Epidemiology of major depression in a predominantly rural health region

Transcription

1 Soc Psychiatry Psychiatr Epidemiol (2003) 38 : DOI /s ORIGINAL PAPER Scott B. Patten Heather L. Stuart Margaret L. Russell Colleen J. Maxwell Julio Arboleda-Flórez Epidemiology of major depression in a predominantly rural health region Accepted: 3 March 2003 SPPE 651 Abstract Background Several large-scale cross-sectional studies have evaluated the prevalence of major depression in Canadian populations. Few studies have employed prospective methods, which are necessary to evaluate incidence, and few have focused on predominantly rural areas. Methods Subjects who had participated in a cross-sectional general health survey were invited to participate in a second wave of data collection 6 months later. These subjects were recontacted using a telephone interview. A brief diagnostic instrument for major depression was used in both waves,and a variety of other variables relevant to the epidemiology of major depression were measured. Results Of 801 subjects initially enrolled, 666 (83.1 %) consented to be recontacted, and 501 (75.2 %) of these were successfully reached. The incidence of major depression was 3.8 %. The incidence was higher in women, although this difference did not attain statistical significance. Having a past history of depression and having a high level of perceived stress were predictors of risk.an exploratory comparison with data collected using similar methods in a nearby urban centre determined that the rural prevalence was lower than urban, and that a variety of factors (street drugs, deficits in Scott B. Patten, MD, PhD ( ) M. L. Russell, MD, PhD C. J. Maxwell, PhD Dept. of Community Health Sciences University of Calgary 3330 Hospital Drive NW Calgary, AB, Canada T2N 4N1 Tel.: / Fax: / patten@ucalgary.ca Website: www: H. L. Stuart, MA, PhD Dept. of Community Health Queens University Kingston, Ontario, Canada J. Arboleda-Flórez, MD, PhD Dept. of Psychiatry Queens University Kingston, Ontario, Canada social support, unemployment, recent life events) may contribute have to this difference. Conclusions The 6- month cumulative incidence of depressive disorder was much higher than that reported by most previous studies.this may be a result of the diagnostic instrument employed, which captures a broader spectrum of depressive morbidity than the instruments used in most previous studies. Also, as the study did not exclude subjects with previous depressive episodes during their lifetime, the incidence rate reflects risks of major depressive episodes rather than major depressive disorders. Key words depressive disorder rural health epidemiology prospective studies risk factors Introduction There is now an extensive international literature concerned with the epidemiology of depressive disorders. This literature confirms that major depression is among the most important sources of health-related disability globally [1].Various Canadian surveys have reported 12- month prevalence estimates in the range of 4 % to 6 % [2, 3,4].Genetic factors are known to play an important role [5] and antidepressant medications are effective treatments [6]. However, strong and consistent associations with a variety of psychosocial factors [7, 8] suggest an important environmental contribution. Associations between depressive disorders and various psychosocial factors are influenced by complex temporal effects. Many of the factors that may play a role in causing depression may also result from depression, including unemployment, changes in marital status and diminished social support. Prospective study designs are required to clarify these various effects, but prospective studies have been infrequent, particularly in Canadian populations. Newman and colleagues conducted a prospective follow-up of a sample from a large survey conducted in Edmonton [9]. However, their analysis suggested that the instrument employed, the Diagnostic

2 Interview Schedule (DIS), potentially led to misleading results because of inconsistencies regarding the timing of episodes. Beaudet [10] has recently reported prospective data from the Canadian National Population Health Survey (NPHS), including incidence estimates and risk factor associations. The incidence estimate from the NPHS varied by age and gender and was higher in all categories than another recent estimate from a Nova Scotia study [11]. With the exception of the Nova Scotia study, all of these investigations used either urban samples, or sampled from geographical areas with a majority of non-rural subjects. Only one study, by Parikh [12], examined urban vs. rural differences. Here, the prevalence was 4.2 % and 3.6 % in urban and rural Ontario residents, respectively; a difference that did not achieve statistical significance. We recently encountered an opportunity to explore these epidemiological relationships using populationbased data and a prospective methodological paradigm. In this paper we (a) describe the epidemiology of major depression in a predominantly rural western Canadian population and (b) compare this to data collected during a methodologically comparable study in an adjacent urban area. Subjects and methods Setting The study was conducted in a predominantly rural health region in Southern Alberta, that stretches from Calgary south and west to the British Columbia border. At the time of the study, the region had a population of approximately 74,000. The predominant industries are agriculture, oil and tourism. For comparison purposes, previously collected data from Calgary, an adjacent city with 850,000 residents, were used. The latter survey was cross-sectional with no prospective component [13]. Study design For purposes of regional health planning, a general health survey was conducted. It included a brief predictive