Functional Limitations as Potential Mediators of the Effects of Self-Reported Vision Status on Fall Risk of Older Adults

Abstract

Objective: To test whether limitations in mobility and large-muscle functioning mediate self-reported vision status to increase fall risk among respondents age 65 and above. Method: This study used two waves from the Health and Retirement Study. We conducted binary logistic and negative binomial regression analyses to test indirect paths leading from self-reported vision status to falls, via indices of mobility and large-muscle functioning. Results: Limited evidence was found for a mediating effect among women; however, large-muscle groups were implicated as partially mediating risk factors for falls among men with fair self-reported vision status. Discussion: Implications of these findings are discussed including the need for prioritizing improved muscle strength of older men and women with poor vision as a preventive measure against falls.

Keywords

falls prevention self-reported vision functional impairment

Amid the large array of factors that affect the health, well-being, and safety of older adults, age-related vision impairments are commonly cited on lists that enumerate risk factors for falls. Many studies that have looked at the association between vision status and fall risk in later life have assessed total effects—most commonly those associated with specific clinical decrements of vision that often occur in acuity, visual fields, binocularity, and contrast sensitivity (De Boer et al., 2004; Ivers, Cumming, Mitchell, & Attebo, 1998; Lord & Dayhew, 2001). A more modest number of studies have examined the contributions of integrated or indirect (i.e., the moderated and/or mediated) effects of poor vision on or by other systems as they relate to falls experienced by older people.

Previous research has suggested that poor vision could moderate declines in other systems, as when age-related limitations in limb functioning are exacerbated by decreases in physical activity due to worsening vision (Lord, Smith, & Menant, 2010; Steinman, 2008; Steinman, Pynoos, & Nguyen, 2009). By definition, a moderator changes the strength of a relationship between two variables. In the example cited above, limb functioning, as a predictor of falls, could be moderated by different levels of vision status (Ray & Wolf, 2008). Moderation is also defined by how it is calculated quantitatively. In the case of the papers by Steinman (2008) and Steinman and colleagues (2009), moderation was calculated by creating interaction variables (i.e., vision by functioning), which were included in regression models.

A second means by which effects of vision status could be associated with other systems is by way of a mediated effect, in which effects of vision status on falls are contingent on declines in other systems. Whereas moderating variables influence the strength of the relationship between two variables, mediating variables are said to intervene between independent variables and dependent variables to produce an outcome that is directly related to the mediator. In other words, mediating variables explain the relationship between two variables. For example, when an older individual with losses in the central visual field trips over unseen objects in the environment, she may be less able to recover and avoid a fall when limb functioning is also reduced. While poor vision directly causes the person to trip, it only indirectly causes the fall. A more direct cause of the fall is impaired limb functioning. In the current study, we assessed mediation of vision by other systems, with respect to fall risk. In this proposed scenario, the total association between vision and falling can be partitioned into (a) effects that are directly related to vision status and (b) indirect effects of vision status that are contingent on other physical losses (unrelated to vision impairment).

The purpose of this study was to examine mediating effects of two indices of physical functioning on self-reported vision status, as it relates to falls in older adults. To address this issue, the study relied on analytical steps outlined by Baron and Kenny (1986). In contrast to earlier studies that tested whether vision status moderates change in upper- and lower-limb functioning to increase fall risk (Steinman, 2008; Steinman et al., 2009), the present study tests the degree to which functioning in various systems (i.e., mobility functioning and large-muscle functioning) contributes to the total effect of vision status in relation to falls. As a theoretical basis, variable selection and analyses for this study were guided by the sociological model of disability originally developed by Nagi (1965, 1976) and later was conceptualized further by Verbrugge and Jette (1994), Crimmins (2004), and others. The model consists of four main stages that progress from (a) risk factors through (b) pathology and impairments to result in (c) limitation of functioning, and eventually (d) disability in daily activities (see Figure 1).

Figure 1.

Adapted model of disability (Nagi, 1965, 1976).

In accordance with Nagi’s (1965, 1976) model, we included controls for sociodemographic risk factors for falls identified in previous research (Masud & Morris, 2001), and covariates representing diagnosed chronic conditions, functional limitations, and indicators of disability in daily self-maintenance activities.

Method

Data Source

Data for this study came from two waves (2006 and 2008) of the University of Michigan, Health and Retirement Study (HRS; 2010), including variables that were cleaned and constructed by the Research and Development Corporation (RAND, 2010). Variables representing sociodemographic characteristics, diagnosed chronic conditions, disability in activities of daily living (ADL), and instrumental activities of daily living (IADL), vision status, and the two indices of functioning are from the 2006 wave of HRS. An indicator of whether HRS participants had fallen between 2006 and 2008 was included in the latter wave of HRS and was used in our study as the main dependent variable.

In 2006, HRS included 19,740 surviving participants (representing an attrition rate of roughly 13% compared with the study’s baseline in 1998). Participants in the 2006 wave were included for analyses in the present study if they were age 65 or above, if their interviews were conducted with the individual (not by proxy), and if they provided responses in both waves used in the present study (2006 and 2008). With these selection criteria applied, a weighted, nationally representative sample of 11,080 participants remained for our analyses.

