Risk Factors for Depression Among Elderly Community Subjects: A Systematic Review and Meta-Analysis
Abstract
OBJECTIVE: The goal of this study was to determine risk factors for depression among elderly community subjects. METHOD: MEDLINE and PsycINFO were searched for potentially relevant articles published from January 1966 to June 2001 and from January 1967 to June 2001, respectively. The bibliographies of relevant articles were searched for additional references. Twenty studies met the following six inclusion criteria: original research reported in an English or French publication, study group of community residents, age of subjects 50 years or more, prospective study design, examination of at least one risk factor, and use of an acceptable definition of depression. The validity of studies was assessed according to the four primary criteria for risk factor studies described by the Evidence-Based Medicine Working Group. Information about group size at baseline and follow-up, age, proportion of men, depression criteria, exclusion criteria at baseline, length of follow-up, number of incident cases of depression, and risk factors was abstracted from each report. RESULTS: Follow-up of the inception cohort was incomplete in most studies. In the qualitative meta-analysis, risk factors identified by both univariate and multivariate techniques in at least two studies each were disability, new medical illness, poor health status, prior depression, poor self-perceived health, and bereavement. In the quantitative meta-analysis, bereavement, sleep disturbance, disability, prior depression, and female gender were significant risk factors. CONCLUSIONS: Despite the methodologic limitations of the studies and this meta-analysis, bereavement, sleep disturbance, disability, prior depression, and female gender appear to be important risk factors for depression among elderly community subjects.
Major depression occurs in 1% to 3% of the general elderly population (1, 2), and an additional 8% to 16% have clinically significant depressive symptoms (1, 3). The prognosis of these depressive states is poor. A meta-analysis of outcomes at 24 months estimated that only 33% of subjects were well, 33% were depressed, and 21% had died (4). Moreover, studies of depressed adults (5, 6) indicate that those with depressive symptoms, with or without depressive disorder, have poorer functioning, comparable to or worse than that of people with chronic medical conditions such as heart and lung disease, arthritis, hypertension, and diabetes (7). In addition to poor functioning, depression increases the perception of poor health (7), the utilization of medical services (8), and health care costs (9).
The preceding findings suggest that depression in elderly community subjects is a serious problem. Nonetheless, probably fewer than 20% of cases are detected or treated (2, 4). Even among those detected and treated, the effectiveness of interventions appears to be modest (10). Escalating health care costs and shrinking health care resources challenge health care professionals to find more effective and less expensive approaches to depression in the elderly.
The success of a program for preventing delirium among elderly medical inpatients (11) offers hope that a similar intervention model may be useful in preventing depression among elderly community subjects. This program involved identification of elderly medical inpatients with at least one of six targeted risk factors for delirium and implementation of standardized intervention protocols for each of the risk factors present. The program attenuated the risk factors and reduced the incidence of delirium by 40%. To develop a similar intervention model for preventing depression among elderly community subjects, risk factors for depression in this population must be defined. Thus, the purpose of this investigation was to determine risk factors for depression among elderly community subjects by systematically reviewing original research on this topic. The review process, modified from the one described by Oxman et al. (12), involved systematic selection of articles, assessment of validity, abstraction of data, and qualitative and quantitative synthesis of results.
Method
Selection of Articles
The selection process involved four steps. First, two computer databases, MEDLINE and PsycINFO, were searched for potentially relevant articles published from January 1966 to June 2001 and from January 1967 to June 2001, respectively. For MEDLINE, the key words “depression,” “risk factor,” and “aged” and the text word “community” were used; for PsychINFO, the same words were used as text words. Second, relevant articles (judged on the basis of the title and abstract) were retrieved for more detailed evaluation. Third, the bibliographies of relevant articles were searched for additional references. Finally, all retrieved articles were screened to determine which met the following six inclusion criteria: 1) original research published in English or French, 2) study group of community residents, 3) subjects age 50 years or older, 4) prospective design that excluded subjects who were depressed at baseline (or controlled for baseline depression in the analysis), 5) study of at least one risk factor for depression, and 6) acceptable definition of depression (either recognized diagnostic criteria or cutoff on a depression rating scale).
Assessment of Validity
To determine validity, the methods of each study were assessed according to the four primary criteria for risk factor studies described by the Evidence-Based Medicine Working Group (13): 1) clearly identified comparison groups that were similar with respect to important determinants of outcome, other than the one of interest (or analysis that controlled for differences in important determinants), 2) measurement of exposures and outcomes in the same way, 3) a sufficiently long follow-up (i.e., 1 year), and 4) a sufficiently complete follow-up (i.e., including 80% of inception cohort). Each study was scored with respect to meeting (+) or not meeting (–) each of the these criteria.
