The Bidirectional Relationship Between Depressive Symptoms and Homebound Status Among Older Adults

The Bidirectional Relationship Between Depressive Symptoms and Homebound Status Among Older Adults Abstract Objectives This study aimed to examine the bidirectional relationship between depressive symptoms and homebound status among older adults. Method The study sample included 7,603 community-dwelling older adults from the National Health and Aging Trends Study. A bivariate latent state-trait model of depressive symptoms and homebound status was estimated via structural equation modeling. Results The model fit the data well (Root Mean Square Error of Approximation = .02, Comparative Fit Index = .97, Standardized Root Mean Square Residual = .06). The relationship between homebound status and depressive symptoms can be decomposed into three parts: a moderate correlation between the stable trait components (r = .56, p <.001); a contemporary association of the state components (b = .17, p <.001); and bidirectional lagged effects between the state components. Change in homebound status was as a stronger predictor of depressive symptoms (b = .19, p < .001) than change in depressive symptoms was of homebound status (b = .06, p < .001; test of difference: Δ scaled χ2(1) = 24.2, p < .001). Discussion Homebound status and depressive symptoms form a feedback loop to influence each other. Improving the outdoor mobility of older adults may have immediate benefits for reducing depressive symptoms. Depression, Disability, Outdoor mobility Epidemiological and community-based studies have consistently documented a high prevalence of depression among older adults who are homebound or receive home-based aging services (Bruce et al., 2002; Choi & McDougall, 2007; Pickett, Raue, & Bruce, 2012; Richardson et al., 2012; Sirey et al., 2008). In a sample of older patients receiving home health care, 13.5% met the diagnostic criteria for major depression (Bruce et al., 2002). A study of participants recruited from aging service settings, many of whom are homebound, found that 27% of the sample met criteria for a current major depressive episode (Richardson et al., 2012). In comparison, the prevalence of 12-month major depressive disorder was 2.7% among the general older adult population (Hasin, Goodwin, Stinson, & Grant, 2005). A recent population-based study reported that nearly 60% of older adults who had not gone outside in the past month had clinically significant depressive symptoms compared with only 10% of non-homebound older adults (Ornstein et al., 2015). Depression and the state of being homebound have further negative consequences for healthy aging. Depression complicates the management of chronic illness, increases risk for institutionalization, and lowers quality of life and life expectancy (Sivertsen, Bjorklof, Engedal, Selbaek, & Helvik, 2015; Xiang & An, 2015). The state of being homebound is marked by the excessive burden of functional disabilities, medical comorbidities, cognitive impairment, and premature mortality (Ornstein et al., 2015; Qiu et al., 2010; Soones, Federman, Leff, Siu, & Ornstein, 2017; Xiang & Brooks, 2017). Fortunately, both conditions are potentially modifiable. Late-life depression can be successfully treated by psychotherapy, medications, or a combination of the two. The state of being homebound can be partially or fully remediated by the availability of personal assistance, assistive devices, modifications to the home and neighborhood environment, and access to transportation (Levasseur et al., 2015). Whether being homebound leads to depressive symptoms or depressive symptoms increase the risk of becoming homebound has not been tested. Studies to date have focused on estimating the contemporary, unidirectional association of homebound status and depression. It is often suggested, without empirical evidence, that being homebound leads to social isolation, increasing the risk for depression (Choi & McDougall, 2007; Pickett et al., 2012). Drawing from the extensive literature that shows the reciprocal relationship between physical disability and depression (Bruce, 2001; Chen et al., 2012; Ormel, Rijsdijk, Sullivan, van Sonderen, & Kempen, 2002), we suspect a similar relationship exists between being homebound and depression, such that the two mutually reinforce each other, creating a vicious cycle. The state of being homebound, however, is distinctive from functional impairment. Homebound older adults are an understudied, hard-to-reach population with complex medical, psychiatric, neurologic, and social needs (Xiang & Brooks, 2017), which means that their needs profile may look different from individuals with functional impairment alone. Being homebound, particularly to the point of never going outside, is an extreme form of social isolation that creates a different experience for the person affected than functional impairment alone. Despite the severity of functional impairment among homebound older adults, the state of being homebound is not inevitable for those with functional impairment. The availability of personal assistance, for example, can partially or fully remediate someone’s physical limitations, preventing the state of being homebound (Ornstein et al., 2015). Because being homebound is modifiable, its mechanisms are particularly important to investigate. The conceptual framework for the bidirectional relationship between depression and being homebound draws on the International Classification of Functioning, Disability and Health (ICF) and the concept of autonomy (Collopy, 1988). Under the ICF model, the state of being homebound is a result of the interactions between body functions and structures, activities and participation, and environmental factors (Kostanjsek, 2011). The development of chronic illness, decline in functional ability, loss of mobility, and reduction in social networks and interactions during the aging process often result in loss of physical and social independence, prohibiting an older person from staying active and engaged and to an extreme extent, resulting in the state of completely homebound. The loss of independence during the process of becoming homebound erodes individual autonomy, defined as a condition of self-government or self-determination (Collopy, 1988). Discourses on autonomy and independence with older adults have revealed that they perceive being able to “do things alone,” such as getting out to social activities, to be an important aspect of autonomy and independence (Hillcoat-Nallétamby, 2014; Plath, 2008). Based on her qualitative interviews, Plath (2008) observed that older adults’ desires for freedom, control, and self-sufficiency drive them to do things alone to attain a sense of individual achievement and competence, and the desire to doing things alone may be particularly strong in societies with liberal, individualist values such as the United States. For homebound older adults, doing things alone is often unattainable due to limited physical and mental capacity, which can cause common symptoms of depression such as a diminished sense of self-control and competence. Conversely, depression can impair body functions and structures, restrict activities and social participation, and limit social interactions and support, all of which increase the risk of becoming homebound. Homebound status and depression, therefore, can form a vicious cycle and mutually reinforce one another over time. This study aims to examine the bidirectional relationship between depressive symptoms and homebound status in older adults over time. We hypothesize that (1) depressive symptoms are positively associated with the extent to which a person is homebound, and are more prevalent among persons with a lower level of autonomy; (2) worsening depressive symptoms are associated with an immediate worsening of the extent to which a person is homebound; and (3) there are cross-lagged associations between worsening depressive symptoms and the extent to which a person is homebound. Method Study Sample Data came from Round 1 (2011) through Round 6 (2016) of the National Health and Aging Trends Study (NHATS) public-use datasets (https://www.nhats.org/). The Johns Hopkins University Institutional Review Board approved the NHATS study, which is a nationally representative panel study of Medicare beneficiaries aged 65 and older, with oversamples of persons in older age groups and African Americans. When a participant was not available to be interviewed, NHATS methodology involved the use of proxy respondents in their stead. Top reasons for having a proxy were that the sample person had dementia or another illness. A total of 7,777 community-dwelling older adults or their proxies completed in-person interviews at Round 1. Annual follow-up interviews were conducted regardless of residential status. In Round 1, 7,603 participants responded to questions regarding homebound status, 583 of whom were proxy respondents. In Rounds 2–6, the number of proxy respondents were 813, 776, 638, 507, and 463, respectively. Measures Homebound status Our conceptualization of homebound status is a continuum of outdoor mobility determined by physical capacity, availability of social support, and degree of autonomy. To this end, four questions were used to determine the extent to which a participant was homebound. Participants or their proxies were asked, “How often did you (participant) go out in the last month?” on a 5-point Likert scale (i.e., “never,” “rarely” [≤1 day], “some days” [2–4 days], “most days” [5–6 days], and “every day”). Respondents who reported “never” were then asked, “Did you ever have to stay in because no one was there to help?” Respondents who reported ever going out were asked, “Did anyone ever help you?” and “How often did you go outside by yourself?” Those who reported ever going out without help were then asked, “How much difficulty did you have leaving the house by yourself?” Figure 1 details the process of determining the level of homebound status. Participants were first grouped into three categories (non-homebound, semi-homebound, and homebound) based on the reported frequency of outdoor mobility. Within each category, participants were grouped into finer categories based on the receipt of assistance going outside and difficulty going outside alone, both of which indicated the participant’s level of independence and autonomy. For instance, we determined that participants had a lower degree of autonomy if they reported never going outside by themselves. This process resulted in eight levels of homebound status, ranging from non-homebound to completely homebound, and moving from the highest level of autonomy to the lowest level. For example, level 1, non-homebound individuals, went outside two or more days per week, received no help going out, and reported no difficulty going outside alone. This group was believed to have the highest level of autonomy. Level 7 and level 8 homebound individuals, on the other hand, reported never having gone outside in the past month; however, the level 8 homebound group was believed to have a lower level of autonomy, because they reported lack of help as a reason for not going outside. In the final model, we reduced the homebound categories to five in consideration of linearity assumption. In addition, levels 2–8 were combined to form a dichotomous indicator of homebound status (non-homebound vs. homebound) for describing patterns of change in homebound status in the descriptive analysis. Figure 1. View largeDownload slide Determining homebound status. Lvl = level. Figure 1. View largeDownload slide Determining homebound status. Lvl = level. Depressive symptoms The Patient Health Questionnaire-2 (PHQ-2), administered to participants and proxy respondents, was used to screen depressive symptoms (Löwe, Kroenke, & Gräfe, 2005). The PHQ-2 measures how often a person has been bothered by “little interest or pleasure in doing things” and “feeling down, depressed or hopeless” over the last month on a 4-point Likert scale, that is, “not at all” (0), “several days” (1), “more than half the days” (2), and “nearly every day” (3). The composite PHQ-2 score, ranging from 0 to 6, with higher scores indicating more severe depressive symptoms, was used in the main analysis. A dichotomous indicator of clinically significant depressive symptoms was also created for descriptive analysis, using a cutoff score of 3. A cutoff score of 3 has a sensitivity of 0.87 and a specificity of 0.78 for major depressive disorder, and has a sensitivity of 0.79 and specificity of 0.86 for any depressive disorder (Löwe et al., 2005). Sociodemographic and functioning characteristics Sociodemographic variables included age, sex (female, male), race/ethnicity (non-Hispanic white, non-Hispanic black, other race, and Hispanic), education (less than high school, high school or GED, some college, college degree or higher), a dichotomous indicator of living alone, and Medicaid–Medicare dual enrollment status (yes or no). A simple count of self-reported chronic diseases included heart attack/heart disease, arthritis, osteoporosis, diabetes, lung disease, stroke, and cancer. Three levels of dementia status (probable dementia, possible dementia, and no dementia) were assessed based on self-reported diagnosis of dementia or Alzheimer’s disease, the AD8 Dementia Screening Interview, and cognitive tests (Kasper, Freedman, & Spillman, 2013). Activities of daily living (ADLs) included eating, bathing, toileting, and dressing. Statistical Analysis Model description The proposed relationships between depression and homebound status were tested in a structural equation model (SEM) (Figure 2). The main model consisted of three parts: (1) a latent state-trait (LST) model for six time points addressing homebound status, (2) a LST model for six time points addressing depressive symptoms, and (3) correlations and regression effects linking the two trait-state models. Figure 2. View largeDownload slide Standardized estimates for the bivariate latent