Study Variables

Controls/covariates

Three sociodemographic risk variables, including age, race, and years of education in 2006, were statistically controlled. Age and education were coded as continuous variables, whereas race was coded as an indicator variable with “White” as the reference category. Other categories were “Black” and “Other.” In addition, gender was controlled on the basis of performing separate analyses for men and women.

A variable representing the number of chronic conditions at baseline was computed by summing the conditions reported by participants, as diagnosed by their doctor. Participants of HRS were asked whether or not they had ever been told by a doctor that they had high blood pressure, diabetes, cancer, lung problems, heart problems, stroke, psychiatric problems, or arthritis. When summed, values for this aggregate continuous variable could range from 0 to 8. With respect to disability, the ADL index was a summation of three items proposed by Wallace and Herzog (1995), including bathing, dressing, and eating. The IADL index was composed of items measuring participants’ difficulty using the phone, managing money, managing medications, shopping, and preparing meals. Participants were coded in the positive direction (has disability = 1) if they indicated any difficulty, or were unable to perform each task. Each self-reported ADL and/or IADL limitation added one to their respective summary measures, such that ADL limitations ranged from 0 to 3, and IADL limitations ranged from 0 to 5.

Finally, each functional index (described below) was also included as a covariate in analyses when they were not being tested for their roles as mediators in respective models. For example, when mobility functioning was tested as a mediator, large-muscle functioning was held constant as a covariate representing physical functioning, and vice versa.

Vision status

Self-reported vision status was assessed in the HRS by asking participants to state whether their vision was excellent, very good, good, fair, or poor when wearing glasses or corrective lenses. Participants could also state that they were legally blind. This variable was recoded into three indicator variables—one combining persons with self-reported poor vision and legal blindness, a second composed of persons with fair vision, and a third composed of persons with good or better vision, which served as the reference category.

Mediators

Summary indices of physical functioning were taken from the RAND HRS data set, and are based on factors compiled and validated by Wallace and Herzog (1995). The two indices represent mobility and large-muscle functioning. Specifically, the mobility index included variables assessing the participant’s self-reported difficulty walking one block, walking several blocks, walking across a room, climbing one flight of stairs, and climbing several flights of stairs. The large-muscle index included items that assessed the participant’s self-reported difficulty sitting for 2 hr; getting up from a chair; stooping, kneeling, or crouching; and pushing or pulling large objects. Participants were coded in the positive direction (has functional limitation = 1) if they indicated any difficulty, or being unable to perform each task. Each functional limitation added one to its respective summary measure, such that mobility could range from 0 to 5 and the large-muscle index could range from 0 to 4.

Dependent variable

The main dependent variable of this study was an indicator of whether or not participants reported a fall in the 2 years prior to the 2008 wave of HRS. Participants were asked whether they had fallen in the previous 2 years (since the 2006 wave of the study). Responses to this item were recoded and analyzed as a binary indicator variable (fall = 1).

Statistical Analyses

Statistical Analysis Software (SAS) version 9.3 was used to calculate basic descriptive statistics, regression models, and the Sobel statistic. Means, standard deviations, and percentages were computed for all sociodemographic characteristics and health and functioning variables, including age, race, education, aggregate chronic conditions, ADL and IADL difficulty, self-reported vision status, the two indices of functioning, and the main outcome variable (whether participants reported falling in the previous 2 years). Univariate analyses of variance (ANOVA) and Pearson chi square (χ²) tests were conducted to evaluate differences between vision status groups by these variables. Eta square (η²) and the phi coefficient (Φ)—measures of effect size for ANOVA and Pearson χ², respectively—were calculated to determine the practical significance of statistical differences by vision status groups. According to Green, Salkind, and Akey (2000), values of .01, .06, and .14 for η², and .10, .30, and .50 for Φ, represent small, medium, and large effect sizes, each, respectively. When appropriate, follow-up tests were conducted to evaluate pairwise differences in means and proportions by vision status. In each ANOVA, variances among the three vision status groups were heterogeneous; therefore, Dunnett’s C assessment of comparisons, a test that does not assume equal variance among groups, was used in post hoc tests. Pairwise differences between vision status groups on non-continuous variables were tested using Pearson χ² separately on each individual pair.

Recall that identifying statistically significant correlations between vision and prospective mediators is necessary as a preliminary step in establishing a mediating relationship between these variables with respect to falls (Baron & Kenny, 1986). Thus, we computed partial correlation coefficients of key explanatory variables and covariates, holding constant sociodemographic variables. A p value of less than .006 (.05/8) was required for significance to control for type 1 error across eight partial correlations. All coefficients were in the expected direction, and statistically significant at p < .0005 (not shown in tables), indicating that there was adequate justification to test for the possible role of mediators.

