Functional ability scores correspond to the marginal predicted proportions from multivariable Poisson regression models. The marginal predicted proportions were divided by the number of survey items making up each of the 3 outcomes to make the scale equivalent across outcomes.
eTable 1. Poisson model of mobility activity scale as a function of dementia and self-reported vision status.
eTable 2. Poisson model of self-care activity scale as a function of dementia and self-reported vision status.
eTable 3. Poisson model of household activity scale as a function of dementia and self-reported vision status.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Patel N, Stagg BC, Swenor BK, Zhou Y, Talwar N, Ehrlich JR. Association of Co-occurring Dementia and Self-reported Visual Impairment With Activity Limitations in Older Adults. JAMA Ophthalmol. 2020;138(7):756–763. doi:10.1001/jamaophthalmol.2020.1562
What is the association of co-occurring dementia and self-reported visual impairment on activity limitations in older US adults?
In this nationally representative study, participants with dementia and self-reported visual impairment reported greater limitations in mobility, self-care, and household activities than would be expected based on the individual associations of each of these health conditions.
In this analysis, activity limitations were more severe in older adults with coexisting dementia and self-reported visual impairment; if causal associations are established, interventions may be warranted across the activity spectrum to address the growing burden of visual impairment and dementia in an aging population
The number of older adults with visual impairment (VI) and dementia is projected to increase in the US because of the aging of the population. Dementia and VI commonly co-occur and are each independently associated with disability. To care for an aging population, it may be important to characterize the association of coexisting dementia and self-reported VI on daily functioning.
To evaluate the association of co-occurring dementia and self-reported VI on daily functioning.
Design, Setting, and Participants
This cross-sectional analysis of a nationally representative cohort study used data from the National Health and Aging Trends Study (NHATS), an annual study of US adults 65 years and older. Participants in the 2015 survey with complete data on outcomes, associated factors, and covariates were included in this study. Data analysis took place from January 2019 to November 2019.
Main Outcomes and Measures
Multivariable Poisson regression was used to model the independent associations and interaction of dementia and self-reported VI status on 3 functional activity scales (self-care, mobility, and household activities). Marginal predicted proportions were calculated. Analyses were adjusted for sociodemographic and medical factors and accounted for the complex survey design.
A total of 7124 participants were included. The weighted proportion of female respondents was 55.3% (95% CI, 54.0%-56.6%), and 56.1% (95% CI, 54.1%-58.1%) were between 65 and 74 years old. Self-reported VI was present in 8.6% (95% CI, 7.8%-9.3%) of participants, while 8.3% (95% CI, 7.8%-8.9%) had possible dementia and 6.3% (95% CI, 5.7%-6.9%) had probable dementia. Self-reported VI was associated with an expected decrease in mobility score of 14.7% (functional scale scores: no VI, 10.82 vs VI, 9.23), self-care score of 9.5% (no VI, 14.54 vs VI, 13.16), and household activity score of 15.2% (no VI, 18.23 vs VI, 15.45), while probable dementia was associated with expected decreases of 27.8% (no dementia, 10.82 vs probable dementia, 7.81), 22.9% (no dementia, 14.54 vs probable dementia, 11.20), and 34.7% (no dementia, 18.23 vs probable dementia, 11.90), respectively. Among those with probable dementia and self-reported VI, there was an expected decrease in mobility score of 50.1% (functional ability scores: no VI and no dementia, 10.82 vs VI and probable dementia, 5.40), self-care score of 42.4% (no VI and no dementia, 14.54 vs VI and probable dementia, 8.38), and household activity score of 52.4% (no VI and no dementia, 18.23 vs VI and probable dementia, 8.68), suggesting that respondents with co-occurring dementia and self-reported VI had lower functional activity scores than would be associated with the independent contributions of these conditions.
Conclusions and Relevance
Older adults with both dementia and self-reported VI may be at high risk for disability, and their co-occurrence may potentiate this risk. These findings suggest that the growing population of older adults with both visual impairment and dementia may benefit from interventions to maximize vision and cognition and promote functioning and independence.