state-trait model of homebound status and depression without covariates. Time corresponds to survey waves. p < .001 all for estimates shown. Unstandardized estimates: a = .19, b = .06, c = .17, d = .09, e = .11, and f = .44. DEP = depressive symptoms; HMB = homebound status; PHQ = Patient Health Questionnaire-2. Figure 2. View largeDownload slide Standardized estimates for the bivariate latent state-trait model of homebound status and depression without covariates. Time corresponds to survey waves. p < .001 all for estimates shown. Unstandardized estimates: a = .19, b = .06, c = .17, d = .09, e = .11, and f = .44. DEP = depressive symptoms; HMB = homebound status; PHQ = Patient Health Questionnaire-2. The LST model of depression assumes that the severity of depressive symptoms at each time point is the function of two latent variables: a trait component and a state component. The trait component is the common factor or time-invariant component representing the general tendency of a person to experience depressive symptoms. The state component is a time-varying factor representing the occasion-specific variation that is not accounted for by the trait factor. The state component reflects the within-subject change in depressive symptoms in a 2-year time period. The across-time structure of the state component (States 1–6 in Figure 2) was modeled as a first-order autoregressive model (paths f). The LST model of homebound status makes a similar assumption such that the extent to which a person is homebound at each time point is the function of a time-invariant trait component and an occasion-specific state component. In longitudinal studies with repeated assessments, correlation between trait factors (i.e., relatively stable dispositions) can obscure the ability to detect associations between more acute, state-level symptoms and behaviors (Hertzog & Nesselroade, 1987). LST modeling decomposes variation into state and trait components and explicitly tests the relationship between these components, and is therefore well suited to study the reciprocal relationship between constructs over time. The model depicted in Figure 2 was obtained by linking the LST models of depression and homebound status such that the latent state variables of homebound status can influence the latent state variables of depressive symptoms, and vice versa. These effects can be contemporary (paths c) and lagged (paths a and b). In addition, the model allows correlation between the two trait factors (path g) and between the two first state factors (path d). The correlation between the trait factors represents common causes of the tendency to become depressed and homebound, such as shared genetic susceptibility and physical functioning. Building upon the main model (Figure 2), we tested an adjusted model by regressing the two trait factors on sociodemographic and clinical factors, as well as an indicator of proxy respondents. The model with trait-level covariates allowed us to examine common correlates of the tendency to become depressed and homebound. Due to the set-up of bivariate LST models, adding trait-level covariates would affect the trait association but would have little impact on the correlations between the state factors. Model specification and identification Model specification was based on the recommendations of Kenny and Zautra (1995) and Ormel et al. (2002): (1) each observed variable was regressed on two latent variables, a latent trait and a latent state variable with factor loadings set to 1.0 (paths h, i, j, and k in Figure 2); (2) autoregressive paths from one state variable to the next (paths f and e) were constrained to be equal within each construct; (3) residual variables for the endogenous latent state variables were constrained to be equal within each construct, and residual variances for observed variables were set to 0; (4) temporary associations between the endogenous homebound and depression state variables were constrained to be equal (paths c); and (5) the cross-lagged paths from homebound status to depressive symptoms (paths b) were constrained to be equal, as were the paths from depressive symptoms to homebound status (paths a). Model estimation and evaluation Model estimation was conducted in Mplus 8 (Muthén & Muthén, 2017). Robust maximum likelihood estimation was used to accommodate non-normal distributions of the variables. Estimation adjusted for the complex survey design of NHATS, using design factors at baseline including sampling weights, strata, and primary sampling units. A good-fitting model has a nonsignificant χ2 statistic, a Root Mean Square Error of Approximation (RMSEA) <.06, Comparative Fit Index (CFI) >.95, and Standardized Root Mean Square Residual (SRMR) <.08 (Hooper, Coughlan, & Mullen, 2008). The χ2 statistic is sensitive to sample size and deviations from multivariate normality (Hooper et al., 2008); it nearly always rejects the model when large samples are used (Jöreskog & Sörbom, 1993). Although an alternative statistic, χ2/df, has been developed, consensus is lacking regarding an acceptable ratio for this statistic (Hooper et al., 2008). Given these limitations of χ2 statistic and our large sample size, evaluation of model fit relied on RMSEA, CFI, and SRMR. The Satorra-Bentler scaled chi-square difference test was used to compare nested models (Satorra & Bentler, 2010). Missing data During the 5-year follow-up period, a substantial proportion of the study sample was lost to follow-up. The number of respondents, including proxy respondents, for Rounds 2–6 was 6,408 (84.3%), 5,285 (69.5%), 4,357 (57.3%), 3,874 (50.9%), and 3,459 (45.5%), respectively. At the time of Round 6, non-response accounted for about two-thirds of the loss at follow-up, and death accounted for one-third. Loss at follow-up was positively associated with age, racial/ethnic minority status, lower levels of education, Medicaid coverage, number of chronic diseases, probable or possible dementia, ADL limitations, depression, a greater degree of home confinement, and loss of independence. Missing data are handled within the analysis model by a full information maximum likelihood (FIML) method, where all available information is used to estimate the model. FIML is an optimal method of handling missing data under the condition of missing at random (Enders & Bandalos, 2001). Sensitivity analysis We performed sensitivity analysis to check the robustness of results against the categorizations of homebound status and different ways of handling missing data. The bivariate LST model (shown in Figure 2) was re-estimated using a homebound measure with six (by combining homebound status in Figure 1) and seven (by combing homebound status 7 and 8 in Figure 1) categories, respectively. This model was also re-estimated using an alternative measure of homebound status, taken from Ornstein et al. (2015). In addition, we re-estimated the model using the pattern mixture models under the condition of not missing at random (NMAR). Results Prevalence of Depression by Homebound Status The unadjusted bivariate relationship between baseline depression and homebound status appeared to be monotonic, as the prevalence of significant depressive symptoms increased significantly with increased homebound levels. One in ten non-homebound older adults (level 1) had significant depressive symptoms; in contrast, one in seven homebound older adults who never went out due to lack of help (level 8) reported significant depressive symptoms. To approximate linearity, homebound levels 3 and 4 were collapsed into a single category, as were levels 5 and 6. Levels 7 and 8 were also collapsed into a single category due to the small number of people qualifying as level 8. The reduced homebound measure therefore had five categories and was used in the SEM models (Table 1). Table 1. Prevalence of Clinically Significant Depressive Symptoms by Homebound Status at Baseline   Unweighted, N  Weighted, %  Weighted % with significant depressive symptoms within each level of homebound status  Homebound status   Non-homebound (level 1)  5,681  79.4  10.0   Semi-homebound (level 2)  612  7.4  23.8   Semi-homebound (level 3)  407  4.4  31.8   Semi-homebound (level 4)  342  3.3  32.7   Homebound (level 5)  196  2.4  36.3   Homebound (level 6)  232  2.1  42.9   Homebound (level 7)  110  0.9  60.3   Homebound (level 8)  23  0.2  68.6    Unweighted, N  Weighted, %  Weighted % with significant depressive symptoms within each level of homebound status  Homebound status   Non-homebound (level 1)  5,681  79.4  10.0   Semi-homebound (level 2)  612  7.4  23.8   Semi-homebound (level 3)  407  4.4  31.8   Semi-homebound (level 4)  342  3.3  32.7   Homebound (level 5)  196  2.4  36.3   Homebound (level 6)  232  2.1  42.9   Homebound (level 7)  110  0.9  60.3   Homebound (level 8)  23  0.2  68.6  View Large Table 1. Prevalence of Clinically Significant Depressive Symptoms by Homebound Status at Baseline   Unweighted, N  Weighted, %  Weighted % with significant depressive symptoms within each level of homebound status  Homebound status   Non-homebound (level 1)  5,681  79.4  10.0   Semi-homebound (level 2)  612  7.4  23.8   Semi-homebound (level 3)  407  4.4  31.8   Semi-homebound (level 4)  342  3.3  32.7   Homebound (level 5)  196  2.4  36.3   Homebound (level 6)  232  2.1  42.9   Homebound (level 7)  110  0.9  60.3   Homebound (level 8)  23  0.2  68.6    Unweighted, N  Weighted, %  Weighted % with significant depressive symptoms within each level of homebound status  Homebound status   Non-homebound (level 1)  5,681  79.4  10.0   Semi-homebound (level 2)  612  7.4  23.8   Semi-homebound (level 3)  407  4.4  31.8   Semi-homebound (level 4)  342  3.3  32.7   Homebound (level 5)  196  2.4  36.3   Homebound (level 6)  232  2.1  42.9   Homebound (level 7)  110  0.9  60.3   Homebound (level 8)  23  0.2  68.6  View Large Sociodemographic and Functioning Characteristics by Homebound Status Compared with non-homebound older adults, semi-homebound and homebound older adults were much older. Nearly three-quarters of homebound (levels 7 and 8) older adults were 80 years or older, whereas less than a quarter of non-homebound older adults were 80 years or older. Individuals who were females, racial/ethnic minorities, less educated, and/or Medicare–Medicaid dual eligible were over-represented in semi-homebound and homebound older adults. Semi-homebound and homebound older adults were sicker, were more likely to have cognitive impairment, and had a higher number of ADL needs compared with non-homebound older adults. Use of proxies increased as homebound level increased, ranging from 2.3% among non-homebound older adults and 51.6% among homebound older adults (levels 7 and 8). Finally, each higher level of homebound status was associated with an approximately 10% increase in prevalence of significant depressive symptoms (Table 2). Table 2. Baseline Sample Characteristics by Homebound Status   Non-homebound (level 1)  Semi-homebound (level 2)  Semi-homebound (levels 3 & 4)  Homebound (levels 5 and 6)  Homebound (levels 7 and 8)  Age groups (%)   65–69 years  30.9  25.1  13.5  10.8  7.6   70–74 years  26.9  21.5  14.6  18.4  9.9   75–79 years  19.4  20.3  19.2  12.9  10.5   80–84 years  13.7  17.7  18.7  17.8  24.1   85–89 years  6.7  11.1  19.7  25.2  26.3   90 years or older  2.4  4.3  14.3  15.0  21.6  Male (%)  46.6  41.5  25.8  22.9  32.1  Race/ethnicity (%)   White, non-Hispanic  82.4  78.4  72.0  68.9  65.9   Black, non-Hispanic  7.4  10.2  10.5  10.8  15.7   Other race, non-Hispanic  4.6  4.3  4.7  4.8  7.7   Hispanic  5.6  7.1  12.8  15.5  10.8  Education (%)   Less than high school  18.0  31.3  37.1  42.7  36.8   High school  27.7  28.2  25.5  27.8  31.9   Some college, no degree  21.9  20.8  20.0  17.2  15.4   College graduate  32.4  19.8  17.5  12.3  16.0  Living alone (%)  28.0  36.5  33.7  41.2  27.0  Medicare–Medicaid dual enrollment (%)  8.7  20.0  24.3  29.9  29.8  Number of chronic diseases (mean)  1.56 (0.02)  2.37 (0.07)  2.48 (0.06)  2.51 (0.09)  2.49 (0.14)  Dementia status (%)   No dementia  85.2  71.8  51.2  44.3  19.9   Possible dementia  9.6  14.5  16.5  14.9  14.8   Probable dementia  5.0  13.7  32.4  40.8  65.4  Number of ADL needs (mean)  0.18 (0.01)  1.15 (0.06)  1.92 (0.07)  1.87 (0.08)  2.84 (0.14)  Proxy respondents  2.3  5.4  22.7  28.6  51.6  PHQ-2 ≥ 3 (%)  10.0  23.8  32.2  39.4  62.0    Non-homebound (level 1)  Semi-homebound (level 2)  Semi-homebound (levels 3 & 4)  Homebound (levels 5 and 6)  Homebound (levels 7 and 8)  Age groups (%)   65–69 years  30.9  25.1  13.5  10.8  7.6   70–74 years  26.9  21.5  14.6  18.4  9.9   75–79 years  19.4  20.3  19.2  12.9  10.5   80–84 years  13.7  17.7  18.7  17.8  24.1   85–89 years  6.7  11.1  19.7  25.2  26.3   90 years or older  2.4  4.3  14.3  15.0  21.6  Male (%)  46.6  41.5  25.8  22.9  32.1  Race/ethnicity (%)   White, non-Hispanic  82.4  78.4  72.0  68.9  65.9   Black, non-Hispanic  7.4  10.2  10.5  10.8  15.7   Other race, non-Hispanic  4.6  4.3  4.7  4.8  7.7   Hispanic  5.6  7.1  12.8  15.5  10.8  Education (%)   Less than high school  18.0  31.3  37.1  42.7  36.8   High school  27.7  28.2  25.5  27.8  31.9   Some college, no degree  21.9  20.8  20.0  17.2  15.4   College graduate  32.4  19.8  17.5  12.3  16.0  Living alone (%)  28.0  36.5  33.7  41.2  27.0  Medicare–Medicaid dual enrollment (%)  8.7  20.0  24.3  29.9  29.8  Number of chronic diseases (mean)  1.56 (0.02)  2.37 (0.07)  2.48 (0.06)  2.51 (0.09)  2.49 (0.14)  Dementia status (%)   No dementia  85.2  71.8  51.2  44.3  19.9   Possible dementia  9.6  14.5  16.5  14.9  14.8   Probable dementia  5.0  13.7  32.4  40.8  65.4  Number of ADL needs (mean)  0.18 (0.01)  1.15 (0.06)  1.92 (0.07)  1.87 (0.08)  2.84 (0.14)  Proxy respondents  2.3  5.4  22.7  28.6  51.6  PHQ-2 ≥ 3 (%)  10.0  23.8  32.2  39.4  62.0  Note. ADL = activities of daily living; PHQ = Patient Health Questionnaire-2. Estimates adjusted for the complex design of NHATS; p < .001 for all comparisons by homebound status. View Large Table 2. Baseline