Thus, in our main analyses, we conducted a series of regressions designed to align with analytical steps outlined by Baron and Kenny (1986) to show mediation between vision and falls by way of impaired functioning. Models were designed to test whether the proposed mediators (mobility and large-muscle functioning in 2006) exercised their influence on two levels of the key independent variable (poor and fair self-reported vision status in 2006—good or better vision, which served as the reference category was withheld from regression models) with respect to its effect on the dependent variable (falls in 2008). Negative binomial regression was used to test the relationship between vision and mediating interval variables, whereas logistic regression was used to test the effects of vision and the mediators on the key binary outcome variable, falls. Because analyses were based on different scales, regression coefficients were standardized using formulas described by MacKinnon and Dwyer (1993). These comparable coefficients were then used to compute Sobel’s test statistic, which compares the strength of the indirect effect to the null hypothesis that the effect equals zero (Preacher & Hayes, 2004). Sobel is a standardized normal test of significance of the difference between a total and direct effect; therefore, values above 1.96 are statistically significant at α = .05. As a rough measure of effect size, we calculated the proportion of change in the total effect due to the mediator (Vittinghoff, Sen, & McCulloch, 2009).

Figure 2 displays theoretical relationships between key variables and covariates included in each of four models. According to Baron and Kenny (1986), a total effect (depicted by line c) is made up of a direct effect of vision on falls (line c’), plus the total indirect effect of vision status, via the indices of physical functioning (the products of a and b). We calculated indirect effects separately for mobility and large-muscle functioning. Statistically significant differences between direct and indirect effects (c − c’) represent the total mediating effect. In addition to the functional indices, four sequential models controlled covariates that have theoretical significance with respect to Nagi’s (1965, 1976) disability model—namely, sociodemographic risk factors, number of diagnosed chronic health conditions, functional limitations, and disability in daily self-care activities (ADLs and IADLs). Model 1 included just sociodemographic variables; Model 2 added chronic conditions; Model 3 included sociodemographics, chronic conditions, and functioning; and Model 4 included sociodemo-graphics, chronic conditions, functioning, and ADL and IADL disability. Each model was computed for the sample overall, and separately for men and women. The theoretical basis for conducting separate analyses comes from literature showing important differences between men and women, with respect to the types of falls experienced (O’Neill et al., 1994), incidence of falling (Bath & Morgan, 1999), and differences in the muscle and bone structures of women and men, and the effects of age on these structures (Frontera, Hughes, Lutz, & Evans, 1991; Hannan et al., 2000).

Figure 2.

Theoretical relationships between self-reported vision status, functional mediators, and falls.

Results

Descriptive statistics and results from univariate ANOVA and χ² tests comparing sociodemographic variables, number of diagnosed chronic diseases, ADLs and IADLs, and falls by vision status groups are displayed in Table 1. In addition, results of follow-up tests, which evaluated pair-wise differences between vision-status groups are also displayed. Overall, the average age of participants was 73.7 years (SD = 6.8). Approximately 58% of the total sample was composed of women, and the vast majority of the sample (roughly 90%) was White; 7% were Black, and about 3% reported some other race. The average years of education were 12.4 (SD = 3.1) years. Compared with persons with good vision, persons with poor vision were statistically older, and greater proportions were women, Black, and had achieved less education; however, calculated effect size measures suggest that these results were based on only small effect sizes.

Table 1.

Descriptive Sociodemographic Characteristics, Number of Diagnosed Chronic Diseases, ADL and IADL Disability, and Falls by Self-Reported Vision Status; Tests of Significance and Pair-Wise Comparisons Between Vision Groups (HRS, 2006 and 2008, Weighted).

	All				Test of significant differences between vision groups
Variables	$\bar{X}$	SD	Range	N	Test	df		p	Effect size
Age	73.7	6.8	65-104	11,080	F = 125.3	2, 11,077		<.0005	η² = .022
Female	58%		0-1	10,785	χ² = 11.2	2		.004	Φ = .03
White	90%		0-1	10,786	χ² = 84.6	2		<.0005	Φ = .09
Black	7%		0-1	10,787	χ² = 32.9	2		<.0005	Φ = .09
Other	3%		0-1	10,787	χ² = 8.3	2		.016	Φ = .03
Education	12.4	3.1	0-17	11,054	F = 280.7	2, 11,051		<.0005	η² = .048
Diseases	2.4	1.4	0-8	11,078	F = 222.4	2, 11,075		<.0005	η² = .039
ADL	0.3	0.7	0-5	11,078	F = 294.7	2, 11,075		<.0005	η² = .051
IADL	0.2	0.6	0-5	11,078	F = 646.1	2, 10,075		<.0005	η² = .104
Fall	38%		0-1	10,762	χ² = 91.7	2		<.0005	Φ = .09
	Good vision				Fair vision				Poor vision
Variables	$\bar{X}$	SD	Range	N	$\bar{X}$	SD	Range	n	$\bar{X}$	SD	Range	n
Age	73.3_a	6.6	65-104	8,747	74.5_b	71	65-98	1,711	77.5_c	7.9	65-97	622
Female	58%_a		0-1	8,537	59%_a		0-1	1,652	65%_b	0.5	0-1	596
White	91%_a		0-1	8,537	85%_b		0-1	1,652	83%_b		0-1	597
Black	6%_a		0-1	8,538	11%_b		0-1	1,652	13%_b		0-1	597
Other	3%		0-1	8,538	4%		0-1	1,652	4%		0-1	597
Education	12.8_a	2.9	0-17	8,728	11.3_b	3.5	0-17	1,705	10.6_c	3.7	0-17	621
Diseases	2.3_a	1.3	0-8	8,745	2.7_b	1.5	0-8	1,711	3.3_c	1.7	0-7	622
ADL	0.2_a	0.6	0-5	8,745	0.4_b	0.9	0-5	1,711	0.9_c	1.3	0-5	622
IADL	0.1_a	0.5	0-5	8,745	0.3_b	0.7	0-5	1,711	1.0_c	1.4	0-5	622
Fall	36%_a		0-1	8,518	43%_b		0-1	1,649	53%_c		0-1	595

Note. Subscripts represent pair-wise differences between vision categories, Significant at <.05. ADL = activities of daily living; IADL = instrumental activities of daily living; HRS = Health and Retirement Study.

On average, respondents reported 2.4 (SD = 1.4) diagnosed chronic diseases; persons with poor vision reported statistically more chronic health conditions $\bar{X}$ = 3.3; SD = 1.7) than those with fair vision $\bar{X}$ = 2.7; SD = 1.5) and good vision $\bar{X}$ = 2.3; SD = 1.3), F (2, 11,078) = 222.4, p < .0005. Calculated η² indicated that around 4% of the variance in health conditions—a small effect—was associated with vision status. With respect to disability indices, on average, all participants reported difficulty with 0.3 (SD = 0.7) ADLs and 0.2 (SD = 0.6) IADLs. Respondents with poor vision reported difficulties with 0.9 (SD = 1.3) ADLs, compared with 0.4 (SD = 0.9) for those with fair vision and 0.2 (SD = 0.6) for those with good vision, F (2, 11,075) = 294.7, p < .0005; η² = .051. Similarly, participants with poor vision reported statistically more difficulty with IADLs, F (2, 11,075) = 646.1, p < .0005; η² = .104; on average, persons with poor vision experienced difficulty with 1.0 (SD = 1.4) IADLs, compared with 0.3 (SD = 0.7) for those with fair vision and 0.1 (SD = 0.5) for those with good vision.

With respect to the study’s dependent variable, whether participants had fallen between the 2006 and 2008 waves, older persons with poor vision reported statistically more falls than persons with fair or good vision (Pearson χ², [2, N = 10,762] = 91.7). On average, more than half (53%) of those with poor vision reported a fall, compared with 43% with fair vision, and 36% of those with good vision, though this finding was based on a very small effect size (Φ = .09).

Table 2 displays descriptive statistics for indices of functioning by vision status, as well as results of univariate ANOVAs conducted to test overall significance and pair-wise differences between vision status groups. Overall, participants reported 1.2 (SD = 1.5) mobility-related difficulties. Vision status was statistically associated with mobility functioning, F (2, 11,077) = 313.8, p < .0005; η² = .054. On average, persons with poor vision experienced difficulty with 2.2 (SD = 1.7) mobility tasks, compared with 1.6 (SD = 1.6) by those with fair vision and 1.0 (SD = 1.4) by those with good vision. With respect to large-muscle limitations, overall, participants reported difficulties with 1.4 (SD = 1.3) large-muscle-related activities. Respondents with relatively worse vision status reported statistically more large-muscle difficulties, F (2, 11,074) = 215.9, p < .0005; η² = .038. On average, persons with poor vision experienced difficulty with 2.1 (SD = 1.3) large-muscle tasks, compared with 1.7 (SD = 1.4) for those with fair vision and 1.2 (SD = 1.2) for those with good vision.

Table 2.

Descriptive Statistics for Indices of Functioning by Self-Reported Vision Status; Test of Significance and Pair-Wise Comparisons Between Vision Groups (HRS, 2006, Weighted).

	All				Test of significant differences between vision groups
Variables	$\bar{X}$	SD	Range	N	Test	df		p	Effect size
Mobility	1.2	1.5	0-5	11,080	F = 313.8	2, 11,077		<.0005	η² = .054
Large muscle	1.4	1.3	0-4	11,077	F = 215.9	2, 11,074		<.0005	η² = .038
	Good vision				Fair vision				Poor vision
Variables	$\bar{X}$	SD	Range	n	$\bar{X}$	SD	Range	n	$\bar{X}$	SD	Range	n
Mobility	1.0_a	1.4	0-5	8,747	1.6_b	1.6	0-5	1,711	2.2_c	1.7	0-5	622
Large muscle	1.2_a	1.2	0-4	8,745	1.7_b	1.4	0-4	1,710	2.1_c	1.3	0-4	622

Note. Subscripts represent pair-wise differences between vision categories, Significant at <.05. HRS = Health and Retirement Study.