Abstraction of Data
Information about the size of the study group at baseline and follow-up, subjects’ age, proportion of men, criteria for depression, exclusion criteria at baseline, length of follow-up, number of incident cases of depression, and risk factors was abstracted from each report.
Data Synthesis
Qualitative
All abstracted information was tabulated. A qualitative meta-analysis was conducted by summarizing, comparing, and contrasting the abstracted data.
Quantitative
A quantitative meta-analysis was conducted for risk factors with usable data from two or more studies. To obtain a pooled estimate of the odds of depression associated with each risk factor, we conducted a meta-analysis using a Bayesian hierarchical (random effects) model (14). In the Bayesian framework, information available before the analysis is combined with the observed data to obtain a posterior distribution for the parameters of interest (14). We assumed no prior information was available. The variance between odds ratios from different studies is a measure of the heterogeneity of the studies. A Bayesian 95% posterior credible interval may be interpreted in a straightforward manner as an interval that contains the parameter of interest with 95% probability given the observed data. We also estimated the probability that the pooled odds ratio was greater than 1.
Results
Selection of Articles
The search strategy yielded 130 potentially relevant studies; 45 were retrieved for more detailed evaluation. Twenty studies (15–34) met the inclusion criteria (Table 1). The other 25 studies were excluded for the following reasons: four did not meet the age criterion, 16 were not prospective, two did not study at least one risk factor, and three did not meet two or more of the inclusion criteria.
Assessment of Validity
Six studies met all of the criteria. Most studies had incomplete follow-up of the inception cohort (Table 2).
Data Synthesis
Qualitative
The 20 included studies (Table 1) involved more than 23,058 subjects at baseline, more than 20,678 subjects at follow-up, and more than 1,694 subjects with incident depression. The numbers of subjects at baseline and follow-up ranged from 141 to 3,747 and from 79 to 5,449 subjects, respectively. The subjects’ mean ages were reported in 13 articles (mean=58–85 years). Nineteen articles included gender distribution: 0%–65% of subjects were men (median=41%). The length of reported follow-up ranged from 3 to 96 months (median=24). Nine studies used DSM or structured interview criteria to diagnose depression, nine used a cutoff on a depression rating scale, and two used both. Among the 17 reports that included the frequency of incident depression, the frequencies ranged from 1.8% to 24.1% (median=12.0%) and were generally higher in studies using cutoffs on rating scales than in those using diagnostic criteria.
Forty-two different risk factors were studied by univariate analysis, 25 in two or more studies and 17 in one study each (Table 3). Disability, being older, female gender, new medical illness, poor health status, sleep disturbance, prior depression, less education, cognitive impairment, new disability, poor self-perceived health, poor social support, bereavement, and vision or hearing impairment were identified as risk factors for depression in at least two studies each.
Forty-three risk factors were studied by multivariate analysis, 15 in two or more studies and 28 in one study each (Table 4). Disability, bereavement, new medical illness, poor health status, female gender, prior depression, sleep disturbance, and poor self-perceived health were identified as risk factors for depression in at least two studies each. Risk factors identified by both univariate and multivariate techniques in at least two studies each were disability, female gender, new medical illness, poor health status, prior depression, sleep disturbance, poor self-perceived health, and bereavement.
Quantitative
Only 13 risk factors had data that could be used in the quantitative meta-analysis (Table 5 and Figure 1). The combined odds ratios ranged from 1.0 to 3.3. Greater heterogeneity was observed among studies evaluating lower education, disability, poor health status, cognitive impairment, prior depression, and new medical illness as risk factors for depression. On the basis of the combined odds ratios (and their 95% credible intervals) and the posterior distributions of the odds ratios (pooled odds ratio >1), the following were found to be significant risk factors for depression: bereavement, sleep disturbance, disability, prior depression, and female gender. Higher age, lower education level, being unmarried, and poor social support did not appear to be risk factors. Poor health, cognitive impairment, living alone, and new medical illness were uncertain risk factors.
Discussion
The combined results of 20 prospective studies of risk factors for depression among elderly community subjects indicate that five factors (bereavement, sleep disturbance, disability, prior depression, and female gender) are significant risk factors for depression. The median interval between the determinations of risk factor status and depression status was 24 months.
