Poor Vision, Functioning, and Depressive Symptoms: A Test of the Activity Restriction Model

Abstract

Purpose: This study tested the applicability of the activity restriction model of depressed affect to the context of poor vision in late life. This model hypothesizes that late-life stressors contribute to poorer mental health not only directly but also indirectly by restricting routine everyday functioning. Method: We used data from a national probability-based sample of older adults (N = 1,178; M = 69.2 years, approximately 50% female). Vision was assessed both subjectively (via self-report) and objectively (via a visual acuity test). Respondents also reported on their levels of physical and driving limitations, feelings of social isolation, and symptoms of depression. Results: Path analyses indicated a strong fit of the data to the activity restriction model for subjective vision. In addition to directly contributing to higher depressive symptomatology, subjective vision contributed to depressive symptoms indirectly by predicting more physical limitations and feelings of social isolation that, in turn, contributed to more symptoms of depression; driving limitations did not mediate the relationship between subjective vision and depressive symptomatology. Objective vision contributed significantly to physical and driving limitations but was unrelated to feelings of social isolation and depressive symptomatology. Implications: Supporting the activity restriction model, poorer self-rated vision in late life contributes to lower mental health directly and also indirectly by restricting individuals’ ability to carry out routine day-to-day physical activities and increasing their feelings of social isolation. Interventions for older adults with vision-related problems could focus on maintaining or enhancing their physical and social functioning in order to promote their adaptation to poor vision.

Good visual capacity is among the most important contributors to the value older individuals attach to their lives (Jopp, Roth, & Oswald, 2008). With advancing age, however, vision-related problems are a common stressor. Recent estimates obtained from the National Health Interview Survey indicate that 17% of individuals 65+ years of age reported having trouble with their vision even with corrective glasses and lenses (Federal Interagency Forum on Aging-Related Statistics, 2008). Much evidence indicates that poor visual function can have adverse effects on the lives and well-being of older adults (e.g., Crews, 1994; Federal Interagency Forum on Aging-Related Statistics, 2008; Horowitz, 2004; O’Donnell, 2005). For example, Wahl, Schilling, Oswald, and Heyl (1999) found that poorer visual ability was related to lower subjective well-being in a sample of older adults recruited through physician practices. Bourque, Leger, Pushkar, and Beland (2007) reported that poor vision was related to lower life satisfaction in a representative sample of older Canadians, and Reinhardt (1996) found a similar relationship in a convenience sample of visually impaired older adults residing in the United States.

An especially well-established finding is that poor vision is linked to lower psychological health or emotional well-being, most notably to higher depressive symptomatology. Research indicates that not only are symptoms of depression elevated in older adults diagnosed with diseases of the eye such as age-related macular degeneration (e.g., Casten & Rovner, 2008; Tolman, Hill, Kleinschmidt, & Gregg, 2005) but also that poorer visual capacity in general is reliably linked to more depressed affect in late life. For example, Lupsakko, Mantyjarvi, Kautiainen, and Sulkava (2002) found that higher levels of depressive symptomatology occurred in a population-based sample of visually impaired older adults relative to their nonimpaired peers, and Crews, Jones, and Kim (2006) reported that mild or moderate levels of depressive symptoms are a common comorbid condition among elders who are visually impaired. Numerous other studies with convenience and probability-based samples also have found that older adults who rated their vision to be poorer had higher levels of depressive symptomatology (e.g., Brody et al., 2001; Casten, Rovner, & Edmonds, 2001; Chou, 2008; Femia, Zarit, & Johansson, 2001; Furner, Wallace, Arguelles, Miles, & Goldberg, 2006; Reinhardt, 1996).

The current study builds on this body of literature by testing three mediators in the link between poor vision and depressive symptomatology. It uses the activity restriction model of depressed affect (Williamson & Christie, 2009; Williamson & Shaffer, 2000) as the theoretical framework, which explains that an important factor in individuals’ psychological adjustment when faced with life stressors is the extent to which the stressor restricts their normal day-to-day activities. According to this model, disruptions in carrying out normal personal and social activities (e.g., shopping, visiting friends) at least partially account for the link between health-related stressors and emotional well-being. The model thus proposes mediational pathways, whereby life stressors such as chronic physical health problems often contribute not only directly to poorer mental health outcomes but also contribute indirectly via their adverse effects on individuals’ ability to carry out day-to-day routine activities. Such activity restriction, in its turn, contributes to poorer mental health and emotional well-being. The activity restriction model has garnered considerable support in studies on late-life stressors such as chronic pain, family caregiving, and living with breast cancer, demonstrating that the link from such stressors to depressive symptomatology is mediated partially or wholly by declines in physical functioning and/or social engagement (Bookwala, Harralson, & Parmelee, 2003; Bookwala & Schulz, 2000; Williamson, 2000a, 2000b; Williamson & Dooley, 2001; Williamson & Shaffer, 2000). However, the activity restriction model of depressed affect has received little attention in the domain of vision impairment. In this study, we propose that the model offers a theoretically viable explanation for the impact of vision impairment on depressed affect in late life. Based on the activity restriction model, we hypothesize that poor vision contributes to depressed affect not only directly but it also results in more limitations in the performance of personal and social activities that, in turn, contribute to higher depressed mood.