Sample Characteristics by Homebound Status   Non-homebound (level 1)  Semi-homebound (level 2)  Semi-homebound (levels 3 & 4)  Homebound (levels 5 and 6)  Homebound (levels 7 and 8)  Age groups (%)   65–69 years  30.9  25.1  13.5  10.8  7.6   70–74 years  26.9  21.5  14.6  18.4  9.9   75–79 years  19.4  20.3  19.2  12.9  10.5   80–84 years  13.7  17.7  18.7  17.8  24.1   85–89 years  6.7  11.1  19.7  25.2  26.3   90 years or older  2.4  4.3  14.3  15.0  21.6  Male (%)  46.6  41.5  25.8  22.9  32.1  Race/ethnicity (%)   White, non-Hispanic  82.4  78.4  72.0  68.9  65.9   Black, non-Hispanic  7.4  10.2  10.5  10.8  15.7   Other race, non-Hispanic  4.6  4.3  4.7  4.8  7.7   Hispanic  5.6  7.1  12.8  15.5  10.8  Education (%)   Less than high school  18.0  31.3  37.1  42.7  36.8   High school  27.7  28.2  25.5  27.8  31.9   Some college, no degree  21.9  20.8  20.0  17.2  15.4   College graduate  32.4  19.8  17.5  12.3  16.0  Living alone (%)  28.0  36.5  33.7  41.2  27.0  Medicare–Medicaid dual enrollment (%)  8.7  20.0  24.3  29.9  29.8  Number of chronic diseases (mean)  1.56 (0.02)  2.37 (0.07)  2.48 (0.06)  2.51 (0.09)  2.49 (0.14)  Dementia status (%)   No dementia  85.2  71.8  51.2  44.3  19.9   Possible dementia  9.6  14.5  16.5  14.9  14.8   Probable dementia  5.0  13.7  32.4  40.8  65.4  Number of ADL needs (mean)  0.18 (0.01)  1.15 (0.06)  1.92 (0.07)  1.87 (0.08)  2.84 (0.14)  Proxy respondents  2.3  5.4  22.7  28.6  51.6  PHQ-2 ≥ 3 (%)  10.0  23.8  32.2  39.4  62.0    Non-homebound (level 1)  Semi-homebound (level 2)  Semi-homebound (levels 3 & 4)  Homebound (levels 5 and 6)  Homebound (levels 7 and 8)  Age groups (%)   65–69 years  30.9  25.1  13.5  10.8  7.6   70–74 years  26.9  21.5  14.6  18.4  9.9   75–79 years  19.4  20.3  19.2  12.9  10.5   80–84 years  13.7  17.7  18.7  17.8  24.1   85–89 years  6.7  11.1  19.7  25.2  26.3   90 years or older  2.4  4.3  14.3  15.0  21.6  Male (%)  46.6  41.5  25.8  22.9  32.1  Race/ethnicity (%)   White, non-Hispanic  82.4  78.4  72.0  68.9  65.9   Black, non-Hispanic  7.4  10.2  10.5  10.8  15.7   Other race, non-Hispanic  4.6  4.3  4.7  4.8  7.7   Hispanic  5.6  7.1  12.8  15.5  10.8  Education (%)   Less than high school  18.0  31.3  37.1  42.7  36.8   High school  27.7  28.2  25.5  27.8  31.9   Some college, no degree  21.9  20.8  20.0  17.2  15.4   College graduate  32.4  19.8  17.5  12.3  16.0  Living alone (%)  28.0  36.5  33.7  41.2  27.0  Medicare–Medicaid dual enrollment (%)  8.7  20.0  24.3  29.9  29.8  Number of chronic diseases (mean)  1.56 (0.02)  2.37 (0.07)  2.48 (0.06)  2.51 (0.09)  2.49 (0.14)  Dementia status (%)   No dementia  85.2  71.8  51.2  44.3  19.9   Possible dementia  9.6  14.5  16.5  14.9  14.8   Probable dementia  5.0  13.7  32.4  40.8  65.4  Number of ADL needs (mean)  0.18 (0.01)  1.15 (0.06)  1.92 (0.07)  1.87 (0.08)  2.84 (0.14)  Proxy respondents  2.3  5.4  22.7  28.6  51.6  PHQ-2 ≥ 3 (%)  10.0  23.8  32.2  39.4  62.0  Note. ADL = activities of daily living; PHQ = Patient Health Questionnaire-2. Estimates adjusted for the complex design of NHATS; p < .001 for all comparisons by homebound status. View Large Patterns of Change in Homebound and Depression Status The most prevalent patterns of change in homebound and depression status were in persistently non-depressed and persistently non-homebound individuals, respectively. Both onset and remission of depression and homebound status occurred, and onset was slightly more frequent than remission overall. About 4% of the sample was persistently depressed in a given 2-year period. The prevalence of being persistently homebound was more frequent, ranging from 14.2% to 16.5% of the sample in a given 2-year period (Table 3). Table 3. Change and Stability in Depression and Homebound Status per Pair of Survey Rounds Patterns of change and stability  T1–T2  T2–T3  T3–T4  T4–T5  T5–T6  Depressiona   00-Persistently non-depressed  78.5  80.7  82.0  82.1  82.6   01-Onset of depression  7.8  7.90  7.2  7.1  7.0   10-Remission of depression  8.9  7.0  6.5  6.4  6.1   11-Persistently depressed  4.8  4.4  4.4  4.4  4.3  Homeboundb   00-Persistently non-homebound  70.5  69.6  68.6  69.2  67.5   01-Onset of being homebound  9.2  9.9  10.6  8.9  11.1   10-Remission of being homebound  6.1  5.8  5.3  5.6  4.9   11-Persistently homebound  14.2  14.7  15.5  16.3  16.5  Patterns of change and stability  T1–T2  T2–T3  T3–T4  T4–T5  T5–T6  Depressiona   00-Persistently non-depressed  78.5  80.7  82.0  82.1  82.6   01-Onset of depression  7.8  7.90  7.2  7.1  7.0   10-Remission of depression  8.9  7.0  6.5  6.4  6.1   11-Persistently depressed  4.8  4.4  4.4  4.4  4.3  Homeboundb   00-Persistently non-homebound  70.5  69.6  68.6  69.2  67.5   01-Onset of being homebound  9.2  9.9  10.6  8.9  11.1   10-Remission of being homebound  6.1  5.8  5.3  5.6  4.9   11-Persistently homebound  14.2  14.7  15.5  16.3  16.5  Notes. Weighted % presented in table, adjusting for the complex survey design of NHATS. aDepressed is defined as ≥3 on the PHQ-2. bHomebound combines levels 2 through 8 on the original homebound measure; non-homebound individuals went outside in the last month, received no help, and had no difficulty going out alone. View Large Table 3. Change and Stability in Depression and Homebound Status per Pair of Survey Rounds Patterns of change and stability  T1–T2  T2–T3  T3–T4  T4–T5  T5–T6  Depressiona   00-Persistently non-depressed  78.5  80.7  82.0  82.1  82.6   01-Onset of depression  7.8  7.90  7.2  7.1  7.0   10-Remission of depression  8.9  7.0  6.5  6.4  6.1   11-Persistently depressed  4.8  4.4  4.4  4.4  4.3  Homeboundb   00-Persistently non-homebound  70.5  69.6  68.6  69.2  67.5   01-Onset of being homebound  9.2  9.9  10.6  8.9  11.1   10-Remission of being homebound  6.1  5.8  5.3  5.6  4.9   11-Persistently homebound  14.2  14.7  15.5  16.3  16.5  Patterns of change and stability  T1–T2  T2–T3  T3–T4  T4–T5  T5–T6  Depressiona   00-Persistently non-depressed  78.5  80.7  82.0  82.1  82.6   01-Onset of depression  7.8  7.90  7.2  7.1  7.0   10-Remission of depression  8.9  7.0  6.5  6.4  6.1   11-Persistently depressed  4.8  4.4  4.4  4.4  4.3  Homeboundb   00-Persistently non-homebound  70.5  69.6  68.6  69.2  67.5   01-Onset of being homebound  9.2  9.9  10.6  8.9  11.1   10-Remission of being homebound  6.1  5.8  5.3  5.6  4.9   11-Persistently homebound  14.2  14.7  15.5  16.3  16.5  Notes. Weighted % presented in table, adjusting for the complex survey design of NHATS. aDepressed is defined as ≥3 on the PHQ-2. bHomebound combines levels 2 through 8 on the original homebound measure; non-homebound individuals went outside in the last month, received no help, and had no difficulty going out alone. View Large Unadjusted SEM Model The bivariate LST model without covariates had a good model fit based on RMSEA = .02, CFI = .97, and SRMR = .06. The model had a significant χ2(65) = 343.6 (p < .001), likely due to the large sample size. Removing the equality constraints on the cross-lagged paths (a and b) did not lead to a significant improvement in overall model fit (Δ scaled χ2(8) = 10.8, p = .213), suggesting that it was appropriate to pose these equality constraints. Standardized estimates are shown in Figure 2, with unstandardized estimates included in the figure caption. The interpretation of unstandardized coefficients for the contemporary (c) and cross-lagged paths (a and b) had clear meaning (i.e., the association of one category change in homebound status and one point change in the PHQ-2), whereas the interpretation of standardized estimates was unclear (i.e., the association of 1 SD change in homebound status and 1 SD change in the PHQ-2). Therefore, we focused on unstandardized estimates for the cross-lagged paths in reporting and discussion. A moderate association existed between the homebound status trait and depressive symptoms trait variable (r = .56, p < .001). The cross-variable associations between homebound state and depression state variables (bc = .17, p < .001) were small but significant. Change in homebound status and change in depressive symptoms had significant contemporary effects on each other. One category of change in homebound status (e.g., from level 1 non-homebound to level 2 semi-homebound) was associated with a .17-point increase in the PHQ-2 score during the same year. This suggests that changes in homebound status and changes in depressive symptoms influenced each other rather quickly. Change in homebound status and change in depressive symptoms also had significant 1-year lagged effects on each other. One category of change in homebound status was associated with a .19-point increase in the PHQ-2 score 1 year later (ba = .19, p < .001). On the other hand, a 1-point change in PHQ-2 level was associated with .06-point increase in the level of homebound status (bb = .06, p < .001). These two sets of lagged effects were significantly different from each other (Δ scaled χ2[1] = 24.2, p < .001). This suggests that the lagged effect of change in homebound status on change in depressive symptoms was stronger than the lagged effect of change in depressive symptoms on change in homebound status. Adjusted SEM Model The bivariate LST model with trait variables regressed on sociodemographic and functioning covariates had a good model fit based on RMSEA = .02, CFI = .96, and SRMR = .04. As expected, adding trait-level covariates had little impact on the cross-lagged effects between the state variables (ba = .18, bb = .06, p < .001 for both). The correlation between homebound trait and depression trait variables was reduced from r = .56 to r = .18 (p < .001). Homebound trait variable was positively associated with age (β = .23, p < .001), Hispanic ethnicity as compared with non-Hispanic white ethnicity (β = .04, p = .006), Medicare–Medicaid dual eligibility (β = .07, p < .001), number of chronic disease (β = .12, p < .001), probable (β = .18, p < .001) or possible (β = .06, p < .001) dementia, number of ADL needs (β = .56, p < .001), and use of proxy respondents (β = .10, p < .001). Male sex (β = –.10, p < .001) and higher educational achievement (βhigh school = –.04, p = .014; βsome college = –.04, p = .012; βcollege degree = –.07, p < .001 as compared with less than high school education) were negatively associated with homebound trait variable. Depression trait variable was negatively associated with age (β = –.03, p = .014) and some college (β = –.09, p < .001) or college degree education (β = –.17, p < .001) as compared with less than high school education. Depression trait variable was positively associated with Hispanic ethnicity as compared with non-Hispanic white ethnicity (β = .07, p = .006), living alone (β = .06, p < .001), Medicare–Medicaid dual eligibility (β = .05, p < .001), number of chronic disease (β = .22, p < .001), probable (β = .14, p < .001) or possible (β = .06, p < .001) dementia, and number of ADL needs (β = .37, p < .001). Sensitivity Analysis Estimates similar to those reported above were obtained by re-estimating the bivariate LST models using different categorizations and measures of homebound status, as well as using the pattern mixture models under the condition of NMAR. These results suggested that the study findings were robust against varying measures of homebound status and methods of handling missing data. Discussion This study suggests that the relationship between homebound status and depressive symptoms in community-dwelling older adults can be decomposed into three parts. First, there is a moderate correlation between the stable trait components of homebound status and depressive symptoms, and this correlation cannot be fully explained away by sociodemographic and functioning factors. Second, there is a contemporary correlation of change in homebound status and depressive symptoms. Finally, there are bidirectional lagged effects between change in homebound status and depressive symptoms, and the lagged effect of change in homebound status on depressive symptoms was stronger than the lagged effect of change in depressive symptoms on homebound status. These results provide the first set of empirical evidence that demonstrate the reciprocal relationship between homebound status and depressive symptoms. Although little research exists to evaluate the reciprocal relationship between homebound status and depressive symptoms, the abundant literature on the relationship between depression and disability provides some insights into the nature of this relationship, because homebound status is closely tied to disability. Several longitudinal studies have examined the bidirectional relationship between depressive symptoms and disability, often measured in terms of ADLs (Chen et al., 2012; Ormel et al., 2002). Using different samples and analytical methods, these studies have consistently showed that disability and depressive symptoms influence each other in terms of a feedback loop, and that change in disability is a stronger predictor of depressive symptoms than change in depressive symptoms is of disability (Chen et al., 2012; Ormel et al., 2002). These results are in line with findings from the present study. It is plausible that disability and physical illness produce instantaneous effects on mental health, whereas the detrimental effects of mental illness on physical health manifest over an extended period (Ormel et al., 2002). The state-trait model showed that individual differences in homebound status and depressive symptoms are, to a large extent, stable among older adults aged 65 and older during a 4-year period. The stable trait component in LST models does not necessarily imply the trait is biological or non-modifiable. Rather, trait variance may be due to stable environments or other factors (Kenny & Zautra, 1995). Traits can be temporally invariant or can also change, albeit slowly (Nesselroade, 1988). The correlation between these stable trait components can be a major contributor to the high prevalence of depression in homebound older adults observed in previous