The model-based predictive mean for each chronic health condition is the linear prediction of EQ-5D index score using the average value of the covariates including age, smoking status, household income, education, occupation, and living situation. The estimated regression coefficient of interaction terms between VI and each chronic health condition and corresponding 95% CIs are provided in each graph. To evaluate the interaction between VI and each chronic health condition, the regression model includes VI, each chronic health condition, the interaction between VI and each chronic health condition, and the listed covariates. OA indicates osteoarthritis; RA, rheumatic arthritis.
eFigure. Model-Based Predictive Mean and 95% CI of VI-Stratified EQ-5D Index Scores for Chronic Health Conditions
eTable 1. Frequencies and Weighted Prevalences of Covariates and Chronic Health Conditions in Included and Excluded Participants
eTable 2. EQ-5D Index Scores and Interaction Terms With VI in Each Covariate and Chronic Health Condition Among Participants Aged =60 Years
eTable 3. EQ-5D Index Scores and Interaction Terms With VI (US Definition) in Each Covariate and Chronic Health Condition
eAppendix 1. Additional Analyses in Addition to the 3-Step Modelling
eAppendix 2. Additional Analyses to Address Possible Sampling Issues Due to Small-Numbered Co-occurring Events
Customize your JAMA Network experience by selecting one or more topics from the list below.
Park SJ, Ahn S, Woo SJ, Park KH. Extent of Exacerbation of Chronic Health Conditions by Visual Impairment in Terms of Health-Related Quality of Life. JAMA Ophthalmol. 2015;133(11):1267–1275. doi:10.1001/jamaophthalmol.2015.3055
Copyright 2015 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
Visual impairment (VI) causes a considerable public health burden and substantial deterioration in health-related quality of life (HRQoL). However, the relative effect of VI on HRQoL compared with other chronic health conditions is unknown as is the additive effect of VI with other conditions.
To investigate whether the impact of chronic health conditions on HRQoL varies according to VI presence.
Design, Setting, and Participants
Cross-sectional study involving 29 639 participants aged 19 years and older and using a multistage, probability-cluster survey, which can produce nationally representative estimates. We analyzed data from the 2008-2012 Korean National Health and Nutrition Examination Survey, which included results for vision assessment and HRQoL, measured using the European Quality of Life–5 Dimensions Questionnaire (EQ-5D). All analyses were conducted in October and November 2014. Visual impairment was defined as the presenting distance best-corrected visual acuity of less than decimal 0.32 (approximate Snellen equivalent 20/63). Linear regression models accounting for sample weights were used to examine interactions between VI and each of 14 chronic health conditions on the EQ-5D index score, adjusted for demographic and socioeconomic covariates.
Main Outcomes and Measures
The EQ-5D index score for participants with VI or 14 chronic health conditions, as well as the effect of the interactions between VI and each condition on the EQ-5D index score.
The EQ-5D index score with VI was substantially lower than without VI (mean difference, −0.158; 95% CI, −0.186 to −0.130; P < .001). In most conditions, the comorbidity with VI resulted in lower EQ-5D index scores; furthermore, participants with stroke, osteoarthritis or rheumatic arthritis, hepatitis B or C, and depression showed lower EQ-5D index scores than expected when they were comorbid with VI, indicating an interaction between VI and each condition. The estimated β coefficients for interaction terms were −0.256 (95% CI, −0.480 to −0.032) for stroke, −0.124 (95% CI, −0.223 to −0.026) for osteoarthritis or rheumatic arthritis, −0.183 (95% CI, −0.327 to −0.038) for hepatitis B or C, and −0.130 (95% CI, −0.229 to −0.032) for depression.
Conclusions and Relevance
These results suggest that VI has a substantial effect on HRQoL, even in the presence of concurrent chronic health conditions, and the combined effect of VI and stroke, osteoarthritis/rheumatic arthritis, hepatitis, or depression on HRQoL was greater. However, because this study group was a representative sampling of South Koreans, generalization to other races/ethnicities and countries should be approached with caution.
