| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
a Center for the Study of Aging and Human Development, Duke University Medical Center, Durham, NC
b Department of Psychiatry and Behavioral Sciences, Duke University Medical Center, Durham, NC
c Department of Community and Family Medicine, Duke University Medical Center, Durham, NC
Correspondence: Celia F. Hybels, PhD, Center for the Study of Aging and Human Development, Duke University Medical Center, Box 3003, Durham, NC 27710. E-mail: cfh{at}geri.duke.edu.
Laurence G. Branch, PhD
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Aging • Depressive symptoms • Epidemiology
Much research over the last decade has focused on residual cases of depression that do not meet criteria according to the nomenclature, but are potentially clinically significant. Wells and colleagues 1989
reported that many patients seen in primary care had depressive symptoms but failed to meet the criteria for major depression or dysthymia. Yet, these patients exhibited a decreased quality of life and more dysfunction and disability than did patients with either hypertension or diabetes. Similarly, Broadhead, Blazer, George, and Tse 1990 reported that individuals in the community with minor depression had 51% more disability days than persons with major depression. The number of days lost from work among those with minor depression was similar to that reported by individuals with major depression. Johnson, Weissman, and Klerman 1992 reported that depressive symptoms were associated with as much service utilization and social morbidity as clinical depression. Philipp and colleagues 1992 applied several existing criteria to a sample of both psychiatric inpatients and outpatients and found the definition for depression too restrictive. Relaxing the time criteria and reducing the necessary symptom count to introduce a category of minor depression reduced the number of Depression Not Otherwise Specified cases by 80%.
Minor depression was classified in the Research Diagnostic Criteria (RDC; Spitzer, Endicott, and Robins 1978), but it did not appear in the Diagnostic and Statistical Manual of Mental Disorders (DSM; American Psychiatric Association 1980) criteria until reintroduced in the appendix to DSM-IV (American Psychiatric Association 1994). The lifetime prevalence of minor depression reported from the National Comorbidity Study using an approximation of the DSM criteria for minor depression was 10% (Kessler, Zhao, Blazer, and Swartz 1997). Judd, Rapaport, Paulus, and Brown 1994 reported that the one-year prevalence of subsyndromal depression (defined as depressive symptomatology not meeting the criteria for major or minor depression or dysthymia) in adults 18 or older in the landmark Epidemiologic Catchment Area (ECA) data was 11.8%, higher than that for all diagnosed DSM-III (American Psychiatric Association 1980) mood disorders combined. While this phenomenon has been referred to by different names including subthreshold, subclinical, minor, mild, and subsyndromal depression, a summary of the research to date concludes that, regardless of label, depressive symptoms not meeting diagnostic criteria are prevalent and associated with morbidity and functional impairment (Pincus, Davis, and McQueen 1999). The specific criteria applied to capture these subthreshold symptoms have uniformly identified persons who do not meet traditional criteria, yet experience dysfunction secondary to depressive symptoms.
The issue of subthreshold depression is particularly relevant for elderly people. As older age often brings a decline in physical health and functioning, decreases in cognitive functioning, bereavement, loss of independence, reduced income and role loss through retirement, and other factors associated with depression, one would assume the prevalence of depression to be high among older adults. In a sample of community-dwelling elders, the prevalence of significant dysphoric symptomatology was 14.7%, whereas the prevalence of major depression was much lower, 3.7% (Blazer and Williams 1980). In the ECA study, the prevalence of DSM-III-defined major depressive disorder among those 65 or older was 1.0% compared to 2.3% in those aged 45–64 and 3.4% in those aged 18–44 (Robins and Regier 1991). These findings have led to suggestions that the current diagnostic criteria for depression may be less applicable to elders and need to be broadened to include depression as seen in older adults (Blazer 1994; Ernst and Angst 1995). Lyness, King, Cox, Yoediono, and Caine 1999 recently reported that subsyndromal depression in older primary care patients was more prevalent than major depression, minor depression, and dysthymia, and was associated with functional disability and medical comorbidity similar to that seen in major or minor depression.