Visual impairment (VI) is common in older adults and estimated to affect approximately 1 in 11 individuals 65 years and older in the US.1 Because of growth and aging of the population in the US and globally, the number affected by VI and blindness is projected to more than double by 2050.2,3 Visual impairment in later life is associated with an increased risk of falls, social isolation, loss of independence, disability, and dementia.4-8 In fact, VI and dementia co-occur in about 4% of older US adults.4 Individuals who experience both poor vision and dementia in late life may be at greater risk for disability than those with only one of these conditions.9-11
There is evidence that both VI12-15 and dementia16-19 are risk factors for declining mobility, activities of daily living (ADLs), and instrumental activities of daily living (IADLs). However, little is known about the association of co-occurring VI and dementia on disability in older adults. A study from North Carolina reported that the odds of mobility, ADL disability, and IADL disability were greatest among those with both self-reported VI and cognitive impairment, but that there was no statistical interaction between vision and cognitive status in prognosticating disability.9 A study of older adults in Singapore largely corroborated this result, failing to detect a statistical interaction in assessing likely mobility or IADL disability while finding a negative interaction in a model assessing likely ADL disability.10 The absence of a positive interaction in these studies suggested that the co-occurrence of VI and dementia was not associated with increases in disability beyond the independent contributions of these 2 conditions. However, these studies both used a binary definition of disability based on the inability to perform 1 or more tasks independently. This approach may underestimate statistical interaction because of floor and ceiling effects. Additionally, these studies did not account for the full range of disablement concepts, such as behavioral and environmental adaptations.20
To our knowledge, the association of combined dementia and self-reported VI on activity limitations has not been studied in a nationally representative cohort of older US adults. The National Health and Aging Trends Study (NHATS) provides an opportunity to study vision, dementia, and disability in a contemporary, nationally representative cohort of older US adults. The detailed assessments of disability in the NHATS allow for a nuanced analysis of the functional outcomes of vision and dementia on mobility, self-care (corresponding to ADLs), and household activities (corresponding to IADLs), while accounting for behavioral adaptations to changes in capacity.
The current study tested the hypothesis that older US adults with combined dementia and self-reported VI have a greater degree of activity limitations than would be expected based on the independent association of each of these 2 conditions. Findings from this study are important for designing informed interventions to promote independence and optimal aging for older adults.
The NHATS is a nationally representative survey that is administered annually to a cohort of US Medicare beneficiaries 65 years and older and has provided important insights on vision and aging.5-7,21 The baseline NHATS sample was recruited in 2011, and the sample was replenished in 2015. At the time of recruitment, all participants were living in settings other than nursing homes. Data in the NHATS were collected by trained examiners who visited the homes of study participants. All NHATS data used in this study, including self-reported and performance measures, were part of the standard NHATS data collection protocol.
This study used data from the replenished 2015 NHATS sample. The University of Michigan institutional review board deemed this study exempt since the NHATS consists of publicly available, nonidentifiable data. Informed consent for NHATS participation was obtained by the NHATS investigators.
The NHATS measures functional ability for mobility, self-care, and household activities. The following mobility activities were assessed: going outside, getting around inside, and getting out of bed. Self-care activities included eating, dressing, toileting, and bathing. Household activities consisted of doing laundry, shopping for groceries, making hot meals, paying bills and banking, and keeping track of medications. For each activity, respondents reported whether they used any assistive devices or environmental modifications to perform the activity in the past month. Those who ever performed the activity without help rated their difficulty performing the activity alone (without devices or environmental modifications). For all activities except getting out of bed, toileting, and eating, they noted if they did the activity less frequently compared with a year ago.
A validated 4-level scale22 was used to classify performance of each activity: (4) fully able: having no limitation in the ability to carry out activities (performing them without devices, reduced frequency, difficulty, or assistance); (3) successful accommodation: accommodating for limitations by using devices (for self-care and mobility activities) or performing activities less frequently but without difficulty or assistance; (2) difficulty: having difficulty performing activities alone even with accommodations, but receiving no assistance; and (1) assistance: receiving assistance from another person (because of health or functioning for household activities) or not doing the activity. For each type of activity (mobility, self-care, and household activities), a summary measure was calculated in which lower scores indicated greater limitations. The range of possible scores was from 3 to 12 for mobility, 4 to 16 for self-care, and 5 to 20 for the household domain.
Respondents were classified as having self-reported VI if they were blind or could not see across the street and/or read newspaper print, even with glasses. Dementia status was determined using the validated NHATS classification that included the following: probable dementia (a report of physician diagnosis of dementia or Alzheimer disease; a score indicating likely dementia on the AD8 Dementia Screening Interview of proxy respondents; or scores ≤1.5 SDs less than the mean on ≥2 cognitive performance tests of memory, orientation, and executive function), possible dementia (a score ≤1.5 SDs less than the mean on 1 cognitive performance test), or no dementia.23
Covariates were included that were conceptually associated with the independent variables and outcomes. Covariates included age in 5-year intervals; sex; race/ethnicity; highest education level; family income in quartiles; body mass index; history of cataract surgery; depressive symptoms; number of self-reported, physician-diagnosed medical comorbidities (arthritis, cancer, diabetes, heart attack, heart disease, hypertension, lung disease, osteoporosis, and stroke); and need for a proxy respondent. The presence of clinically significant depressive symptoms was determined based on a score of 3 or more on the Patient Health Questionnaire 2.24 Income quartiles were defined based on the income distribution of the sample, and multiple imputation values provided by NHATS were used to fill in missing income data.25
The weighted proportions of participants for each covariate and dementia classification stratified by self-reported VI status were calculated. Unadjusted P values from Pearson χ2 and t tests are reported, and P values less than .05 were considered significant.
Unweighted models were examined for each outcome, and the χ2 goodness-of-fit test demonstrated that the data fit a Poisson distribution. Generalized linear models with a Poisson distribution and log-link function were used to test the associations of dementia classification and self-reported VI (independent variables) with the mobility, self-care, and household activity functional ability hierarchies (outcomes), adjusted for all aforementioned covariates. An interaction term between dementia classification and self-reported VI was then entered into the models. After running weighted models, the distributions of residuals were reinspected to confirm good fit with the Poisson distribution. The marginal predicted proportions for each outcome were calculated based on the weighted distributions of independent variables and covariates in the study sample. To provide a graphical representation of results in which the scale was identical across the 3 outcomes, marginal predicted proportions scores were divided by the number of items making up each outcome. All models accounted for the complex NHATS survey design, including sample weights, units, and strata. Analyses were performed using SAS version 9.4 (SAS Institute).
Two sensitivity analyses were performed to test the robustness of model assumptions. First, the main analysis was rerun while excluding any data provided by proxy respondents. Second, the definition of self-reported VI was varied to include only those who self-reported blindness or difficulty seeing at distance.
There were 7124 participants included in this study. A total of 7576 respondents completed the study in 2015, and 452 were excluded because of missing data for 1 or more independent variable, outcome, or covariate.
The weighted proportion of female respondents was 55.3% (95% CI, 54.0%-56.6%), and 56.1% (95% CI, 54.1%-58.1%) were 65 to 74 years old. The overall weighted proportion of study participants with self-reported VI was 8.6% (95% CI, 7.8%-9.3%), while 8.3% (95% CI, 7.8%-8.9%) had possible dementia and 6.3% (95% CI, 5.7%-6.9%) had probable dementia. Table 1 presents summary characteristics of the study sample, stratified by self-reported VI status; self-reported VI was more common among those who had dementia, were older, female, nonwhite, less educated, had lower incomes, a history of cataract surgery, depressive symptoms, more medical comorbidities, or required a proxy respondent.
Table 2 presents the results of the main analyses, testing the interaction of dementia and self-reported VI status in assessing likely functional ability for mobility, self-care, and household activities. There was an interaction detected between possible and probable dementia and self-reported VI across all 3 outcomes. eTables 1, 2, and 3 in the Supplement present the results of full interaction models containing all covariates.
The marginal predicted proportions in Table 2 display the expected functional score for an individual given any combination of dementia and self-reported VI statuses while accounting for the distribution of covariates in the analytic sample. Those with self-reported VI had lower functional scores compared with those without self-reported VI for all 3 activity scales. In older adults without dementia, self-reported VI was associated with an expected decrease in mobility score of 14.7% (functional ability scores: no VI, 10.82 vs VI, 9.23), self-care score of 9.5% (no VI, 14.54 vs VI, 13.16), and household activity score of 15.2% (no VI, 18.23 vs VI, 15.45) compared with those without self-reported VI.
With respect to dementia categories, those with probable dementia had the lowest expected functional ability scores, followed by those with possible dementia, and those with no dementia. In older adults without self-reported VI, possible dementia was associated with an expected decrease in mobility score of 9.5% (functional ability scores: no dementia, 10.82 vs possible dementia, 9.79), self-care score of 6.0% (no dementia, 14.54 vs possible dementia, 13.67), and household activity score of 8.4% (no dementia, 18.23 vs possible dementia, 16.69), compared with those with no dementia. Probable dementia was associated with expected decreases of 27.8% (no dementia, 10.82 vs probable dementia, 7.81), 22.9% (no dementia, 14.54 vs probable dementia, 11.20), and 34.7% (no dementia, 18.23 vs probable dementia, 11.90) for mobility, self-care, and household activity scores, respectively, compared with those with no dementia.
Expected scores for respondents with both possible or probable dementia and self-reported VI were lower than would be calculated by the independent contributions of VI and dementia status, as shown in the Figure. Among those with both possible dementia and self-reported VI, compared with those with neither condition, there was an expected decrease in mobility score of 25.6% (functional ability scores: no VI and no dementia, 10.82 vs VI and possible dementia, 8.05), self-care score of 19.3% (no VI and no dementia, 14.54 vs VI and possible dementia, 11.73), and household activity score of 33.1% (no VI and no dementia, 18.23 vs VI and possible dementia, 12.20). Among those with both probable dementia and self-reported VI, compared with those with neither condition, there was an expected decrease in mobility score of 50.1% (no VI and no dementia, 10.82 vs VI and probable dementia, 5.40), self-care score of 42.4% (no VI and no dementia, 14.54 vs VI and probable dementia, 8.38), and household activity score of 52.4% (no VI and no dementia, 18.23 vs VI and probable dementia, 8.68).
In sensitivity analyses, models were first rerun excluding all data provided by proxy respondents. In a separate sensitivity analysis, a more specific definition of self-reported VI was used that required self-report of blindness or difficulty seeing at distance. Altering these model specifications did not have a measurable outcome on the regression coefficients or their statistical significance. Interactions between possible and probable dementia and self-reported VI remained across all 3 outcomes.
This study examined a nationally representative cohort of older US adults to determine the association of co-occurring dementia and self-reported VI on disability. The results of this study suggest that when dementia and self-reported VI occur together, the capacity to carry out mobility, self-care, and household activities is decreased to a greater degree than would be expected based on the independent associations of these 2 health conditions. This is the only US study to our knowledge that provides national data on the magnitude of disability in coexisting dementia and self-reported VI. By accounting for the role of behavioral and environmental adaptations, this study may offer a more nuanced examination of disability than many prior investigations. Furthermore, findings from this study suggest the need for interventions to optimize visual and cognitive performance and promote functioning across the activity spectrum to maximize independence and optimal aging for a growing population of older adults with visual disability and coexisting dementia.
The results of this study are nationally representative of the Medicare-eligible population in the US 65 years and older. Prior studies in which samples were nonrepresentative or geographically constrained may not reflect the true disability burden in the US population. This may be the case when the distributions of relevant sociodemographic and medical factors like age, sex, race/ethnicity, education, or multimorbidity do not reflect the larger population. Accordingly, the insights that this study provides on the nature and magnitude of disability in co-occurring dementia and self-reported VI may be useful to policy makers and public health actors aiming to quantify and address this growing burden at a population level.