Notably, three of these risk factors are potentially modifiable, namely, bereavement, sleep disturbance, and disability. Based on the pooled odds ratios data in this meta-analysis, the attributable risks for these three risk factors were 69.4% (95% credible interval=42.2–79.5), 57.0% (95% credible interval=35.7–73.3), and 56.5% (95% credible interval=20.4–83.5), respectively. Thus, a large proportion of depression among elderly people in the community may be attributed to one of these risk factors. Because these risk factors are frequent in elderly community subjects, their modification could be expected to have an important public health impact.
Elderly populations could be screened to identify individuals at high risk of depression (e.g., bereaved women with prior depression, disability, and sleep disturbance). Subsequently, these individuals could be targeted for interventions to abate the three potentially modifiable risk factors and reduce the risk of depression. Such interventions might include education about the significance of the risk factors, bereavement counseling and support (35), new skills training, “maintenance of routines” protocols (36), enhancement of social supports (37), individual or group therapy to facilitate adjustment to loss of function (38), and sleep enhancement protocols (39).
These five risk factors may serve two other purposes (40). First, they could identify whole populations at high risk of depression and aid the development of population-based interventions to reduce the frequency of depression. Second, they could focus treatment on the most important putative contributing factors (e.g., bereavement, loss of function, sleep disturbance).
The finding that bereavement is an important risk factor for depression contradicts the results of the Epidemiologic Catchment Area (ECA) study (41), which indicated low rates of bereavement-related depression in the elderly. However, it has been argued that the ECA study probably failed to diagnose the low-level symptomatic forms of depression experienced by many elderly (42).
This review has 10 potential limitations. First, the search of the literature was conducted by one author only. Second, the search was limited to articles published in English or French. Third, we did not assess publication bias, although it is unlikely that this bias influences publication of risk factor studies. Fourth, the data were abstracted by one author only. Fifth, follow-up of the enrolled cohort was incomplete in most studies; however, the results of studies with and without complete follow-up were similar. Sixth, examination of depression status was complicated by differences in the length of follow-up; nonetheless, there were no consistent differences in reported risk factors by length of follow-up. Seventh, the examination of the results of the univariate and multivariate analyses was complicated by differences in the definitions of some risk factors from one study to the next, and the examination of the results of the multivariate analyses was complicated by adjustments for different variables in different studies. Eighth, we have identified with some confidence five factors that increase the risk of depression and four factors (higher age, lower education level, being unmarried, poor social support) that do not appear to increase the risk of depression; however, many potential risk factors have not been studied adequately. Ninth, in this meta-analysis, we could not determine whether the simultaneous presence of multiple risk factors results in a cumulative increase in the risk of depression; however, the results of four studies included in this meta-analysis (15, 18, 19, 29) suggest that different risk factors play both additive and interactive roles. Finally, there was heterogeneity in the results for some risk factors (i.e., lower education level, disability, poor health status, cognitive impairment, prior depression, new medical illness), perhaps related to different definitions of these variables in different studies and small study groups in some studies; consequently, the results of the meta-analysis for these risk factors must be interpreted cautiously.
To conclude, five risk factors for depression among elderly community subjects include bereavement, sleep disturbance, disability, prior depression, and female gender. Despite the methodologic limitations of the studies and this meta-analysis, these findings may guide efforts to develop programs to prevent depression in this population.
 |
 |
 |
 |
 |
Received April 16, 2002; revisions received Oct. 8 and Nov. 25, 2002; accepted Dec. 2, 2002. From the Department of Psychiatry, St. Mary’s Hospital and McGill University; the Department of Clinical Epidemiology and Community Studies, St. Mary’s Hospital; and the Department of Epidemiology and Biostatistics, McGill University, Montreal. Address reprint requests to Dr. Cole, Department of Psychiatry, St. Mary’s Hospital, 3830 Lacombe Ave., Montreal, Quebec H3T 1M5, Canada.
Figure 1. Individual and Combined Odds Ratios and 95% Credible Intervals in Prospective Studies of Risk Factors for Depression Among the Elderly
aCredible interval extends to 0.09.
bCredible interval extends to 14.9.
cCredible interval extends to 0.07.
dCredible interval extends to 11.7.
eCredible interval extends to 12.8.
fCredible interval extends to 10.5.
gCredible interval extends to 14.5.
hCredible interval extends to 10.1.