measure of major depression and measures of a variety of other variables. At the time of participation in the survey, subjects were asked if they would provide permission to be recontacted by telephone. Those in agreement were recontacted 6 months later and additional data collection was carried out, including a re-evaluation of major depression status. The initial survey utilized the Mitofsky-Waksberg procedure for random digit dialing (RDD), adopting a modification and weighting procedure reported by Potthoff [14] to select a sample of 801 subjects. From each household selected by RDD, one resident was selected using the last birthday method [15]. The interviews were conducted by telephone, using baccalaureate trained interviewers all of whom had at least 1 year of experience conducting epidemiological telephone surveys. An identical sampling procedure had been used in the urban survey. In each study, data collection was supervised by an experienced study coordinator. Data collection and measures The survey questionnaire was a composite instrument including items from previous Canadian surveys such as the Labour Force Survey, Health Promotion Survey and the NPHS. These items evaluated 361 age, education, marital and employment status, family income, general health, chronic medical conditions, recent life events and medication use. Ten recent life events were covered in the interview: physical assault, unwanted pregnancy, abortion or miscarriage, financial crisis, failure at school, change in job for a worse one, demotion or cut in pay, increased arguments with partner, going on welfare or having a child move back into the house.other items were adapted from similar or related questions employed in these or other surveys to evaluate alcohol ingestion, street drug use, perceived stress and clinical diagnoses of depression. Also, some items were formulated specifically for the survey in response to the information and planning needs of the health region.a series of questions asked whether the subject had been diagnosed by a health professional with one or more of a series of long-term medical conditions, and this was followed by an openended question about long-term medical conditions other than those listed.a single item asked each subject whether they had ever received a diagnosis of depression from a health professional. Items referring to early parental loss and having three or more children under the age of 14 at home [16] were also included in the interview. The questions asked during the follow-up interview closely resembled those of the initial survey, except that demographic information (e. g. gender) was not repeated and a brief semi-structured interview designed to evaluate family history of depression [17] was included. Depression was evaluated using the WHO Composite International Diagnostic Interview (CIDI) Short Form for major depression [18]. This is a brief predictive instrument initially designed to identify episodes of major depression occurring in the preceding 12 months. For example, the initial question in the CIDI-SFMD interview is: During the past 12 months, was there ever a time when you felt sad,blue or depressed for 2 weeks or more in a row?.the CIDI-SFMD was used in an unmodified form in the initial stage of data collection in this project, thereby providing a 12-month period prevalence estimate. In the follow-up interview, which occurred 6 months later, all of the questions were revised in order that they refer to a 6-month,rather than 12-month, time period. As such, it was possible to identify episodes occurring during the 6-month follow-up period. For example, the initial item was modified to: During the past 6 months, was there ever a time when you felt sad, blue or depressed for 2 weeks or more in a row?. The proportion of subjects who were free of major depressive episodes in the year preceding the baseline interview who subsequently reported an episode provided an estimate of incidence. It should be emphasized that this concept of incidence refers to the onset of new episodes, rather than the first-ever onset of a major depressive episode. As such, the proportions reported do not estimate the risk of onset of major depressive disorder. The CIDI-SFMD has excellent face validity for major depressive episode when a score of five is used to identify these episodes. This score indicates the presence of five symptoms listed in the DSM-IV A criteria for major depression, at least one of these being a depressed mood or loss of interest.all symptoms must occur during the same 2- week period. However, the interview does not include all of the items evaluating the clinical significance of symptoms that are found in the CIDI, nor does it include all of the etiological probes contained in that instrument. The CIDI Short Form, therefore, identifies a somewhat broader spectrum of depressive morbidity than does the full CIDI interview [19]. Data management and analysis Epi Info was used to enter the study data and merge the databases from the initial and follow-up surveys [20]. Graphical and tabular displays were then used to screen for coding errors and review the descriptive characteristics of the data. Sampling weights were calculated as described by Potthoff [14], and were subsequently adjusted to account for an over-representation of women in the sample.the C-Sample command in Epi Info [20] was used to calculate weighted estimates of incidence and prevalence. Odds ratios deriving from contingency table analysis and logistic regression were not weighted. Version 7 of Stata [21] was used for logistic regression. In order to facilitate comparisons with the urban data, direct standardization was used with the Canadian national population being the standard population. This part of the analysis was conducted using Stata [21].