The conceptualization of disability in the NHATS provides for a detailed assessment of function across 12 mobility, self-care, and household activities. Rather than classifying individuals as independent or disabled, as some prior studies have done,9,10 the assessment of disability in the NHATS accounts for adaptations, such as the use of assistive devices and performance of an activity less often or with difficulty. There are several advantages to this approach. First, the NHATS functional ability hierarchies may lend themselves to detecting statistical interactions that can be obfuscated by floor or ceiling effects in dichotomous data. Additionally, these hierarchies make it possible to quantify more precisely the functional outcome of the independent and joint associations of self-reported VI and dementia on disability.
This study demonstrated an interaction between dementia and self-reported VI across all 3 functional hierarchies. In each case, those with probable dementia had lower calculated functional ability scores compared with those with possible dementia, indicating a dose-response association between cognitive performance and functional ability. The association with functional ability was greatest when self-reported VI co-occurred with dementia, resulting in an outcome that was significantly larger than the sum of the independent contributions of self-reported VI and either possible or probable dementia.
Not surprisingly, the largest independent and joint associations were seen for household activities, followed by mobility, and then self-care activities. This is consistent with prior research, which has demonstrated relatively preserved self-care ADL function, even as IADL and mobility functions decline in participants with mild to moderate cognitive impairment.24 These findings may inform the design and implementation of future interventions. For example, integrated vision rehabilitation programs for older adults with VI and mild cognitive impairment may focus on the types of activities that are most likely to be affected by further visual and cognitive decline in this population.
This study had several limitations. Survey data are prone to recall bias, and this may be particularly true among those with dementia. Although self-reported data on visual function have been widely used in prior population-based studies,5,25,26 self-reports of VI may be less reliable among those with cognitive impairment, which could have biased the results. Additional investigations are needed using objectively measured visual function to determine the association between disability and changes in vision and cognitive function over time. Finally, it is not possible to determine causality from these data.
In conclusion, this study demonstrated in a national sample of older US adults that dementia and self-reported VI were independently associated with decreased mobility, self-care, and household activity function. Furthermore, there was an interaction between dementia and self-reported VI that potentiated declines in all 3 functional domains when these conditions co-occurred. Findings from this study may be relevant to designing and implementing clinical and public health interventions to promote independence and optimal aging for a rapidly growing population of older US adults.
Accepted for Publication: March 27, 2020.
Corresponding Author: Joshua R. Ehrlich, MD, MPH, Department of Ophthalmology and Visual Sciences, Kellogg Eye Center, 1000 Wall St, Ann Arbor, MI 48105 (email@example.com).
Published Online: May 14, 2020. doi:10.1001/jamaophthalmol.2020.1562
Author Contributions: Dr Ehrlich had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Stagg, Ehrlich.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Patel, Stagg, Zhou, Ehrlich.
Critical revision of the manuscript for important intellectual content: Stagg, Swenor, Zhou, Talwar.
Statistical analysis: Stagg, Swenor, Zhou, Talwar, Ehrlich.
Obtained funding: Ehrlich.
Administrative, technical, or material support: Patel, Ehrlich.
Conflict of Interest Disclosures: Dr Ehrlich is a co-investigator of the National Health and Aging Trends Study and reported grants from National Institutes of Health during the conduct of the study. Dr Stagg reported grants from Research to Prevent Blindness outside the submitted work. No other disclosures were reported.
Funding/Support: This research was supported by grants from the National Eye Institute (K23 EY027848 [Dr Ehrlich]) and unrestricted grants from Research to Prevent Blindness to the Department of Ophthalmology and Visual Sciences at the University of Michigan and the Department of Ophthalmology and Visual Sciences at the University of Utah.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Meeting Presentation: This work was presented in part at the Association for Research in Vision and Ophthalmology meeting; April 30, 2019; Vancouver, British Columbia, Canada.
Additional Contributions: The authors would like to acknowledge Sara Weiss, MPH, Department of Ophthalmology and Visual Sciences, University of Michigan, for her assistance editing the manuscript. She was compensated for her contribution.