Visual impairment (VI) has emerged as an increasing public health concern, given the aging population.1-5 Approximately 285 million persons worldwide are visually impaired (visual acuity [VA] <6/18 in the better-seeing eye),5 and the prevalence of VI has increased particularly rapidly in the past 2 decades.1,6 Visual impairment often results in functional disability and health-related concerns such as decreased mobility,7 increased injuries,8 greater risk for depression,9,10 and increased mortality.11-13 Moreover, VI is associated with an increased risk for other concurrent chronic health conditions.14 Consequently, VI has a negative effect on health-related quality of life (HRQoL),3,4,15-17 resulting in a burden on the public health care system18 that is likely equivalent to or greater than that of hypertension, diabetes mellitus, and obesity, which are considered major chronic health problems.18
Compared with individuals without VI, those with VI require more formal and informal, as well as paid and unpaid, support19,20; in addition, VI-related services (eg, rehabilitation and counseling) and physical aids for limited vision are often required.20,21 As a result, individuals with VI and their families and communities endure an economic burden and decreased QoL.22,23 The growing VI burden has led to ever-increasing demands on health care systems in developed countries.23 Because of their limited availability, resources in public health care systems are often prioritized based on the effect of chronic health conditions on HRQoL; however, to our knowledge, there is a paucity of data regarding the effect of VI on public health burden, particularly HRQoL, in relation to other health conditions. Therefore, the present study aimed to determine the independent and concurrent effects of VI and chronic health conditions on HRQoL.
We investigated whether the impact of chronic diseases on health-related quality of life (HRQoL) varies according to the presence of visual impairment (VI).
Concerns have emerged that some chronic diseases might affect individuals’ HRQoL more devastatingly when they were comorbid with VI.
The results showed that VI has substantial effects on HRQoL, over and above the presence of other concurrent chronic diseases.
Furthermore, the effect of stroke, arthritis, hepatitis, and depression on HRQoL was greater when VI was also present.
The results underscore the importance of VI in public health, as well as in managing patients with chronic diseases.
We analyzed data collected in a 5-year period (2008-2012) as part of the Korean National Health and Nutrition Examination Survey (KNHANES). Detailed information of the survey has been described previously.24-27 Briefly, the KNHANES is an ongoing, government-led, population-based, cross-sectional survey in South Korea.25-27 The 2008-2012 KNHANES randomly selected 3840 to 4600 households on an annual basis in 192 to 200 postal codes representing the civilian, noninstitutionalized Korean population using rolling sampling designs, which involved a multistage, probability-cluster survey. The KNHANES consists of a Health Interview Survey, Health Examination Survey, and Nutrition Survey25-27; we analyzed data from the first 2 surveys. All analyses were conducted in October and November 2014. The institutional review board of the Seoul National Bundang Hospital approved this study and provided a waiver of informed consent because this study used anonymous patient data (X-1506-304-912).
Quiz Ref IDAll participants aged 5 years and older underwent both VA and autorefraction testing (KR8800; Topcon). Uncorrected and/or corrected VA (decimal VA), with the individual’s own glasses/lenses, was measured at a distance of 4 m using an international standard vision chart based on the LogMAR Scale (Jin’s vision chart).24,28 Participants with VA less than decimal 0.8 underwent best-corrected VA (BCVA) testing using autorefraction results. If participants could not read any number at 4 m, VA was measured at a distance of 1 m. Visual impairment was defined as a presenting distance BCVA less than decimal 0.32 (0.25 or worse) (approximate Snellen equivalent 20/63) in the better-seeing eye, as defined by the World Health Organization.29 In addition, another definition of VI widely used in the United States was applied: presenting distance BCVA less than decimal 0.5 in the better-seeing eye.2,30 The ophthalmic examination quality was verified by the Epidemiologic Survey Committee of the Korean Ophthalmological Society.24-27
For HRQoL, all participants aged 19 years and older were asked to complete the validated Korean version of the European Quality of Life–5 Dimensions Questionnaire (EQ-5D),31-33 which records 3 levels of self-reported problems (no problem, some problems, or extreme problems) within each of 5 dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression).33 The results were scored using the responses to the 5 dimensions and the Korean population–based EQ-5D index score, which is based on the preference weights for Koreans derived from representative samples using the time trade-off method (−0.257 [worst health state] to 1.00 [perfect health state]).32
Based on previous related studies, demographic/socioeconomic covariates were determined as follows: age, smoking status (never smoker, ex-smoker, or current smoker), equalized gross household income (>50% or ≤50% household income), education (≥high school or ≤ middle school), residence (urban or rural area), occupation (white collar, blue collar, or unemployed), and living situation (living alone or living with others).26,27