The Longitudinal Aging Study Amsterdam (LASA) of community-dwelling adults aged 55–85 reported a one-month prevalence of minor depression of 12.9% (Beekman, Deeg, Van Tilberg, et al. 1995). Data from the LASA have shown that chronic physical illness, perceived poorer health, functional limitation, and lower scores on cognitive functioning tests are associated with minor depression. The LASA investigators defined minor depression as a score of 16 or more on the Center for Epidemiologic Studies–Depression scale (CES–D; Radloff 1977) but not meeting DSM-III criteria (American Psychiatric Association 1980) for major depression.
The objective of these analyses was to broaden the definition of subthreshold or minor depression and to examine the prevalence and correlates of subthreshold depression in a sample of community-dwelling elders. We hypothesized that: (1) The prevalence of subthreshold depression in older adults would be higher than that of CES–D-defined depression; (2) Subthreshold depression would be more prevalent in women, and the prevalence in both genders would increase with age; and (3) The symptom patterns and associations with demographic and social and physical health correlates observed in elders with subthreshold depression would be similar to those observed in individuals with CES–D-defined depression.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
). Briefly, the sample is a stratified multistage area probability sample in which African Americans were oversampled. A total of 4,000 interviews were completed with sample members and an additional 162 interviews were completed with proxy informants, for an overall baseline survey response rate of 80%. The baseline survey data were collected through in-person interviews. Written consent was obtained, and all data collection and processing were conducted according to a prescribed protocol that had been approved by the institutional review boards at participating institutions. Only sample members who participated themselves are included in these analyses because depression questions were not asked of proxy informants.
Subthreshold Depression and CES–D Depression.
The EPESE measured depressive symptomatology using the CES–D scale (Radloff 1977). The CES–D is a 20-item scale that asks the respondent to indicate to what extent he or she had a particular feeling the previous week. A modified version was used at the Duke EPESE site, where respondents were asked to indicate whether or not the feeling had been present the previous week. In its original form, the CES–D has a range of scores from 0 to 60 with a score of 16 or greater considered depressed. Beekman, Deeg, Van Limbeek, et al. 1997 studied the validity of the CES–D in an elderly community-based sample. Using the one-month prevalence of major depression derived from the Diagnostic Interview Schedule (Robins, Helzer, Croughan, and Radcliff 1981) as criterion, the CES–D had a positive predictive value of 13.2%. Although the majority of those depressed according to the CES–D did not fulfill the DSM-III criteria for major depression, the authors concluded the validity of the CES–D was satisfactory in their sample of older adults.
As described elsewhere (Blazer et al. 1991), a score of 9 or more symptoms on the modified scale used in the Duke EPESE was determined to be comparable to a score of 16 or greater on the original scale. In Fig. 1 we present the cumulative frequency by gender of CES–D-defined depressive symptoms observed in our sample. As expected, men were more likely to have fewer symptoms than women. However, as indicated by the smooth curves, no clearly defined cutpoints for CES–D-scale depression or subthreshold depression were observed for either gender. For purposes of these analyses, subthreshold depression was then arbitrarily defined as a score of 6 to 8 depressive symptoms on the modified scale to lower the threshold from the usual cutpoint for CES–D-scale depression. Sample members with fewer than six symptoms were classified as nondepressed. The categories are therefore mutually exclusive.
|
Physical Functioning and Disability.
Self-perceived health was classified as excellent, good, fair, or poor. The four-level variable was modeled as a continuous variable in the multivariable analysis, with "excellent" as the reference group. To measure chronic disease, we used a measure of health status (Fillenbaum, Leiss, Pieper, and Cohen 1998) that summed all chronic conditions present from those listed (heart problems, hypertension, diabetes, stroke, and cancer), with each condition weighted according to the estimated impact on overall health status. For the bivariate analyses, we used a dichotomous variable created from the full sample comparing the upper 35% to the lower 65%. For the regression analyses, we used the continuous variable with a range of 0 to 206, with 0 being no chronic disease. We used number of doctor visits as a continuous variable in the regression model, with a range of 0 to 100 visits. We measured self-reported disability and physical limitations in two ways. First, we asked, "During the past 3 months did you ever have to cut down on things you usually do because of illness or injury (not counting days in bed)?" to capture disability days. We also used a summary measure incorporating the seven activities of daily living (ADL) tasks identified by Katz, Downs, Cash, and Grotz 1970, three of the six items reported by Rosow and Breslau 1966, and the seven items from the instrumental ADL scale used in the Older Americans Resources and Services (OARS) survey (Duke University Center for the Study of Aging and Human Development 1978). Subjects having difficulty or needing help with two or more activities were classified in the bivariate analyses as having some limitations in physical functioning. We used the continuous variable with a range of 0 to 13 in the regression model, with 0 indicating no difficulty. The summary measure was an attempt to capture an overall rating of physical functioning.