Poor vision among older adults has consistently been found to have adverse effects on a wide variety of day-to-day functional domains. Numerous studies have found that poor vision diminishes older adults’ ability to carry out everyday physical activities (e.g., Branch, Horowitz, & Carr, 1989; Crews et al., 2006; Lindő & Nordholm, 1999; Horowitz, 1994; Laforge, Spector, & Sternberg, 1992; Travis, Boerner, Reinhardt, & Horowitz, 2004; West et al., 1997). For example, Gunnel and Nordholm found that reading, writing, and watching television were common functional problems associated with poor vision. In addition, mobility-related problems, such as difficulty with moving about outdoors, driving or using public transportation, and conducting bank-related business, were commonly reported by their visually impaired elderly sample. Travis and colleagues found that older adults with impaired vision reported higher levels of difficulty with instrumental activities of daily living (e.g., selecting/locating and identifying clothing, food items, and money; using a telephone; writing checks; and taking medications) and that these individuals identified these visual problems rather than other health problems as the source of these difficulties. Crews and colleagues also found that older adults who reported visual impairment were more likely to report difficulty performing everyday tasks, such as walking, climbing steps, and shopping relative to their age-matched peers.

A functional domain in late life that is particularly adversely affected by poor vision is driving ability (e.g., Horowitz, Boerner, & Reinhardt, 2002; Keay et al., 2009; Satariano, MacLeod, Cohn, & Ragland, 2004). Indeed, studies have reported that older adults often cite problems with their eyesight as the reason they limited their driving or avoided driving altogether (Persson, 1993; Ragland, Satariano, & MacLeod, 2004). An emerging literature also indicates that in addition to contributing to limitations in physical function and driving ability, poor vision can have an adverse impact on social functioning domains, such as social network size, relationship maintenance, and social integration (O’Donnell, 2005; Reinhardt, Boerner, & Benn, 2003; Wahl & Tesch-Romer, 2001). Accordingly, studies have found that poor vision is correlated with lower levels of social integration (Femia et al., 2001), more difficulty engaging in social relationships (Crews et al., 2006), and higher levels of social isolation (Femia et al., 2001). Despite widespread evidence from individual studies linking poor vision in older adults to either their functioning (physical, driving-related, or social) or depressive symptoms, there has been little attempt to examine the interrelationships among these variables simultaneously. We address this gap in the literature by testing these relationships in a single model using the activity restriction model of depressed affect as our guiding framework, which posits that the contribution of stressors such as poor vision to depressive symptomatology is explained at least partially by the impact of poor vision on multiple domains of functioning.

In the present study, we tested the role of three mediators—physical limitations, driving limitations, and feelings of social isolation—in the link from vision impairment to depressive symptomatology using both subjective (self-report) and objective (visual acuity test) indicators of vision impairment. Difficulties with performing day-to-day activities have been used as indicators of activity restriction by Williamson (2000b), and feelings of social isolation have received attention in other vision research (Femia et al., 2001). Subjective and objective measures of vision are included in the study because they provide complementary information on visual ability among older adults (Horowitz, 2004) and including both can offer a more comprehensive understanding of the effects of vision on psychological health. We drew data for this study from the National Social Life, Health, and Aging Project (NSHAP; O’Muircheartaigh, Eckman, & Smith, 2009; Smith et al., 2009), a survey-based study of a nationally representative sample of older adults residing in the United States. Consistent with the activity restriction model of Williamson (2000) and Williamson and Shaffer (2000), we tested (a) the direct link from subjective and objective vision to depressive symptoms and (b) the extent to which this association is mediated by functional impairment and social isolation known to result from poor vision. We hypothesized that poor vision—assessed both subjectively and objectively—will be associated with greater depressive symptomatology and that more physical and driving limitations and more feelings of social isolation will at least partially account for this association. We predicted that these relationships would exist after controlling for sociodemographic variables that are known to be related to functioning and depressive symptomatology in older samples such as ours, including age, gender, education level, race, and marital status.