studies (Bruce et al., 2002; Choi & McDougall, 2007; Pickett et al., 2012; Richardson et al., 2012; Sirey et al., 2008). The temporary and lagged associations between the state components play a weaker role in explaining the relationship between homebound status and depression. It is likely that most people already experienced an episode of being homebound or depressive symptoms prior to study enrollment. Changes in homebound status and depressive symptoms may be uncommon afterwards. Alternatively, people could have developed strategies to cope with these changes because of prior experiences, making them less vulnerable to the detrimental effects of the changes. Future studies following a younger cohort for an extended period may shed light on the reciprocal relationship between homebound status and depressive symptoms. To inform interventions for depression and home confinement in older adults, future studies should examine the common causes of these conditions and identify individual, interpersonal, and environment factors amendable to intervention. Study findings suggested a positive feedback loop between homebound status and depressive symptoms that may lead to a vicious cycle of worsening depressive symptoms and higher level of confinement to the home. However, the prevalence of clinically significant depressive symptoms decreased over time, and the percentage of people who were not homebound only moderately increased in this sample. One possible explanation is attrition and survival bias. Depression and the state of being homebound are both associated with reduced life expectancy (Sivertsen et al., 2015; Soones et al., 2017); therefore, as time goes by, the number of people who are homebound or depressed is reduced in the sample. Limitations Interpretation of the study results should consider the following limitations. All measures were self-reported and subject to recall bias and reporting errors. PHQ-2 is a brief screening tool and requires follow-up with PHQ-9 to further probe depression severity. PHQ-2 is not a diagnostic test and is prone to ceiling effects. In addition, proxy respondents answered PHQ-2 questions on behalf of sample persons unavailable for interviews, and the reliability and validity of administering PHQ-2 to a proxy has not been established. Proxies accounted for half of homebound respondents and only a small percentage of non-homebound older adults. Systematic reporting differences may exist between proxies and self-respondents, which could bias the study estimates. For example, proxies may over-report the severity of homebound status, resulting in an overestimate of the impact of homebound status on depression. Finally, the homebound measure was restricted to activities in the last month and had various skip patterns. The skip patterns prevented us from gaining a comprehensive understanding of the causes of homebound status. Implications Reversing the state of being homebound may provide immediate benefits for improving the mental health of older adults. Prospective studies should confirm whether strategies to improve the outdoor mobility of older adults, such as removing environmental barriers and increasing assistance with mobility needs, can alleviate depressive symptoms. Although the lagged effect of change in depressive symptoms on homebound status was very small, improving access to depression treatment may still be a cost-effective approach for improving outdoor mobility among older adults in the long term. This is because depression can be successfully treated, with the possibility of complete remission (DeRubeis, Siegle, & Hollon, 2008), whereas it is unclear to what extent the state of being homebound is reversible. Targeting depression is effective because clinical trials have shown that depression treatment reduces functional impairments in older adults (Lin et al., 2000). Alleviating depressive symptoms also improves other aspects of quality of life (Menza et al., 2009), which may help improve outdoor mobility and social engagement. Conclusion Homebound status and depressive symptoms influence each other and form a feedback loop. However, change in homebound status is a stronger predictor of depressive symptoms than change in depressive symptoms is of homebound status. That is, the prevalence of depressive symptoms in older adults increases as their outdoor mobility and the degree of autonomy decreases. Improving the outdoor mobility of older adults may have immediate benefits for reducing depressive symptoms. Funding This study was supported by a grant from the National Institutes of Health, P30 AG015281, and the Michigan Center for Urban African American Aging Research. Conflict of Interest The authors declare no conflict of interest. Acknowledgments The authors thank Allison L. Goldstein and Ashley Zuverink for editing the manuscript. X. Xiang conceptualized the study, conducted the analysis, and drafted the manuscript. R. An assisted the lead author in developing the models and drafting the manuscript. H. Oh provided feedback and contributed to the writing of the manuscript. References Bruce, M. L. ( 2001). Depression and disability in late life: Directions for future research. The American Journal of Geriatric Psychiatry , 9, 102– 112. doi: 10.1097/00019442-200105000-00003 Google Scholar CrossRef Search ADS PubMed  Bruce, M. L., McAvay, G. J., Raue, P. J., Brown, E. L., Meyers, B. S., Keohane, D. J., … Weber, C. ( 2002). Major depression in elderly home health care patients. The American Journal of Psychiatry , 159, 1367– 1374. doi: 10.1176/appi.ajp.159.8.1367 Google Scholar CrossRef Search ADS PubMed  Chen, C. M., Mullan, J., Su, Y. Y., Griffiths, D., Kreis, I. A., & Chiu, H. C. ( 2012). The longitudinal relationship between depressive symptoms and disability for older adults: A population-based study. The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences , 67, 1059– 1067. doi: 10.1093/gerona/gls074 Google Scholar CrossRef Search ADS   Choi, N. G., & McDougall, G. J. ( 2007). Comparison of depressive symptoms between homebound older adults and ambulatory older adults. Aging & Mental Health , 11, 310– 322. doi: 10.1080/13607860600844614 Google Scholar CrossRef Search ADS PubMed  Collopy, B. J. ( 1988). Autonomy in long term care: Some crucial distinctions. The Gerontologist , 28( Suppl.), 10– 17. doi: 10.1093/geront/28.Suppl.10 Google Scholar CrossRef Search ADS PubMed  DeRubeis, R. J., Siegle, G. J., & Hollon, S. D. ( 2008). Cognitive therapy versus medication for depression: Treatment outcomes and neural mechanisms. Nature Reviews: Neuroscience , 9, 788– 796. doi: 10.1038/nrn2345 Google Scholar CrossRef Search ADS PubMed  Enders, C. K., & Bandalos, D. L. ( 2001). The relative performance of full information maximum likelihood estimation for missing data in structural equation models. Structural Equation Modeling , 8, 430– 457. doi: 10.1207/S15328007SEM0803_5 Google Scholar CrossRef Search ADS   Hasin, D. S., Goodwin, R. D., Stinson, F. S., & Grant, B. F. ( 2005). Epidemiology of major depressive disorder: Results from the national epidemiologic survey on alcoholism and related conditions. Archives of General Psychiatry , 62, 1097– 1106. doi: 10.1001/archpsyc.62.10.1097 Google Scholar CrossRef Search ADS PubMed  Hertzog, C., & Nesselroade, J. R. ( 1987). Beyond autoregressive models: Some implications of the trait-state distinction for the structural modeling of developmental change. Child Development , 58, 93– 109. doi: 10.2307/1130294 Google Scholar CrossRef Search ADS PubMed  Hillcoat-Nallétamby, S. ( 2014). The meaning of “independence” for older people in different residential settings. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 69, 419– 430. doi: 10.1093/geronb/gbu008 Google Scholar CrossRef Search ADS   Hooper D., Coughlan J., & Mullen M. R. ( 2008). Structural equation modeling: Guidelines for determining model fit. Electronic Journal of Business Research Methods , 6, 53– 60. Retrieved from www.ejbrm.com Jöreskog, K., & Sörbom, D. ( 1993). LISREL 8: Structural equation modeling with the SIMPLIS command language . Chicago, IL: Scientific Software International Inc. Kasper, J. D., Freedman, V. A., & Spillman, B. ( 2013). Classification of persons by dementia status in the National Health and Aging Trends Study (Technical Paper #5) . Baltimore, MD: Johns Hopkins University School of Public Health. Retrieved from www.NHATS.org Kenny, D. A., & Zautra, A. ( 1995). The trait-state-error model for multiwave data. Journal of Consulting and Clinical Psychology , 63, 52. doi: 10.1037/0022-006X.63.1.52 Google Scholar CrossRef Search ADS PubMed  Kostanjsek, N. ( 2011). Use of the International Classification of Functioning, Disability and Health (ICF) as a conceptual framework and common language for disability statistics and health information systems. BMC Public Health , 11 ( Suppl. 4), S3. doi: 10.1186/1471-2458-11-S4-S3 Google Scholar CrossRef Search ADS PubMed  Levasseur, M., Généreux, M., Bruneau, J. F., Vanasse, A., Chabot, É., Beaulac, C., & Bédard, M. M. ( 2015). Importance of proximity to resources, social support, transportation and neighborhood security for mobility and social participation in older adults: Results from a scoping study. BMC Public Health , 15, 503. doi: 10.1186/s12889-015-1824-0 Google Scholar CrossRef Search ADS PubMed  Lin, E. H., VonKorff, M., Russo, J., Katon, W., Simon, G. E., Unützer, J., … Ludman, E. ( 2000). Can depression treatment in primary care reduce disability? A stepped care approach. Archives of Family Medicine , 9, 1052– 1058. doi: 10.1001/archfami.9.10.1052 Google Scholar CrossRef Search ADS PubMed  Löwe, B., Kroenke, K., & Gräfe, K. ( 2005). Detecting and monitoring depression with a two-item questionnaire (PHQ-2). Journal of Psychosomatic Research , 58, 163– 171. doi: 10.1016/j.jpsychores.2004.09.006 Google Scholar CrossRef Search ADS PubMed  Menza, M., Dobkin, R. D., Marin, H., Mark, M. H., Gara, M., Buyske, S., … Dicke, A. ( 2009). The impact of treatment of depression on quality of life, disability and relapse in patients with Parkinson’s disease. Movement Disorders , 24, 1325– 1332. doi: 10.1002/mds.22586 Google Scholar CrossRef Search ADS PubMed  Muthén & Muthén.( 2017). Mplus user’s guide  ( 8th ed.). Los Angeles, CA: Muthén & Muthén. Nesselroade, J. R. ( 1988). Some implications of the trait-state distinction for the study of development over the life span: The case of personality. In P. B. Baltes, D. L. Featherman, & R. M. Lerner (Eds.), Life-span development and behavior  (Vol. 8, pp. 163–189). Hillsdale, NJ: Erlbaunm. Ormel, J., Rijsdijk, F. V, Sullivan, M., van Sonderen, E., & Kempen, G. I. J. M. ( 2002). Temporal and reciprocal relationship between IADL/ADL disability and depressive symptoms in late life. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 57, P338– P347. doi: 10.1093/geronb/57.4.P338 Google Scholar CrossRef Search ADS   Ornstein, K. A., Leff, B., Covinsky, K. E., Ritchie, C. S., Federman, A. D., Roberts, L., … Szanton, S. L. ( 2015). Epidemiology of the homebound population in the United States. JAMA Internal Medicine , 175, 1180– 1186. doi: 10.1001/jamainternmed.2015.1849 Google Scholar CrossRef Search ADS PubMed  Pickett, Y., Raue, P. J., & Bruce, M. L. ( 2012). Late-life depression in home healthcare. Aging Health , 8, 273– 284. doi: 10.2217/ahe.12.28 Google Scholar CrossRef Search ADS PubMed  Plath, D. ( 2008). Independence in old age: The route to social exclusion? British Journal of Social Work , 38, 1353– 1369. doi: 10.1093/bjsw/bcm045 Google Scholar CrossRef Search ADS   Qiu, W. Q., Dean, M., Liu, T., George, L., Gann, M., Cohen, J., & Bruce, M. L. ( 2010). Physical and mental health of homebound older adults: An overlooked population. Journal of the American Geriatrics Society , 58, 2423– 2428. doi: 10.1111/j.1532-5415.2010.03161.x Google Scholar CrossRef Search ADS PubMed  Richardson, T. M., Friedman, B., Podgorski, C., Knox, K., Fisher, S., He, H., & Conwell, Y. ( 2012). Depression and its correlates among older adults accessing aging services. The American Journal of Geriatric Psychiatry , 20, 346– 354. doi: 10.1097/JGP.0b013e3182107e50 Google Scholar CrossRef Search ADS PubMed  Satorra, A., & Bentler, P. M. ( 2010). Ensuring positiveness of the scaled difference chi-square test statistic. Psychometrika , 75, 243– 248. doi: 10.1007/s11336-009-9135-y Google Scholar CrossRef Search ADS PubMed  Sirey, J. A., Bruce, M. L., Carpenter, M., Booker, D., Reid, M. C., Newell, K. A., & Alexopoulos, G. S. ( 2008). Depressive symptoms and suicidal ideation among older adults receiving home delivered meals. International Journal of Geriatric Psychiatry , 23, 1306– 1311. doi: 10.1002/gps.2070 Google Scholar CrossRef Search ADS PubMed  Sivertsen, H., Bjørkløf, G. H., Engedal, K., Selbæk, G., & Helvik, A. S. ( 2015). Depression and quality of life in older persons: A review. Dementia and Geriatric Cognitive Disorders , 40, 311– 339. doi: 10.1159/000437299 Google Scholar CrossRef Search ADS PubMed  Soones, T., Federman, A., Leff, B., Siu, A. L., & Ornstein, K. ( 2017). Two-year mortality in homebound older adults: An analysis of the National Health and Aging Trends Study. Journal of the American Geriatrics Society , 65, 123– 129. doi: 10.1111/jgs.14467 Google Scholar CrossRef Search ADS PubMed  Xiang, X., & An, R. ( 2015). The impact of cognitive impairment and comorbid depression on disability, health care utilization, and costs. Psychiatric Services , 66, 1245– 1248. doi: 10.1176/appi.ps.201400511 Google Scholar CrossRef Search ADS PubMed  Xiang, X., & Brooks, J. ( 2017). Correlates of depressive symptoms among homebound and semi-homebound older adults. Journal of Gerontological Social Work , 60, 201– 214. doi: 10.1080/01634372.2017.1286625 Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png The Journals of Gerontology Series B: Psychological Sciences and Social Sciences Oxford University Press

The Bidirectional Relationship Between Depressive Symptoms and Homebound Status Among Older Adults

Loading next page...
 
/lp/ou_press/the-bidirectional-relationship-between-depressive-symptoms-and-SxCc2vbPDL
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
1079-5014
eISSN
1758-5368
D.O.I.
10.1093/geronb/gbx180
Publisher site
See Article on Publisher Site

Abstract

Abstract Objectives This study aimed to examine the bidirectional relationship between depressive symptoms and homebound status among older adults. Method The study sample included 7,603 community-dwelling older adults from the National Health and Aging Trends Study. A bivariate latent state-trait model of depressive symptoms and homebound status was estimated via structural equation modeling. Results The model fit the data well (Root Mean Square Error of Approximation = .02, Comparative Fit Index = .97, Standardized Root Mean Square Residual = .06). The relationship between homebound status and depressive symptoms can be decomposed into three parts: a moderate correlation between the stable trait components (r = .56, p <.001); a contemporary association of the state components (b = .17, p <.001); and bidirectional lagged effects between the state components. Change in homebound status was as a stronger predictor of depressive symptoms (b = .19, p < .001) than change in depressive symptoms was of homebound status (b = .06, p < .001; test of difference: Δ scaled χ2(1) = 24.2, p < .001). Discussion Homebound status and depressive symptoms form a feedback loop to influence each other. Improving the outdoor mobility of older adults may have immediate benefits for reducing depressive symptoms. Depression, Disability, Outdoor mobility Epidemiological and community-based studies have consistently documented a high prevalence of depression among older adults who are homebound or receive home-based aging services (Bruce et al., 2002; Choi & McDougall, 2007; Pickett, Raue, & Bruce, 2012; Richardson et al., 2012; Sirey et al., 2008). In a sample of older patients receiving home health care, 13.5% met the diagnostic criteria for major depression (Bruce et al., 2002). A study of participants recruited from aging service settings, many of whom are homebound, found that 27% of the sample met criteria for a current major depressive episode (Richardson et al., 2012). In comparison, the prevalence of 12-month major depressive disorder was 2.7% among the general older adult population (Hasin, Goodwin, Stinson, & Grant, 2005). A recent population-based study reported that nearly 60% of older adults who had not gone outside in the past month had clinically significant depressive symptoms compared with only 10% of non-homebound older adults (Ornstein et al., 2015). Depression and the state of being homebound have further negative consequences for healthy aging. Depression complicates the management of chronic illness, increases risk for institutionalization, and lowers quality of life and life expectancy (Sivertsen, Bjorklof, Engedal, Selbaek, & Helvik, 2015; Xiang & An, 2015). The state of being homebound is marked by the excessive burden of functional disabilities, medical comorbidities, cognitive impairment, and premature mortality (Ornstein et al., 2015; Qiu et al., 2010; Soones, Federman, Leff, Siu, & Ornstein, 2017; Xiang & Brooks, 2017). Fortunately, both conditions are potentially modifiable. Late-life depression can be successfully treated by psychotherapy, medications, or a combination of the two. The state of being homebound can be partially or fully remediated by the availability of personal assistance, assistive devices, modifications to the home and neighborhood environment, and access to transportation (Levasseur et al., 2015). Whether being homebound leads to depressive symptoms or depressive symptoms increase the risk of becoming homebound has not been tested. Studies to date have focused on estimating the contemporary, unidirectional association of homebound status and depression. It is often suggested, without empirical evidence, that being homebound leads to social isolation, increasing the risk for depression (Choi & McDougall, 2007; Pickett et al., 2012). Drawing from the extensive literature that shows the reciprocal relationship between physical disability and depression (Bruce, 2001; Chen et al., 2012; Ormel, Rijsdijk, Sullivan, van Sonderen, & Kempen, 2002), we suspect a similar relationship exists between being homebound and depression, such that the two mutually reinforce each other, creating a vicious cycle. The state of being homebound, however, is distinctive from functional impairment. Homebound older adults are an understudied, hard-to-reach population with complex medical, psychiatric, neurologic, and social needs (Xiang & Brooks, 2017), which means that their needs profile may look different from individuals with functional impairment alone. Being homebound, particularly to the point of never going outside, is an extreme form of social isolation that creates a different experience for the person affected than functional impairment alone. Despite the severity of functional impairment among homebound older adults, the state of being homebound is not inevitable for those with functional impairment. The availability of personal assistance, for example, can partially or fully remediate someone’s physical limitations, preventing the state of being homebound (Ornstein et al., 2015). Because being homebound is modifiable, its mechanisms are particularly important to investigate. The conceptual framework for the bidirectional relationship between depression and being homebound draws on the International Classification of Functioning, Disability and Health (ICF) and the concept of autonomy (Collopy, 1988). Under the ICF model, the state of being homebound is a result of the interactions between body functions and structures, activities and participation, and environmental factors (Kostanjsek, 2011). The development of chronic illness, decline in functional ability, loss of mobility, and reduction in social networks and interactions during the aging process often result in loss of physical and social independence, prohibiting an older person from staying active and engaged and to an extreme extent, resulting in the state of completely homebound. The loss of independence during the process of becoming homebound erodes individual autonomy, defined as a condition of self-government or self-determination (Collopy, 1988). Discourses on autonomy and independence with older adults have revealed that they perceive being able to “do things alone,” such as getting out to social activities, to be an important aspect of autonomy and independence (Hillcoat-Nallétamby, 2014; Plath, 2008). Based on her qualitative interviews, Plath (2008) observed that older adults’ desires for freedom, control, and self-sufficiency drive them to do things alone to attain a sense of individual achievement and competence, and the desire to doing things alone may be particularly strong in societies with liberal, individualist values such as the United States. For homebound older adults, doing things alone is often unattainable due to limited physical and mental capacity, which can cause common symptoms of depression such as a diminished sense of self-control and competence. Conversely, depression can impair body functions and structures, restrict activities and social participation, and limit social interactions and support, all of which increase the risk of becoming homebound. Homebound status and depression, therefore, can form a vicious cycle and mutually reinforce one another over time. This study aims to examine the bidirectional relationship between depressive symptoms and homebound status in older adults over time. We hypothesize that (1) depressive symptoms are positively associated with the extent to which a person is homebound, and are more prevalent among persons with a lower level of autonomy; (2) worsening depressive symptoms are associated with an immediate worsening of the extent to which a person is homebound; and (3) there are cross-lagged associations between worsening depressive symptoms and the extent to which a person is homebound. Method Study Sample Data came from Round 1 (2011) through Round 6 (2016) of the National Health and Aging Trends Study (NHATS) public-use datasets (https://www.nhats.org/). The Johns Hopkins University Institutional Review Board approved the NHATS study, which is a nationally representative panel study of Medicare beneficiaries aged 65 and older, with oversamples of persons in older age groups and African Americans. When a participant was not available to be interviewed, NHATS methodology involved the use of proxy respondents in their stead. Top reasons for having a proxy were that the sample person had dementia or another illness. A total of 7,777 community-dwelling older adults or their proxies completed in-person interviews at Round 1. Annual follow-up interviews were conducted regardless of residential status. In Round 1, 7,603 participants responded to questions regarding homebound status, 583 of whom were proxy respondents. In Rounds 2–6, the number of proxy respondents were 813, 776, 638, 507, and 463, respectively. Measures Homebound status Our conceptualization of homebound status is a continuum of outdoor mobility determined by physical capacity, availability of social support, and degree of autonomy. To this end, four questions were used to determine the extent to which a participant was homebound. Participants or their proxies were asked, “How often did you (participant) go out in the last month?” on a 5-point Likert scale (i.e., “never,” “rarely” [≤1 day], “some days” [2–4 days], “most days” [5–6 days], and “every day”). Respondents who reported “never” were then asked, “Did you ever have to stay in because no one was there to help?” Respondents who reported ever going out were asked, “Did anyone ever help you?” and “How often did you go outside by yourself?” Those who reported ever going out without help were then asked, “How much difficulty did you have leaving the house by yourself?” Figure 1 details the process of determining the level of homebound status. Participants were first grouped into three categories (non-homebound, semi-homebound, and homebound) based on the reported frequency of outdoor mobility. Within each category, participants were grouped into finer categories based on the receipt of assistance going outside and difficulty going outside alone, both of which indicated the participant’s level of independence and autonomy. For instance, we determined that participants had a lower degree of autonomy if they reported never going outside by themselves. This process resulted in eight levels of homebound status, ranging from non-homebound to completely homebound, and moving from the highest level of autonomy to the lowest level. For example, level 1, non-homebound individuals, went outside two or more days per week, received no help going out, and reported no difficulty going outside alone. This group was believed to have the highest level of autonomy. Level 7 and level 8 homebound individuals, on the other hand, reported never having gone outside in the past month; however, the level 8 homebound group was believed to have a lower level of autonomy, because they reported lack of help as a reason for not going outside. In the final model, we reduced the homebound categories to five in consideration of linearity assumption. In addition, levels 2–8 were combined to form a dichotomous indicator of homebound status (non-homebound vs. homebound) for describing patterns of change in homebound status in the descriptive analysis. Figure 1. View largeDownload slide Determining homebound status. Lvl = level. Figure 1. View largeDownload slide Determining homebound status. Lvl = level. Depressive symptoms The Patient Health Questionnaire-2 (PHQ-2), administered to participants and proxy respondents, was used to screen depressive symptoms (Löwe, Kroenke, & Gräfe, 2005). The PHQ-2 measures how often a person has been bothered by “little interest or pleasure in doing things” and “feeling down, depressed or hopeless” over the last month on a 4-point Likert scale, that is, “not at all” (0), “several days” (1), “more than half the days” (2), and “nearly every day” (3). The composite PHQ-2 score, ranging from 0 to 6, with higher scores indicating more severe depressive symptoms, was used in the main analysis. A dichotomous indicator of clinically significant depressive symptoms was also created for descriptive analysis, using a cutoff score of 3. A cutoff score of 3 has a sensitivity of 0.87 and a specificity of 0.78 for major depressive disorder, and has a sensitivity of 0.79 and specificity of 0.86 for any depressive disorder (Löwe et al., 2005). Sociodemographic and functioning characteristics Sociodemographic variables included age, sex (female, male), race/ethnicity (non-Hispanic white, non-Hispanic black, other race, and Hispanic), education (less than high school, high school or GED, some college, college degree or higher), a dichotomous indicator of living alone, and Medicaid–Medicare dual enrollment status (yes or no). A simple count of self-reported chronic diseases included heart attack/heart disease, arthritis, osteoporosis, diabetes, lung disease, stroke, and cancer. Three levels of dementia status (probable dementia, possible dementia, and no dementia) were assessed based on self-reported diagnosis of dementia or Alzheimer’s disease, the AD8 Dementia Screening Interview, and cognitive tests (Kasper, Freedman, & Spillman, 2013). Activities of daily living (ADLs) included eating, bathing, toileting, and dressing. Statistical Analysis Model description The proposed relationships between depression and homebound status were tested in a structural equation model (SEM) (Figure 2). The main model consisted of three parts: (1) a latent state-trait (LST) model for six time points addressing homebound status, (2) a LST model for six time points addressing depressive symptoms, and (3) correlations and regression effects linking the two trait-state models. Figure 2. View largeDownload