Quiz Ref IDWe collected information regarding the presence of 14 chronic health conditions that could considerably affect HRQoL,34 including 12 chronic diseases (hypertension, diabetes mellitus, dyslipidemia, stroke, myocardial infarction or ischemic heart disease [MI/IHD], osteoarthritis or rheumatic arthritis [OA/RA], pulmonary tuberculosis, asthma, renal failure, hepatitis B or C, depression, and any cancer [cancer survivors]), obesity (body mass index ≥30, calculated as weight in kilograms divided by height in meters squared), and anemia (hemoglobin level [XE-2100D, Sysmex] <13 g/dL in men or <12 g/dL in women; to convert to grams per liter, multiply by 10). The presence of 12 chronic diseases was defined using the Health Interview Survey, and obesity and anemia were defined using the Health Examination Survey.25-27
We restricted the analyses data from participants aged 19 years and older who had both VA and EQ-5D data. For all analyses, the KNHANES sample weights were applied to adjust for survey design, nonresponse, and poststratification to generate nationally representative, population-based results.25,26 To evaluate the impact of each chronic health condition on HRQoL and the additional effect of VI with each chronic health condition on HRQoL, weighted means of EQ-5D index scores appropriately adjusted for the other effects in the model were calculated for each chronic health condition and then stratified by VI. To evaluate interactions between VI and chronic health conditions for the EQ-5D index score, the following linear regression analysis models were conducted for each chronic health condition as well as for HRQoL: model 1 included effects for VI and each of the chronic health conditions and the interaction term between them; model 2 added age (as a continuous variable) and sex effects to model 1; and model 3 added all prespecified covariates to model 2. We displayed the model-based predictive mean and its 95% CI for the VI-stratified EQ-5D index score for each chronic health condition; these are the linear predictions of the EQ-5D index score using the average value of the covariates. Linear model assumptions were examined through regression diagnostics. The same set of analyses was repeated for sensitivity analyses; the first set was conducted in participants aged 60 years and older and the second set was conducted using the US VI definition in all included participants.
Statistical analyses were conducted using SAS version 9.3 (SAS Institute, Inc) with the PROC SURVEY procedures to account for the complex survey sampling design and R software version 3.0.2 (R Foundation for Statistical Computing) using the ggplot2 library to present the adjusted weighted means plot. The not-missing-completely-at-random option was implemented for variability of participants with missing values. Statistical significance was set at P < .05.
Of the 29 639 persons aged 19 years and older who participated in the 2008-2012 KNHANES, 1257 participants were excluded owing to a lack of VA data (689 participants) or EQ-5D results (743 participants), resulting in 28 382 participants in the analyses (eTable 1 in the Supplement). Using the World Health Organization definition, 173 participants had VI, with a lower index score (weighted mean [SE], 0.781 [0.023]) than those without VI (n = 28 209; mean [SE], 0.948 [0.001]). The mean difference was −0.158 (95% CI, −0.186 to −0.130; P < .001). The index scores were substantially lower in participants with VI and each chronic health condition compared with those without VI and each of the chronic health conditions (range differences were from −0.010 in hepatitis B/C to −0.144 in stroke), and the index score of VI individuals was lower than that of all other conditions (Table).
Quiz Ref IDWhen each of the chronic health conditions was stratified by VI, the index scores were lower with VI and the chronic health condition combined than with the chronic health condition alone, except for MI/IHD. In addition, 4 conditions (stroke, OA/RA, hepatitis B or C, and depression) resulted in substantially lower index scores than expected when they were comorbid with VI. The interaction terms between VI and each of the 4 conditions were statistically significant after adjustment for covariates (model 3 β coefficients for the interaction terms were −0.256; 95% CI, −0.480 to −0.032 for stroke; −0.124; 95% CI, −0.223 to −0.026 for OA/RA; −0.183; 95% CI, −0.327 to −0.038 for hepatitis B or C; and −0.130; 95% CI, −0.229 to −0.032 for depression), indicating greater effects on HRQoL when combined with VI. In addition, some covariates (ie, older age groups, lower household income, lower education level, rural residence, and nonwhite collar employment) were associated with lower index scores, while age group, household income (model 3 β coefficient for the interaction term, −0.112; 95% CI, −0.185 to −0.039), and residence (−0.118; 95% CI, −0.209 to −0.026) had statistically significant interactive relationships with VI in model 3, indicating that the effect of VI on HRQoL was modified by these covariates (Table; Figure; and eFigure in the Supplement).
The sensitivity analyses showed similar results to the primary analysis. The first sensitivity analysis included 8984 participants aged 60 years and older and showed that the impact of 3 conditions (stroke, hepatitis B or C, and depression) on HRQoL was intensified when comorbid with VI (eTable 2 in the Supplement); model 3 β coefficients for the interaction terms were −0.301 (95% CI, −0.516 to −0.086), −0.203 (95% CI, −0.398 to −0.008), and −0.158 (95% CI, −0.290 to −0.026), respectively. The second sensitivity analysis using the US definition included 522 participants with VI who had a lower index score (mean [SE], 0.791 [0.014]) than those without VI (n = 27 860; mean [SE], 0.950 [0.001]). The effects of concurrent VI and each chronic health condition on HRQoL were similar to the main results. However, only the concurrent effect of OA/RA was intensified when it was comorbid with VI (model 3 β coefficient for the interaction terms were −0.067; 95% CI, −0.122 to −0.012) (eTable 3 in the Supplement).