Methods

Sample

Our sample consisted of older adults who participated in the NSHAP (O’Muircheartaigh et al., 2009; Smith et al., 2009). The NSHAP is a large-scale survey study that assessed components of health, social relationships, and well-being in older adults aged 57–85 years using face-to-face interviews and self-administered questionnaires. The NSHAP data were collected in 2005–2006 for which eligible cases were identified as part of a larger national area probability sample of households (O’Muircheartaigh et al., 2009). The NSHAP sample was balanced on age and gender subgroups and oversampled African Americans and Latinos. Approximately 50% of the original NSHAP sample (N = 3,005) was randomly selected to receive the objective vision assessment (N = 1,506). To be eligible for the present analyses, respondents who were sampled to receive the objective assessment had to have no missing data on the study variables. This yielded a final sample of 1,170 older adults (77.7% of NSHAP respondents who were randomly selected for the objective vision assessment). On average, respondents were approximately 69.2 years of age (SD = 7.9) and 50.7% of the sample was female (N = 593). The majority of the sample had at least a high-school diploma or equivalent (79.3%, N = 928). More than 80% of the sample described their ethnic background as White (80.3%, N = 939), 13.1% (N = 153) as Black, and 6.7% (N = 78) as other. Almost two thirds of the sample were married or living with a partner (65.9%, N = 770), 10.2% (N = 120) was separated or divorced, almost one fifth was widowed (19.7%, N = 231), and 4.2% (N = 49) was never married.

Measures

Vision.—

Self-reported vision (subjective vision) was assessed in the NSHAP using a single-item measure with a 5-point rating scale (Schumm et al., 2009). Respondents were asked to rate their eyesight using the item “With your glasses or contact lenses if you wear them, is your eyesight poor, fair, good, very good, or excellent?” Responses were recoded such that higher scores represented poorer self-reported vision. The mean rating for self-reported vision was 2.54 (SD = 1.05). Slightly less than half the sample subjectively rated their vision to be poor (4.3%, N = 50), fair (12.2%, N = 143), or good (33.4%, N = 391). The remaining respondents rated their vision to be very good (33.0%, N = 386) or excellent (17.1%, N = 200). Visual acuity (objective vision) was assessed in both eyes together at a distance of 3 m using a chart with Sloan optotypes manufactured by Precision Vision (catalog number 2104; Schumm et al., 2009). NSHAP interviewers were trained to follow a detailed protocol to ensure consistent distance from the chart (via use of a premeasured string laid out on the floor), line of sight (the interviewer seated the respondent and held the chart at the respondent’s eye level), and lighting (sufficient light for reading with low glare or strong backlighting). Respondents who normally wore glasses or contact lenses for driving or distance vision were instructed to wear them during the test and asked to begin reading the smallest discernible line. Success or not on this task resulted in the respondent being instructed to read the successive line directly below or above the just-read line, respectively, until the smallest discernible line that could be successfully read was determined. Performance on the visual acuity test was coded using standard guidelines (Schumm et al., 2009). Scores of 20/20 vision or better were coded as “normal or better” vision, between 20/40 and 20/20 were coded as “good” vision, between approximately 20/60 and 20/40 were coded as “moderately decreased” vision, and those with worse scores were coded as “poor” vision. Respondents who were unable to read the largest line on the Sloan chart at 3 m represented visual acuity worse than 20/200; these individuals were coded to have 20/200 vision. The mean score on the visual acuity test was 1.14 (SD = 0.76). Based on their performance on this test, 16.3% (N = 191) had “normal or better” vision, more than 60% had “good” vision (60.7%, N = 710), 16.1% (N = 188) had “moderately decreased” vision, and almost 7% (6.9%, N = 81) had “poor” vision (N = 27 scored 20/80, N = 20 scored 20/100, and N = 34 scored 20/125 or worse).