Finally, Tables 3 and 4 report results of regressions calculated to test mediating effects of mobility and large-muscle functional impairments on self-reported vision as it effects falls. Coefficients are reported for total and direct effects of vision status on falls, along with the percentage effect size, and Sobel test of significance of the mediating effects. These tables show results for the entire sample, as well as for men and women separately, by vision status. The Sobel test is used to determine whether a reduction in the effect of an independent variable, after a mediator is included in the model, is statistically different from zero and therefore whether the mediation effect is statistically significant. Specifically, we used the Sobel statistic to determine whether functioning domains significantly mediated the influence of vision status on falls, when covariates chosen to represent Nagi’s (1965, 1976) dimensions of disability were sequentially added and held constant in four models (see Figure 2). Thus, the most comprehensive models depicted in Tables 3 and 4 included sociodemographics, diagnosed chronic conditions, functioning, and ADL and IADL disability (Model 4).

Table 3.

Mediating Effect of Mobility Function on the Association Between Vision Status and Falls.

	Mobility function
	Total effect	Direct effect	% Mediation	Sobel	P
All
Poor vision
Model 1	0.5216	0.2926	44	2.31	*
Model 2	0.3426	0.2675	22	2.01	*
Model 3	0.2675	0.2050	23	1.53	n.s.
Model 4	0.1543	0.1411	9	0.81	n.s.
Fair vision
Model 1	0.2610	0.1455	44	2.93	**
Model 2	0.1877	0.1181	37	2.15	*
Model 3	0.1316	0.1041	21	1.87	^†
Model 4	0.1202	0.1016	15	1.28	n.s.
Male
Poor vision
Model 1	0.7421	0.5074	32	2.02	*
Model 2	0.5960	0.4519	24	1.61	^†
Model 3	0.4935	0.4317	13	1.17	n.s.
Model 4	0.3856	0.3748	3	0.11	n.s.
Fair vision
Model 1	0.3259	0.2075	36	2.78	*
Model 2	0.2442	0.1724	29	2.34	*
Model 3	0.1635	0.1439	12	1.45	n.s.
Model 4	0.1545	0.1400	9	1.23	n.s.
Female
Poor vision
Model 1	0.4041	0.1870	54	1.25	n.s.
Model 2	0.2118	0.1035	51	0.96	n.s.
Model 3	0.1574	0.0966	39	0.84	n.s.
Model 4	0.0397	0.0262	34	0.63	n.s.
Fair vision
Model 1	0.2154	0.1011	53	1.54	n.s.
Model 2	0.1495	0.0804	46	0.85	n.s.
Model 3	0.1061	0.0721	32	0.77	n.s.
Model 4	0.0917	0.0693	24	0.2	n.s.

Note. Model 1 includes sociodemographic variables; Model 2 includes sociodemographics and diagnosed chronic conditions; Model 3 includes sociodemographics, diagnosed chronic conditions, and functioning; Model 4 includes sociodemographics, diagnosed chronic conditions, functioning, and ADL/IADL disability. n.s. not significant. ADL = activities of daily living; IADL = instrumental activities of daily living.

†

p < .1. *p ≤ .05. **p < .005. ***p < .0005.

Table 4.

Mediating Effect of Large-Muscle Function on the Relationship Between Vision Status and Falls.

	Large-muscle function
	Total effect	Direct effect	% Mediation	Sobel	P
All
Poor vision
Model 1	0.5216	0.3671	30	2.61	**
Model 2	0.3426	0.2675	22	1.95	^†
Model 3	0.2165	0.2050	5	0.25	n.s.
Model 4	0.1444	0.1411	2	0.05	n.s.
Fair vision
Model 1	0.2610	0.1658	36	3.62	***
Model 2	0.1877	0.1316	30	2.62	**
Model 3	0.1181	0.1041	12	1.79	^†
Model 4	0.1141	0.1016	11	1.27	n.s.
Male
Poor vision
Model 1	0.7421	0.5670	24	1.94	^†
Model 2	0.5960	0.4935	17	1.52	n.s.
Model 3	0.4519	0.4317	4	0.34	n.s.
Model 4	0.3821	0.3748	2	−0.004	n.s.
Fair vision
Model 1	0.3259	0.2004	39	3.16	**
Model 2	0.2442	0.1635	33	2.70	**
Model 3	0.1724	0.1439	17	2.01	*
Model 4	0.1670	0.1400	16	1.78	^†
Female
Poor vision
Model 1	0.4041	0.2706	33	1.68	^†
Model 2	0.2118	0.1574	26	1.20	n.s.
Model 3	0.1035	0.0966	7	0.12	n.s.
Model 4	0.0268	0.0262	2	−0.06	n.s.
Fair vision
Model 1	0.2154	0.1367	37	2.19	*
Model 2	0.1495	0.1061	29	1.27	n.s.
Model 3	0.0804	0.0721	10	0.70	n.s.
Model 4	0.0765	0.0693	9	0.19	n.s

†

p < .1. *p ≤ .05. **p < .005. ***p < .0005.