3 362 Results In the initial survey, a total of 8,107 telephone numbers were called. Of these, 2,679 (33 %) were disconnected or not working, 603 (7.4 %) were fax numbers, 1,102 (13.6 %) were businesses, 1,714 (21.1 %) were refusals at the household level and 882 (10.9 %) could not be classified (answering machines, language problem, etc.). The remaining 1,127 were residential households who allowed the selection of one member for the study (last birthday method). The selected household member could not be reached in 139 cases (12.3 %) and refused in 184 cases (16.3 %). There were two incomplete (terminated) interviews. In other words, there were 988 potential respondents who were personally contacted about the study, and 804 (81.4 %) consented to participate. Data collection was successfully completed for 801 (99.6 %) of these subjects. Of these subjects, 666 (83.1 %) agreed to be recontacted 6 months later,and 501 (75.2 %) were successfully recontacted. Considering both the consent rate and the successful recontact rate, the overall response rate for the 6-month follow-up was 501/801 (62.5 %). Subjects included in the prospective component of the study closely resembled the initial sample in terms of demographic features and frequency of depression (Table 1).The weighted 12-month prevalence at baseline was 10.4 % (95 % CI ), with point estimates of prevalence being higher in women and those aged 45 and younger (Table 2). However, the 95 % confidence intervals for the sex- and age-specific prevalence estimates were indicative of low precision. Table 2 Age- and gender-specific prevalence, 6-month incidence rates and extrapolated* to 12-month incidence rates, weighted data 12-month 6-month Extrapolated prevalence incidence 12-month (95% Cl) (95% Cl) incidence* Gender Male 7.4% ( ) 2.5% ( ) 4.9% Female 13.2% ( ) 5.0% ( ) 9.8% Age Age < % ( ) 3.3% ( ) 6.5% Age % ( ) 4.2% ( ) 8.2% * In order to extrapolate the 6-month incidence (P 6 ) to a 12-month rate (P 12 ), the following formula was used: P 12 = 1 (1 P 6 ) 2 Fifty-three subjects who were depressed at the initial interview were excluded from the calculation of incidence, as was one subject with missing data at baseline. Hence, there were 447 subjects who were considered to be at risk from the point of view of the incidence calculation. During the next 6 months, 21 of these (4.7 %) developed the major depressive syndrome. The weighted incidence estimate was 3.8 % (95 % CI ). In order to facilitate comparison with other studies, a monthly rate was calculated from the 6-month rate and then extrapolated over a 12-month period (see the footnote to Table 2 for the method of extrapolation), the resulting estimate of 12-month incidence was 7.5 %. The following variables were not associated with incident depressive episodes: marital status, family history of depression, alcohol consumption, street drug use, daily smoking, deficient social support, unemployment, educational level (less than high school graduation), re- Table 1 Demographic features and major depression prevalence: unweighted data Initial sample Providing Successful MDE prevalence MDE baseline MDE prevalence consent for recontact initial sample prevalence: at second recontact successfully interview recontacted sample N = 801 (%) N = 666 (%) N = 501 (%) N = 801 (%) N = 501 (%) N = 501 (%) Age < (16.7) 112 (16.8) 69 (13.8) 14 (10.4) 5 (7.2) 2 (2.9) (32.6) 222 (33.3) 175 (34.9) 33 (12.6) 22 (12.6) 20 (11.4) (33.8) 225 (33.8) 181 (36.1) 31 (11.4) 20 (11.0) 17 (9.4) (16.9) 107 (16.1) 76 (15.2) 7 (5.2) 6 (7.9) 6 (7.9) Gender Male 275 (34.3) 232 (34.8) 162 (32.3) 24 (8.7) 14 (8.6) 11 (6.8) Female 526 (65.7) 434 (65.2) 339 (67.7) 61 (11.6) 39 (11.5) 34 (10.0) Marital Status Married* 510 (63.7) 434 (65.2) 348 (69.5) 41 (8.0) 28 (8.0) 26 (7.5) Single 133 (16.6) 110 (16.5) 69 (13.8) 21 (15.8) 9 (13.0) 7 (10.1) Previously married** 158 (19.7) 122 (18.3) 84 (16.8) 23 (14.6) 16 (19.0) 12 (14.3) Education < High school diploma 111 (13.9) 86 (12.9) 63 (12.6) 18 (16.2) 11 (17.5) 5 (7.9) High school diploma 202 (25.2) 160 (24.0) 115 (23.0) 19 (9.4) 15 (13.0) 14 (12.2) Some post secondary 307 (38.3) 264 (39.6) 195 (38.9) 34 (11.1) 19 (9.7) 21 (10.8) University degree 180 (22.5) 155 (23.3) 128 (25.5) 14 (7.8) 8 (6.3) 5 (3.9) Missing 1 (0.1) 1 (0.2) 0 (0.0) 0 (0.0) 0 (0.0) Working*** Yes 544 (67.9) 462 (69.4) 346 (69.1) 56 (10.3) 35 (10.1) 30 (8.7) No 257 (32.1) 204 (30.6) 155 (30.9) 29 (11.3) 18 (11.6) 15 (9.7) * Married, common law or living with a partner; ** Divorced, widowed, separated; *** Currently working for pay or profit