Functional limitations were assessed in terms of difficulty with day-to-day physical function and difficulty with driving. Seven items assessing the amount of difficulty that respondents experienced related to walking (e.g., walking a street block, walking across a room) and self-care activities (e.g., dressing, eating, using the toilet, bathing or showering, getting in and out of bed) were summed to yield a measure of physical limitations. Driving limitations were measured via the sum of responses to two items, difficulty experienced driving during the day and driving at night. Responses to items assessing physical and driving limitations were made on a 4-point scale (1 = no difficulty, 2 = some difficulty, 3 = much difficulty, and 4 = unable to do). Activities that had never been performed were scored as “no difficulty ”; for the driving items, respondents could indicate that they no longer drove, which was coded as “unable to do.” For both measures, responses were scored such that higher scores represented greater difficulty. The physical limitations measure obtained a Cronbach’s alpha of .82, and the driving limitations measure obtained a Cronbach’s alpha of .83. Mean scores on physical limitations and driving limitations were 8.1 (SD = 2.20) and 2.94 (SD = 1.75), respectively.

Feelings of social isolation were assessed using three items (how often do you feel isolated from others, feel that you lack companionship, and feel left out?) adapted from the UCLA Loneliness Scale (Russell, Peplau, & Cutrona, 1980). Similar items were used by Femia and colleagues (2001) in their study of outcomes associated with vision impairment to measure feelings of social isolation in older adults. Responses to these items were made using a 3-point scale ranging from hardly ever or never to often, and scores were summed such that higher scores represented more feelings of social isolation (Cronbach’s α = .80). The sample mean for social isolation was 3.97 (SD = 1.4).

Depressive symptomatology was measured using an 11-item version of the Center for Epidemiological Studies-Depression scale (CES-D; Radloff, 1977) that assesses the severity of depressive symptoms over a 1-week recall period. Items (e.g., “I could not ‘get going’” and “I felt that everything I did was an effort”) were answered on a 4-point scale ranging from rarely/none of the time to most of the time. (It should be noted that the response scale in the NSHAP, which ranged from 1 to 4, was recoded for this study to match the original CES-D scale, which ranges from 0 to 3.) Higher scores reflected greater depressive symptomatology, and the scale was internally consistent (Cronbach’s α = .79). Mean depressive symptomatology was 5.44 (SD = 5.16).

Control Variables.—

Sociodemographic variables included age, gender, education level, race, and marital status; these variables were measured in the NSHAP using standard self-report items. In the model testing, race was coded as a dichotomous variable (White vs. non-White) as was marital status (married vs. not); gender was coded as 1 = male and 2 = female.

Data Analyses

Descriptive statistics and zero-order correlations between study variables were computed first. Model testing was conducted via path analytic procedures using the EQS statistical software program (Bentler & Wu, 1995) and maximum likelihood estimation of path coefficients. All variables in the model were treated as observed variables in the analysis. Direct paths were specified from subjective and objective vision to the three mediator variables (physical limitations, driving limitations, and feelings of social isolation) and depressive symptomatology; indirect paths from subjective and objective vision to depressive symptomatology via the mediator variables also were estimated. Sociodemographic variables (age, education level, gender, marital status, and race) were included as control variables in the model testing as follows: (a) when significant bivariate correlations (p < .01) between the control and study variables were obtained, these relationships were included as paths in the model and (b) covariances were estimated between control variables that were significantly correlated with each other (p < .01). The Wald test (which identifies parameters that can be dropped without significantly compromising model fit) and the Lagrange test (which identifies parameters that, if added, could significantly improve model fit) were used to identify improvements for the overall fit of the model and to generate a more parsimonious model.

To determine the fit of the model to the data, we computed multiple goodness-of-fit indices, including the χ² goodness-of-fit index, the comparative fit index (CFI), the incremental fit index (IFI), normed and non-normed fit indices (NFI and NNFI), the standardized mean-square residual (SRMR), and the root mean-square error of approximation (RMSEA). Multiple fit indices are recommended because the χ² goodness-of-fit statistic is sensitive to sample size, and thus, in tests with even reasonably sized samples, relying solely on the χ² goodness-of-fit index can result in the rejection of a theoretically meaningful model (Bentler, 1990). Acceptable cutoffs for these fit indices include a nonsignificant χ² goodness of fit (p > .05); values of .95 or larger for CFI, IFI, NFI, and NNFI; and SRMR and RMSEA values of .06 or lower (Hu & Bentler, 1999).

Results

Zero-order correlations between study variables are listed in Table 1. A small but statistically significant correlation was obtained between self-reported vision and scores on the visual acuity test (r = .29, p < .001). Consistent with expectations, in general, poorer self-reported and objectively assessed vision was related to more physical and driving limitations, feelings of social isolation, and symptoms of depression. As Table 1 indicates, the bivariate relationships between subjective vision and these variables were somewhat stronger than seen with objective vision.

Table 1.