Cognitive Functioning.
Cognitive impairment was assessed using the Short Portable Mental Status Questionnaire (SPMSQ; Pfeiffer 1975). We compared sample members with three or more errors to those with less than three, consistent with the cutpoint used by Pfeiffer to indicate mild to moderate impairment. Items marked "Can't do" or "Refused" were counted as errors. We had very few persons with six or more errors, most likely because the SPMSQ was used as a screening tool in the EPESE. Individuals with many errors participated by proxy because of cognitive reasons; they were not included in these analyses because depression was not ascertained through proxy respondents. These scores were not adjusted for race and education as done by Pfeiffer, because these demographic variables were included in the final regression model.
Use of Psychotropic Medications.
The coding of the medication data has been described elsewhere (Hanlon et al. 1992). The field interviewer obtained detailed information concerning prescription medications taken within the past 2 weeks or prescribed to be taken as needed. In these analyses, psychotropic medications included antidepressants, sedatives, hypnotics, and antianxiety and antipsychotic medications.
Perceived Social Support.
The Duke EPESE included multiple measures of social support. Previous research has shown that, although social support is multidimensional, perceived social support is significantly associated with adverse health outcomes (e.g., Blazer 1982). We used two questionnaire items to measure perceived social support, specifically, the degree to which a respondent feels he or she has someone to turn to if needed. The items were: "In times of trouble, can you count on at least some of your family or friends most of the time, some of the time, or hardly ever?" and "Can you talk about your deepest problems with at least some of your family or friends most of the time, some of the time, or hardly ever?" The responses to these items were summed, with a resulting range of 2 to 6. We classified a score of less than 5 as impaired for the bivariate analyses because the majority of the sample had a score of 6, and used the continuous measure for the regression with 2 being the reference group.
Data Analysis.
Weighted data were used for all analyses as well as for significance testing. The use of weights adjusted for the unequal probabilities of selection for each sample member. All analyses were first run using Statistical Analysis System (SAS) software (SAS Institute 1990) with a weight statement attached. The sample weights were downweighted to the original sample size for the purpose of significance testing in the initial analyses. The analyses were then run using SUrvey DAta ANalysis (SUDAAN) software (Research Triangle Institute 1997) to adjust for the clustering effect in the sample design.
We first conducted general descriptive analyses and looked at the bivariate associations between selected factors and each of the levels of depression. To simultaneously control for the effects of various factors, we employed ordinal logistic regression with a three-level depression variable as the outcome and the social and physical health variables as independent predictors of depression status. Depression data were available for 3,996 of the 4,000 participants. Data from 3,674 of these sample members were complete for all independent variables. Nearly all of the predictors had at least some missing values, but those with the most missing data were physician visits (4%) and limitations in physical functioning (2%). Persons with missing data on one or more of the variables used in the logistic regression analyses were removed from all analyses.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
|
|
To be sure the model assumptions of linearity were not violated, each of the continuous variables to be used in the model was tested individually in the full model. Specifically, each continuous variable was included along with higher order terms up to the fifth power. We used the log likelihood ratio test to see if the additional higher order terms as a group were significant. The only two variables with significant nonlinear relationships with depression were number of doctor visits and number of limitations in physical functioning. By progressively dropping out the highest order terms, we found physical functioning was only significant to the second power. We plotted the summary of the odds of the continuous variable and its squared term and found the odds of depression increased as the number of limitations increased to an approximate score of 7 and then began to decrease. To adjust for this nonlinear relationship, we included the squared term in the final model. The number of doctor visits was significant for the highest order term. Again, we plotted the summary of these odds and found the odds of depression increased as the number of doctor visits increased to an approximate value of 16 and then decreased to a value of 51, increased again to 62, and then decreased. To accommodate this nonlinear relationship, we included the four higher order terms in the model.