1. NIH Consensus Development Conference: Diagnosis and treatment of depression of late life. JAMA 1992; 268:1018-1029Crossref, Medline, Google Scholar
2. Cole MG, Yaffe MJ: Pathway to psychiatric care of the elderly with depression. Int J Geriatr Psychiatry 1996; 11:157-161Crossref, Google Scholar
3. Blazer D: Depression in the elderly. N Engl J Med 1989; 320:164-166Crossref, Medline, Google Scholar
4. Cole MG, Bellavance F, Mansour A: Prognosis of depression in elderly community and primary care populations: a systematic review and meta-analysis. Am J Psychiatry 1999; 156:1182-1189Abstract, Google Scholar
5. Gurland BJ, Wilkder DE, Berkman C: Depression and disability in the elderly: reciprocal relations and changes with age. Int J Geriatr Psychiatry 1988; 3:163-179Crossref, Google Scholar
6. von Korff M, Ormel J, Katon W, Lin EH: Disability and depression among high utilizers of health care: a longitudinal analysis. Arch Gen Psychiatry 1992; 49:91-100Crossref, Medline, Google Scholar
7. Wells KB, Burman MA: Caring for depression in America: lessons learned from early findings of the Medical Outcomes Study. Psychiatr Med 1991; 9:503-519Medline, Google Scholar
8. Katon W, von Korff M, Lin E, Bush T, Ormel J: Adequacy and duration of antidepressant treatment in primary care. Med Care 1992; 30:67-76Crossref, Medline, Google Scholar
9. Unutzer J, Patrick DL, Simon G, Grembowski D, Walker E, Rutter C, Katon W: Depressive symptoms and the cost of health services in HMO patients aged 65 and over: a 4-year prospective study. JAMA 1997; 277:1618-1623Crossref, Medline, Google Scholar
10. McCusker J, Cole MG, Keller E, Bellavance F, Berard A: Effectiveness of treatments of depression in older ambulatory patients. Arch Intern Med 1998; 158:705-712Crossref, Medline, Google Scholar
11. Inouye SH, Bogartus ST, Charpentier PA, Summers L, Acampora D, Halford TR: A multicomponent intervention to prevent delirium in hospitalized older patients. N Engl J Med 1999; 340:669-676Crossref, Medline, Google Scholar
12. Oxman AD, Cook DJ, Guyatt GH: Users’ guides to the medical literature, VI: how to use an overview. JAMA 1994; 272:1367-1371Crossref, Medline, Google Scholar
13. Levine M, Walter S, Lee H, Haines T, Holbrook A, Moyer L: Users’ guides to the medical literature, IV: how to use an article about harm. JAMA 1994; 271:1615-1619Crossref, Medline, Google Scholar
14. Gelman A, Carlin JB, Stern HS, Rubin DV: Bayesian Data Analysis. Chapman & Hall, 1995Google Scholar
15. Phifer JF, Murrell SA: Etiologic factors in the onset of depressive symptoms in older adults. J Abnorm Psychol 1986; 95:282-291Crossref, Medline, Google Scholar
16. McHorney CA, Mor V: Predictors of bereavement depression and its health services consequences. Med Care 1988; 26:882-893Crossref, Medline, Google Scholar
17. Kennedy GJ, Kelman HR, Thomas C: The emergence of depressive symptoms in late life: the importance of declining health and increasing disability. J Community Health 1990; 15:93-104Crossref, Medline, Google Scholar
18. Harlow SD, Goldberg EL, Comstock GW: A longitudinal study of risk factors for depressive symptomatology in elderly widowed and married women. Am J Epidemiol 1991; 134:526-536Crossref, Medline, Google Scholar
19. Russell DW, Cutrona CE: Social support, stress and depressive symptoms among the elderly: test of a process model. Psychol Aging 1991; 6:190-201Crossref, Medline, Google Scholar
20. Green BH, Copeland JRM, Dewey ME, Sharma V, Saunders PA, Davidson IA, Sullivan C, McWilliam C: Risk factors for depression in elderly people: a prospective study. Acta Psychiatr Scand 1992; 86:213-217Crossref, Medline, Google Scholar
21. Livingston G, Blizard B, Mann A: Does sleep disturbance predict depression in elderly people? a study in inner London. Br J Gen Pract 1993; 43:445-448Medline, Google Scholar
22. Mendes de Leon CF, Kasl SV, Jacobs S: A prospective study of widowhood and the changes in symptoms of depression in a community sample of the elderly. Psychol Med 1994; 24:613-624Crossref, Medline, Google Scholar
23. Beekman ATF, Deeg DJH, Smit JH, van Tilburg W: Predicting the course of depression in the older population: results from a community-based study in the Netherlands. J Affect Disord 1995; 34:41-49Crossref, Medline, Google Scholar