Results suggested that in most cases, both mobility and large-muscle functional measures were significant mediators of self-reported vision status in relation to falls, when just sociodemographic characteristics were controlled (Model 1). The sole exception to this finding was in regard to mobility difficulties experienced by women with poor and fair self-reported vision. Mobility limitations were not significant mediators in any of the models tested for this group. Nevertheless, in general, statistically significant mediating effects of mobility and large-muscle functioning were reduced or eliminated completely as the models became more complex. Among the entire sample, limitations of both mobility and large-muscle functioning remained significant mediators of fair vision status until ADLs and IADLs were added in the final model (Model 4). Only large-muscle functioning by men with fair vision status continued to trend toward significance when all covariates were included in the model (Sobel = 1.78; p < .1).

Discussion

In previous studies examining the relationship between vision status and falls, researchers have, by and large, assumed that fall risk associated with vision status is mainly a direct effect—for example, when older persons with vision impairment trip and fall over unseen obstacles in their path. By comparison, this study acknowledged the potential for vision status to operate through other systems to result in increased fall risk. Mediating effects of functional limitations, such as mobility and large-muscle limitations, are important to consider because of the increased likelihood of experiencing functional difficulties concomitantly with age-related vision impairment, as well as other health problems that are more likely to occur after around age 65. In addition, fully understanding how vision-related falls may be contingent on losses in limb functioning may inform fall prevention programs that address multiple fall-risk factors in their interventions (Steinman, Nguyen, Pynoos, & Leland, 2011).

In this study, we hypothesized that part or all of the influence of self-reported vision status on falls would be contingent on mobility and large-muscle limitations. The rationale behind our a priori expectations was that individuals who experienced losses in limb functioning would be less able to correct or stabilize themselves following a vision-related misstep, trip, or stumble. Thus, falls that are instigated by direct vision effects could be mitigated or avoided altogether if other systems were fit and intact. This premise is supported by the large body of research showing the beneficial effects of bone and muscle toning and maintenance via exercise as an efficacious fall prevention strategy (Rose & Hernandez, 2010).

Nevertheless, results of this study indicated that mediation effects of functional measures—at least of the indices we assessed—were somewhat limited, and differed according to whether vision status was fair or poor, and by gender. With respect to the total sample, mediation effects by functional limitations were most apparent among participants who reported fair vision status. Self-reported measures of vision status are, by definition, a subjective account of the respondent’s ability to see; however, our results demonstrated a robust ordinal relationship between vision status categories with respect to diagnosed chronic disease, functional limitations, and disability dimensions described by Nagi (1965, 1976). Based on these results, we assume that on average, older persons with good or better vision or fair vision status are generally in better health and more active than individuals who report poor vision status, and that these differences may result in differential fall risk associated with some functional limitations. Severe vision impairments may be somewhat protective, with respect to falls, in that they reduce the performance of functional activities that provide quantitatively more opportunities to fall. Conversely, individuals who may have experienced some functional decline, but are relatively less limited by their vision status (i.e., those with fair vision), may have greater risk of falling while performing functional activities as part of their daily routines.

Thus, one possible explanation for the mediating effects of functional limitations among older people (particularly men) who reported fair vision status is that they are more active than those with poor vision status, and therefore are afforded more opportunities to fall. This perspective is supported by results of the progressive sequence of models that we analyzed. In most cases, a mediation effect was completely eliminated in the fourth model, which included all covariates, including ADL and IADL capacity (see Tables 3 and 4). For the entire sample, the mediation effects of both mobility and large-muscle functioning were significant until ADL and IADLs were included. This suggests that measures of disability, rather than functional limitation, may be significant mediating factors. In addition, our results support Nagi’s (1965, 1976) contention that these dimensions represent quantitatively different constructs. The disability construct was conceptualized based on the capacity of respondents to carry out essential personal and household activities that often depend on adequate functional capacity, but that are unique and specific in nature. In future research studies, assessment of daily living activities as well as participation in community and leisure activities could provide a productive path for exploration, in regard to their integrated influence on falls among older individuals who have vision impairments.

With respect to differences by gender, no statistically significant mediation effects were found for women when models included all covariates, suggesting that no or only a limited portion of the effects of self-reported vision status on falls is contingent on limitations in mobility and large-muscle functioning. By contrast, only large-muscle functioning was a statistically significant mediator for men when all covariates were included in the model. This finding points to the importance of maintaining muscle strength and functioning, especially for men with vision impairments, as a falls prevention strategy. Previous studies have indicated that reduced limb functioning has a direct influence on fall risk for both men and women (Steinman, 2008; Steinman et al., 2009). In addition, results of this study suggest that declines in the large-muscle groups, including the chest, legs, and back may mediate the effects of vision status on falls. That is to say, a small portion of the effects of self-reported vision status as a predictor of falls is through declines in large muscles. Older men who experience vision loss may be less likely to avoid a full fall (e.g., after tripping over an unseen object) if their muscles are not fit, and they are unable to regain stability. In this scenario, decrements in vision status may cause the primary event, which sets off a series of responses within multiple other systems that are designed to prevent a fall or protect from injury due to a fall—including positioning one’s arms to regain balance, or widening the base of one’s stance, to increase stability. However, because of declines in a secondary system, in this case, reduced strength in arms and leg muscles, a fall could occur, which might otherwise have been avoided if strength had been maintained.