Finally, we assessed the model for collinearity between the independent variables and found that the social and physical health variables were relatively independent. Self-rated health had the lowest tolerance (0.69). Interaction terms between both age and gender and each of the social and physical health correlates were not significant.
The resulting odds ratio for the association between each of the independent variables and depression in our final logistic model compares the change in the relative odds for being in a more depressed group (CES–D-defined depression and subthreshold depression vs nondepressed and CES–D-scale depression vs subthreshold depression and nondepressed) for a one-unit change in the factor of interest. The associations of the sociodemographic variables of age, race/ethnicity, and education with depression were not significant (Table 5 ). Multiplying the beta coefficient for age, 0.0031, by 10 and exponentiating the result, we found the odds of being in a more depressed group are 1.03 for each decade-increase in age. Gender (p < .05) and marital status (p < .01) were associated with depression. The odds of a higher level of depression were 1.28 in women compared to men and 1.50 in unmarried men and women compared to married. Chronic health problems were not significantly associated with a higher depression group. The association between the number of doctor visits and depression group depends on the number of visits.
|
Impaired perceived social support and psychotropic medication use were also significantly associated with depression group (p < .001), whereas the association between cognitive impairment and depression group was not significant. The reference group for perceived social support was those with a score of 2 on the scale. The odds of belonging to a group with higher depression scores decreased as perceived social support increased. That is, perceived social support is protective. For example, relative to the lowest support group, the odds of belonging to a more depressed group were 0.67 for those with a value of 3 and 0.21 for those with a value of 6, the highest value on our scale. Finally, the odds of being in a more depressed group were 1.93 in those using psychotropic medication in the last week compared to nonusers.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
In the controlled analyses, age was not significantly related to either level of depression, suggesting that perhaps much of the relationship between age and depression is indirect through some other variable(s) such as physical functioning. Female gender was associated with depression, whereas race/ethnicity and level of education were not. Our finding that depressive symptoms were associated with functional limitations and disability days was consistent with previous research (Beekman, Deeg, Van Tilberg, et al. 1995; Beekman, Deeg, Braam, Smit, and Van Tilberg 1997; Blazer and Williams 1980; Broadhead et al. 1990; Maier, Gansicke, and Weiffenbach 1997; Wells et al. 1989). Because this is a cross-sectional analysis, whether depression leads to a decrease in physical functioning as a result of decreased activity or whether decreased functioning leads to depression is unclear and will be the subject of future investigation. In controlled analyses, depression group was not associated with chronic disease, but was correlated with self-rated health, consistent with the findings of Beekman, Kriegsman, Deeg, and Van Tilberg 1995 that the more subjective aspects of illness were more strongly associated with depression than disease categories.
Depression group was not independently associated with cognitive functioning in controlled analyses, a finding in contrast to previous work (Beekman, Deeg, Van Tilberg, et al. 1995; Fichter, Bruce, Schroppel, Meller, and Merikangas 1995). This finding may be due, in part, to our excluding from the analyses those with more severe cognitive impairment. Although an association between impaired social support and depression in elders has been documented (Blazer 1983), the association with sub-threshold depression is less known. Our finding that being unmarried and having impaired perceived social support were associated with depression is consistent with that of Beekman, Deeg, Van Tilberg, and colleagues (1995), who found that minor depression was associated with smaller network size, less instrumental support given, and more emotional support received. Finally, our finding that psychotropic medication use was correlated with subthreshold depression is consistent with the findings of Lyness and colleagues 1999, who reported that in a sample of primary care patients, the proportion of subjects with subsyndromal depression taking an antidepressant (38.5%) was comparable to the proportion of major depressives (47.8%) and minor depressives (46.2%) taking an antidepressant.