slide Standardized estimates for the bivariate latent state-trait model of homebound status and depression without covariates. Time corresponds to survey waves. p < .001 all for estimates shown. Unstandardized estimates: a = .19, b = .06, c = .17, d = .09, e = .11, and f = .44. DEP = depressive symptoms; HMB = homebound status; PHQ = Patient Health Questionnaire-2. Figure 2. View largeDownload slide Standardized estimates for the bivariate latent state-trait model of homebound status and depression without covariates. Time corresponds to survey waves. p < .001 all for estimates shown. Unstandardized estimates: a = .19, b = .06, c = .17, d = .09, e = .11, and f = .44. DEP = depressive symptoms; HMB = homebound status; PHQ = Patient Health Questionnaire-2. The LST model of depression assumes that the severity of depressive symptoms at each time point is the function of two latent variables: a trait component and a state component. The trait component is the common factor or time-invariant component representing the general tendency of a person to experience depressive symptoms. The state component is a time-varying factor representing the occasion-specific variation that is not accounted for by the trait factor. The state component reflects the within-subject change in depressive symptoms in a 2-year time period. The across-time structure of the state component (States 1–6 in Figure 2) was modeled as a first-order autoregressive model (paths f). The LST model of homebound status makes a similar assumption such that the extent to which a person is homebound at each time point is the function of a time-invariant trait component and an occasion-specific state component. In longitudinal studies with repeated assessments, correlation between trait factors (i.e., relatively stable dispositions) can obscure the ability to detect associations between more acute, state-level symptoms and behaviors (Hertzog & Nesselroade, 1987). LST modeling decomposes variation into state and trait components and explicitly tests the relationship between these components, and is therefore well suited to study the reciprocal relationship between constructs over time. The model depicted in Figure 2 was obtained by linking the LST models of depression and homebound status such that the latent state variables of homebound status can influence the latent state variables of depressive symptoms, and vice versa. These effects can be contemporary (paths c) and lagged (paths a and b). In addition, the model allows correlation between the two trait factors (path g) and between the two first state factors (path d). The correlation between the trait factors represents common causes of the tendency to become depressed and homebound, such as shared genetic susceptibility and physical functioning. Building upon the main model (Figure 2), we tested an adjusted model by regressing the two trait factors on sociodemographic and clinical factors, as well as an indicator of proxy respondents. The model with trait-level covariates allowed us to examine common correlates of the tendency to become depressed and homebound. Due to the set-up of bivariate LST models, adding trait-level covariates would affect the trait association but would have little impact on the correlations between the state factors. Model specification and identification Model specification was based on the recommendations of Kenny and Zautra (1995) and Ormel et al. (2002): (1) each observed variable was regressed on two latent variables, a latent trait and a latent state variable with factor loadings set to 1.0 (paths h, i, j, and k in Figure 2); (2) autoregressive paths from one state variable to the next (paths f and e) were constrained to be equal within each construct; (3) residual variables for the endogenous latent state variables were constrained to be equal within each construct, and residual variances for observed variables were set to 0; (4) temporary associations between the endogenous homebound and depression state variables were constrained to be equal (paths c); and (5) the cross-lagged paths from homebound status to depressive symptoms (paths b) were constrained to be equal, as were the paths from depressive symptoms to homebound status (paths a). Model estimation and evaluation Model estimation was conducted in Mplus 8 (Muthén & Muthén, 2017). Robust maximum likelihood estimation was used to accommodate non-normal distributions of the variables. Estimation adjusted for the complex survey design of NHATS, using design factors at baseline including sampling weights, strata, and primary sampling units. A good-fitting model has a nonsignificant χ2 statistic, a Root Mean Square Error of Approximation (RMSEA) <.06, Comparative Fit Index (CFI) >.95, and Standardized Root Mean Square Residual (SRMR) <.08 (Hooper, Coughlan, & Mullen, 2008). The χ2 statistic is sensitive to sample size and deviations from multivariate normality (Hooper et al., 2008); it nearly always rejects the model when large samples are used (Jöreskog & Sörbom, 1993). Although an alternative statistic, χ2/df, has been developed, consensus is lacking regarding an acceptable ratio for this statistic (Hooper et al., 2008). Given these limitations of χ2 statistic and our large sample size, evaluation of model fit relied on RMSEA, CFI, and SRMR. The Satorra-Bentler scaled chi-square difference test was used to compare nested models (Satorra & Bentler, 2010). Missing data During the 5-year follow-up period, a substantial proportion of the study sample was lost to follow-up. The number of respondents, including proxy respondents, for Rounds 2–6 was 6,408 (84.3%), 5,285 (69.5%), 4,357 (57.3%), 3,874 (50.9%), and 3,459 (45.5%), respectively. At the time of Round 6, non-response accounted for about two-thirds of the loss at follow-up, and death accounted for one-third. Loss at follow-up was positively associated with age, racial/ethnic minority status, lower levels of education, Medicaid coverage, number of chronic diseases, probable or possible dementia, ADL limitations, depression, a greater degree of home confinement, and loss of independence. Missing data are handled within the analysis model by a full information maximum likelihood (FIML) method, where all available information is used to estimate the model. FIML is an optimal method of handling missing data under the condition of missing at random (Enders & Bandalos, 2001). Sensitivity analysis We performed sensitivity analysis to check the robustness of results against the categorizations of homebound status and different ways of handling missing data. The bivariate LST model (shown in Figure 2) was re-estimated using a homebound measure with six (by combining homebound status in Figure 1) and seven (by combing homebound status 7 and 8 in Figure 1) categories, respectively. This model was also re-estimated using an alternative measure of homebound status, taken from Ornstein et al. (2015). In addition, we re-estimated the model using the pattern mixture models under the condition of not missing at random (NMAR). Results Prevalence of Depression by Homebound Status The unadjusted bivariate relationship between baseline depression and homebound status appeared to be monotonic, as the prevalence of significant depressive symptoms increased significantly with increased homebound levels. One in ten non-homebound older adults (level 1) had significant depressive symptoms; in contrast, one in seven homebound older adults who never went out due to lack of help (level 8) reported significant depressive symptoms. To approximate linearity, homebound levels 3 and 4 were collapsed into a single category, as were levels 5 and 6. Levels 7 and 8 were also collapsed into a single category due to the small number of people qualifying as level 8. The reduced homebound measure therefore had five categories and was used in the SEM models (Table 1). Table 1. Prevalence of Clinically Significant Depressive Symptoms by Homebound Status at Baseline   Unweighted, N  Weighted, %  Weighted % with significant depressive symptoms within each level of homebound status  Homebound status   Non-homebound (level 1)  5,681  79.4  10.0   Semi-homebound (level 2)  612  7.4  23.8   Semi-homebound (level 3)  407  4.4  31.8   Semi-homebound (level 4)  342  3.3  32.7   Homebound (level 5)  196  2.4  36.3   Homebound (level 6)  232  2.1  42.9   Homebound (level 7)  110  0.9  60.3   Homebound (level 8)  23  0.2  68.6    Unweighted, N  Weighted, %  Weighted % with significant depressive symptoms within each level of homebound status  Homebound status   Non-homebound (level 1)  5,681  79.4  10.0   Semi-homebound (level 2)  612  7.4  23.8   Semi-homebound (level 3)  407  4.4  31.8   Semi-homebound (level 4)  342  3.3  32.7   Homebound (level 5)  196  2.4  36.3   Homebound (level 6)  232  2.1  42.9   Homebound (level 7)  110  0.9  60.3   Homebound (level 8)  23  0.2  68.6  View Large Table 1. Prevalence of Clinically Significant Depressive Symptoms by Homebound Status at Baseline   Unweighted, N  Weighted, %  Weighted % with significant depressive symptoms within each level of homebound status  Homebound status   Non-homebound (level 1)  5,681  79.4  10.0   Semi-homebound (level 2)  612  7.4  23.8   Semi-homebound (level 3)  407  4.4  31.8   Semi-homebound (level 4)  342  3.3  32.7   Homebound (level 5)  196  2.4  36.3   Homebound (level 6)  232  2.1  42.9   Homebound (level 7)  110  0.9  60.3   Homebound (level 8)  23  0.2  68.6    Unweighted, N  Weighted, %  Weighted % with significant depressive symptoms within each level of homebound status  Homebound status   Non-homebound (level 1)  5,681  79.4  10.0   Semi-homebound (level 2)  612  7.4  23.8   Semi-homebound (level 3)  407  4.4  31.8   Semi-homebound (level 4)  342  3.3  32.7   Homebound (level 5)  196  2.4  36.3   Homebound (level 6)  232  2.1  42.9   Homebound (level 7)  110  0.9  60.3   Homebound (level 8)  23  0.2  68.6  View Large Sociodemographic and Functioning Characteristics by Homebound Status Compared with non-homebound older adults, semi-homebound and homebound older adults were much older. Nearly three-quarters of homebound (levels 7 and 8) older adults were 80 years or older, whereas less than a quarter of non-homebound older adults were 80 years or older. Individuals who were females, racial/ethnic minorities, less educated, and/or Medicare–Medicaid dual eligible were over-represented in semi-homebound and homebound older adults. Semi-homebound and homebound older adults were sicker, were more likely to have cognitive impairment, and had a higher number of ADL needs compared with non-homebound older adults. Use of proxies increased as homebound level increased, ranging from 2.3% among non-homebound older adults and 51.6% among homebound older adults (levels 7 and 8). Finally, each higher level of homebound status was associated with an approximately 10% increase in prevalence of significant depressive symptoms (Table 2). Table 2. Baseline Sample Characteristics by Homebound Status   Non-homebound (level 1)  Semi-homebound (level 2)  Semi-homebound (levels 3 & 4)  Homebound (levels 5 and 6)  Homebound (levels 7 and 8)  Age groups (%)   65–69 years  30.9  25.1  13.5  10.8  7.6   70–74 years  26.9  21.5  14.6  18.4  9.9   75–79 years  19.4  20.3  19.2  12.9  10.5   80–84 years  13.7  17.7  18.7  17.8  24.1   85–89 years  6.7  11.1  19.7  25.2  26.3   90 years or older  2.4  4.3  14.3  15.0  21.6  Male (%)  46.6  41.5  25.8  22.9  32.1  Race/ethnicity (%)   White, non-Hispanic  82.4  78.4  72.0  68.9  65.9   Black, non-Hispanic  7.4  10.2  10.5  10.8  15.7   Other race, non-Hispanic  4.6  4.3  4.7  4.8  7.7   Hispanic  5.6  7.1  12.8  15.5  10.8  Education (%)   Less than high school  18.0  31.3  37.1  42.7  36.8   High school  27.7  28.2  25.5  27.8  31.9   Some college, no degree  21.9  20.8  20.0  17.2  15.4   College graduate  32.4  19.8  17.5  12.3  16.0  Living alone (%)  28.0  36.5  33.7  41.2  27.0  Medicare–Medicaid dual enrollment (%)  8.7  20.0  24.3  29.9  29.8  Number of chronic diseases (mean)  1.56 (0.02)  2.37 (0.07)  2.48 (0.06)  2.51 (0.09)  2.49 (0.14)  Dementia status (%)   No dementia  85.2  71.8  51.2  44.3  19.9   Possible dementia  9.6  14.5  16.5  14.9  14.8   Probable dementia  5.0  13.7  32.4  40.8  65.4  Number of ADL needs (mean)  0.18 (0.01)  1.15 (0.06)  1.92 (0.07)  1.87 (0.08)  2.84 (0.14)  Proxy respondents  2.3  5.4  22.7  28.6  51.6  PHQ-2 ≥ 3 (%)  10.0  23.8  32.2  39.4  62.0    Non-homebound (level 1)  Semi-homebound (level 2)  Semi-homebound (levels 3 & 4)  Homebound (levels 5 and 6)  Homebound (levels 7 and 8)  Age groups (%)   65–69 years  30.9  25.1  13.5  10.8  7.6   70–74 years  26.9  21.5  14.6  18.4  9.9   75–79 years  19.4  20.3  19.2  12.9  10.5   80–84 years  13.7  17.7  18.7  17.8  24.1   85–89 years  6.7  11.1  19.7  25.2  26.3   90 years or older  2.4  4.3  14.3  15.0  21.6  Male (%)  46.6  41.5  25.8  22.9  32.1  Race/ethnicity (%)   White, non-Hispanic  82.4  78.4  72.0  68.9  65.9   Black, non-Hispanic  7.4  10.2  10.5  10.8  15.7   Other race, non-Hispanic  4.6  4.3  4.7  4.8  7.7   Hispanic  5.6  7.1  12.8  15.5  10.8  Education (%)   Less than high school  18.0  31.3  37.1  42.7  36.8   High school  27.7  28.2  25.5  27.8  31.9   Some college, no degree  21.9  20.8  20.0  17.2  15.4   College graduate  32.4  19.8  17.5  12.3  16.0  Living alone (%)  28.0  36.5  33.7  41.2  27.0  Medicare–Medicaid dual enrollment (%)  8.7  20.0  24.3  29.9  29.8  Number of chronic diseases (mean)  1.56 (0.02)  2.37 (0.07)  2.48 (0.06)  2.51 (0.09)  2.49 (0.14)  Dementia status (%)   No dementia  85.2  71.8  51.2  44.3  19.9   Possible dementia  9.6  14.5  16.5  14.9  14.8   Probable dementia  5.0  13.7  32.4  40.8  65.4  Number of ADL needs (mean)  0.18 (0.01)  1.15 (0.06)  1.92 (0.07)  1.87 (0.08)  2.84 (0.14)  Proxy respondents  2.3  5.4  22.7  28.6  51.6  PHQ-2 ≥ 3 (%)  10.0  23.8  32.2  39.4  62.0  Note. ADL = activities of daily living; PHQ = Patient Health Questionnaire-2. Estimates adjusted for the complex design of NHATS; p < .001 for all comparisons