| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Department of Epidemiology and Biostatistics, University of Illinois at Chicago.
2 Department of Family and Geriatric Medicine, University of Louisville, Kentucky.
3 Department of Epidemiology, University of Washington, Seattle.
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
DEPRESSION among older adults is a major public health problem that has a deleterious effect on morbidity, mortality, and quality of life (Blazer, 2003
; Blazer, Hughes, & George, 1987; Lyness, 2004). The prevalence of mood disorders among community-dwelling individuals who are 65 years of age or older ranges from 1% to 4% for major depressive disorder (Beekman et al., 1995; Bland, Newman, & Orn, 1988; Blazer & Williams, 1980; Henderson et al., 1993; Newman, Sheldon, & Bland, 1998; Regier et al., 1988), 4% to 13% for minor depressive disorder (Beekman et al.; Newman et al.), and 9% to 26% for depressive symptoms (Black, Markides, & Miller, 1998; Blazer, Burchett, Service, & George, 1991; Blazer & Williams).
Depression among older adults is associated with a loss of physical function (Alexopoulos et al., 1996; Bruce, 2001; Kennedy, Kelman, & Thomas, 1990; Ormel, Rijsdijk, Sullivan, van Sonderen, & Kempen, 2002; Penninx & Leveille, 1999; Zeiss, Lewinsohn, Rohde, & Seeley, 1996), poor self-reported health status (Dorfman et al., 1995; Mulsant, Ganguli, & Seaberg, 1997), hospitalization (Huang et al., 2000), suicide (Turvey et al., 2002), and mortality (Penninx et al., 1999, 2001; Schoevers et al., 2000; Unutzer, Patrick, Marmon, Simon, & Katon, 2002).
Depression is more prevalent in women than men (Blazer et al., 1991; Chen, Eaton, Gallo, Nestadt, & Crum, 2000; Dorfman et al., 1995; Ried & Planas, 2002), regardless of age. The prevalence of major depressive disorder in African Americans ranges from 3% to 9% (Brown, Ahmed, Gary, & Milburn, 1995; Dunlop, Song, Lyons, Manheim, & Chang, 2003). Despite comparable prevalence between the two groups (Blazer, Landerman, Hays, Simonsick, & Saunders, 1998), African Americans are less likely to receive treatment for depression than are their Caucasian counterparts (Brown, Salive, et al., 1995; Hanlon et al., 1992; Sirey et al., 1999). African Americans are also less likely to report mental health problems to their physician (Gallo, Cooper-Patrick, & Lesikar, 1998). Thus, mental health research among African Americans has been limited.
Our purpose in this study is to examine the association of demographics, health behaviors, perceptions of health and functional status, and comorbid health problems with depressive symptoms in 180 pairs of African-American female twins who are 65 years of age and older. In this unique sample, we examined if associations between risk factors and depressive symptoms were confounded by familial or genetic factors. To our knowledge, this is the first study of depressive symptoms in older African-American female twins.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
; Goldberg et al., 1997). With this algorithm, researchers identified 606 paired records for African-American older women by matching records on race, date of birth, and the first seven digits of the social security number. These represented potential twin pairs. After confirming twin identity and removing twin pairs in which only one twin responded or one was deceased, researchers conducted a 45-minute telephone interview with each twin. The interview collected information on childhood similarity, health history, current physical function, selective lifestyle factors, cognitive function, and depressive symptoms. We did not include subjects with low cognitive function (according to the Telephone Instrument for Cognitive Status; Gallo & Breitner, 1995) or other health problems that excluded them from the interview process in the current sample analysis. The final analytic sample included 180 twin pairs.
Measures
Zygosity
We assigned twin zygosity on the basis of questions about childhood similarity and self-assessment. This method is widely used in population-based studies of twins and correctly classifies zygosity with an accuracy of between 95% and 98%, compared with biological markers (Eisen, Neuman, Goldberg, Rice, & True, 1989; Torgensen, 1979). Zygosity classification is important because monozygotic (MZ) pairs have identical DNA whereas while dizygotic (DZ) twins share 50% of their DNA on average. In this study, when each twin responded yes to the similarity question and that she was an identical twin, we considered her to be MZ. When each twin answered no to the similarity question and that she was a fraternal twin, we considered her to be DZ. We considered twins with different responses to the similarity and self-assessment questions to be DZ.