The CES–D scale (Radloff 1977) was developed for use in community studies; it was not designed to elicit clinical diagnoses, but to screen for clinically significant depressive symptoms. Most of the validity literature has assessed the validity of the CES–D-scale to capture clinically significant depression. For example, Lyness and colleagues 1997 have suggested that the cutpoint should be raised to 21 to capture only clinically significant depression. As Beekman, Deeg, Van Limbeek, and associates (1997) note, the positive predictive value of the CES–D scale is low. That is, many individuals identified as depressed do not meet DSM criteria for major depression. Yet, Beekman, Deeg, Van Tilberg, and colleagues (1995) found that those with CES–D-criteria depression not meeting DSM criteria had a similar risk factor profile to those with DSM major depression. In other words, the ability of the CES–D scale to validate a diagnosis of major depression does not capture the strength of the CES–D scale to identify community-dwelling elders with clinically significant depression. Therefore, the validity of the CES–D with DSM criteria is less relevant here, as we are interested in depressive symptomatology. In fact, most of the literature regarding subthreshold depression is not so much concerned with establishing whether such an entity exists, but rather exploring the public health burden of depressive symptoms that do not meet typical diagnostic or screening criteria. Rather than creating a new nosological category of depression, our intent was to ask: If the threshold were lowered for elders, would we see the risk factors and associations seen with more symptomatic depression? We found that depressive symptomatology below the threshold of the CES–D cutpoint was associated with impairments in functioning in older adults.
As in the original CES–D, our cutpoint was decided arbitrarily in the absence of obvious breaks. Our hypothesis comparing the prevalence in the two groups, therefore, cannot be tested, but the finding of a large proportion of elders with depressive symptomatology below the current CES–D threshold can be noted. These results suggest that depression appears to exist along a continuum, with demographic and physical health predictors of subthreshold depression similar to predictors of CES–D-criteria depression.
The sample was drawn from community-dwelling elders who may not be representative of older adults who go to physicians for treatment or reside in long-term care institutions. In particular, no depression data were available for sample members who participated by proxy (approximately 4% of the baseline sample). The CES–D scale does not measure duration of symptoms; therefore, DSM criteria for major or minor depression or dysthymia cannot be applied to these data. The depression data were obtained through self-report, which could be viewed as a limitation by clinicians. Nevertheless, self-report data are the norm for community-based studies.
Research is needed to further characterize subthreshold depression in these older adults, particularly its course and relationship to a course of more symptomatic depression. Beekman, Deeg, Smit, and Van Tilberg 1995 reported that, after one year, 32% of subjects with minor depression relapsed, 25% remitted but relapsed later, and 43% were chronically depressed. Beekman, Deeg, VanTilberg, and colleagues (1995) also found that a history of major depression was associated with current minor depression, supporting the hypotheses that the subtypes are different manifestations of the same illness. Questions remain concerning the longitudinal course of subthreshold depression in older adults, particularly whether it predicts more symptomatic depression and whether the course is affected by treatment. We also plan to look at the depression continuum and test the threshold against an outcome such as mortality.
Finally, these results have public health implications with regard to recognizing depression in older adults and preventing undertreatment. Clinicians and researchers should recognize that not only older individuals who do not meet DSM criteria for depression, but also those who fall below the threshold on instruments such as the CES–D, may experience symptoms of depression that deserve attention because of their potential to be associated with adverse health consequences.
|
Acknowledgments |
|---|
Received for publication June 13, 2000. Accepted for publication December 11, 2000.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  Z. D. Gellis, J. McGinty, L. Tierney, C. Jordan, J. Burton, and E. Misener Randomized Controlled Trial of Problem-Solving Therapy for Minor Depression in Home Care Research on Social Work Practice, November 1, 2008; 18(6): 596 - 606. [Abstract] [PDF]
|
|
|
|
 |
|
 J. R. Piazza, S. T. Charles, and D. M. Almeida Living With Chronic Health Conditions: Age Differences in Affective Well-Being J. Gerontol. B. Psychol. Sci. Soc. Sci., November 1, 2007; 62(6): P313 - P321. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. McCusker, M. Cole, A. Ciampi, E. Latimer, S. Windholz, and E. Belzile Does Depression in Older Medical Inpatients Predict Mortality? J. Gerontol. A Biol. Sci. Med. Sci., September 1, 2006; 61(9): 975 - 981. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. I. Cohen, C. Magai, R. Yaffee, and L. Walcott-Brown Racial Differences in Syndromal and Subsyndromal Depression in an Older Urban Population Psychiatr Serv, December 1, 2005; 56(12): 1556 - 1563. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. G. Blazer Guest Editorial The Prevalence of Depressive Symptoms J. Gerontol. A Biol. Sci. Med. Sci., March 1, 2002; 57(3): M150 - 151. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
