Gender and remission of mental illness.Gender differences in the prevalence of mental illness are well documented, (1-6) but whether gender also influences the timing of remission is unclear. Do the factors that contribute to a higher prevalence of illness among females also translate into a gender gap in the remission of illness? There is a good rationale to anticipate that gender is a factor in remission. For example, the literature suggests that gender dissimilarities in response to depression could lead to differences in the alleviation, complication or persistence of symptoms. (7-13) In addition, gender differences in the clinical features of illness could also influence remission. Yet, the literature provides inconsistent conclusions and largely focuses on mood disorders.
One group of studies argues that gender has a non-significant effect on remission. For example, Benedetti et al. (14) investigated whether gender influences the course of bipolar disorder. Their research showed that gender is non-significant in terms of the reduction of symptoms and number of recurrences of bipolar disorder. Benedetti et al. remarked that the effect of gender observed in other studies could be a result of ignoring dissimilarities in medical treatment. However, gender does not moderate the effect of pharmacological treatment for depression, according to Grubbe Hildebrandt et al. (15) Their research demonstrates that, given equivalent therapies, gender is a non-significant factor in the post-treatment outcomes of depression.
Other studies indicate that gender has a significant effect. (7-13) Riise and Lund (13) demonstrated that depressed females have a higher risk of depression at long-term follow-up. The authors observed that the risk of chronic depression among females increased after baseline level of illness had been adjusted for. Riise and Lund's findings could reflect a higher rate of relapse among females, not a lower remission rate, but they at least demonstrated that depression is a more persistent condition for females and that gender can influence long-term prognosis. Other studies confirm that females experience higher rates of chronic depression and relapse. (9-11)
Bland et al. (16) examined remission of several psychiatric disorders in a Canadian population and observed potential gender patterns that warrant the present research. However, their attention was directed toward age patterns of remission, and their analysis did not report whether there were statistically significant differences in gender patterns of remission. That said, Bland et al. presented some interesting gender-specific findings in 1-year remission rates. Their cross-tabulations suggest that there could be a female disadvantage. In general, about 26% of females achieve remission within 1 year, in contrast with 40% of males. Bland et al. also suggested that this female disadvantage is loaded on specific illnesses rather than representing a generalized effect.
The present research examines the effect of gender on the timing of remission (length of medical treatment) for mental illness. The analysis considers remission of a wide spectrum of mental disorders. Almost all previous studies have focused on a particular disorder or class of disorders, and most targeted depression. Hence, this research is an improvement on previous studies because it considers the general and disorder-specific effects of gender on remission and controls for the effects of comorbid conditions. Another limitation of previous studies is the reliance on data from a particular source (e.g., a specific psychiatric practice). Our data represent an entire population under treatment for a clinical mental illness in British Columbia. These data reduce the influence of potential site-specific biases, because they consist of hundreds of different treatment sites and service providers.
Our analysis used longitudinal data from the British Columbia Linked Health Database (BCLHD), which includes datasets on physician and specialist visits, hospitalizations and hospital separations, and tertiary and extended care. (17) The datasets are linked to Medical Service Plan (MSP) records. The MSP is a single-payer medical insurance plan that conforms with the Canada Health Act, guaranteeing universal and comprehensive medical coverage for all "medically necessary" hospitalizations, outpatient treatments and extended care. About 95% of BC residents are enrolled in the MSP. (18) Persons with a diagnosis of a clinical mental illness (ICD-9 diagnostic codes 290-314) are eligible to receive public direct health care, as provided through provincial/ regional agencies, private practitioners and hospitals. The MSP does not offer comprehensive coverage for milder conditions (e.g., subsyndromal depression), but general practitioners often treat these, and such treatment is also an MSP-billable service.
The target population was all BC residents aged 18 and older who started treatment for an ICD-9 diagnosis of mental illness in 1990. The study followed a 10% random sample of these patients from 1990 to 2001. It excluded cases missing core variables (e.g., care episode and gender). The final study sample consisted of 5,118 female and 2,470 male patients, and a total of 10,137 care episodes (cases). These cases were complete for patients with admission dates in 1990 and discharge dates before or in 2001. The information for ongoing (censored) cases was unavailable.
Our dependent variable was length of treatment for a mental illness. It was time-invariant at the episode level but time-variant at the individual level when the respondent experienced multiple, non-concurrent care episodes (separate cases) during the period of observation. A "care episode" refers to MSP-billable contact with a health care professional, and it represents the formal diagnosis of illness and commencement of treatment. We subtracted the date of discontinuation of treatment from the date of first contact to measure the timing of remission (symptom resolution) in terms of days of treatment.