Bivariate Correlations for Study Variables (N = 1,178)

	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
Self-rated vision (1)	.29	.22	.27	.20	.32	.08	.06	−.27	.22	.13
Visual acuity (2)		.16	.23	.08	.16	.30	.12	−.18	.05	.20
Physical limitations (3)			.41	.15	.38	.11	.12	−.20	.09	.09
Driving limitations (4)				.13	.25	.20	.21	−.21	.12	.18
Social isolation (5)					.50	−.06	.13	−.07	.12	.20
Depressive symptomatology (6)						.05	.14	−.19	.08	.20
Age (7)							.06	−.16	−.10	.21
Gender (8)								−.11	.02	.28
Education level (9)									−.26	−.13
Race (10)										.11
Marital status (11)

Notes: |r| ≥ .08 were significant at p < .01 or better. Higher scores represent poorer vision; more physical limitations, driving limitations, and feelings of social isolation; higher depressive symptomatology; and being older, female, more educated, non-White, and not married.

Next, we conducted a path analysis using the EQS software to test the activity restriction model of depressed affect in the context of poor vision in which we hypothesized that the path from poor vision to depressive symptomatology would be at least partially mediated by more physical limitations, driving limitations, and social isolation. In other words, we anticipated that poorer vision would predict more difficulties with physical function and driving and more feelings of social isolation that, in turn, would contribute to more symptoms of depression. To this end, the model estimated (a1) paths from the exogenous variables of subjective and objective vision to the four endogenous variables of physical limitations, driving limitations, feelings of social isolation, and depressive symptoms and (b) paths from physical limitations, driving limitations, and feelings of social isolation to depressive symptoms (see Figure 1). In addition, because of the significant bivariate correlation between the exogenous variables of subjective and objective vision, they were allowed to covary; the error variances for physical limitations, driving limitations, and feelings of social isolation also were permitted to covary. Paths also were estimated as needed from the sociodemographic variables to each endogenous variable (physical limitations, driving limitations, social isolation, and depressive symptomatology) based on significant bivariate correlations that were obtained (p < .01, see Table 1). These paths were as follows: from age to physical limitations and driving limitations; gender, marital status, and race to physical limitations, driving limitations, social isolation, and depressive symptoms; and education level to physical limitations, driving limitations, and depressive symptoms. Covariances between significantly correlated (p < .01) control variables and subjective and objective vision, pairs of control variables that were significantly correlated, and the error terms for physical limitations, driving limitations, and feelings of social isolation also were estimated in the initial test of the model.

Figure 1.

Final path model testing physical limitations, driving limitations, and social isolation as mediators of the vision-depressive symptomatology link. Notes: **Standardized path coefficients significant at p < .01 or better. Significant paths from the control variables to the endogenous variables (not shown in Figure 1) were as follows: Being female (β = .09, z = 2.99, p < .001) and less educated (β = −.13, z = −4.34, p < .001) were associated with more physical limitations, and being older (β = .12, z = 4.36, p < .001), female (β = .16, z = 6.05, p < .001), less educated (β = −.09, z = −2.98, p < .001), and non-White (β = .06, z = 2.18, p < .05) were associated with more driving limitations. Being younger (β = −.12, z = −3.83, p < .001), female (β = .08, z = 2.61, p < .05), and not married (β = .17, z = 5.71, p < .001) were significantly associated with more feelings of social isolation. Being less educated (β = −.07, z = −2.93, p < .01) and being White (β = −.05, z = −2.00, p < .05) were linked to more depressive symptoms.