by homebound status. View Large Table 2. Baseline Sample Characteristics by Homebound Status   Non-homebound (level 1)  Semi-homebound (level 2)  Semi-homebound (levels 3 & 4)  Homebound (levels 5 and 6)  Homebound (levels 7 and 8)  Age groups (%)   65–69 years  30.9  25.1  13.5  10.8  7.6   70–74 years  26.9  21.5  14.6  18.4  9.9   75–79 years  19.4  20.3  19.2  12.9  10.5   80–84 years  13.7  17.7  18.7  17.8  24.1   85–89 years  6.7  11.1  19.7  25.2  26.3   90 years or older  2.4  4.3  14.3  15.0  21.6  Male (%)  46.6  41.5  25.8  22.9  32.1  Race/ethnicity (%)   White, non-Hispanic  82.4  78.4  72.0  68.9  65.9   Black, non-Hispanic  7.4  10.2  10.5  10.8  15.7   Other race, non-Hispanic  4.6  4.3  4.7  4.8  7.7   Hispanic  5.6  7.1  12.8  15.5  10.8  Education (%)   Less than high school  18.0  31.3  37.1  42.7  36.8   High school  27.7  28.2  25.5  27.8  31.9   Some college, no degree  21.9  20.8  20.0  17.2  15.4   College graduate  32.4  19.8  17.5  12.3  16.0  Living alone (%)  28.0  36.5  33.7  41.2  27.0  Medicare–Medicaid dual enrollment (%)  8.7  20.0  24.3  29.9  29.8  Number of chronic diseases (mean)  1.56 (0.02)  2.37 (0.07)  2.48 (0.06)  2.51 (0.09)  2.49 (0.14)  Dementia status (%)   No dementia  85.2  71.8  51.2  44.3  19.9   Possible dementia  9.6  14.5  16.5  14.9  14.8   Probable dementia  5.0  13.7  32.4  40.8  65.4  Number of ADL needs (mean)  0.18 (0.01)  1.15 (0.06)  1.92 (0.07)  1.87 (0.08)  2.84 (0.14)  Proxy respondents  2.3  5.4  22.7  28.6  51.6  PHQ-2 ≥ 3 (%)  10.0  23.8  32.2  39.4  62.0    Non-homebound (level 1)  Semi-homebound (level 2)  Semi-homebound (levels 3 & 4)  Homebound (levels 5 and 6)  Homebound (levels 7 and 8)  Age groups (%)   65–69 years  30.9  25.1  13.5  10.8  7.6   70–74 years  26.9  21.5  14.6  18.4  9.9   75–79 years  19.4  20.3  19.2  12.9  10.5   80–84 years  13.7  17.7  18.7  17.8  24.1   85–89 years  6.7  11.1  19.7  25.2  26.3   90 years or older  2.4  4.3  14.3  15.0  21.6  Male (%)  46.6  41.5  25.8  22.9  32.1  Race/ethnicity (%)   White, non-Hispanic  82.4  78.4  72.0  68.9  65.9   Black, non-Hispanic  7.4  10.2  10.5  10.8  15.7   Other race, non-Hispanic  4.6  4.3  4.7  4.8  7.7   Hispanic  5.6  7.1  12.8  15.5  10.8  Education (%)   Less than high school  18.0  31.3  37.1  42.7  36.8   High school  27.7  28.2  25.5  27.8  31.9   Some college, no degree  21.9  20.8  20.0  17.2  15.4   College graduate  32.4  19.8  17.5  12.3  16.0  Living alone (%)  28.0  36.5  33.7  41.2  27.0  Medicare–Medicaid dual enrollment (%)  8.7  20.0  24.3  29.9  29.8  Number of chronic diseases (mean)  1.56 (0.02)  2.37 (0.07)  2.48 (0.06)  2.51 (0.09)  2.49 (0.14)  Dementia status (%)   No dementia  85.2  71.8  51.2  44.3  19.9   Possible dementia  9.6  14.5  16.5  14.9  14.8   Probable dementia  5.0  13.7  32.4  40.8  65.4  Number of ADL needs (mean)  0.18 (0.01)  1.15 (0.06)  1.92 (0.07)  1.87 (0.08)  2.84 (0.14)  Proxy respondents  2.3  5.4  22.7  28.6  51.6  PHQ-2 ≥ 3 (%)  10.0  23.8  32.2  39.4  62.0  Note. ADL = activities of daily living; PHQ = Patient Health Questionnaire-2. Estimates adjusted for the complex design of NHATS; p < .001 for all comparisons by homebound status. View Large Patterns of Change in Homebound and Depression Status The most prevalent patterns of change in homebound and depression status were in persistently non-depressed and persistently non-homebound individuals, respectively. Both onset and remission of depression and homebound status occurred, and onset was slightly more frequent than remission overall. About 4% of the sample was persistently depressed in a given 2-year period. The prevalence of being persistently homebound was more frequent, ranging from 14.2% to 16.5% of the sample in a given 2-year period (Table 3). Table 3. Change and Stability in Depression and Homebound Status per Pair of Survey Rounds Patterns of change and stability  T1–T2  T2–T3  T3–T4  T4–T5  T5–T6  Depressiona   00-Persistently non-depressed  78.5  80.7  82.0  82.1  82.6   01-Onset of depression  7.8  7.90  7.2  7.1  7.0   10-Remission of depression  8.9  7.0  6.5  6.4  6.1   11-Persistently depressed  4.8  4.4  4.4  4.4  4.3  Homeboundb   00-Persistently non-homebound  70.5  69.6  68.6  69.2  67.5   01-Onset of being homebound  9.2  9.9  10.6  8.9  11.1   10-Remission of being homebound  6.1  5.8  5.3  5.6  4.9   11-Persistently homebound  14.2  14.7  15.5  16.3  16.5  Patterns of change and stability  T1–T2  T2–T3  T3–T4  T4–T5  T5–T6  Depressiona   00-Persistently non-depressed  78.5  80.7  82.0  82.1  82.6   01-Onset of depression  7.8  7.90  7.2  7.1  7.0   10-Remission of depression  8.9  7.0  6.5  6.4  6.1   11-Persistently depressed  4.8  4.4  4.4  4.4  4.3  Homeboundb   00-Persistently non-homebound  70.5  69.6  68.6  69.2  67.5   01-Onset of being homebound  9.2  9.9  10.6  8.9  11.1   10-Remission of being homebound  6.1  5.8  5.3  5.6  4.9   11-Persistently homebound  14.2  14.7  15.5  16.3  16.5  Notes. Weighted % presented in table, adjusting for the complex survey design of NHATS. aDepressed is defined as ≥3 on the PHQ-2. bHomebound combines levels 2 through 8 on the original homebound measure; non-homebound individuals went outside in the last month, received no help, and had no difficulty going out alone. View Large Table 3. Change and Stability in Depression and Homebound Status per Pair of Survey Rounds Patterns of change and stability  T1–T2  T2–T3  T3–T4  T4–T5  T5–T6  Depressiona   00-Persistently non-depressed  78.5  80.7  82.0  82.1  82.6   01-Onset of depression  7.8  7.90  7.2  7.1  7.0   10-Remission of depression  8.9  7.0  6.5  6.4  6.1   11-Persistently depressed  4.8  4.4  4.4  4.4  4.3  Homeboundb   00-Persistently non-homebound  70.5  69.6  68.6  69.2  67.5   01-Onset of being homebound  9.2  9.9  10.6  8.9  11.1   10-Remission of being homebound  6.1  5.8  5.3  5.6  4.9   11-Persistently homebound  14.2  14.7  15.5  16.3  16.5  Patterns of change and stability  T1–T2  T2–T3  T3–T4  T4–T5  T5–T6  Depressiona   00-Persistently non-depressed  78.5  80.7  82.0  82.1  82.6   01-Onset of depression  7.8  7.90  7.2  7.1  7.0   10-Remission of depression  8.9  7.0  6.5  6.4  6.1   11-Persistently depressed  4.8  4.4  4.4  4.4  4.3  Homeboundb   00-Persistently non-homebound  70.5  69.6  68.6  69.2  67.5   01-Onset of being homebound  9.2  9.9  10.6  8.9  11.1   10-Remission of being homebound  6.1  5.8  5.3  5.6  4.9   11-Persistently homebound  14.2  14.7  15.5  16.3  16.5  Notes. Weighted % presented in table, adjusting for the complex survey design of NHATS. aDepressed is defined as ≥3 on the PHQ-2. bHomebound combines levels 2 through 8 on the original homebound measure; non-homebound individuals went outside in the last month, received no help, and had no difficulty going out alone. View Large Unadjusted SEM Model The bivariate LST model without covariates had a good model fit based on RMSEA = .02, CFI = .97, and SRMR = .06. The model had a significant χ2(65) = 343.6 (p < .001), likely due to the large sample size. Removing the equality constraints on the cross-lagged paths (a and b) did not lead to a significant improvement in overall model fit (Δ scaled χ2(8) = 10.8, p = .213), suggesting that it was appropriate to pose these equality constraints. Standardized estimates are shown in Figure 2, with unstandardized estimates included in the figure caption. The interpretation of unstandardized coefficients for the contemporary (c) and cross-lagged paths (a and b) had clear meaning (i.e., the association of one category change in homebound status and one point change in the PHQ-2), whereas the interpretation of standardized estimates was unclear (i.e., the association of 1 SD change in homebound status and 1 SD change in the PHQ-2). Therefore, we focused on unstandardized estimates for the cross-lagged paths in reporting and discussion. A moderate association existed between the homebound status trait and depressive symptoms trait variable (r = .56, p < .001). The cross-variable associations between homebound state and depression state variables (bc = .17, p < .001) were small but significant. Change in homebound status and change in depressive symptoms had significant contemporary effects on each other. One category of change in homebound status (e.g., from level 1 non-homebound to level 2 semi-homebound) was associated with a .17-point increase in the PHQ-2 score during the same year. This suggests that changes in homebound status and changes in depressive symptoms influenced each other rather quickly. Change in homebound status and change in depressive symptoms also had significant 1-year lagged effects on each other. One category of change in homebound status was associated with a .19-point increase in the PHQ-2 score 1 year later (ba = .19, p < .001). On the other hand, a 1-point change in PHQ-2 level was associated with .06-point increase in the level of homebound status (bb = .06, p < .001). These two sets of lagged effects were significantly different from each other (Δ scaled χ2[1] = 24.2, p < .001). This suggests that the lagged effect of change in homebound status on change in depressive symptoms was stronger than the lagged effect of change in depressive symptoms on change in homebound status. Adjusted SEM Model The bivariate LST model with trait variables regressed on sociodemographic and functioning covariates had a good model fit based on RMSEA = .02, CFI = .96, and SRMR = .04. As expected, adding trait-level covariates had little impact on the cross-lagged effects between the state variables (ba = .18, bb = .06, p < .001 for both). The correlation between homebound trait and depression trait variables was reduced from r = .56 to r = .18 (p < .001). Homebound trait variable was positively associated with age (β = .23, p < .001), Hispanic ethnicity as compared with non-Hispanic white ethnicity (β = .04, p = .006), Medicare–Medicaid dual eligibility (β = .07, p < .001), number of chronic disease (β = .12, p < .001), probable (β = .18, p < .001) or possible (β = .06, p < .001) dementia, number of ADL needs (β = .56, p < .001), and use of proxy respondents (β = .10, p < .001). Male sex (β = –.10, p < .001) and higher educational achievement (βhigh school = –.04, p = .014; βsome college = –.04, p = .012; βcollege degree = –.07, p < .001 as compared with less than high school education) were negatively associated with homebound trait variable. Depression trait variable was negatively associated with age (β = –.03, p = .014) and some college (β = –.09, p < .001) or college degree education (β = –.17, p < .001) as compared with less than high school education. Depression trait variable was positively associated with Hispanic ethnicity as compared with non-Hispanic white ethnicity (β = .07, p = .006), living alone (β = .06, p < .001), Medicare–Medicaid dual eligibility (β = .05, p < .001), number of chronic disease (β = .22, p < .001), probable (β = .14, p < .001) or possible (β = .06, p < .001) dementia, and number of ADL needs (β = .37, p < .001). Sensitivity Analysis Estimates similar to those reported above were obtained by re-estimating the bivariate LST models using different categorizations and measures of homebound status, as well as using the pattern mixture models under the condition of NMAR. These results suggested that the study findings were robust against varying measures of homebound status and methods of handling missing data. Discussion This study suggests that the relationship between homebound status and depressive symptoms in community-dwelling older adults can be decomposed into three parts. First, there is a moderate correlation between the stable trait components of homebound status and depressive symptoms, and this correlation cannot be fully explained away by sociodemographic and functioning factors. Second, there is a contemporary correlation of change in homebound status and depressive symptoms. Finally, there are bidirectional lagged effects between change in homebound status and depressive symptoms, and the lagged effect of change in homebound status on depressive symptoms was stronger than the lagged effect of change in depressive symptoms on homebound status. These results provide the first set of empirical evidence that demonstrate the reciprocal relationship between homebound status and depressive symptoms. Although little research exists to evaluate the reciprocal relationship between homebound status and depressive symptoms, the abundant literature on the relationship between depression and disability provides some insights into the nature of this relationship, because homebound status is closely tied to disability. Several longitudinal studies have examined the bidirectional relationship between depressive symptoms and disability, often measured in terms of ADLs (Chen et al., 2012; Ormel et al., 2002). Using different samples and analytical methods, these studies have consistently showed that disability and depressive symptoms influence each other in terms of a feedback loop, and that change in disability is a stronger predictor of depressive symptoms than change in depressive symptoms is of disability (Chen et al., 2012; Ormel et al., 2002). These results are in line with findings from the present study. It is plausible that disability and physical illness produce instantaneous effects on mental health, whereas the detrimental effects of mental illness on physical health manifest over an extended period (Ormel et al., 2002). The state-trait model showed that individual differences in homebound status and depressive symptoms are, to a large extent, stable among older adults aged 65 and older during a 4-year period. The stable trait component in LST models does not necessarily imply the trait is biological or non-modifiable. Rather, trait variance may be due to stable environments or other factors (Kenny & Zautra, 1995). Traits can be temporally invariant or can also change, albeit slowly (Nesselroade, 1988). The correlation between these stable trait components can be a major contributor to the high prevalence of depression in homebound