4 Table 3 Logistic regression model for the incidence of the major depressive syndrome, unweighted data Variable name Regression Odds Ratio p-value coefficient (95% CI) Constant 4.07 Sex (female) ( ) 0.55 Past history of depression ( ) Perceived stress ( ) Table 4 Logistic regression model: urban vs. rural comparison, with adjustment for selected variables, unweighted data Variable name Regression Odds Ratio p-value coefficient (95% CI) Constant 2.86 Residence (urban) Age Street drug use Deficits in social support < Recent life event(s) < cent life events, long-term medical conditions, early parental loss and having three or more children under the age of 14 at home. Having a past history of depression was significantly associated with incident depressive episodes (OR = 4.96, 95 % CI ), as was perceived stress (OR = 3.68, 95 % CI ). Both past history of depression (OR = 4.26, 95 % CI ) and perceived stress (OR = 4.89, 95 % CI 2.99 = 8.05) were also associated with the 12-month measure of major depression prevalence at baseline. Several additional variables were associated with major depressive episodes in the 12 months preceding the baseline interview (but not incident episodes): street drug use (OR = 2.42, 95 % CI ), daily smoking (OR = 3.31, 95 % CI ), deficits in social support (OR = 3.46, 95 % CI ) and recent life events (OR for one or more events = 2.01, 95 % CI ). Table 3 presents the results of a logistic regression analysis for incident episodes. Despite a lack of statistical significance, gender was included in the model for two reasons: it improved the overall fit and it reflected an expected association. Both perceived stress and past history remained significantly associated with major depression incidence, but other variables and interaction terms were not. The prevalence of the depressive syndrome in our rural sample was compared to that of a similar study completed in an adjacent urban area [13]. The unweighted crude prevalence in the rural region was lower than that of the urban region: 10.6 % compared to 17.1 %. The odds ratio for urban vs. rural residence was 1.74 (95 % CI ). Since this could be due to demographic differences, direct standardization by age and gender was carried out. This did not result in a substantial change: the adjusted rates were 10.5 % (95 % CI 8.2 % 12.7 %) and 15.2 % (95 % CI 13.9 % 16.6 %), respectively. In order to explore possible reasons for the urban/rural difference in prevalence, the unweighted proportions in each sample reporting potential determinants were tabulated, but only in instances where an identical measure was used in each survey.a higher proportion of the urban subjects were 45 years of age or younger (65.7 % vs %), more urban subjects reported illicit drug use (6.1 % vs. 3.2 %), deficient social support (16.5 % vs %) and occurrence of recent life events (45.7 % vs %). Table 4 presents a logistic regression model incorporating these variables. The urban-rural difference was weakened and no longer attained statistical significance after adjustment for these variables. In the urban survey, 4.9 % reported taking antidepressants at the time of the interview. This included 18.2 % of those reporting an episode of major depression in the preceding year. In the rural sample, 3.8 % reported taking antidepressants, including 11.8 % of those reporting an episode of major depression in the preceding year. Discussion This study was opportunistic in that it represented a follow-up to a general health survey that was conducted for health-planning purposes. Although the general survey included an item seeking permission for recontact, the study did not include the various measures (multiple contact mechanism, tracing strategies and record linkages) often employed to increase the rate of successful follow-up in prospective epidemiological studies. Perhaps as a result, the response rate was modest, with attrition occurring both at the stage of obtaining consent for recontact, and at the stage of reaching those persons who provided consent for recontact. Selection bias might have been introduced into the estimated odds ratios if the non-respondents differed both in terms of disease and exposure status [22]. The demographic and clinical features of those subjects not providing consent and those subjects who were not successfully