The present study showed that VI is associated with decreased HRQoL, and the impact of VI on HRQoL was comparable with or higher than that of the other evaluated conditions. Furthermore, VI resulted in lower HRQoL, over and above other chronic health conditions that also negatively affect HRQoL. In addition, the effects of stroke, OA/RA, hepatitis B or C, and depression on HRQoL were significantly greater when VI was also present. To our knowledge, this is the first study to simultaneously present the independent effects of VI and chronic health conditions on HRQoL in a large general population and to assess the combined effects of VI and chronic health conditions on HRQoL. Individuals with VI experience decreased mobility7 and health care use19,35,36; increased need for support19,20 and services,20,21 including the use of medicines37; and limitations in driving.38 These difficulties may affect HRQoL, although the effect may differ based on the exact nature of concurrent conditions.
Stroke had the greatest effect on HRQoL among the evaluated conditions, which increased more with the presence of VI. Stroke causes substantial difficulties in every phase of life, and VI likely intensifies these difficulties.39,40 Numerous studies have recommended careful vision assessments in stroke patients because concurrent VI is associated with negative rehabilitation and HRQoL outcomes.39,40 The interaction between VI and stroke might reflect, at least partially, stroke severity and/or other concurrent stroke-related visual problems, although bilateral vision deterioration rarely occurs with stroke.40 The result underscores the need for better strategies in treating stroke-related vision loss.
Similarly, a bidirectional association between VI and depression has been reported,41,42 which might explain their interaction. The reported effects of VI, including social isolation,43 feelings of loneliness,44 and lower life satisfaction,45 might collectively exacerbate declines of HRQoL in patients with depression. The results provide additional evidence for the need for vision assessment in patients with depression and for the monitoring of depression in patients with VI.
Regarding the relationships between VI and OA/RA or hepatitis in the present study, there was no inherent shared pathophysiology. Rather, similar to VI, OA/RA causes immediate adverse effects on mobility, health care use, and the need for support/services, although for different reasons.46 Because OA/RA is one of the most common disabilities (12.5% of the general population aged ≥19 years) and has the second greatest impact on HRQoL, OA/RA itself is associated not only with a substantial decline in HRQoL, but also with a significant public health burden.34,46 Therefore, the observed interaction between OA/RA and VI on HRQoL underscores the importance of VI management in patients with OA/RA. Hepatitis per se does not result in decreased HRQoL. However, in the present study, individuals with both hepatitis and VI showed lower HRQoL than expected, which might be partially explained by hepatitis severity; the degree of VI may vary by hepatitis severity (eg, development of diabetes mellitus in end-stage hepatitis47).
Lastly, although not statistically significant, the results suggest that VI might have an interaction effect on HRQoL with renal failure, pulmonary tuberculosis, and a history of cancer (Figure, G and H, and eFigure, F in the Supplement). Further investigations are warranted to determine whether the effects of these conditions vary according to VI presence. The sensitivity analyses presented not only similar results, but also severity-response trends in the effects on HRQoL, suggesting the robustness of the results. Under the US VI definition, only OA/RA had an interaction with VI, while stroke, OA/RA, hepatitis, and depression had an interaction in the primary analysis, suggesting that the threshold of VA to effectuate interactions between VI and stroke, hepatitis, and depression is somewhere between a better-seeing eye BCVA of decimal 0.32 (World Health Organization definition) (approximate Snellen equivalent 20/63) and decimal 0.5 (US definition) (approximate Snellen equivalent 20/40). This threshold may differ according to legal, racial/ethnic, and cultural backgrounds.
Quiz Ref IDTwo covariates (household income and residence) also modified effects of VI on HRQoL; the effect of VI on HRQoL was greater in individuals with low household income or living in rural areas compared with those with high household income or living in urban areas, respectively. They might affect accessibility and use of formal/informal and paid/unpaid support as well as health care services.48-50 However, because VI could undermine incomes and residence locations directly, further in-depth investigations are warranted.