For our purposes, a psychiatrist grouped the numerous specific diagnoses of mental illness contained in BCLHD data into nine distinct classes of illness: 1) alcohol/substance abuse, 2) delirium, 3) psychoses, 4) mood disorders, 5) anxiety disorders, 6) adjustment disorders, 7) dementia, 8) conditions needing counseling (e.g., bereavement, relationship difficulties, school-related problems) and 9) other disorders (e.g., sexual disorders, sleep disorders, pain disorders). This categorical variable was introduced to control for the effect of type of illness and the effect of comorbid illness on the timing of remission. Table 1 presents the definitions and descriptive statistics for all selected variables.
The regression models included controls for several other control variables. The analysis controlled for age, marital status, Aboriginal status, geographic location and socio-economic status, which have well-established effects on the prevalence of mental illness, remission of illness and access to services. (5,16,19,20)
The generalized estimating equations (GEE) method (marginal models) was used to estimate average group-level (male-female) differences in length of treatment. Length of treatment was measured as a discrete count variable (days of treatment), which was assumed to follow the Poisson distribution. Because care episodes at the patient level are sequential, we treated them as repeated measurements in the longitudinal design. The GEE method is well suited for analyzing repeated measurements. (21,22) The GEE model allows the number and spacing of the repeated measurements to vary among individuals. It assumes that observations for each individual are correlated, though observations among individuals are assumed to be independent. We assumed that the correlation is constant (exchangeable) between any two observation times and used an exchangeable correlation model. The GEE models were estimated using the GEN-MOD procedure in SAS (Statistical Analysis System) version 9.1.
Table 1 presents a bivariate examination of gender differences in remission from mental illness. Length of treatment (symptom resolution) serves as a proxy for remission. These initial results illustrate an important and encouraging finding: gender appears to be a non-significant factor in remission from mental illness in general. The average length of treatment is about 208 days for females and 203 days for males, a small but non-significant difference in the timing of remission. This finding implies that gender disparities in mental illness do not complicate treatment or prolong remission. Hence, even though gender still could influence responses to mental illness, this does not seem to have an effect on the duration of illness.
Table 1 also illustrates the gender-specific frequencies of the types of mental illnesses that were treated. In general, mood disorders accounted for a large proportion of all treatment received and represented the most prevalent illness in this respect. About 24% of females and 21% of males were treated for a mood disorder. This represents a significant difference at the p<0.001 level. There are similar gender differences in the treatment of anxiety and adjustment disorders. About 13% of females and 7% of males were treated for an anxiety disorder, and 17% of females and 13% of males were treated for an adjustment disorder. Again, these differences are significant at the 0.001 level. There are also significant gender differences in the treatment of alcohol/substance dependencies, psychoses, delirium and dementia, as a higher proportion of males than females were treated for these illnesses.
As Table 2 suggests, gender differences in remission could depend on the type of illness. The average length of treatment for mood disorders was 264 days for females and 223 days for males, a difference of about 6 additional weeks of treatment for females. The length of treatment for females also appears to be longer for anxiety disorders (18 additional days), adjustment disorders (45 additional days) and other illnesses. These findings warrant concern, and the objective of the subsequent regression analysis was to determine whether the differences contributed to significant overall or illness-specific gender differences in the timing of remission.
Table 3 presents GEE results for the effects of gender and other selected variables on remission. Model 1 considers the effect of gender alone. There was no gender difference according to this model, which indicates that, on average, the timing of remission is not longer for females than for males. In consideration of the gender differences in length of treatment shown in Table 2, Model 2 estimates the effect of type of illness on remission and controls for comorbidities. In comparison with counseling-needed conditions, all illnesses except addictions showed increased length of treatment. However, type of illness is a non-significant factor in terms of gender differences in the timing of remission. Model 3 introduces demographic and socio-economic variables to those considered previously. A modest but significant (p<0.001) gender difference emerges in this model, which demonstrates that gender is indeed a determinant of remission, net of demographic and socio-economic characteristics.
Table 4 explores whether there are gender differences in remission in terms of specific types of illness. Although we observed a small general difference in the timing of remission between females and males, this aggregated analysis could mask important gender differences in remission from specific illness. As Table 4 shows, however, the general pattern observed in Table 3 is accurate. There are no gender differences in the timing of remission of specific illnesses, and it is only in cumulative terms that a gender difference in remission is observable.