24. Zeiss AM, Lewinsohm PM, Rohde P, Seeley JR: Relationship of physical disease and functional impairment to depression in older people. Psychol Aging 1996; 11:572-581Crossref, Medline, Google Scholar
25. Kivela SL, Kongas-Saviard P, Kimmo P, Kesti E, Laippala P: Health, health behaviour and functional ability predicting depression in old age: a longitudinal study. Int J Geriatr Psychiatry 1996; 11:871-877Crossref, Google Scholar
26. Prince MJ, Harwood RH, Thomas A, Mann AH: A prospective population-based cohort study of the effects of disablement and social milieu on the onset and maintenance of late-life depression: the Gospel Oak Project VII. Psychol Med 1998; 28:337-350Crossref, Medline, Google Scholar
27. Turvey CL, Carney C, Arndt S, Wallace RB, Herzog R: Conjugal loss and syndromal depression in a sample of elders aged 70 years or older. Am J Psychiatry 1999; 156:1596-1601Link, Google Scholar
28. Livingston G, Watkin V, Milne B, Manela MV, Katona C: Who becomes depressed? the Islington study of older people. J Affect Disord 2000; 58:125-133Crossref, Medline, Google Scholar
29. Schoevers RA, Beekman ATF, Deeg DJH, Geerlings MI, Jonker C, van Tilburg W: Risk factors for depression in later life: results of a prospective community based study (AMSTEL). J Affect Disord 2000; 59:127-137Crossref, Medline, Google Scholar
30. Geerlings SW, Beekman ATF, Deeg DJH, van Tilburg W: Physical health and the onset of and persistence of depression in older adults: an eight-wave prospective community-based study. Psychol Med 2000; 30:369-380Crossref, Medline, Google Scholar
31. Forsell Y: Predictors of depression, anxiety and psychotic symptoms on a very elderly population: data from a 3-year follow-up study. Soc Psychiatry Psychiatr Epidemiol 2000; 35:259-263Crossref, Medline, Google Scholar
32. Paterniti S, Verdier-Taillefer M, Geneste C, Bisserbe J, Alperrovitch A: Low blood pressure and risk of depression in the elderly. Br J Psychiatry 2000; 176:464-467Crossref, Medline, Google Scholar
33. Roberts RE, Shema SJ, Kaplan GA, Strawbridge WJ: Sleep complaints and depression in an aging cohort: a prospective perspective. Am J Psychiatry 2000; 157:81-88Link, Google Scholar
34. Kritz-Silverstein D, Barrett-Connor E, Corbeau C: Cross-sectional and prospective study of exercise and depressed mood in the elderly. Am J Epidemiol 2001; 153:596-603Crossref, Medline, Google Scholar
35. Marmar CR, Horowitz MJ, Weiss DS, Wilner NR, Kaltreider NS: A controlled trial of brief psychotherapy and mutual-help group treatment of conjugal bereavement. Am J Psychiatry 1988; 145:203-209Link, Google Scholar
36. Brown LF, Reynolds CF, Monk TH, Prigerson HG, Dew MA, Houck PR, Mezumders S, Buysse DJ, Hach CC, Kupfer DJ: Social rhythm stability following late-life spousal bereavement: associations with depression and sleep impairment. Psychiatry Res 1996; 62:161-169Crossref, Medline, Google Scholar
37. Morgan DL: Adjusting to widowhood: do social networks really make it easier? Gerontologist 1989; 29:101-107Crossref, Medline, Google Scholar
38. Anderson BL: Psychological interventions for cancer patients to enhance quality of life. J Consult Clin Psychol 1992; 60:552-568Crossref, Medline, Google Scholar
39. Morin CM, Colecchi C, Stone J, Sood R, Brink D: Behavioural and pharmacological therapies for late-life insomnia: a randomized controlled trial. JAMA 1999; 281:991-999Crossref, Medline, Google Scholar
40. Cole MG: Public health models of mental health care for elderly populations. Int Psychogeriatr 2002; 14:3-6Crossref, Medline, Google Scholar
41. Myers JK, Weisman MM, Tischler GD, Halzer CE, Leaf PU, Onaschel H, Anthony JC, Boyd JH, Burke JD, Kramer M, Stoltzman R: Six-month prevalence of psychiatric disorders in three communities. Arch Gen Psychiatry 1984; 41:959-967Crossref, Medline, Google Scholar
42. Tweed DC, Blazer DG, Gazlo JA: Psychiatric epidemiology in the elderly population, in The Epidemiologic Study of the Elderly. Edited by Wallace RB, Woolson RF. New York, Oxford University Press, 1992, pp 213-233Google Scholar