Most of the goodness-of-fit estimates yielded by the initial test of the model indicated a strong fit of the model to the data (NFI = .982, NNFI = .901, CFI = .986, IFI = .986, SRMR = .025, RMSEA = .053, ${\chi }_8^2$ = 33.98, p < .001). The Wald test results were examined to determine if the model could be further improved by dropping paths specified in the initial test of the conceptual model. Eight paths could be dropped without compromising model fit, five of which involved paths from a control variable to an endogenous variable (from marital status to physical limitations, driving limitations, and depressive symptoms; from race to physical limitations; and from gender to depressive symptoms). Because estimates of these paths were not directly relevant to the theoretical model, these five parameters were dropped in the next iteration of model testing. The remaining three paths were retained in the retest of the model (from objective vision to feelings of social isolation and depressive symptoms and from driving limitations to depressive symptoms) because of their centrality to the theoretical model. The retest of this more parsimonious model incorporating the earlier respecifications yielded even stronger fit indices (NFI = .983, NNFI = .951, CFI = .989, IFI = .990, SRMR = .023, RMSEA = .037, ${\chi }_{12}^2$ = 31.31, p < .01). The Lagrange multivariate test (which identifies parameters that could be added to the model to improve model fit) in this model run indicated that the fit of the model could be significantly improved by estimating a path from age to social isolation. Thus, this path was added and the model retested. All fit indices yielded at this point by the model testing (NFI = .991, NNFI = .984, CFI = .997, IFI = .997, SRMR = .020, RMSEA = .021, ${\chi }_{11}^2$ = 16.75, p > .10) exceeded acceptable cutoff levels (Hu & Bentler, 1999). The $\Delta {\chi }^2$ test, which tests the improvement in model fit offered by the final model over the original model, indicated that the more parsimonious final model provided a significantly better fit to the data over the original model ($\Delta {\chi }_3^2$ = 17.23, p < .001). Overall, the final model explained 8.2% of the variance in physical limitations, 16.5% in driving limitations, 9.0% in social isolation, and 37.8% in depressive symptoms. Figure 1 presents the final model with standardized regression coefficients for the key paths tested in the model. Significant paths from the control variables to the endogenous variables are listed in the notes that accompany Figure 1.

As hypothesized, poorer self-reported vision (β = .16, z = 5.31, p < .001) and poorer visual acuity (β = .07, z = 2.14, p < .05) contributed to more physical limitations; likewise, poorer self-reported vision and visual acuity contributed to more difficulty experienced with driving (β = .17, z = 5.98, p < .001 and β = .16, z = 5.41, p < .001, respectively). In other words, both subjectively and objectively assessed vision predicted lower capacity to carry out day-to-day tasks and drive oneself. With regard to feelings of social isolation, poorer subjective vision predicted greater social isolation (β = .17, z = 5.63, p < .001); objectively assessed vision did not contribute significantly to feelings of social isolation, however (β = .02, z < 1, ns). As Figure 1 also shows, subjectively assessed vision had a direct effect on depressive symptomatology (β = .15, z = 5.89, p < .001) such that poorer self-reported vision predicted more depressive symptoms; objectively assessed vision, however, did not contribute significantly to depressive symptomatology in the model (β = .02, z < 1, ns). More physical limitations (β = .26, z = 10.22, p < .001) and social isolation (β = .43, z = 17.90, p < .001) were significantly associated with greater depressive symptomatology as well, but difficulty with driving was unrelated to symptoms of depression (β = .03, z = 1.01, ns).

Next, we examined the extent to which the effects of vision on depressive symptomatology were mediated by functional impairment and social isolation. The EQS test of the path model yielded tests for the overall indirect effects of both subjective and objective vision on depressive symptoms via the set of three mediator variables. These tests revealed that the collective indirect effect of self-reported vision on depressive symptoms via physical limitations, driving limitations, and social isolation was significant (β = .12, z = 6.99, p < .001), but objectively assessed vision did not have an indirect effect on symptoms of depression via the three mediators (β = .03, z = 1.90, p > .05). Follow-up Sobel tests were performed subsequently to separately test the indirect effect of subjectively assessed vision on depressive symptomatology via each individual mediator. These tests indicated that physical limitations (z = 4.71, p < .001) and feelings of social isolation (z = 5.40, p < .001) individually mediated the link between self-reported vision and depressive symptoms. Driving limitations, however, did not significantly mediate the link between poor vision and depressive symptomatology (z = 1.00, ns). Collectively, these results indicate that in addition to contributing directly to depressive symptoms, poorer self-reported vision also contributes to more depressive symptoms indirectly by predicting more physical limitations and feelings of social isolation that, in turn, predict more depressive symptoms. Individual Sobel tests were not performed for the mediation of the path from objectively assessed vision to depressive symptomatology because the overall model did not support such mediation.

Discussion

In this study, we examined the interrelationships among objectively and subjectively assessed vision, functioning, and depressive symptoms in a representative sample of community-dwelling older adults who participated in the National Social Life, Health, and Aging Project. We tested the applicability of the activity restriction model of depressed affect (Williamson & Christie, 2009; Williamson & Shaffer, 2000) in the context of poor vision in late life. According to Williamson’s model, a chronic health-related stressor can contribute to higher depressive symptomatology both directly as well as indirectly through a cascading mechanism whereby the stressor also contributes to poorer overall functioning which, in turn, contributes to more depressive symptoms. Such a cascading mechanism is consistent with research linking other chronic conditions to disability and, in turn, disability to poor psychological adaptation (see Kahana, Kahana, Namazi, Kercher, & Stange, 1997). Our findings show that the activity restriction model of depressed affect provided an excellent fit to the data within the context of poor visual function in late life. Specifically, direct effects of poor self-reported vision on depressive symptomatology and its indirect effects on depressive symptomatology via physical limitations and feelings of social isolation were seen. The data indicated that poorer self-reported vision directly predicted higher depressive symptomatology. In addition, poorer self-reported vision predicted both greater physical limitations and more social isolation, and these, in turn, predicted more depressive symptoms. The indirect (or mediated) effect of poor self-reported vision on depressive symptomatology via each of these two mediators was statistically significant. These findings are noteworthy more so because the tested model controlled for objective vision as assessed by performance on a test of visual acuity. The data did not indicate, however, that limitations in driving during the day or night mediate the link between poor self-reported vision and symptoms of depression. Although both subjective and objective vision predicted more driving limitations, the latter did not contribute significantly to depressive symptomatology in the overall model. It is plausible that, after controlling for the effects of poor vision, restricting their driving activity may serve to control fears and anxiety about driving that older adults can experience (Taylor, Alpass, Stephens, & Towers, 2011) especially when their vision is not optimal (Persson, 1993; Ragland et al., 2004) and that this offsets any adverse role of restricted driving activity in depressed affect.

That restriction in physical and social function serve as intervening variables that partially account for the contribution of poor vision to depressive symptomatology underscores the adverse impact such restriction can have on psychological adjustment under stress. Williamson and Dooley (2001) offer that the restriction of activities—be they physical or social in nature—may function as an effective coping behavior that individuals use when faced with a stressor. For example, in the case of poor vision, such activity restriction can prevent falls or accidents. Paradoxically, however, the very restriction of such routine activities intended to assist with coping with the stressor can result in poorer psychological adjustment. Thus, when a health-related stressor results in activity restriction, the adverse effects of the stressor may become even further exacerbated. Our findings in support of the activity restriction model for vision are generally consistent with research on the applicability of this model to understanding the effects of other health-related stressors, such as chronic pain, family caregiving, and cancer (Bookwala & Schulz, 2000; Bookwala et al., 2003; Williamson, 2000a; Williamson & Dooley, 2001; Williamson & Schulz, 1992; Williamson & Shaffer, 2000; Williamson, Shaffer, & Schulz, 1998).

Although the complete model obtained an excellent fit to the data, it is important to note that the activity restriction model (Williamson & Christie, 2009; Williamson & Shaffer, 2000) did not obtain support in the case of objectively assessed vision. When all model components were considered simultaneously, performance on the visual acuity test did not significantly predict depressive symptoms directly or indirectly via the mediators. The differential applicability of the activity restriction model of depressed affect (Williamson & Christie, 2009; Williamson & Shaffer, 2000) for subjectively versus objectively assessed vision may be partially explained by the small (albeit significant) correlation between self-reported vision and the visual acuity test, suggesting low concordance between subjective and objective assessments of vision. Low concordance between subjective and objective vision has been reported in earlier research (Fors, Thorslund, & Parker, 2006) and may signify that self-reported vision is a broader more global construct that is likely also to be based on comparisons with past visual ability and/or more vision-related dimensions rather than on distance vision alone (e.g., ease of reading printed material). Supporting evidence comes from studies that have found that difficulty with reading and taking medication pills are frequently experienced by older adults who have poor vision (Raasch & Rubin, 1993; Windham et al., 2005).

Subjective and objective measures of vision also are viewed as offering complementary information on visual ability among older adults, and each may serve to counter the limitations of the other (Horowitz, 2004). Horowitz points out that a performance-based measure of visual acuity does not assess impairments that may be experienced due to visual field, contrast sensitivity, and depth perception problems and thus can offer an objective but more exclusive estimate of visual problems. In contrast, she notes that a self-reported measure of visual ability is likely to capture more global evaluations of vision, including trouble seeing, difficulty reading newspaper print, and difficulty seeing a familiar person across a room’s length that provide insight into the day-to-day implications of poor vision. As such, self-reported data on visual function are likely to offer a subjective but more inclusive estimate of visual problems. Research on other self-reported aspects of health as measured via global indicators also supports this idea. For example, self-rated health is routinely measured using a single global item. Krause and Jay (1994) found that this standard self-rated health item taps different dimensions ranging from general health to specific health behaviors. Future research may gain from assessing multiple aspects of visual function objectively (e.g., impairment in depth perception, extreme contrast or brightness sensitivity in addition to visual acuity as assessed in NSHAP) and subjectively (e.g., self-reported trouble with reading newspaper print and recognizing people at a small distance in addition to the global assessment of one’s vision as assessed in NSHAP) to more clearly evaluate the direct and indirect link between vision and its effects on depressive symptomatology.

Our study offers important and new insights to inform current understanding about the psychological impact of poor vision. It indicates that the link between vision and psychological adjustment in late life is complex rather than simple and merits further investigation. It identifies that the activity restriction model of depressed affect may offer a viable explanation for the effects of poor vision—especially self-reported vision—on adaptation to this stressor. Based on our findings, we recommend that clinicians and other health care practitioners who work with older adults with compromised visual ability evaluate its impact on their clients’ physical limitations and feelings of social isolation in addition to the impact on mental health. Multifocused clinical and community-based interventions that can maintain or enhance physical and social functioning levels are also likely to be more effective in promoting adaptation to poor vision and better mental health. For example, interventions that focus on making transportation services available and providing home visits from individuals in the community to assist with daily chores or engage with the visually impaired socially may be especially effective for better psychological adjustment.

Using a nationally representative sample of community-dwelling older adults increases the significance of our findings because of their enhanced generalizability to the larger population of elders in the United States. Moreover, the use of objective and subjective assessments of vision in examining the psychological impact of poor vision is an important strength of the study. Despite these strengths, it should be noted that the data are cross-sectional in nature and, as such, preclude causal interpretations of relationships. For example, it is plausible that symptoms of depression contribute to poorer self-rated vision rather than vice versa or, alternatively, that there is a bidirectional or even cyclical association between these variables. It is important to note, however, that the activity restriction model of depressed affect, which describes that health-related stressors such as poor vision contribute to depressive symptomatology, is a well-established theoretical model in the realm of adaptation to stress and that this model provided an excellent fit to the data in the current study. Another limitation tied to the cross-sectional nature of the data is that the present study cannot speak to the issue of vision loss experienced by the sample because prior levels of vision are unknown. As such, it is not possible to assess the extent to which deterioration in vision plays a role in well-being and the extent to which physical limitations, driving limitations, and social isolation mediate the impact of such deterioration on depressive symptomatology. Likewise, the present data do not allow an examination of the impact of driving cessation on depressive symptomatology. As future waves of the NSHAP become available, however, it will be possible to assess longitudinal relationships among deterioration in vision, the resulting loss of function (e.g., the cessation of driving, decline in physical, and social functioning), and psychological adjustment. As indicated earlier, the current study also was limited in that only one type of visual ability was assessed objectively, namely visual acuity; the NSHAP did not include other objective tests of visual function. However, research has indicated that in addition to visual acuity, age-related declines are seen in contrast sensitivity, glare, and visual field tests (Rubin et al., 1997). The current study also did not assess the impact of diseases of the eye (e.g., age-related macular degeneration and glaucoma) that are known to have an adverse impact on older adults’ quality of life (Crossland, Gould, Helman, Feely, & Rubin, 2007; Freeman, Munoz, West, Jampel, & Friedman, 2008). Future research would benefit from an inclusion of objective performance tests related to various aspects of vision and a closer examination of the effects of eye diseases when studying the extent to which functioning is an intervening variable in the link between poor vision and depressive symptomatology. In addition, the present study used feelings of social isolation as an indicator of social functioning. This variable has been assessed by other studies on late life vision (e.g., Femia et al., 2001) and may be the result of restricted social activities, but it does not assess social engagement per se. It also should be noted that the measure of depressive symptomatology (the CES-D; Radloff, 1977) relied on self-report. It may be valuable for future studies to focus on social engagement specifically and to use a clinician-based measure of depressive symptomatology, such as the Hamilton Rating Scale for Depression (Hamilton, 1960). Finally, the applicability of the activity restriction model may vary across different groups (e.g., based on age, gender, or race), and psychosocial resources are likely to mitigate the adverse effects of poor vision on psychological adjustment (Bookwala, Resubmitted); however, these issues were not examined in the present study. Further research is recommended on such moderating variables in the psychological adjustment to poor vision. Despite the limitations that are noted, the present study is characterized by several strengths and makes important and novel contributions to our understanding of the linkages among vision, physical and social functioning, and depressive symptomatology in late life.

Funding

The NSHAP is supported by the National Institutes of Health—the National Institute on Aging, Office of Womens Health Research, Office of AIDS Research, and the Office of Behavioral and Social Science Research (5R01AG021487).