Study Limitations and Conclusion

The results of this study reflect important relationships between self-reported vision status, functional limitations, falls, and other risk factors that could lead to falls. Nevertheless, there are limitations to our study, which are worthy of acknowledgment. First is the duration of time for which respondents were expected to report fall episodes. Survey respondents of HRS were asked if they had experienced a fall in the previous 2 years. This lengthy period is troublesome because of difficulties some participants may have had in accurately recalling falls that occurred during that time—especially falls that did not result in an injury, which may be easier to remember (Ganz, Higashi, & Rubenstein, 2005). Although some selection criteria used for this study may have biased the sample in the favor of a relatively healthier, cognitively intact, and more robust group (e.g., no proxy respondents; no mortality during the study period), reports of falls in this study could potentially represent an underestimate of the true number of falls experienced by the older population at large.

A related limitation pertains to the operationalization of the study’s independent and dependent variables. Self-reported measures convey the perceptions of respondents regarding their sensory and functional circumstances, and likely reflect how well participants are able to use their residual capacities; however, the subjective nature of key variables could call into question the reliability of our findings. It is possible that a similar study that used observable clinical measures of vision and musculoskeletal functioning would report stronger or weaker relationships relative to the results reported here. Similarly, the subjective nature of the falls measure raises some concern. One commonly accepted definition set forth by the International Classification of Diseases (ICD-10) defines a fall as “an unexpected event where a person falls to the ground from an upper level or the same level” (Masud & Morris, 2001). Nevertheless, no definition was provided to HRS participants, so it is uncertain whether participants interpreted the item in a consistent manner.

Finally, we acknowledge the existence of additional variables that could influence the meditational relationships explored in this study, but that were not assessed. For instance, modifiable contextual factors, including environmental considerations (e.g., home modifications, assistive technology), are likely to improve functional capacity by increasing access and/or removing physical/behavioral barriers associated with functional limitations and sensory impairments (Lawton & Nahemow, 1973; Pynoos, Steinman, Nguyen, & Bressette, 2012). Indeed, work by Allen, Foster, and Berg (2001) demonstrated that older people with disabilities who used mobility equipment (such as canes and crutches) and home modifications needed less human assistance and were more autonomous in their daily activities than those who did not use assistive technology. Our covariates were chosen based on a theoretical model that did not emphasize similar contextual factors. Therefore, future studies could benefit by using a more comprehensive theoretical framework. One such framework is the World Health Organization’s (WHO) International Classification of Functioning (ICF), which includes dimensions related to environmental context, personal attributes, and community participation, in addition to comparable health dimensions assessed in this study (WHO, 2001).

Despite these limitations, implications of this study point to the continued need for programs and services designed to maintain functioning and reduce disability as a means to reduce falls among older men and women who acquire vision impairments in later life. This important goal can be achieved through interventions that focus on improving confidence and strength of older individuals after they experience vision loss, via participation in exercise programs, environmental assessment and modification, medical risk assessment, and education about falls risk (Steinman et al., 2011). In addition, broader dissemination of current programs that train older adults with vision impairments to move safely through their environments, and exercise programs that take into consideration the special needs of older people with vision impairments, may help reduce the number of falls and serious outcomes such as hip fractures and/or mortality related to falls among this high-risk segment of the older population. Finally, more research that explores integrated relationships between multiple known fall-risk factors is merited and needed to better understand the complexities of fall risk, and to inform new cutting-edge research-based interventions designed to improve the well-being and safety of older people.

Footnotes

Acknowledgements

The authors extend their appreciation to Susan Enguidanos, James Gauderman, and Merril Silverstein for their comments on earlier drafts of this manuscript.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Work on this manuscript was completed with support from training grants provided by the National Institute on Aging (5T32AG0037) and the Agency for Healthcare Research and Quality (5T32HS000011), as well as the Archstone Foundation.

References

Allen

S. M.

Foster

Berg

(2001). Receiving help at home. The Journals of Gerontology, Series B: Psychological Sciences & Social Sciences, 56, S374-S382.

Baron

R. M.

Kenny

D. A.

(1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173-1182.

Bath

P. A.

Morgan

(1999). Differential risk factor profiles for indoor and outdoor falls in older people living at home in Nottingham, UK. European Journal of Epidemiology, 15, 65-73.

Crimmins

E. M.

(2004). Trends in the health of the elderly. Annual Review of Public Health, 25, 79-98.

De Boer

M. R.

Pluijm

S. M. F.

Lips

Moll

A. C.

Volker-Dieben

H. J.

Deeg

D. J. H.

van Rens

G. H. M. B.

(2004). Different aspects of visual impairment as risk factors for falls and fractures in older men and women. Journal of Bone and Mineral Research, 19, 1539-1547.

Frontera

W. R.

Hughes

V. A.

Lutz

K. J.

Evans

W. J.

(1991). A cross-sectional study of muscle strength and mass in 45- to 78-yr-old men and women. Journal of Applied Physiology, 71, 644-650.

Ganz

D. A.

Higashi

Rubenstein

L. Z.

(2005). Monitoring falls in cohort studies of community-dwelling older people: Effect of the recall interval. Journal of the American Geriatrics Society, 53, 2190-2194.

Green

S. B.

Salkind

N. J.

Akey

T. M.

(2000). Using SPSS for windows: Analyzing and understanding data (2nd ed.). Upper Saddle River, NJ: Prentice Hall.

Hannan

M. T.

Felson

D. T.

Dawson-Hughes

Tucker

K. L.

Cupples

L. A.

Wilson

P. W. F.

Kiel

D. P.

(2000). Risk factors for longitudinal bone loss in elderly men and women: The Framingham Osteoporosis Study. Journal of Bone and Mineral Research, 15, 710-720.

10.

Ivers

R. Q.

Cumming

R. G.

Mitchell

Attebo

(1998). Visual impairment and falls in older adults: The Blue Mountains Eye Study. Journal of the American Geriatrics Society, 46, 58-64.

11.

Lawton

M. P.

Nahemow

(1973). Ecology and the aging process. In Eisdorfer

Lawton

M. P.

(Eds.), The psychology of adult development and aging (pp. 619 – 674). Washington, DC: American Psychological Association.

12.

Lord

S. R.

Dayhew

(2001). Visual risk factors for falls in older people. Journal of the American Geriatrics Society, 49, 508-515.

13.

Lord

S. R.

Smith

S. T.

Menant

J. C.

(2010). Vision and falls in older people: Risk factors and intervention strategies. Clinics in Geriatric Medicine, 2, 6, 569-581.

14.

Mackinnon

D. P.

Dwyer

J. H.

(1993). Estimating mediated effects in prevention studies. Evaluation Review, 17, 144-158.

15.

Masud

Morris

R. O.

(2001). Epidemiology of falls. Age and Ageing, 30(Suppl. 4), 3-7.

16.

Nagi

S. Z.

(1965). Some conceptual issues in disability and rehabilitation. In Sussman

M. B.

(Ed.), Sociology and rehabilitation (pp. 100-113). Washington, DC: American Sociological Association.

17.

Nagi

S. Z.

(1976). An epidemiology of disability among adults in the United States. The Milbank Memorial Fund Quarterly: Health and Society, 54, 439-467.

18.

O’Neill

T. W.

Varlow

Silman

A. J.

Reeve

Reid

D. M.

Todd

Woolf

A. D.

(1994). Age and sex influences on fall characteristics. Annals of the Rheumatic Diseases, 53, 773-775.

19.

Preacher

K. J.

Hayes

A. F.

(2004). SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behavior Research Methods, 36, 717-731.

20.

Pynoos

Steinman

B. A.

Nguyen

A. Q. D.

Bressette

(2012). Assessing and adapting the home environment to reduce falls and meet the changing capacity of older adults. Journal of Housing for the Elderly, 26, 136-154.

21.

RAND. (2010, March). RAND HRS Data. Version I. Produced by the Rand Center for the Study of Aging, with funding from the National Institute on Aging and the Social Security Administration. Santa Monica, CA.

22.

Ray

C. T.

Wolf

S. L.

(2008). Review of intrinsic factors related to fall risk in individuals with visual impairments. Journal of Rehabilitation Research & Development, 45, 1115-1124.

23.

Rose

D. J.

Hernandez

(2010). The role of exercise in fall prevention for older adults. Clinics in Geriatric Medicine, 26, 607-631.

24.

Steinman

B. A.

(2008). Self-reported vision, upper/lower limb disability, and fall risk in older adults. Journal of Applied Gerontology, 27, 406-423.

25.

Steinman

B. A.

Nguyen

A. Q. D.

Pynoos

Leland

(2011). Falls-prevention interventions for older adults who are blind or visually impaired. InSight, 4, 83-91.

26.

Steinman

B. A.

Pynoos

Nguyen

A. Q. D.

(2009). Fall risk in older adults: The role of self-rated vision, home modification, and limb function. Journal of Aging and Health, 21, 655-676.

27.

University of Michigan, HRS. (2010). The Health and Retirement Study: A longitudinal study of health, retirement, and aging sponsored by the National Institute on Aging. Retrieved from http://hrsonline.isr.umich.edu/

28.

Verbrugge

L. M.

Jette

(1994). The disablement process. Social Science & Medicine, 39, 1-14.

29.

Vittinghoff

Sen

McCulloch

C. E.

(2009). Sample size calculations for evaluating mediation. Statistics in Medicine, 28, 541-557.

30.

Wallace

R. B.

Herzog

A. R.

(1995). Overview of the health measures in the Health and Retirement Study. Journal of Human Resources, 30, S84-S107.

31.

World Health Organization. (2001). International classification of functioning, disability and health. Geneva, Switzerland: Author.