Center for Epidemiologic Studies–Depression scale
We measured depressive symptoms with a shortened form of the Center for Epidemiologic Studies–Depression scale (CES-D). The CES-D is a short, self-report scale that measures current depression symptoms in the general population. The original version of the CES-D includes 20 items representing the symptoms of depression within four domains: depressed affect, positive affect, somatic–retarded activity, and interpersonal (Radloff, 1977). Each of the 20 items is scored on a scale that ranges from 0 to 3 points. When summed, the original scale ranges from 0 to 60 points.
We adapted the CES-D for use in the BETS. We maintained the same four depressive symptom domains, but we reduced the number of items in each domain. In total, we included 12 question items with the same response categories as the original CES-D. Table 1 displays the items from the original CES-D and those used in the BETS survey. We selected the BETS shortened version because older persons reportedly perform better on shortened versions of the CES-D because of the reduced administration time (Kohout, Berkman, Evans, & Cornoni-Huntley, 1993).
|
= 0.85 in the general population, = 0.90 in psychiatry patients, = 0.86 in frail elderly persons, and = 0.86 in older African Americans (Davidson, Feldman & Crawford, 1994; Long Foley, Reed, Murtan, & DeVellis, 2002; Radloff, 1977). Other shortened versions of the CES-D retain high internal consistency (ranging 0.76–0.80) among elderly populations (Kohout et al., 1993). Because of the high internal consistency of the CES-D, shortened versions of the scale can be used without substantial loss of reliability (Kohout et al. Shrout & Yager, 1989). As shown in Table 1, in this population of elderly African-American twins, internal consistency was good, with the standardized Cronbach's alpha ( = 0.74) for the shortened CES-D.
We created an ordinal outcome variable based on the CES-D. In the original CES-D, scores of 16 or greater are indicative of depressive symptoms (Radloff, 1977; Weissman, Sholomskas, Pottenger, Prusoff, & Locke 1977). However, it has been reported that higher cutoff scores ranging from 20 to 23 have been more appropriate for the determination of clinically relevant depression in older adults (Haringsma, Beekman, & Spinhoven, 2004; Himmelfarb & Murrell, 1983; Husaini & Neff, 1980; Lyness et al., 1997) and that cutoff scores 
20 yield greater sensitivity and specificity with elderly populations (Husaini & Neff; Myers & Weissman, 1980).
We used the proportional equivalent of the cutoff score for a shortened version of the CES-D scale by multiplying the cutoff score of interest by n/20, where n is the number of items on the shortened scale (Shrout & Yager, 1989). With the suggested cutoff points in the literature taken into account, a BETS CES-D score of 12 or greater (equivalent to the cutoff score of >22 on the original CES-D) was an indication of high levels of depressive symptoms. Medium depressive symptoms (equivalent to scores of 16–21 on the original scale) were indicated by scores that ranged from 9 to 12 on the BETS CES-D. Low levels of depressive symptoms were indicated by scores of less than 9 on the BETS CES-D (equivalent to scores of <16 on the original scale).
Risk factor domains
We considered a number of risk factors within five domains that are associated with depressive symptoms in the older population. These domains include demographics, health behaviors, perceptions of health and functional status, and comorbid health problems.
Demographic factors collected in the study included education level, marital status, and parity. Health behaviors included cigarette smoking, alcohol consumption, and the use of replacement estrogens. Physical activity included two measures: infrequent walking (taking a walk outside fewer than 3 days per week), and no exercise to sweat within the past week.
Researchers asked twins about their perceptions of health and functional status. Respondents rated their own current health as excellent, very good, good, fair, or poor. We measured physical function by the use of questions regarding activities of daily living (ADLs) and instrumental activities of daily living (IADLs). Researchers asked seven individual ADL questions: washing and drying the body, getting in and out of the tub, dressing oneself, bending or picking up clothes, standing from a seated position, walking, and cutting meat (McDowell & Newell, 1996). We summed these items to create an overall ADL scale. Similarly, researchers asked six IADL items about the following activities: reaching or getting a 5-lb (2.26 kg) object, opening jars, using a pen or pencil, walking 0.25 mile (0.40 km), paying bills, and carrying groceries.
We deemed respondents who had difficulty doing at least one ADL or IADL activity, could not do the activity without help, or could not do it at all to have an ADL or IADL limitation.
The chronic diseases included a lifetime history of arthritis, diabetes, hypertension, and heart attack. Other physical health problems included major hearing problems, as indicated by the use of a hearing aid, and major vision problems, based on a report of blindness in one or both eyes, blurred vision caused by cataracts, or glaucoma. We also considered obesity, defined as a current body mass index (BMI) of greater than 30, to be a physical health problem.
Statistical Methods
Our initial analysis examined the prevalence of demographic and health characteristics according to the level of depressive symptomology. We grouped the CES-D scale into three categories: low (<9), medium (9–12), and high (>12) depressive symptoms. This initial analysis treated all of the twins as independent observations (n = 360) and provided a descriptive portrait of the sample.
Our formal statistical analysis used clustered data-analysis methods for twin pairs as suggested by Carlin and colleagues (Carlin, Gurrin, Sterne, Morley, & Dwyer, 2005). In particular, we used the mixed-effects ordinal logistic regression model for twin data to account for the lack of independence between members of a twin pair (Hu, Goldberg, Hedeker, & Henderson, 1998). We fit a model that includes a random effect for twin pair and provides estimated odds ratios (ORs) and confidence intervals (CIs) for the relationship of each separate risk factor with depressive symptoms. Following this, we fit an overall multivariable mixed-effects model; factors were included in this analysis on the basis of significance in the univariate analyses. We then fit a second multivariable model that partitioned the contribution of a specific risk factor into within- and between-pair effects. This analytic approach is critical for proper interpretation, because in the overall analysis the OR estimates are a weighted average of the within and between effects. The within-pair portion of the analysis removes confounding influences that are shared by both members of the pair, such as age, early life environment, and genetics. ORs from a within-pair analysis are those that isolate the effects of individual differences in risk factors such as education and physical health on depressive symptoms. Conversely, the between-pair portion of the analysis captures pair-to-pair differences in risk factors, such as age and family environment, on depressive symptoms.
We conducted likelihood ratio chi-square goodness-of-fit tests comparing the between- and within-pair effects with the overall model. We also attempted to estimate separate within-pair ORs for MZ and DZ pairs; this analysis indicated that there were no differences by zygosity, and we therefore only present the between- and within-pair estimates for all twins combined.
We performed all analyses by using the SAS System for Windows, version 8.
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Table 2 presents the prevalence of each risk factor according to levels of depressive symptoms. There were a greater percentage of twins in the two categories representing medium and high depressive symptoms than in the low-symptom group. Risk factors that demonstrated a trend toward increasing magnitude with increasing depressive symptoms were as follows: having less than a high school education; having been married; having a history of smoking, arthritis, hypertension, diabetes, two or more chronic diseases, hearing problems, vision problems, one or more ADL limitation, or one or more IADL limitation; and poor or fair self-reported health. The most common health-related risk factors among twins with high levels of depressive symptoms were using replacement estrogen (71%), having arthritis (77%), being hypertensive (74%), and having two or more chronic diseases (72%).
|
|
|
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
). In our study, we found that chronic health conditions such as fractures, vision problems, and hearing impairments were associated with depression in this population. Other studies in older adult fracture patients found that depression was common among this patient population (as it occurred in 25% to 49% of these individuals; see Mast, MacNeill, & Lichtenberg, 1999; also see Billing, Ahmed, Kenmore, Amaral, & Shakhashiri, 1986 and Kuhn, Bell, Netscher, Seligson, & Kuhn, 1989). Studies in older adult patients with sensory loss also report higher levels of depression in this group of individuals than in the general population (Capella-McDonnall, 2005; Lupsakko, Mantyjarvi, Kautiainen, & Sulkava 2002; Rovner, Zisselman, & Schmuely-Dulitzky, 1996).
Depression rates among African-American women in the United States have not been widely reported in the literature. In a study of 865 community-dwelling African Americans, the prevalence of major depressive disorder was reported to be 3.2% among women (3.1% among both sexes; see Brown, Ahmed, et al., 1995). In another study of 7,690 preretirement-age adults (1,268 were African American), the prevalence of major depressive disorder among African Americans was found to be 8.9% (Dunlop et al., 2003). In our study, we measured depressive symptoms rather than major depressive disorder, which may explain the greater prevalence.
The prevalence of older persons (aged 65 years or older) with depression in the United States is reported to be 16% (this includes all forms: major depressive disorder, dysthymic disorder, and bipolar disorder; see National Institute of Mental Health, 1999). However, there are many older persons with significant depressive symptoms that are below the severity threshold for a diagnosis of depression according to the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (Unutzer et al., 2002). In this study we examined depression symptoms in a unique sample of older African-American female twins. We found that, after we controlled for familial or genetic factors, the presence of ADL limitations and vision problems is strongly associated with levels of depressive symptoms.
Other adult twin studies have assessed genetic and environmental influences on depression. Heritability estimates from these studies have ranged from 13% to 48%. The conclusions from the majority of these studies indicate that depression is moderately heritable, and that shared environmental factors play a significant role in the etiology of depression and mood disorders (Carmelli et al., 2000; Gatz, Pedersen, Plomin, Nesselroade, & McClearn, 1992; Jansson et al., 2004; Kendler, Neale, Kessler, Heath, & Eaves, 1993; Lyons et al., 1998; McGue & Christensen, 1997; Takkinen et al., 2004). The published twin studies have been among American female twin pairs (Kendler et al.), Scandinavian Caucasian twins (Gatz et al.; Jansson et al.; McGue & Christensen; Takkinen et al.), and American male twin pairs (Carmelli et al.; Lyons et al.). Four of the aforementioned studies were done in older populations (Jansson et al.; Lyons et al.; McGue & Christensen; Takkinen et al.). None of the twin studies published to date have reported data on risk factors for depressive symptoms in a specifically African-American or minority older adult population.
The risk factors we identified are similar to those found in previous studies of older adults. A number of studies have shown that ADL limitations and health status are related to depression (Alexopoulos et al., 1996; Bruce, 2001; Dorfman et al., 1995; Kennedy et al., 1990; Mulsant et al., 1997; Ormel et al., 2002; Penninx & Leveille, 1999; Zeiss et al., 1996).
The prevalence of depression is not overwhelming among the older age groups (Blazer & Williams, 1980), but the presence of depression is strongly associated with morbidity and mortality in older people. The relationship of age with depression is attenuated when physical limitations are controlled for (Harris et al., 2003; Roberts, Kaplan, Shema & Strawbridge, 1997). This suggests that disease processes associated with age are of importance to depression, not that depression is a problem that inevitably occurs in tandem with normal aging (Penninx & Leveille, 1999). Our study further supports this concept, while controlling for familial or genetic factors.
There are limitations to this study that should be noted. The twins were relatively young, and all were female. We may be missing the frail segment of the population who were unable to participate in the telephone interview. The telephone interview collected cross-sectional data. With these data it was not possible to establish the temporal sequence between the risk factors and depressive symptoms limiting causal inference. However, we have included factors known to be associated with depression, and our findings were consistent with those of previous studies (Alexopolous et al., 1996; Bruce et al., 2001; Kennedy et al., 1990; Ormel et al., 2002; Penninx & Leveille, 1999; Zeiss et al., 1996).
Another limitation is the absence of a diagnosis of depression. We collected data on depressive symptomology by using an adapted version of the CES-D rather than by using a structured or semistructured diagnostic instrument such as the Composite International Diagnostic Interview (CIDI) (Semler et al., 1987). However, the CES-D shows modest correlation with a formal diagnosis of depression (Myers & Weissman 1980, Roberts & Vernon, 1983) and has excellent psychometric properties (Davidson et al., 1994; Long Foley et al., 2002; Radloff, 1977). We also do not know if the twins were being treated for depression, though it is unlikely that many older Black females were receiving treatment for depression at the time the survey was administered. Lastly, our sample is of modest size, and this resulted in wide confidence intervals for some estimates and suggests that our findings should be interpreted cautiously.
The incidence of depression is likely to increase as the proportion of the individuals in the general population who are aged 65 years and older grows (Blazer, 2003; Blazer et al., 1987; Lyness, 2004). The causes of depression in the older population, particularly among minority groups, are poorly understood. Our findings suggest that factors such as ADL limitations and vision problems are associated with depressive symptoms. Further studies of the relationship of physical and mental health in older populations are likely to be productive. Knowledge of these relationships may help researchers create targeted interventions designed to reduce the prevalence of depressive symptoms in older women.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received for publication July 1, 2005. Accepted for publication April 24, 2006.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