older adults observed in previous studies (Bruce et al., 2002; Choi & McDougall, 2007; Pickett et al., 2012; Richardson et al., 2012; Sirey et al., 2008). The temporary and lagged associations between the state components play a weaker role in explaining the relationship between homebound status and depression. It is likely that most people already experienced an episode of being homebound or depressive symptoms prior to study enrollment. Changes in homebound status and depressive symptoms may be uncommon afterwards. Alternatively, people could have developed strategies to cope with these changes because of prior experiences, making them less vulnerable to the detrimental effects of the changes. Future studies following a younger cohort for an extended period may shed light on the reciprocal relationship between homebound status and depressive symptoms. To inform interventions for depression and home confinement in older adults, future studies should examine the common causes of these conditions and identify individual, interpersonal, and environment factors amendable to intervention. Study findings suggested a positive feedback loop between homebound status and depressive symptoms that may lead to a vicious cycle of worsening depressive symptoms and higher level of confinement to the home. However, the prevalence of clinically significant depressive symptoms decreased over time, and the percentage of people who were not homebound only moderately increased in this sample. One possible explanation is attrition and survival bias. Depression and the state of being homebound are both associated with reduced life expectancy (Sivertsen et al., 2015; Soones et al., 2017); therefore, as time goes by, the number of people who are homebound or depressed is reduced in the sample. Limitations Interpretation of the study results should consider the following limitations. All measures were self-reported and subject to recall bias and reporting errors. PHQ-2 is a brief screening tool and requires follow-up with PHQ-9 to further probe depression severity. PHQ-2 is not a diagnostic test and is prone to ceiling effects. In addition, proxy respondents answered PHQ-2 questions on behalf of sample persons unavailable for interviews, and the reliability and validity of administering PHQ-2 to a proxy has not been established. Proxies accounted for half of homebound respondents and only a small percentage of non-homebound older adults. Systematic reporting differences may exist between proxies and self-respondents, which could bias the study estimates. For example, proxies may over-report the severity of homebound status, resulting in an overestimate of the impact of homebound status on depression. Finally, the homebound measure was restricted to activities in the last month and had various skip patterns. The skip patterns prevented us from gaining a comprehensive understanding of the causes of homebound status. Implications Reversing the state of being homebound may provide immediate benefits for improving the mental health of older adults. Prospective studies should confirm whether strategies to improve the outdoor mobility of older adults, such as removing environmental barriers and increasing assistance with mobility needs, can alleviate depressive symptoms. Although the lagged effect of change in depressive symptoms on homebound status was very small, improving access to depression treatment may still be a cost-effective approach for improving outdoor mobility among older adults in the long term. This is because depression can be successfully treated, with the possibility of complete remission (DeRubeis, Siegle, & Hollon, 2008), whereas it is unclear to what extent the state of being homebound is reversible. Targeting depression is effective because clinical trials have shown that depression treatment reduces functional impairments in older adults (Lin et al., 2000). Alleviating depressive symptoms also improves other aspects of quality of life (Menza et al., 2009), which may help improve outdoor mobility and social engagement. Conclusion Homebound status and depressive symptoms influence each other and form a feedback loop. However, change in homebound status is a stronger predictor of depressive symptoms than change in depressive symptoms is of homebound status. That is, the prevalence of depressive symptoms in older adults increases as their outdoor mobility and the degree of autonomy decreases. Improving the outdoor mobility of older adults may have immediate benefits for reducing depressive symptoms. Funding This study was supported by a grant from the National Institutes of Health, P30 AG015281, and the Michigan Center for Urban African American Aging Research. Conflict of Interest The authors declare no conflict of interest. Acknowledgments The authors thank Allison L. Goldstein and Ashley Zuverink for editing the manuscript. X. Xiang conceptualized the study, conducted the analysis, and drafted the manuscript. R. An assisted the lead author in developing the models and drafting the manuscript. H. Oh provided feedback and contributed to the writing of the manuscript. References Bruce, M. L. ( 2001). Depression and disability in late life: Directions for future research. The American Journal of Geriatric Psychiatry , 9, 102– 112. doi: 10.1097/00019442-200105000-00003 Google Scholar CrossRef Search ADS PubMed  Bruce, M. L., McAvay, G. J., Raue, P. J., Brown, E. L., Meyers, B. S., Keohane, D. J., … Weber, C. ( 2002). Major depression in elderly home health care patients. The American Journal of Psychiatry , 159, 1367– 1374. doi: 10.1176/appi.ajp.159.8.1367 Google Scholar CrossRef Search ADS PubMed  Chen, C. M., Mullan, J., Su, Y. Y., Griffiths, D., Kreis, I. A., & Chiu, H. C. ( 2012). The longitudinal relationship between depressive symptoms and disability for older adults: A population-based study. The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences , 67, 1059– 1067. doi: 10.1093/gerona/gls074 Google Scholar CrossRef Search ADS   Choi, N. G., & McDougall, G. J. ( 2007). Comparison of depressive symptoms between homebound older adults and ambulatory older adults. Aging & Mental Health , 11, 310– 322. doi: 10.1080/13607860600844614 Google Scholar CrossRef Search ADS PubMed  Collopy, B. J. ( 1988). Autonomy in long term care: Some crucial distinctions. The Gerontologist , 28( Suppl.), 10– 17. doi: 10.1093/geront/28.Suppl.10 Google Scholar CrossRef Search ADS PubMed  DeRubeis, R. J., Siegle, G. J., & Hollon, S. D. ( 2008). Cognitive therapy versus medication for depression: Treatment outcomes and neural mechanisms. Nature Reviews: Neuroscience , 9, 788– 796. doi: 10.1038/nrn2345 Google Scholar CrossRef Search ADS PubMed  Enders, C. K., & Bandalos, D. L. ( 2001). The relative performance of full information maximum likelihood estimation for missing data in structural equation models. Structural Equation Modeling , 8, 430– 457. doi: 10.1207/S15328007SEM0803_5 Google Scholar CrossRef Search ADS   Hasin, D. S., Goodwin, R. D., Stinson, F. S., & Grant, B. F. ( 2005). Epidemiology of major depressive disorder: Results from the national epidemiologic survey on alcoholism and related conditions. Archives of General Psychiatry , 62, 1097– 1106. doi: 10.1001/archpsyc.62.10.1097 Google Scholar CrossRef Search ADS PubMed  Hertzog, C., & Nesselroade, J. R. ( 1987). Beyond autoregressive models: Some implications of the trait-state distinction for the structural modeling of developmental change. Child Development , 58, 93– 109. doi: 10.2307/1130294 Google Scholar CrossRef Search ADS PubMed  Hillcoat-Nallétamby, S. ( 2014). The meaning of “independence” for older people in different residential settings. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 69, 419– 430. doi: 10.1093/geronb/gbu008 Google Scholar CrossRef Search ADS   Hooper D., Coughlan J., & Mullen M. R. ( 2008). Structural equation modeling: Guidelines for determining model fit. Electronic Journal of Business Research Methods , 6, 53– 60. Retrieved from www.ejbrm.com Jöreskog, K., & Sörbom, D. ( 1993). LISREL 8: Structural equation modeling with the SIMPLIS command language . Chicago, IL: Scientific Software International Inc. Kasper, J. D., Freedman, V. A., & Spillman, B. ( 2013). Classification of persons by dementia status in the National Health and Aging Trends Study (Technical Paper #5) . Baltimore, MD: Johns Hopkins University School of Public Health. Retrieved from www.NHATS.org Kenny, D. A., & Zautra, A. ( 1995). The trait-state-error model for multiwave data. Journal of Consulting and Clinical Psychology , 63, 52. doi: 10.1037/0022-006X.63.1.52 Google Scholar CrossRef Search ADS PubMed  Kostanjsek, N. ( 2011). Use of the International Classification of Functioning, Disability and Health (ICF) as a conceptual framework and common language for disability statistics and health information systems. BMC Public Health , 11 ( Suppl. 4), S3. doi: 10.1186/1471-2458-11-S4-S3 Google Scholar CrossRef Search ADS PubMed  Levasseur, M., Généreux, M., Bruneau, J. F., Vanasse, A., Chabot, É., Beaulac, C., & Bédard, M. M. ( 2015). Importance of proximity to resources, social support, transportation and neighborhood security for mobility and social participation in older adults: Results from a scoping study. BMC Public Health , 15, 503. doi: 10.1186/s12889-015-1824-0 Google Scholar CrossRef Search ADS PubMed  Lin, E. H., VonKorff, M., Russo, J., Katon, W., Simon, G. E., Unützer, J., … Ludman, E. ( 2000). Can depression treatment in primary care reduce disability? A stepped care approach. Archives of Family Medicine , 9, 1052– 1058. doi: 10.1001/archfami.9.10.1052 Google Scholar CrossRef Search ADS PubMed  Löwe, B., Kroenke, K., & Gräfe, K. ( 2005). Detecting and monitoring depression with a two-item questionnaire (PHQ-2). Journal of Psychosomatic Research , 58, 163– 171. doi: 10.1016/j.jpsychores.2004.09.006 Google Scholar CrossRef Search ADS PubMed  Menza, M., Dobkin, R. D., Marin, H., Mark, M. H., Gara, M., Buyske, S., … Dicke, A. ( 2009). The impact of treatment of depression on quality of life, disability and relapse in patients with Parkinson’s disease. Movement Disorders , 24, 1325– 1332. doi: 10.1002/mds.22586 Google Scholar CrossRef Search ADS PubMed  Muthén & Muthén.( 2017). Mplus user’s guide  ( 8th ed.). Los Angeles, CA: Muthén & Muthén. Nesselroade, J. R. ( 1988). Some implications of the trait-state distinction for the study of development over the life span: The case of personality. In P. B. Baltes, D. L. Featherman, & R. M. Lerner (Eds.), Life-span development and behavior  (Vol. 8, pp. 163–189). Hillsdale, NJ: Erlbaunm. Ormel, J., Rijsdijk, F. V, Sullivan, M., van Sonderen, E., & Kempen, G. I. J. M. ( 2002). Temporal and reciprocal relationship between IADL/ADL disability and depressive symptoms in late life. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences , 57, P338– P347. doi: 10.1093/geronb/57.4.P338 Google Scholar CrossRef Search ADS   Ornstein, K. A., Leff, B., Covinsky, K. E., Ritchie, C. S., Federman, A. D., Roberts, L., … Szanton, S. L. ( 2015). Epidemiology of the homebound population in the United States. JAMA Internal Medicine , 175, 1180– 1186. doi: 10.1001/jamainternmed.2015.1849 Google Scholar CrossRef Search ADS PubMed  Pickett, Y., Raue, P. J., & Bruce, M. L. ( 2012). Late-life depression in home healthcare. Aging Health , 8, 273– 284. doi: 10.2217/ahe.12.28 Google Scholar CrossRef Search ADS PubMed  Plath, D. ( 2008). Independence in old age: The route to social exclusion? British Journal of Social Work , 38, 1353– 1369. doi: 10.1093/bjsw/bcm045 Google Scholar CrossRef Search ADS   Qiu, W. Q., Dean, M., Liu, T., George, L., Gann, M., Cohen, J., & Bruce, M. L. ( 2010). Physical and mental health of homebound older adults: An overlooked population. Journal of the American Geriatrics Society , 58, 2423– 2428. doi: 10.1111/j.1532-5415.2010.03161.x Google Scholar CrossRef Search ADS PubMed  Richardson, T. M., Friedman, B., Podgorski, C., Knox, K., Fisher, S., He, H., & Conwell, Y. ( 2012). Depression and its correlates among older adults accessing aging services. The American Journal of Geriatric Psychiatry , 20, 346– 354. doi: 10.1097/JGP.0b013e3182107e50 Google Scholar CrossRef Search ADS PubMed  Satorra, A., & Bentler, P. M. ( 2010). Ensuring positiveness of the scaled difference chi-square test statistic. Psychometrika , 75, 243– 248. doi: 10.1007/s11336-009-9135-y Google Scholar CrossRef Search ADS PubMed  Sirey, J. A., Bruce, M. L., Carpenter, M., Booker, D., Reid, M. C., Newell, K. A., & Alexopoulos, G. S. ( 2008). Depressive symptoms and suicidal ideation among older adults receiving home delivered meals. International Journal of Geriatric Psychiatry , 23, 1306– 1311. doi: 10.1002/gps.2070 Google Scholar CrossRef Search ADS PubMed  Sivertsen, H., Bjørkløf, G. H., Engedal, K., Selbæk, G., & Helvik, A. S. ( 2015). Depression and quality of life in older persons: A review. Dementia and Geriatric Cognitive Disorders , 40, 311– 339. doi: 10.1159/000437299 Google Scholar CrossRef Search ADS PubMed  Soones, T., Federman, A., Leff, B., Siu, A. L., & Ornstein, K. ( 2017). Two-year mortality in homebound older adults: An analysis of the National Health and Aging Trends Study. Journal of the American Geriatrics Society , 65, 123– 129. doi: 10.1111/jgs.14467 Google Scholar CrossRef Search ADS PubMed  Xiang, X., & An, R. ( 2015). The impact of cognitive impairment and comorbid depression on disability, health care utilization, and costs. Psychiatric Services , 66, 1245– 1248. doi: 10.1176/appi.ps.201400511 Google Scholar CrossRef Search ADS PubMed  Xiang, X., & Brooks, J. ( 2017). Correlates of depressive symptoms among homebound and semi-homebound older adults. Journal of Gerontological Social Work , 60, 201– 214. doi: 10.1080/01634372.2017.1286625 Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.

Journal

The Journals of Gerontology Series B: Psychological Sciences and Social SciencesOxford University Press

Published: Jan 25, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off