This study examined gender differences in remission from a spectrum of mental illnesses. The analysis indicates that more females received treatment for an illness than males. This higher rate of treatment could correspond to a greater need among females, but it could also involve an underutilization of service among males. A greater proportion of females received treatment for mood disorders, anxiety disorders and adjustment disorders. A greater proportion of males received treatment for addictions, delirium and psychoses. There are also gender differences in average length of treatment for each type of illness considered. In several instances, females appear to need a longer course of treatment. Although our regression analysis confirms that the timing of remission is somewhat longer for females it provides no clear explanation for this finding. However, a possible candidate is marital status, for there are a disproportionate number of single (never married) males in our sample. Prior research demonstrates that singlehood represents a remission disadvantage. (23) To test this conjecture (unreported analysis), we removed marital status from Model 3 in Table III. As anticipated, gender becomes non-significant in the revised model (p=0.10), suggesting that this demographic characteristic (more single males) suppresses a slight comparative disadvantage in remission among females. These conclusions are limited inasmuch as these data include only individuals treated for serious clinical illnesses and thus may not be generalizable to less severe disorders.
Acknowledgements: The authors acknowledge Chi Zheng and Ruth Kampen for their research assistance and the Canadian Institutes of Health Research, Institute of Neurosciences, Mental Health, and Addiction, for financial support.
Received: February 9, 2009
Accepted: August 6, 2009
(1.) Carter-Snell C, Hegadoren K. Stress disorders and gender: Implications for theory and research. Can J Nurs Res 2003;35:34-55.
(2.) Denton M, Prus S, Walters V. Gender differences in health: A Canadian study of the psychosocial, structural, and behavioural determinants of health. Soc Sci Med 2004;58:2585-600.
(3.) Linzer M, Spitzer R, Kroenke K, Williams JB, Hahn S, Brody G, et al. Gender, quality of life, and mental disorders in primary care: Results from the PRIME-MD study. Am J Med 1996;101:526-33.
(4.) Hopcroft RL, Burr-Bradley D. The sex difference in depression across 29 countries. Soc Forces 2007;85:1484-507.
(5.) Health Canada. A Report on Mental Illness in Canada. Ottawa, ON: Health Canada, 2002.
(6.) Romans SE, Tyas J, Cohen MM, Silverstone T. Gender differences in the symptoms of major depressive disorder. J Nerv Ment Dis 2007;195:905-11.
(7.) Enns MW, Cox BJ. Psychosocial and clinical predictors of symptom persistence vs remission in major depressive disorder. Can J Psychiatry 2005;50:769-77.
(8.) Frank E, Rucci P, Katon W, Barrett J, Williams JW Jr., Oxman T, et al. Correlates of remission in primary care patients treated for minor depression. Gen Hosp Psychiatry 2002; 24:12-19.
(9.) Kessler R, McGonagle K, Swartz M, Blazer D, Nelson C. Sex and depression in the National Comorbidity Survey. I: Lifetime prevalence, chronicity, and reoccurrence. J Affect Disord 1993;29:85-96.
(10.) Kornstein S. The evaluation and management of depression in women across the life span. J Clin Psychiatry 2001;62:11-17.
(11.) Kuehner C. Gender differences in the short-term course of unipolar depression in a follow-up sample of depressed inpatients. J Affect Disord 1999;56:127-39.
(12.) Oldehinkel AJ, Ormel J, Neeleman J. Predictors of time to remission from depression in primary care patients: Do some people benefit more from positive life change than others? J Abnorm Psychol 2000;109:299-307.
(13.) Riise T, Lund A. Prognostic factors in major depression: A long-term follow-up study of 323 patients. J Affect Disord 2001;65:297-306.
(14.) Benedetti A, Fagiolini A, Casamassima F, Mian MS, Adamovit A, Musettit L, et al. Gender differences in bipolar disorder type 1: A 48-week prospective follow-up of 72 patients treated in an Italian tertiary care center. J Nerv Ment Dis 2007;1995:93-96.
(15.) Hildebrandt MG, Steyerberg EW, Stage KB, Passchier J, Kragh-Soerensen P. Are gender differences important for the clinical effects of antidepressants? Am J Psychiatry 2003;160:1643-50.
(16.) Bland RC, Newman SC, Orn H. Age and remission of psychiatric disorders. Can J Psychiatry 1997;42:722-29.
(17.) Chamberlayne R, Green B, Barer ML, Hertzman C, Lawrence WJ, Sheps SB. Creating a population-based linked health database: A new resource for health services research. Can J Public Health 1998;89:270-73.
(18.) Michalak EE, Goldner EM, Jones W, Oetter HM, Lam RW. The management of depression in primary care: Current state and a new team approach. BC Med J 2002;44:408-11.
(19.) Kirmayer LJ, Gill K, Ternar Y, Boothroyd Quesney C, Smith A, Ferrara N, Hayton B. Emerging Trends in Research on Mental Health among Canadian Aboriginal Peoples. A report prepared for the Royal Commission on Aboriginal Peoples. Ottawa, ON: Government of Canada, 1994.
(20.) Maddess R. Mental health care in rural British Columbia. BC Med J 2006;48:172-73.
(21.) Diggle PJ, Heagerty K, Liang S, Zeger L. The Analysis of Longitudinal Data. Oxford: Oxford University Press, 2002.
(22.) Liang K, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73:13-22.
(23.) Wu Z, Penning MJ, Zheng C, Schimmele CM. Marital status and mental health convalescence. Paper presented at the American Sociological Association Annual Meeting; August 11-14, 2007, New York.
Christoph M. Schimmele, MA, Zheng Wu, PhD, Margaret J. Penning, PhD
Department of Sociology, University of Victoria, BC
Correspondence and reprint requests: Christoph M. Schimmele, P.O. Box 3050 STN CSC, Cornett A333, 3800 Finnerty Rd., Victoria, BC V8W 3P5, E-mail: email@example.com.
Table 1. Definitions and Gender Differences in the Variables Used in the Analysis: Canadian Adults (Age 18+), 1990 Variable Definition Females Care episode in days Length of treatment in 207.6 days (range 0-4,178 days) Type of illness Alcohol/substance Dummy indicator (1 = yes, 0 = no) 2.4% Delirium Dummy indicator (1 = yes, 0 = no) 3.6% Psychoses Dummy indicator (1 = yes, 0 = no) 4.1% Mood disorders Dummy indicator (1 = yes, 0 = no) 24.0% Anxiety disorders Dummy indicator (1 = yes, 0 = no) 12.5% Adjustment Dummy indicator (1 = yes, 0 = no) 16.5% disorders Dementia Dummy indicator (1 = yes, 0 = no) 5.1% Other conditions Dummy indicator (1 = yes, 0 = no) 16.0% Counseling-needed Reference group 15.9% conditions Age Age at diagnosis (range 18-105) 47.44 Marital status Single Dummy indicator (1 = yes, 0 = no) 17.9% Separated/divorced Dummy indicator (1 = yes, 0 = no) 23.3% Widowed Dummy indicator (1 = yes, 0 = no) 16.2% Married/cohabiting Reference group 42.7% Aborginal Dummy indicator (1 = yes, 0 = no) 3.8% Rurality Urban fringe Dummy indicator (1 = yes, 0 = no) 3.7% Rural fringe Dummy indicator (1 = yes, 0 = no) 7.0% Urban areas outside Dummy indicator (1 = yes, 0 = no) 15.5% CMAs/CAs ([dagger]) Rural areas outside Dummy indicator (1 = yes, 0 = no) 9.8% CMAs/CAs ([dagger]) Urban core Reference group 63.9% Work outside home Dummy indicator (1 = employed full/ 28.8% part time, 0 = otherwise) Household income Household income in decile 4.98 (range: 1-10) 5,118 N Variable Males p value * Care episode in days 202.5 0.576 Type of illness Alcohol/substance 7.0% <0.001 Delirium 5.8% <0.001 Psychoses 8.3% <0.001 Mood disorders 20.6% 0.001 Anxiety disorders 7.4% <0.001 Adjustment 12.8% <0.001 disorders Dementia 6.4% 0.019 Other conditions 17.5% 0.089 Counseling-needed 14.2% 0.064 conditions Age 49.48 <0.001 Marital status Single 28.8% <0.001 Separated/divorced 20.6% 0.009 Widowed 8.2% <0.001 Married/cohabiting 42.4% 0.828 Aborginal 3.7% 0.954 Rurality Urban fringe 3.9% 0.807 Rural fringe 7.0% 0.962 Urban areas outside 17.4% 0.038 CMAs/CAs ([dagger]) Rural areas outside 10.2% 0.572 CMAs/CAs ([dagger]) Urban core 61.5% 0.040 Work outside home 29.2% 0.673 Household income 5.04 0.359 2,470 N * p values are obtained from bivariate logit models of gender and each of the explanatory variables. ([dagger]) CMA, census metropolitan area; CA, census agglomeration Table 2. Average Length of Treatment (in Days) by Type of Illness and Gender: Canadian Adults (Age 18+), 1990 * Female Males Type of Illness Mean SD Mean SD Alcohol/substance 141.9 262.4 107.0 210.6 Delirium 185.3 275.4 218.4 315.1 Psychoses 571.3 722.8 558.2 810.1 Mood disorders 264.0 424.5 223.1 342.6 Anxiety disorders 205.6 337.9 187.5 293.0 Adjustment disorders 174.9 310.3 129.3 212.4 Dementia 237.6 299.3 195.8 272.1 Other conditions 145.3 259.6 154.4 281.9 Counseling needed 132.1 272.4 141.7 342.7 conditions N 5,118 2,470 * Self-weighted data Table 3. Generalized Estimating Equations for Effect of Gender and Selected Explanatory Variables on Length of Treatment, by Model: Canadian Adults (Age 18+), 1990-2001 Variable Model 1 Model 2 Model 3 Female (1 = yes) -0.003 0.060 0.124 *** Type of illness Alcohol/substance - -0.025 -0.045 Delirium - 0.368 *** 0.274 *** Psychoses - 1.297 *** 1.144 *** Mood disorders - 0.647 *** 0.618 *** Anxiety disorders - 0.389 *** 0.399 *** Adjustment disorders - 0.188 ** 0.236 *** Dementia - 0.450 *** 0.360 *** Other conditions - 0.080 0.078 Counseling-needed conditions ([dagger]) Age - - 0.038 *** Age square (x 100) - - -0.030 *** Marital status Single - - 0.382 *** Separated/divorced - - -0.072 Widowed - - -0.057 Married/cohabiting ([dagger]) Aboriginal (1 = yes) - - -0.223 * Rurality Urban fringe - - 0.070 Rural fringe - - -0.016 Urban areas outside - - 0.089 CMAs/CAs ([double dagger]) Rural areas outside - - -0.114 * CMAs/CAs ([double dagger]) Urban core ([dagger]) Work outside home (1 = yes) - - -0.168 *** Household income - - -0.025 *** Intercept 5.317 *** 4.846 *** 3.911 *** Log likelihood 10778255 10778496 10778579 [DELTA] Log likelihood 240.7 *** 83.4 *** d.f. - 8 12 * p<0.05, ** p<0.01, *** p<0.001 (two-tailed test) ([dagger]) Reference category ([double dagger]) CMA, census metropolitan area; CA, census agglomeration Table 4. Generalized Estimating Equations of Effect of Selected Explanatory Variables on Length of Treatment, by Gender: Canadian Adults (Age 18+), 1990-2001 Variable Females Males p value Type of illness Alcohol/substance -0.006 -0.085 0.716 Delirium 0.231 * 0.293 0.757 Psychoses 1.149 *** 1.080 *** 0.689 Mood disorders 0.662 *** 0.539 *** 0.364 Anxiety disorders 0.429 *** 0.349 * 0.616 Adjustment disorders 0.306 *** 0.070 0.112 Dementia 0.446 *** 0.206 0.204 Other conditions 0.102 0.008 0.530 Counseling-needed conditions ([dagger]) Age 0.044 *** 0.026 ** 0.110 Age square (x 100) 0.000 *** 0.000 * 0.157 Marital status Single 0.358 *** 0.384 *** 0.821 Separated/divorced -0.020 -0.219 ** 0.047 Widowed -0.024 -0.092 0.614 Married/ cohabiting ([dagger]) Aboriginal (1 = yes) -0.153 -0.413 * 0.232 Rurality Urban fringe 0.077 0.046 0.864 Rural fringe -0.136 0.211 0.030 Urban areas outside 0.082 0.107 0.803 CMAs/CAs [double dagger] Rural areas outside -0.077 -0.193 * 0.330 CMAs/CAs ([double dagger]) Urban core ([dagger]) Work outside home -0.116 * -0.257 *** 0.118 (1 = yes) Household income -0.023 ** -0.029 ** 0.683 Intercept 3.832 *** 4.318 *** 0.107 Log likelihood 7279438 3499159 * p<0.05, ** p<0.01, *** p<0.001 (two-tailed test) ([dagger]) Reference category ([double dagger]) CMA, census metropolitan area; CA, census agglomeration