followed up resembled those of the subjects from whom complete data were collected. However, this does not exclude the possibility of selection bias. Measuring the incidence of major depression in community samples has been challenging, particularly for studies using the DIS. For example, a previous Canadian study that employed the DIS identified important measurement problems relating to the estimation of incidence. The DIS (like the lifetime version of the CIDI) first determines whether a subject has experienced episodes of major depression in his or her lifetime and, if so, then seeks to specify the timing of these episodes. Many subjects who were considered to be free of major depression at the baseline interview subsequently reported episodes (during the follow-up interview) that

5 364 apparently occurred prior to the date of the baseline assessment [9]. The current study used an instrument that differed from the DIS in several important respects. The CIDI-SFMD directs all questions toward the same 2- week period in the preceding 12 months. In the followup interview, all responses were keyed to a 2-week period in the preceding 6 months. While this is a potential advantage of the CIDI-SFMD, inaccuracy of subjective recollections about the timing of episodes may still be a factor, and argue for caution in interpretation of the incidence estimates presented here. Existing validation data suggest that the CIDI Short Form is somewhat non-specific [19]. The depressive syndrome identified by the Short Form likely includes some more mild episodes than a full version of the structured interview would, and may also capture a more etiologically heterogeneous group of disorders [19]. This may partially explain the high incidence observed in this study relative to other Canadian investigations. The Stirling County Study, also conducted in a predominantly rural area in Canada and using an entirely different approach to the identification of disorders,recently reported a much lower annual incidence of 3.7/1,000 per year [11]. However, subjects with any past history of depressive episodes were excluded from this calculation so that this study evaluated the incidence of major depressive disorders rather than episodes. Recent incidence estimates from the NPHS, which used the same predictive instrument used here and did not exclude those with a prior lifetime episode of major depression, have yielded more comparable rates [10]. Using logistic modeling, we produced much lower fitted values for incidence in subjects without a past history of depression. Hence, the estimates arising out of the NPHS [10] and this study are not as discordant from the Stirling County study as they initially appear to be. However, the characteristics of the measurement instrument may have contributed to the differences in estimated incidence. The CIDI Short Form requires symptoms to be persistent over a 2-week period whereas the instrument employed in the Stirling County Study had a 4-week requirement. Epidemiological analyses focusing on the incidence of major depression have the potential to identify risk factors that are modifiable or remediable. From a public health point of view, these factors may ultimately prove useful for identifying high-risk groups, for screening, or for secondary prevention. Given the relatively small sample size in this research, the statistical procedures may have been vulnerable to Type II error. Therefore, negative results should not be interpreted as evidence that these factors are not predictive of depression. Finally, this study has compared the prevalence of major depression in a rural and urban health region and explored factors that may account for the higher prevalence of depression among urban dwellers. These findings run contrary to the widely held belief that mental illnesses are more prevalent in rural regions in Canada as a consequence of factors such as poor transportation systems, physical isolation, spatial distribution of populations, poverty, unemployment, lack of mental health specialists, less developed health services, and greater stigma [23]. Our findings may be, in part, a reflection of migration, where people who are most in need of mental health treatment may move to urban areas where there are greater opportunities for access to specialized services, temporary housing, job training and employment opportunities, as well as networking opportunities with other individuals having similar conditions. In part, they may also reflect the higher prevalence of known risk factors for depression in urban areas such as street drug use, deficits in social support, unemployment, and life events. Conclusions Geographic differences in the prevalence of mental disorders have important implications for service planning and delivery, particularly where the goals are to achieve equity in resource allocations and link population needs to available resources. Conceptual and statistical models for assessing geographic variations in both the burden of mental health and in health service requirements are still in their early stages [24]. Further research will be important to identify factors that promote geographical patterns in prevalence and to replicate our results. Acknowledgment This project was supported by a research and development grant from the Calgary Health Region and a grant-inaid from the Headwaters Regional Health Authority. Dr. Patten is a Population Health Investigator with the Alberta Heritage Foundation for Medical Research. References 1. Murray CJL (1996) The Executive Summary of The Global Burden of Disease and Injury Series. Harvard School of Public Health, Boston, pp Beaudet MP (1996) Depression. Health Reports 7: Bland RC, Newman SC, Orn H (1988) Period prevalence of psychiatric disorders in Edmonton. Acta Psychiatr Scand (Suppl) 338: Offord DR, Boyle MH, Campbell D, Goering P, Lin E, Wong M, Racine YA (1996) One-year prevalence of psychiatric disorder in Ontarians years of age. Can J Psychiatry 41: Kendler KS, Kessler RC,Walters EE, MacLean C, Neale MC, Heath AC, Eaves LJ (1995) Stressful life events, genetic liability, and onset of an episode of major depression in women. Am J Psychiatry 152: Joffe R, Sokolov S, Streiner D (1996) Antidepressant treatment of depression: a meta-analysis. Can J Psychiatry 41: Weissman MM, Bland RC, Canino GJ, Faravelli C, Greenwald S, Hwu HG,Joyce PR,Karam EG,Lee CK,Lellouch J,Lepine JP,Newman SC, Rubio-Stipec M, Wells EJ, Wickramaratne PJ, Wittchen HU, Yeh EK (1996) Cross-national epidemiology of major depression and bipolar disorder. JAMA 276: Blazer DG, Kessler RC, McGonagle KA, Swartz MS (1994) The prevalence and distribution of Major Depression in a national community sample: the National Comorbidity Survey. Am J Psychiatry 151:

6 9. Newman SC, Bland RC (1998) Incidence of mental disorders in Edmonton: estimates of rates and methodological issues. J Psychiatr Res 32: Beaudet MP (1999) Psychological health depression. Health Reports 11: Murphy JM, Laird NM, Monson RR, Sobol AM, Leighton AH (2000) Incidence of depression in the Stirling County Study: historical and comparative perspectives. Psychol Med 30: Parikh SV, Wasylenki D, Goering P, Wong J (1996) Mood disorders: rural/urban differences in prevalence, health care utilization, and disability in Ontario. J Affect Disord 38: Patten SB (2000) Major depression prevalence in Calgary. Can J Psychiatry 45: Potthoff RF (1994) Telephone sampling in epidemiological research: to reap the benefits, avoid the pitfalls. Am J Epidemiol 139: Watson EK, Firman DW, Heywood A, Hauquitz AC, Ring I (1995) Conducting regional health surveys using a computer-assisted telephone interviewing method. Austral J Pub Hlth 19: Roy A (1987) Five risk factors for depression. Br J Psychiatry 150: Andreasen NC, Endicott J, Spitzer RL, Winokur G (1977) The family history method using diagnostic criteria. Reliability and validity. Arch Gen Psychiatry 34: Kessler RC,Andrews G, Mroczek D, Ustun B,Wittchen HU (1998) The World Health Organization Composite International Diagnostic Interview Short Form (CIDI-SF). International Journal of Methods in Psychiatric Research 7: Patten SB, Brandon-Christie J, Devji J, Sedmak B (2000) Performance of the Composite International Diagnostic Interview Short Form for major depression in a community sample. Chron Dis Can 21: Epi Info,Version 6.04b (1997) Centers for Disease Control/World Health Organization, Atlanta, Geneva 21. Stata Version 7.0 (2001) Stata Corporation, College Station Texas 22. Kleinbaum DG, Morgenstern H, Kupper LL (1981) Selection bias in epidemiological studies. Am J Epidemiol 113: Henderson CCVNP (1991) A Canadian model for developing mental health services in rural communities through linkages with urban centres. The Journal of Mental Health Administration 18: Holley HL (1998) Geography and mental health: a review. Soc Psychiatry Psychiatr Epidemiol 33: