estimates for mood and anxiety disorders in Canada are available, butvarious methodological approaches have produced inconsistent results. Simulation studiesinvolve careful examination of available data by an expert modelling team working togetherwith subject matter experts. Simulation can integrate datasets and literature-based estimatesfrom various sources into a coherent mathematical representation of the underiying totalpopulation epidemiology.

Methods: Supported by the Mental Health Commission of Canada, a simulation modellingproject for mental disorders in Canada was recently undertaken. The modelling was carriedout by RiskAnalytica using their Life at Risk platform. Specification and calibration of the modeloccurred in consultation with national and international experts.

Results: To reconcile estimates of incidence and prevalence, recall bias needed to berepresented in the model. This suggests that the population prevalence of mood and anxietydisorders has been underestimated by population surveys and may explain a discrepancyobserved in the age-specific prevalence in population surveys as compared with studies usingadministrative data. The number of Canadians with mood and anxiety disorders is projectedto increase in upcoming decades as a result of population growth, but, based on conservativeassumptions, an increased prevalence proportion is not anticipated.

Conclusions: Simulation models can act as a platform for economic analyses andepidemiologic projections and can support the rapid exploration of what-if scenarios, therebyinforming policy decisions. This first national-level simulation provides a high level overview ofmood and anxiety disorder epidemiology in Canada.

Objectif : Des estimations de la prévalence des troubles anxieux et de l'humeur au Canadasont disponibles, mais diverses approches méthodologiques ont produit des résultatsincohérents. Les études en simulation font appel à un examen minutieux des donnéesdisponibles par une équipe experte en modélisation qui collabore avec des experts en lamatière. La simulation peut intégrer des ensembles de données et des estimations fondéessur la littérature issus de sources diverses dans une représentation mathématique cohérentede l'épidémiologie sous-jacente dans la population totale.

iVIéthodes : Avec l'appui de la Commission de la santé mentale du Canada, un projet demodélisation par simulation pour les troubles mentaux au Canada a été récemment entrepris.La modélisation a été exécutée par RiskAnalytica, au moyen de leur plateforme Life at Risk.La spécification et le calibrage du modèle ont été effectués en consultation avec des expertsnationaux et internationaux.

Résultats : Afin de concilier les estimations de l'incidence et de la prévalence, il a fallureprésenter des biais de rappel dans le modèle. Cela laisse entendre que la prévalence destroubles anxieux et de l'humeur dans la population a été sous-estimée par les enquêtesauprès de la population et peut expliquer l'écart observé dans la prévalence par âge des

www.IheCJP.ca TheCanadian Journal of Psychiatry, Vol 57, No 12, December 2012 • 765

Original Research

enquêtes dans la population comparativement aux études qui utilisent des données administratives.II est prévu que le nombre de Canadiens souffrant de troubles anxieux et de l'humeur augmenteradans les décennies à venir par suite de la croissance de la population, mais d'après des hypothèsesprudentes, une proportion accrue de la prévalence n'est pas prévue.

Conclusions : Les modèles par simulation peuvent servir de plateforme aux analyses économiqueset aux prévisions épidémiologiques, et peuvent aussi soutenir l'exploration rapide de scénarios paranticipation, éclairant ainsi les décisions de politiques. Cette première simulation à l'échelle nationaleprocure un aperçu de haut niveau de l'épidémiologie des troubles anxieux et de l'humeur au Canada.

Despite nearly 40 years of research in psychiatricepidemiology in Canada,'"' some basic aspects

of the epidemiology of mental disorders remainpoorly understood. This uncertainty is partially due toinconsistencies involving the diagnostic criteria for thesedisorders, the inclusion of different sets of disorders invarious surveys, a preponderance of cross-sectional (asopposed to longitudinal) studies, and variability acrossprovinces in health administrative databases. For example,Canadian surveys have generally reported that mood andanxiety disorder prevalence diminishes with age'; a patternthat has not been observed in administrative data.'" Whereasmajor depression lifetime prevalence estimates in the rangeof 10% have generally been reported by surveys"-' alongwith a low frequency of treatment, a study" using Albertaadministrative data found that the 3-year treated prevalenceof mood disorders in the general population of Alberta was16%. Data from developed countries participating in theWorld Mental Health surveys have raised the possibility thatcohort effects may produce large increases in prevalencein upcoming decades.'- However, these findings have beendisputed, as recall bias is another explanation for the satnepattems in the data.'̂ """ Such uncertainties slow progressin epidemiologic research and leave policy-makers with anincoherent evidence base.

Inconsistencies in the epidemiologic literature couldtheoretically be settled by conducting long-term studiesemploying detailed and highly accurate measurementstrategies. Until such studies are conducted in the realworld, simulation offers the promise of delineating atentative solution to these controversies. In Canada, theMHCC has advanced progress on this front by sponsoringthe development of a simulation model for mental disorders.

Changes to the population and societal burden of diseasecaused by mental disorders, including mood and anxietydisorders, may be infiuenced by various factors (forexample, population aging and growth). Simulationapproaches are capable of representing such complexsystems, including those characterized by dynamic,nonlinear relations." Simulation methods can also combine

AbbreviationsCCHS 1.2 Canadian Community Health Survey: Mental Health

and Weil-Being

MHCC Mental Health Commission of Canada

SUD substance use disorder

Clinical Implications

• An internally consistent model of mood and anxietydisorders in Canada indicates that lifetime prevaience isunderestimated in older populations.

• The number of Canadians with mood or anxietydisorders is expected to increase owing to populationgrowth, particularly for people aged 70 years and older.

• The model provides a consistent framework toquantitatively evaluate the impact of potentialinterventions.

Limitations

These results are simulated. They are not directepidemiologic estimates.

• Development of a baseline model was guided by a panelof subject matter experts. Erroneous interpretationscould lead to inaccurate representation.

• Concise differentiation of mood and anxiety disorderswas not possible.

very disparate types and forms of data in a breadth that isunprecedented. For example, information from focusedstudies of health conditions in population groups can becombined with total population demographic, health andsocial service use, and disability and economic data (suchas labour force statistics) to examine illness dynamics ina total societal context. Computer simulation is thereforean importatit research tool for health policy decision-making as there is currently no other research method withsimilar capabilities." Computer simulation is essentiallya tool for exploring possibilities through the model'sdynamic behaviour which is, in turn governed by a set ofwell-defined operating rules.""''' Simulation fills gaps leftby standard research methods and unexplored researchquestions, allowing for better-informed decision making. Itprovides an opportunity to explore future possibilities, aswell as to investigate the impact of different interventionsthrough exploration of what-if scenarios.

There has been a historical reluctance for the health caresector to apply computer simulation as a research tool.^"Nevertheless, this appears to be changing. For example,simulation was used extensively by Smetanin et aP' andStiff et aP- during the 2009 pandemic to better understandthe ability of intensive care units to handle the widespreadre-emergence of the swine-origin H INI influenza virus.

The current modelling initiative on mental disorders wascarried out by RiskAnalytica, using a platform called Life

766 '4' La Revue canadienne de psychiatrie, vol 57, no 12, décembre 2012

Synthesis Through Simulation: Insights on the Epidemiology of Mood and Anxiety Disorders in Canada

at Risk. Development of the model incorporated directinput from the MHCC as well as a series of consultationswith Canadian subject matter experts and internationalcollaborators. The resulting base model strives to representthe epidemiology of mental disorders in the population.The model synthesizes the best available data fromvarious sources into an internally consistent and crediblerepresentation of the epidemiology. A detailed report-^ isavailable. Our objective is to report on the findings of themodel through peer-reviewed literature to inform the stateof knowledge of mental health of the Canadian population.

MethodsThe Life at Risk platform provides a flexible, but powerful,means to model the impact of illness on the overall healthand economics of a population. For the analysis of mentaldisorders in Canada, the Life at Risk compartmentalmodel was used, which divides the population into a setof unique, nonoverlapping population cells. For example,at a minimum, the population is divided by age and sex.The population can then be further divided into healthstates. This division of population cells can continue untilthe required level of specificity is achieved. Once thepopulation is divided, the rate at which people can transitionbetween cells is defined. For example, people can transitionfrom a younger age group to an older age group throughaging, or from a healthy state to a mental disorder statethrough incidence. In addition to these transitions betweenpopulation cells, some processes allow people from outsidethe model to enter into the population (that is, birth orimmigration), and others remove people from the system(that is, death or emigration).

The population in our study was initially divided bydemographic characteristics consisting of age, sex, andimmigration status (immigrant or nonimmigrant). Thepopulation groups were then further subdivided by possiblehealth states. Both mental illness (including SUDs) and2 chronic medical diseases (heart disease and type IIdiabetes) were included in the model. Because insufficientpopulation-based data on illnesses occurring below theage of 9 years were available, the model was based on aconservative simplifying assumption that children aged 8years and younger had no mental illness or chronic disease.Children aged 9 to 12 years (inclusive) could have up to5 mental illnesses, with each illness defined as active orin remission. In addition, to account for increased riskof adolescent illness for those who have had a childhoodillness, the adolescent population groups were divided byprevious illness. Because of the lack of longitudinal data onchild and adolescent disorders in Canada, meta-analysis ofdata from 3 longitudinal cohort studies from New Zealandand the United States was used to estimate the transitionpatterns from the initial 5 disorders in childhood, throughany subsequent disorder in adolescence and, in turn, fromadolescent disorders through any subsequent disorder inadulthood.

The resulting base model strove to represent the impact ofmental illness in Canada from a national perspective butdid not include all disorders, model specific populationgroups, such as First Nations, Inuit, and Métis (owingto unavailability of data), or specific jurisdictions, suchas provinces and territories (owing to initial decisionson scope). However, the model is capable of furtherspecification on these or related topics if sufficientlydetailed data become available.

While the modelling considered many data sources, amajor challenge to producing an optimal model was thelimited availability of data. Data gaps and limitations areoutlined in a detailed report,-' and (usually conservative)assumptions used to address data are also described therein detail. These assumptions were approved by the project'spanel of subject matter experts.

The model included dementia and schizophrenia as well aschildhood conditions, but the focus in our article is on moodand anxiety disorders. Major sources of mood and anxietydisorder data were the Mental Health Supplement of theOntario Health Study'' (analyses of the original datasetwere carried out) and a Manitoba report by Martens andthe Manitoba Centre for Health Policy.-'' The demographicdata used to determine birth, death, and migration ratescame from Statistics Canada. The relative risk of deathassociated with adult mood and anxiety disorders wastaken from a meta-analysis of international studies.-'Incidence data for mood and anxiety disorders for the 18to 64 year age group came from a major prospective studyconducted in the Netherlands,-' whereas, for the 65 yearsand older age group, the Canadian Study of Health andAging-^ was a major data source. Comorbidity data for 2categories of physical condition—heart disease and type IIdiabetes (both from national data sources and along withpublished literature)—were incorporated in the model,along with their expected effects on mood and anxietydisorder incidence and vice versa.-^'^

Analyzing data from a New Zealand birth cohort study,Moffitt et al'^ observed that repeatedly measured annualdisorder prevalence resulted in a cumulative prevalence thatexceeded retrospectively assessed lifetime prevalence, anobservation that has also been made in Canada." Consistentwith these observations, an adjustment for recall bias wasmade so that mortality-corrected lifetime prevalence wouldincrease with age in the model. The healthy immigranteffect on mental disorder prevalence was represented usingfindings reported by Menezes et al.''' A bibliography ofmany additional data sources may be found in the detailedreport.-'

Several simplifying assumptions were employed. In thisliterature, age- and sex-stratified estimates were generallynot available, thus a single sex ratio was assumed for eachillness category. A single excess risk of mortality was usedfor each category, and this was assumed to be age- and sex-independent. Figure 1 presents a general schematic of theLife at Risk model.

wmi.TheCJP.ca The Canadian Joumal of Psychiatry, Vol 57, No 12, December 2012 4 ' 767

Original Research

Figure 1 A schematic representation of the progression of a single diseasein the model

Population withoutdisease

ncidence rate dependingon comorbid conditions

Population withdisease

Remission rate dependingon comorbid conditions

r

Population with diseasein remission

Mortality due toother causes

Excess mortalitydue to the disease

Figure 2 Estimated (2011) and projected (to 2041) prevalence counts for mood andanxiety disorders, by age and sex

550 000500 000450 000

I 400 000o 350 000^ 300 000 -ï 250 000•° 200 000E

150 000

100 00050 000

<V

550 000500 000450 000

g 400 000I 350 000\ 300 000..o 250 000I 200 000^ 150 000

100 00050 000

12011 D2021 «2031 «2041 12011 0 2021 «2031 «2041

Figure 3 Estimated (2011) and projected (to 2041) annual prevalence proportions (%) formood and anxiety disorders, by age and sex

c

o

oEfN

25.0

20.0

15.0

10.0

5.0

0.00°)

25.0

20.0

12011 0 2021 «2031 «2041 12011 D2021 «2031 «2041

768 + LaRevuecanadiennedepsychiatrie, vol 57, no 12, décembre2012 www.LaRCP.ca

Synthesis Through Simulation: Insights on the Epidemiology of Mood and Anxiety Disorders in Canada

Figure 4 Estimated (2011) and projected (to 2041) lifetime prevalence proportion (%) formood and anxiety disorders, by age and sex

>nce

(w

omen

), <

ne p

reva

lefe

tir

1

40.0 -

35.0 -

30.0 -

25.0

20 0

15.0

10.0

5.0 -

0.0

• 201

ill1 D2021 ^2031

lenc

e (m

en),

%im

e pr

eva

Life

t

• 2041

40.0 -̂

35.0 -

30.0 -

25.0 -

yn 0

15.0 -

10.0

5.0 -

nn

0.0 I I 1 1 . - . -

• 2011 D2021 ^2031

i\

MM

• 2041

Figure 5 Projected reduction in annual prevalence counts of mood and anxiety disordersin 2041 under 2 hypothetical interventions

coun

t

OJ

u

Mie

Ó.Q.01

inth

Red

uctio

n

90 000 n

80 000 -

70 000 '

60 000izn nnn -

An nnn -

30 000 -

20 000 -

10 000 -

• •• •• •• •• _•1 •1 • 11 H HbtàMM:

N ^ ^ -0) _0)

0) '^ o o'̂ Ç\ M/ v"̂ ^^

• 10% increase in remission

• 1^1 • • •H H 1liai.V* v̂ (Q ' ' \ % ^

10% reduction in incidence

ResultsIn an examination of administrative data from severalprovinces, a large degree of variability was observed inthe frequencies of specific diagnostic codes for mood andanxiety disorders separately, but combined prevalence wasfound to be consistent across provinces. This consistencyhas been observed before in Canada." As a result, moodand anxiety disorders were generally modelled together,rather than as separate categories. The resulting datasynthesis irnplicitly attributes the differing pattern of age-specific prevalence seen in Canadian surveys (prevalencegenerally declines with age) and Canadian administrativedata (prevalence does not decline with age)'° to a distortionof survey results by recall bias.

In 2011, it was estimated that there are more than 4 millionpeople living with a mood or anxiety disorder (includingyouth) in Canada. By 2041, the annual number of peopleliving with a mood or anxiety disorder is expected to

increase by 22.9%, reaching more than 4.9 million peopleor 11.4% of the total Canadian population. This increaseis primarily driven by population growth and the agingpopulation demographics.

Given the nature of the model and its underlying assumptions,annual prevalence output was considered more meaningful toreport than lifetime prevalence. Figure 2 shows projectionsfor the number of people with a mood or anxiety disorderby age and sex during the next 3 decades. Figure 3 presentssimilar projections for prevalence proportions. Age-specificprevalence is projected to be stable, but changing populationdemographics will lead to increased numbers of cases,particularly in older age categories.

The model emphasizes some of the limitations of lifetimeprevalence as a population-health metric.'^ This parametercontinues to be frequently reported and an adjustmentfactor for recall bias is built into the model, allowing the

wmi.TheCJP.ca TheCanadianJournalofPsychiatry, Vol 57, No 12, December 2012 * ' 769

Original Research

model to produce estimates of lifetime prevalence. Lifetimeprevalence of mood and anxiety disorders in the baselinemodel, stratified by age and sex, is presented in Figure 4.Prevalence is seen to increase until about age 50 years,and then to level off. This contrasts with declining age-specific lifetime prevalence, as has typically been observedin population surveys. The declining annual prevalence isdue to inclusion of a remission rate in the model, a ratethat was calibrated for consistency with incidence, annualprevalence, and mortality data.

As mentioned above, a major use of this sort of model is toexplore what-if scenarios. In the future, a detailed assessmentof specific health policy options may be feasible. Here, anexample of the general approach is presented. Using thecalibrated base model, the impact of various interventionscenarios were evaluated to help estimate their potentialimpact. For example, a hypothetical intervention wherethe remission rates are increased by 10% would result in259 000 fewer people living with mood and anxietydisorders by 2041. In contrast, a hypothetical interventionthat reduces incidence by 10% would result in 326 000fewer people living with mood and anxiety disordersby the same year. However, as shown in Figure 5, bothhypothetical interventions are complementary and have thegreatest impact in different age groups.

DiscussionThis project had limitations. There are many known riskfactors for mental illnesses that were not included inthe modelling, such as socioeconomic status, adversechildhood events, and smoking. Owing to data and timeconstraints, the model only included age, sex, type IIdiabetes, heart disease, immigrant status, and previous(including childhood) or concurrent mental illness as riskfactors. The long-term projections are therefore vulnerableto unexpected changes in various determinants of mood andanxiety disorders that were either not included in the modelor where the projections (for example, immigration rates)tum out not to be true.

The detailed Life at Risk model did produce projectionsfor mood and anxiety disorders separately. These areavailable in the detailed report.̂ ^ However, in view of thegreater consistency seen in Canadian surveillance datathat combines the 2 groups, the combined estimates werepresented in our report. One of the main applications ofsimulation modelling is the projection of future costs.For interested readers, the detailed report includes costprojections for these disorders. Finally, the detailed reportincludes estimates for various childhood disorders, SUDs,and schizophrenia, as well as estimates of persistence ofchildhood and adolescent disorders into adulthood.

Estimates arising from this simulation modelling projectare national in scope, but most health policy decisions inCanada are made at the provincial level. Province-specificmodelling would provide a better link to policy decisions.The model can be expanded or modified in the future to

accommodate new data, including province-specificdata. The base model will provide a useful platform forintegration of new results with well-established ones.

Psychiatric epidemiology has gone through variousphases—what has been called the first, second, and thirdgeneration of studies.'* The third generation label is appliedto large, sophisticated population surveys, which havebeen available in Canada from regional studies since the1980s'" and nationally since the 2002 CCHS 1.2.** Somebelieve that a next generation of studies will use large, tmlypopulation-based (that is, including the entire population)health administrative databases to overcome the limitationsof survey data. RiskAnalytica did use such infonnationin creating these estimates. However, analyses based onadministrative data have their own limitations. Within thebroader domain of science and technology, simulation isoften viewed as a logical progression from the accrual ofindividual estimates from various types of studies. In a 2007lecture," Jim Gray commented that scientists have beenempirically describing natural phenomena for thousandsof years, and using models to extend and generalize theseobservations during the past few hundred years, but thecomputational branch of science (using simulation torepresent complex phenomena) has been an activity onlyof the past few decades. Simulation is an approach to datasynthesis that can help to integrate available data, providinga useful representation of real-world systems. This project,the first of its kind in Canada, has provided a high-levelperspective on several important issues related to theepidemiology of mood and anxiety disorders.

AcknowledgementsThe project received support from the MHCC. Dr Pattenis a Senior Health Scholar with Alberta Innovates, HealthSolutions. He received an honorarium for reviewinginvestigator-drive research grants submitted to Pfizer, andfor reviewing such a grant submitted to Lundbeck. Dr Pattenreceived payment as a member of an advisory board forServier and as a member of a roundtable meeting sponsoredby the WOCO Foundation, and he received speaking feesfrom Teva for a conference presentation on depression andanxiety in multiple sclerosis. Dr Martens acknowledges thecareer award support of the Canadian Institutes of HealthResearch and Public Health Agency of Canada AppliedPublic Health Chair. Dr Stiff and Dr Smetanin were paidconsultants for the MHCC to conduct mental healthpopulation-based modelling on which the results of thispaper are based. Dr Adair was in a consulting role with theMHCC and provided the major content area support forthis project and received hourly contract remuneration forthe work and other unrelated projects. Costs for additionalanalysis of Ontario Mental Health Supplement data neededby RiskAnalytica were reimbursed by the MHCC.

This project was led by the MHCC, and was made possiblethrough a financial contribution from Health Canada to theMHCC. The views expressed herein solely represent theauthors.

770 • La Revue canadienne de psyehiatrie, vol 57, no 12, décembre 2012 www.LaRCP.ca

Synthesis Through Simulation: Insights on the Epidemiology of Mood and Anxiety Disorders in Canada

Referencest. Murphy JM, Laird NM, Monson RR, et al. Incidence of depression

in the Stirling County Study: historical and comparativeperspectives. Psychol Med. 2000;30:505-514.

2. Murphy JM, Laird NM, Monson RR, et al. A 40-year perspectiveon the prevalence of depression. Arch Gen Psychiatry.2000;57:209-215.

3. Murphy JM, Horton NJ, Laird NM, et al. Anxiety anddepression: a 40-year perspective on relationships regardingprevalence, distribution, and comorbidity. Acta Psychiatr Scand.2004;109(5):355-375.

4. Bland RC, Om H, Newman SC. Lifetime prevalence ofpsychiatric disorders in Edmonton. Acta Psychiatr Scand Suppl.1988;338:24-32.

5. Bland RC, Newman SC, Orn H. Period prevalence ofpsychiatric disorders in Edmonton. Acta Psychiatr Scand Suppl.1988;338:33^2.

6. Offord DR, Boyle MH, Campbell D, et al. One-year prevalenceof psychiatric disorder in Ontarians 15 to 64 years of age.Can J Psychiatry. 1996;41:559-563.

7. Patten SB, Wang JL, Williams JV, et al. Descriptive epidemiology ofmajor depression in Canada. Can J Psychiatry. 2006;51 (2):84—90.

8. Gravel R, Béland Y. The Canadian Community Health Survey:mental health and well-being. Can J Psychiatry. 2005;50:573-579.

9. Beaudet MP. Psychological health—depression. Health Rep.1999;ll:63-75.

10. Kisely S, Lin E, Gilbert C, et al. Use of administrative data for thesurveillance of mood and anxiety disorders. Aust N Z J Psychiatry.2009;43(12):l 118-1125.

lL Slomp M, Bland R, Patterson S, et al. Three-year physician treatedprevalence rate of mental disorders in Alberta. Can J Psychiatry.2009;54(3):199-203.

12. Kessler RC, Üstün TB. The WHO World Mental Health Surveys.Global perspectives on the epidemiology of mental disorders.New York (NY): Cambridge tjniversity Press; 2008.

13. Giuffra LA, Risch N. Diminished recall and the cohorteffect of major depression: a simulation study. Psychol Med.1994;24:375-383.

14. Moffitt TE, Caspi A, Taylor A, et al. How common are commonmental disorders? Evidence that lifetime prevalence rates aredoubled by prospective versus retrospective ascertainment.Psychol Med. 2010;40{6):899-909.

15. Patten SB. Recall bias and major depression lifetime prevalence.Soc Psychiatry Psychiatr Epidemiol. 2003;38:290-296.

16. Patten SB, Gordon-Brown L, Meadows G. Simulation studies ofage-specific lifetime major depression prevalence. BMC Psychiatry.2010;10(l):85.

17. Veazie PJ. Using computer simulation to investigate diseasemanagement. Endocinologist. 2001;ll(3):209-216.

18. Hanneman R, Patrick S. On the uses of computer-assisted simulationmodeling in the social sciences. Sociological Research Online.1997;2(2).

19. Pritzker A, Pedgen C. Introduction to simulation and SLAM.New York (NY): John Wiley & Sons Inc; 1979.

20. Wilson JCT. Implementation of computer simulation projects inhealth care. J Oper Res Soc. 1981;32(9):825-832.

21. Smetanin P, Stiff D, Kumar A, et al. Potential intensive careunit ventilator demand/capacity mismatch due to novel swine-origin H INI in Canada. Can J Infect Dis Med Microbiol.2009;20(4):ell5-el23.

22. Stiff D, Kumar A, Kissoon N, et al. Potential pédiatrie intensive careunit demand/capacity mismatch due to novel pHlNl in Canada.Pediatr Crit Care Med. 2011;12(2):e51-e57.

23. Smetanin P, Stiff D, Briante C, et al. The life and economic impactof major mental illnesses in Canada: 2011 to 2041. Toronto (ON):RiskAnalytica, on behalf of the Mental Health Commission ofCanada; 2012.

24. Martens P, Manitoba Centre for Health Policy. Patterns of regionalmental illness disorder diagnoses and service use in Manitoba:a population based study. Winnipeg (MB): Manitoba Centre forHealth Policy; 2004.

25. Harris EC, Barraclough B. Excess mortality of mental disorder.Br J Psychiatry. 1998;I73:Il-53.

26. Bijl RV, de Graaf R, Ravelli A, et al. Gender and age-specificfirst incidence of DSM-III-R psychiatric disorders in the generalpopulation. Results from the Netherlands Mental Health Survey andIncidence Study (NEMESIS). Soc Psychiatry Psychiatr Epidemiol.2002;(37):372-379.

27. Canadian Study of Health and Aging Working Group. The Canadianstudy of health and aging: study methods and prevalence ofdementia. CMAJ. 1994;I50(6):899-912.

28. Rugulies R. Depression as a predictor for coronary heart disease.A review and meta-analysis. Am J Prev Med. 2002;23:51-6I.

29. Patten SB, Beck CA, Kassam A, et al. Long-term medical conditionsand major depression: strength of association for specific conditionsin the general population. Can J Psychiatry. 2005;50:195-202.

30. Statistics Canada. CANISM Table 102-0524: deaths, by cause,chapter IV: endocrine, nutritional and metabolic diseases (EOO toE90), age group and sex, Canada, annual (number). Ottawa (ON):Statistics Canada; 2010 [cited 2011 Oct 19]. Available from:http://www5.statcan.gc.ca/cansim/aO5?lang=eng&id=l 020524.

31. Public Health Agency of Canada (PHAC). National diabetessurveillance system. Ottawa (ON): PHAC; 2010 [2011 Oct 19].Available from: http://www.phac-aspc.gc.ca/ccdpc-cpcmc/ndss-snsd/english/index-eng.php.

32. Chow CM, Donovan L, Manuel D, et al. Regional variation in self-reported heart disease prevalence in Canada. Can J Cardiol.2005;21(I4):1265-1271.

33. Patten SB. Accumulation of major depressive episodes over time ina prospective study indicates that retrospectively assessed lifetimeprevalence estimates are too low. BMC Psychiatry. 2009;9:19.

34. Menezes NM, Georgiades K, Boyle MH. The infiuence of immigrantstatus and concentration on psychiatric disorder in Canada: a multi-level analysis. Psychol Med. 2011 ;41 ( 10):2221-2231.

35. Streiner DL, Patten SB, Anthony JC, et al. Has lifetime prevalencereached the end of its life? A review of the concept. Int J MethodsPsychiatr Res. 2010; 18:221-228.

36. Dohrenwend BP. A psychosocial perspective on the past and futureof psychiatric epidemiology. Am J Epidemiol. 1998;147:222-231.

37. Gray J. Jim Gray on eScience: a transformed scientific method[Internet]. Based on the transcript of a talk given by Jim Gray tothe NRC-CSTBl in Mountain View, CA, Januai-y 11, 2007 [cited2012 May 3]. Available from: http://research.microsoft.com/en-us/collaboration/fourthparadigm/4th_paradigmbookjim gray transcript.pdf.

www.r/ieCJP.ca TheCanadianJournalofPsychiatry, Vol 57, No 12, December 2012 -*• 771

Copyright of Canadian Journal of Psychiatry is the property of Canadian Psychiatric Association and its content

may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express

written permission. However, users may print, download, or email articles for individual use.