The 2008-2012 KNHANES includes detailed information regarding VA and EQ-5D results in addition to demographic/socioeconomic status and health conditions in every participant aged 19 years and older; it provides sufficient power for analyses of conditions with a low prevalence and allows a rigorous modeling approach for investigating interactions. Low rates of missing data in the variables, apart from anemia, in the present study allowed us to avoid related problems. Another strength was the definition of VI using bilateral BCVA results, which was measured by trained ophthalmologists under guidance of the academic society.24-27 Accurate BCVA results allowed the analyses using 2 popular VI definitions and assessments of the effects of uncorrectable VI. Although refractive error is the leading cause of VI both in developed and developing countries,2,5 VI due to uncorrected refractive error can be easily improved by providing appropriate refractive correction.2
South Korea has provided universal health care coverage since 1988,51 and there is a large number of certified ophthalmologists (about 3000) across the country, providing easy access to vision-related health care services. The KNHANES might be the most relevant population to analyze HRQoL in relation to VI because of the universal coverage and high accessibility, as well as its representation of the Korean population. Moreover, the EQ-5D index score was generated by the Korean version of preference weights, which is largely different across countries and cultural backgrounds.32
However, the present study had several limitations. First, the KNHANES did not include institutionalized individuals in whom VI and chronic health conditions are prevalent,25-27 and we excluded participants without VA or EQ-5D results; therefore, the results might not be generalizable to hospitalized or institutionalized individuals. The generalization to other races/ethnicities/countries might also be limited, given the inclusion of only Koreans living in South Korea. Second, the analyses did not consider the severity of each health condition owing to a lack of information, which might also explain our results. In addition, the 12 chronic diseases were defined by the Health Interview Survey, which might underestimate the number of cases with each condition. Third, VI was defined with only the better-seeing eye, which is likely to underestimate the effect of unilateral VI.52 In addition, other VI dimensions, including abnormalities in visual field, color vision, stereovision, and contrast sensitivity, were not analyzed. Fourth, EQ-5D has poor sensitivity to VA loss because it consists of generic constructs and lacks vision-specific dimensions.53,54 However, EQ-5D has been previously demonstrated as a good instrument for measuring HRQoL in individuals with VI.15,18,52-54 Fifth, the model only included measured covariates and the 1-way interaction between chronic health conditions; therefore, potential residual bias might have existed despite the multistep modeling approach and model validity, as examined through a regression diagnostic. Therefore, we conducted the additional analysis to address the observed interactions in covariates (eg, age, house income, and residence), which is provided in eAppendix 1 and eAppendix 2 in the Supplement. Lastly, in some analyses, the number of participants having both VI and each chronic condition was very small, although we analyzed 30 000 representative participants. To address possible unreliable sampling, we implemented 2 analytic methods: (1) the bootstrapping-based 95% CI for each final multivariable model and (2) the ancillary analysis on the status of any comorbidity. These are also provided in eAppendix 1 and eAppendix 2 in the Supplement.
In summary, the present study showed that VI substantially affects HRQoL even with concurrent chronic health conditions, and the presence of certain health conditions in people with VI might result in greater effects on HRQoL. The results underscore the importance of VI in public health and provide information for the prioritization of VI management in health care systems, especially for individuals with stroke, OA/RA, hepatitis B or C, and depression.
Corresponding Author: Kyu Hyung Park, MD, PhD, Department of Ophthalmology, Seoul National University Bundang Hospital, #300, Gumi-dong, Bundang-gu, Seongnam, Gyeonggi-do 463-707, Republic of Korea (email@example.com).
Submitted for Publication: April 12, 2015; final revision received July 7, 2015; accepted July 16, 2015.
Published Online: September 17, 2015. doi:10.1001/jamaophthalmol.2015.3055.
Author Contributions: Drs Ahn and K. H. Park had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs S. J. Park and Ahn contributed equally to this work.
Study concept and design: All authors.
Acquisition, analysis, or interpretation of data: S. J. Park, Ahn, K. H. Park.
Drafting of the manuscript: S. J. Park, Ahn.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: S. J. Park, Ahn.
Administrative, technical, or material support: All authors.
Study supervision: All authors.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: This study was supported by a National Research Foundation of Korea grant funded by the Ministry of Education, Science, and Technology (grant NRF-2013R1A2A2A04015829) and a grant funded by the Seoul National University Bundang Hospital Fund (grant 02-2014-015). Dr Ahn received a grant from the Seoul National University Bundang Hospital Fund.
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.
Additional Contributions: We are indebted to J. Patrick Barron, PhD (professor emeritus, Tokyo Medical University, and adjunct professor, Seoul National University Bundang Hospital), for his pro bono editing of this manuscript.
Create a personal account or sign in to: