[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address 54.161.216.242. Please contact the publisher to request reinstatement.
[Skip to Content Landing]
Download PDF
Figure 1.
Structural equation modeling path for the relationship between noncorrectable visual impairment, risk markers, and all-cause mortality in all ages, showing hazard ratios (HRs) and 95% confidence intervals (CIs) (total indirect: HR, 1.68; 95% CI, 1.21-2.33). Covariates are corrected for age and sex. BMI indicates body mass index (calculated as weight in kilograms divided by height in meters squared). *Other covariates include the mortality risk markers angina, myocardial infarction, stroke, cancer, hypertension, smoking, diabetes mellitus, home ownership, tertiary education, and self-rated health. †There was no significant relationship between visual impairment and self-rated health independent of walking disability.

Structural equation modeling path for the relationship between noncorrectable visual impairment, risk markers, and all-cause mortality in all ages, showing hazard ratios (HRs) and 95% confidence intervals (CIs) (total indirect: HR, 1.68; 95% CI, 1.21-2.33). Covariates are corrected for age and sex. BMI indicates body mass index (calculated as weight in kilograms divided by height in meters squared). *Other covariates include the mortality risk markers angina, myocardial infarction, stroke, cancer, hypertension, smoking, diabetes mellitus, home ownership, tertiary education, and self-rated health. †There was no significant relationship between visual impairment and self-rated health independent of walking disability.

Figure 2.
Structural equation modeling path for the relationship between presenting visual impairment, risk markers, and all-cause mortality in all ages, showing hazard ratios (HRs) and 95% confidence intervals (CIs) (total indirect: HR, 1.40; 95% CI, 1.13-1.73). Covariates are corrected for age and sex. *Other covariates include the mortality risk markers angina, myocardial infarction, stroke, cancer, hypertension, smoking, diabetes mellitus, home ownership, tertiary education, and self-rated health. †There was no significant relationship between visual impairment and self-rated health independent of walking disability.

Structural equation modeling path for the relationship between presenting visual impairment, risk markers, and all-cause mortality in all ages, showing hazard ratios (HRs) and 95% confidence intervals (CIs) (total indirect: HR, 1.40; 95% CI, 1.13-1.73). Covariates are corrected for age and sex. *Other covariates include the mortality risk markers angina, myocardial infarction, stroke, cancer, hypertension, smoking, diabetes mellitus, home ownership, tertiary education, and self-rated health. †There was no significant relationship between visual impairment and self-rated health independent of walking disability.

Figure 3.
Structural equation modeling path for the relationship between correctable visual impairment, risk markers, and all-cause mortality in all ages, showing hazard ratios (HRs) and 95% confidence intervals (CIs) (total indirect: HR, 1.27; 95% CI, 1.00-1.61). Covariates are corrected for age and sex. *Other covariates include the mortality risk markers angina, myocardial infarction, stroke, cancer, hypertension, smoking, diabetes mellitus, home ownership, tertiary education, and self-rated health. †There was no significant relationship between visual impairment and self-rated health independent of walking disability.

Structural equation modeling path for the relationship between correctable visual impairment, risk markers, and all-cause mortality in all ages, showing hazard ratios (HRs) and 95% confidence intervals (CIs) (total indirect: HR, 1.27; 95% CI, 1.00-1.61). Covariates are corrected for age and sex. *Other covariates include the mortality risk markers angina, myocardial infarction, stroke, cancer, hypertension, smoking, diabetes mellitus, home ownership, tertiary education, and self-rated health. †There was no significant relationship between visual impairment and self-rated health independent of walking disability.

Table 1. 
Prevalence of Mortality Risk Markers in Participants of the Blue Mountains Eye Study 10-Year Follow-up Examination by Visual Impairment
Prevalence of Mortality Risk Markers in Participants of the Blue Mountains Eye Study 10-Year Follow-up Examination by Visual Impairment
Table 2. 
Association of VI and All-Cause Mortality Assessed Using Cox Regression by VI Categorya
Association of VI and All-Cause Mortality Assessed Using Cox Regression by VI Categorya
Table 3. 
Association of VI and All-Cause Mortality Assessed Using Structural Equation Modeling by VI Categorya
Association of VI and All-Cause Mortality Assessed Using Structural Equation Modeling by VI Categorya
Table 4. 
Total Indirect Effects of VI to All-Cause Mortality Using Structural Equation Modeling Pathway Analysis Stratified by Pathwaya
Total Indirect Effects of VI to All-Cause Mortality Using Structural Equation Modeling Pathway Analysis Stratified by Pathwaya
1.
Cugati  SCumming  RGSmith  WBurlutsky  GMitchell  PWang  JJ Visual impairment, age-related macular degeneration, cataract, and long-term mortality: the Blue Mountains Eye Study. Arch Ophthalmol 2007;125 (7) 917- 924
PubMedArticle
2.
Freeman  EEEgleston  BLWest  SKBandeen-Roche  KRubin  G Visual acuity change and mortality in older adults. Invest Ophthalmol Vis Sci 2005;46 (11) 4040- 4045
PubMedArticle
3.
Lee  DJGomez-Marin  OLam  BLZheng  DD Visual acuity impairment and mortality in US adults. Arch Ophthalmol 2002;120 (11) 1544- 1550
PubMedArticle
4.
Klein  RKlein  BEMoss  SE Age-related eye disease and survival: the Beaver Dam Eye Study. Arch Ophthalmol 1995;113 (3) 333- 339
PubMedArticle
5.
McCarty  CANanjan  MBTaylor  HR Vision impairment predicts 5 year mortality. Br J Ophthalmol 2001;85 (3) 322- 326
PubMedArticle
6.
Clemons  TEKurinij  NSperduto  RDAREDS Research Group, Associations of mortality with ocular disorders and an intervention of high-dose antioxidants and zinc in the Age-Related Eye Disease Study: AREDS Report No. 13. Arch Ophthalmol 2004;122 (5) 716- 726
PubMedArticle
7.
Wang  JJMitchell  PSimpson  JMCumming  RGSmith  W Visual impairment, age-related cataract, and mortality. Arch Ophthalmol 2001;119 (8) 1186- 1190
PubMedArticle
8.
Knudtson  MDKlein  BEKlein  R Age-related eye disease, visual impairment, and survival: the Beaver Dam Eye Study. Arch Ophthalmol 2006;124 (2) 243- 249
PubMedArticle
9.
Lee  DJGomez-Marin  OLam  BLZheng  DD Visual impairment and unintentional injury mortality: the National Health Interview Survey 1986-1994. Am J Ophthalmol 2003;136 (6) 1152- 1154
PubMedArticle
10.
Kulmala  JEra  PParssinen  O  et al.  Lowered vision as a risk factor for injurious accidents in older people. Aging Clin Exp Res 2008;20 (1) 25- 30
PubMedArticle
11.
Kulmala  JEra  PTormakangas  TParssinen  ORantanen  THeikkinen  E Visual acuity and mortality in older people and factors on the pathway. Ophthalmic Epidemiol 2008;15 (2) 128- 134
PubMedArticle
12.
Evans  JRFletcher  AEWormald  RP Depression and anxiety in visually impaired older people. Ophthalmology 2007;114 (2) 283- 288
PubMedArticle
13.
Thiagarajan  MEvans  JRSmeeth  LWormald  RPFletcher  AE Cause-specific visual impairment and mortality: results from a population-based study of older people in the United Kingdom. Arch Ophthalmol 2005;123 (10) 1397- 1403
PubMedArticle
14.
Smith  WMitchell  PLeeder  SRWang  JJ Plasma fibrinogen levels, other cardiovascular risk factors, and age-related maculopathy: the Blue Mountains Eye Study. Arch Ophthalmol 1998;116 (5) 583- 587
PubMedArticle
15.
Klein  BEKlein  RLee  KECruickshanks  KJ Performance-based and self-assessed measures of visual function as related to history of falls, hip fractures, and measured gait time: the Beaver Dam Eye Study. Ophthalmology 1998;105 (1) 160- 164
PubMedArticle
16.
Wang  JJMitchell  PSmith  W Vision and low self-rated health: the Blue Mountains Eye Study. Invest Ophthalmol Vis Sci 2000;41 (1) 49- 54
PubMed
17.
Thompson  JRGibson  JMJagger  C The association between visual impairment and mortality in elderly people. Age Ageing 1989;18 (2) 83- 88
PubMedArticle
18.
Mooijaart  SPKoeijvoets  KMSijbrands  EJDaha  MRWestendorp  RG Complement Factor H polymorphism Y402H associates with inflammation, visual acuity, and cardiovascular mortality in the elderly population at large. Exp Gerontol 2007;42 (11) 1116- 1122
PubMedArticle
19.
Hennis  AWu  SYLi  XNemesure  BLeske  MCBarbados Eye Study Group, Lens opacities and mortality: the Barbados Eye Studies. Ophthalmology 2001;108 (3) 498- 504
PubMedArticle
20.
West  SKMunoz  BIstre  J  et al.  Mixed lens opacities and subsequent mortality. Arch Ophthalmol 2000;118 (3) 393- 397
PubMedArticle
21.
Ditlevsen  SChristensen  ULynch  JDamsgaard  MTKeiding  N The mediation proportion: a structural equation approach for estimating the proportion of exposure effect on outcome explained by an intermediate variable. Epidemiology 2005;16 (1) 114- 120
PubMedArticle
22.
Rothman  KJGreenland  S Modern Epidemiology. 2nd ed. Philadelphia, PA Lippincott-Raven1998;
23.
Bentler  PMStein  JA Structural equation models in medical research. Stat Methods Med Res 1992;1 (2) 159- 181
PubMedArticle
24.
Christ  SLLee  DJLam  BLZheng  DDArheart  KL Assessment of the effect of visual impairment on mortality through multiple health pathways: structural equation modeling. Invest Ophthalmol Vis Sci 2008;49 (8) 3318- 3323
PubMedArticle
25.
Attebo  KMitchell  PSmith  W Visual acuity and the causes of visual loss in Australia: the Blue Mountains Eye Study. Ophthalmology 1996;103 (3) 357- 364
PubMedArticle
26.
Powers  JBall  JAdamson  LDobson  A Effectiveness of the National Death Index for establishing the vital status of older women in the Australian Longitudinal Study on Women's Health. Aust N Z J Public Health 2000;24 (5) 526- 528
PubMedArticle
27.
Whitworth  JAWorld Health Organization, International Society of Hypertension Writing Group, 2003 World Health Organization (WHO)/International Society of Hypertension (ISH) statement on management of hypertension. J Hypertens 2003;21 (11) 1983- 1992
PubMedArticle
28.
Muthen  LKMuthen  BO Mplus User's Guide. 5th ed. Los Angeles, CA Muthen & Muthen2007;
29.
Weverling-Rijnsburger  AWBlauw  GJLagaay  AMKnook  DLMeinders  AEWestendorp  RG Total cholesterol and risk of mortality in the oldest old. Lancet 1997;350 (9085) 1119- 1123
PubMedArticle
30.
Langer  RDGaniats  TGBarrett-Connor  E Paradoxical survival of elderly men with high blood pressure. BMJ 1989;298 (6684) 1356- 1357
PubMedArticle
31.
van Bemmel  TGussekloo  JWestendorp  RGBlauw  GJ In a population-based prospective study, no association between high blood pressure and mortality after age 85 years. J Hypertens 2006;24 (2) 287- 292
PubMedArticle
32.
Ivers  RQCumming  RGMitchell  PSimpson  JMPeduto  AJ Visual risk factors for hip fracture in older people. J Am Geriatr Soc 2003;51 (3) 356- 363
PubMedArticle
33.
Chong  EWSinclair  AJGuymer  RH Facts on fats. Clin Experiment Ophthalmol 2006;34 (5) 464- 471
PubMedArticle
34.
Bone  RALandrum  JTGuerra  LHRuiz  CA Lutein and zeaxanthin dietary supplements raise macular pigment density and serum concentrations of these carotenoids in humans. J Nutr 2003;133 (4) 992- 998
PubMed
35.
Seddon  JMCote  JRosner  B Progression of age-related macular degeneration: association with dietary fat, transunsaturated fat, nuts, and fish intake. Arch Ophthalmol 2003;121 (12) 1728- 1737
PubMedArticle
36.
Chiu  CJTaylor  A Nutritional antioxidants and age-related cataract and maculopathy. Exp Eye Res 2007;84 (2) 229- 245
PubMedArticle
37.
Gale  CRHall  NFPhillips  DIMartyn  CN Plasma antioxidant vitamins and carotenoids and age-related cataract. Ophthalmology 2001;108 (11) 1992- 1998
PubMedArticle
38.
Ames  BNShigenaga  MKHagen  TM Oxidants, antioxidants, and the degenerative diseases of aging. Proc Natl Acad Sci U S A 1993;90 (17) 7915- 7922
PubMedArticle
39.
Heidemann  CSchulze  MBFranco  OHvan Dam  RMMantzoros  CSHu  FB Dietary patterns and risk of mortality from cardiovascular disease, cancer, and all causes in a prospective cohort of women. Circulation 2008;118 (3) 230- 237
PubMedArticle
40.
van Dam  RMWillett  WCRimm  EBStampfer  MJHu  FB Dietary fat and meat intake in relation to risk of type 2 diabetes in men. Diabetes Care 2002;25 (3) 417- 424
PubMedArticle
41.
van Dam  RMRimm  EBWillett  WCStampfer  MJHu  FB Dietary patterns and risk for type 2 diabetes mellitus in U.S. men. Ann Intern Med 2002;136 (3) 201- 209
PubMedArticle
42.
Giugliano  DCeriello  AEsposito  K The effects of diet on inflammation: emphasis on the metabolic syndrome. J Am Coll Cardiol 2006;48 (4) 677- 685
PubMedArticle
Epidemiology
October 12, 2009

Direct and Indirect Effects of Visual Impairment on Mortality Risk in Older PersonsThe Blue Mountains Eye Study

Author Affiliations

Author Affiliations: Centre for Vision Research, Department of Ophthalmology, and Westmead Millennium Institute (Drs Karpa, Mitchell, and Wang, Mr Beath, and Ms Rochtchina), and School of Public Health and Centre for Education and Research on Ageing, Concord Hospital (Dr Cumming), University of Sydney, Sydney, and Retinal Vascular Imaging Centre, Centre for Eye Research Australia, Department of Ophthalmology, University of Melbourne, Melbourne (Dr Wang), Australia.

Arch Ophthalmol. 2009;127(10):1347-1353. doi:10.1001/archophthalmol.2009.240
Abstract

Objective  To investigate pathways from visual impairment to increased all-cause mortality in older persons.

Methods  The Blue Mountains Eye Study examined 3654 persons 49 years and older (82.4% response) during 1992-1994 and after 5 and 10 years. Australian National Death Index data confirmed deaths until 2005. Visual impairment was defined as presenting, correctable, and noncorrectable, using better-eye visual acuity. Associations between visual impairment and mortality risk were estimated using Cox regression and structural equation modeling.

Results  After 13 years, 1273 participants had died. Adjusting for mortality risk markers, higher mortality was associated with noncorrectable visual impairment (hazard ratio [HR], 1.35; 95% confidence interval [CI], 1.04-1.75). This association was stronger for ages younger than 75 years (HR, 2.58; 95% CI, 1.42-4.69). Structural equation modeling revealed greater effects of noncorrectable visual impairment on mortality risk (HR, 5.25; 95% CI, 1.97-14.01 for baseline ages <75 years), with both direct (HR, 2.16; 95% CI, 1.11-4.23) and indirect (HR, 2.43; 95% CI, 1.17-5.03) effects. Of mortality risk markers examined, only disability in walking demonstrated a significant indirect pathway for the link between visual impairment and mortality.

Conclusions  Visual impairment predicted mortality by both direct and indirect pathways, particularly for persons younger than 75 years with noncorrectable visual impairment. Disability in walking, which can substantially influence general health, represented a major indirect pathway.

Visual impairment has consistently been associated with a higher risk of dying.18 Visual impairment is also reportedly associated with many factors also linked to increased mortality. These include unintentional injury,9,10 reduced walking speed,9,11 depression,2,11,12 lower body mass index (BMI),13,14 increased risk of falls,9,15 self-rated health,16 self-reported difficulty in physical activity,11,17 systemic inflammation,14,18 cardiovascular disease,11,14 dementia,17 and cancer.19,20 Correction for these “confounders” has been found to attenuate the association between visual impairment and mortality, but the mechanisms behind the association between visual impairment and mortality remain to be determined.

Because of the complex interactions of other mortality risk factors with visual impairment, correcting for these covariates using traditional regression techniques could underestimate the total effect of visual impairment on mortality.21 For example, persons with visual impairment may be more likely to use walking aids because of an increased risk or fear of falling. If true, then adjustment for the use of walking aids, an independent marker of mortality, would underestimate the effect of visual impairment. This covariate is an intermediate variable, a variable that lies on the causal pathway between visual impairment and mortality22 so that simple adjustment for such variables in a traditional statistical model is not appropriate.

Structural equation modeling (SEM) is a modern statistical method that permits modeling of complex relationships that are difficult to estimate using traditional regression techniques.23 Structural equation modeling facilitates the examination and quantification of direct pathways and indirect pathways via intermediate variables. Estimates for such variables can be summated to determine the total indirect effect of the variable of interest on the outcome. Adding the indirect and direct effects then estimates the total effect of the variable of interest on the outcome. To our knowledge, only 1 study has used SEM to examine the associations between visual impairment and mortality.24 This large population survey relied on self-reporting or proxy reporting of visual impairment and comorbidities. The authors reported that in addition to a direct effect on mortality, visual impairment increased mortality risk indirectly through intermediate variables self-rated health and disability.24 To confirm the findings by Christ et al,24 we aimed, in an older Australian population-based cohort, to examine associations between visual impairment, mortality risk markers, and the 13-year risk of mortality using a SEM approach.

METHODS
STUDY POPULATION

The Blue Mountains Eye Study (BMES) is a population-based cohort study of vision and common eye diseases in a suburban Australian population 49 years and older at baseline. The study was approved by the Human Research Ethics Committee of the University of Sydney and was conducted adhering to the tenets of the Declaration of Helsinki. Signed informed consent was obtained from all the participants at each examination.

DATA COLLECTION

Survey methods and procedures have been described previously.25 Briefly, baseline examinations of 3654 residents older than 49 years were conducted during 1992-1994 (BMES 1, 82.4% participation rate). Of the baseline participants, 2335 (75.1% of survivors) returned for 5-year follow-up examinations during 1997-1999 (BMES 2), and 1952 participants (53.4% of the original cohort, or 76.6% of survivors) returned for 10-year follow-up examinations during 2002-2004 (BMES 3). Medical and smoking histories were determined by interviewer-administered questionnaire at baseline. A history of angina, myocardial infarction, diabetes mellitus, hypertension, stroke, or cancer was determined by responses to questions starting with “Has a doctor advised you that you have . . . ?” History of smoking was defined as never, past, or current smoking. Current smokers included those who had stopped smoking within the past year. Weight in kilograms, height in meters, and systolic and diastolic blood pressures were also recorded at baseline.

To identify and confirm persons who died after the baseline examination, demographic information, including surname, first and second names, sex, and date of birth of the 3654 participants, was cross-matched with Australian National Death Index data for deaths to the end of 2005.26 A probabilistic record linkage package was used, adopting a multiple-pass procedure in which both data sets were grouped based on different characteristics (eg, date of birth, name, sex) each time. Matches were divided into exact and nonexact. All nonexact matched records were examined manually and accepted if there was only 1 nonexact matched characteristic that was not critical. Information provided by family members during follow-up was also included if the participant was reported to have died on or before December 2005.

DEFINITIONS

Classification of hypertension was based on the 2003 World Health Organization/International Society of Hypertension guidelines.27 Participants were classified as having hypertension stage 1 if systolic blood pressure was 140 to 159 mm Hg or if diastolic blood pressure was 90 to 99 mm Hg. Participants were classified as having hypertension stage 2 if they were previously diagnosed with hypertension and were using antihypertensive medications, if systolic blood pressure was 160 mm Hg or greater, or if diastolic blood pressure was 100 mm Hg or greater at examination. Body mass index was calculated as weight in kilograms divided by height in meters squared, with less than 20 defined as low. Disability in walking at baseline was assessed as present if the participant was observed by a trained examiner to have walking difficulties or used walking aids or a wheelchair.

Visual acuity (VA) was assessed using a logMAR chart. Visual acuity was initially assessed with patients wearing their current eyeglasses. If initial VA was less than 54 letters read correctly (<20/20 Snellen equivalent), refraction was performed. Visual impairment was categorized as presenting visual impairment (PVI), VA less than 20/40 Snellen equivalent (<39 letters read correctly) in the better eye using current glasses; correctable visual impairment (CVI), PVI correctable to 20/40 Snellen equivalent or better by subjective refraction; and noncorrectable visual impairment (NCVI), PVI correctable to worse than 20/40 Snellen equivalent in the better eye after subjective refraction.

STATISTICAL ANALYSIS

Statistical analyses were performed using SAS software version 9.13 (SAS Institute, Cary, North Carolina) and Mplus.28 Simple statistics included t tests for comparing means and χ2 tests for comparing proportions. Cox regression models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Multivariable-adjusted models included variables found significantly associated with mortality after age adjustment. These were history of acute myocardial infarction, stroke, angina and hypertension, current smoking, low BMI, cancer, diabetes, walking disability, home ownership, tertiary education, and self-rated health. Additional stratified analyses were conducted by age group (<75 years vs ≥75 years) to assess whether the impact from visual impairment was stronger on premature mortality of the relatively younger age group at baseline (<75 years). A P value of less than .05 was considered statistically significant.

Structural equation modeling pathway analysis23 was used to model the relationship between visual impairment, survival, and covariables found significantly associated with mortality by Cox regression. The SEM was fit using the Mplus28 statistical package with maximum likelihood and Monte Carlo integration methods. Standard errors were calculated using the delta method and hazard rates obtained from the coefficients by exponentiation. The covariables used in the model were history of acute myocardial infarction, stroke, angina and hypertension, current smoking, low BMI, cancer, diabetes, walking disability, home ownership, tertiary education, and self-rated health. The multiple potential pathways to mortality are shown in Figure 1. Each variable was adjusted for age and sex. The indirect effect of visual impairment was then calculated for each covariable by multiplying the effects of that covariable on mortality and visual impairment on the covariable. Models were simplified by removing indirect pathways for individual covariables that were not significant at the P value level of .10. The total indirect effect of visual impairment could then be calculated by summating the coefficients of the estimated indirect effects of each mediating variable and then converting to HRs. The total estimated effect of visual impairment on mortality was calculated by summating the coefficients of the indirect and direct effects and then converting to HRs.

RESULTS
STUDY POPULATION

As of December 31, 2005, 1273 BMES participants had died. Table 1 shows the distribution of known mortality risk markers in persons with and without visual impairment. Compared with those with normal vision, participants with NCVI at baseline were more likely to be female, older (age ≥75 years), and underweight. Persons with CVI were more likely to be older (age ≥75 years), but there was no difference in the proportions of women or persons with low BMI. They were more likely to have a self-reported history of angina, myocardial infarction, stroke, cancer, low self-rated health, and an observed difficulty in walking or use of walking aids. They were less likely to have tertiary education or to own their home. There were no significant differences in the proportions of current smokers or history of hypertension or diabetes between the groups with and without visual impairment.

ASSOCIATION BETWEEN VISUAL IMPAIRMENT AND MORTALITY

Table 2 shows all-cause mortality rates and mortality risk after age and sex adjustment, which were higher in persons with visual impairment compared with those with normal vision. This difference was greater in persons younger than 75 years compared with those 75 years and older at baseline. Persons younger than 75 years with visual impairment were at greater risk of PVI, CVI, and NCVI (PVI: HR, 1.69; 95% CI, 1.32-2.17; CVI: HR, 1.60; 95% CI, 1.23-2.09; NCVI: HR, 2.58; 95% CI, 1.42-4.69). The associations for all ages combined remained either significant or marginally significant after multivariable adjustment (Table 2), although they were substantially attenuated in magnitude (PVI: HR, 1.29; 95% CI, 1.09-1.52; CVI: HR, 1.26; 95% CI, 1.04-1.53; NCVI: HR, 1.35; 95% CI, 1.04-1.75). In persons younger than 75 years, only PVI and NCVI were statistically associated with mortality, while for those 75 years and older, only PVI was significantly associated with mortality, after multivariable adjustment.

Structural equation modeling pathway analysis confirmed that visual impairment influenced mortality by both direct and indirect pathways. Table 3 shows the HRs and 95% CIs using this model. The pattern was similar to the Cox regression models, except that the HRs were higher when estimated using SEM.

Of the risk markers investigated by SEM, only disability in walking represented a significant indirect pathway from visual impairment to mortality. Table 4 lists the HRs and 95% CIs for the indirect effects of disability in walking on mortality. Figures 1, 2, and 3 illustrate the detailed pathways that could explain the associations found between visual impairment and mortality. There was a significant indirect pathway from visual impairment to mortality via disability in walking, through poorer self-rated health, to mortality (Table 4) (Figures 1, 2, and 3). An indirect pathway through lower BMI bordered on significance for NCVI in all age groups (HR, 1.21; 95% CI, 0.99-1.48; P = .06) but was not significant for other forms of visual impairment or after age stratification (data not shown). There was no indirect pathway through any other mortality risk markers (eg, history of angina, acute myocardial infarction, stroke, hypertension, diabetes, cancer, or smoking). There were also no indirect pathways from visual impairment to mortality through lower home ownership or education levels. There was no indirect pathway from visual impairment to mortality through poorer self-rated health that was independent of disability in walking.

COMMENT

In agreement with previous reports,18 we observed that the presence of visual impairment predicted mortality in older persons. Using Cox regression models, PVI predicted mortality independent of age, sex, and the presence of known mortality risk markers. Using SEM analysis, we confirmed that visual impairment increased mortality via both direct and indirect pathways. The relationship between visual impairment and mortality was strongest for NCVI among persons with baseline ages younger than 75 years.

Compared with estimates from SEM, Cox regression appears to underestimate the effect of visual impairment on mortality by overcorrecting for intermediate variables that were associated with both visual impairment and mortality. To determine which indirect effects were important in predicting mortality for visually impaired persons, we modeled each covariate as a pathway to mortality (Table 4). Of the mortality risk markers we assessed, only disability in walking represented a significant indirect pathway. For this association, 2 possible pathways were identified; 1 pathway involved only disability in walking, and the second also involved an effect through low self-rated health (Table 4) (Figure 1). Because of reduced power through fewer events, we were unable to differentiate indirect effects via age. The age-related trend for mortality, however, suggests that in persons younger than 75 years at baseline, the pathway through self-rated health may be more important than the pathway directly from disability in walking to mortality (Table 4).

Weaker associations were consistently found in persons 75 years and older at baseline and are not unexpected. Many studies have reported no associations between visual impairment and mortality after correcting for covariates in older populations or in analyses that included all ages in 1 model.1,10,13 Similar findings are reported for cardiovascular risk factors where risk factors lose their predictive power after ages of 80 years and older.1,2931 One explanation is a ceiling effect, where mortality risk is already high in persons 75 years and older because of multiple mortality risk factors so that there is a limit to the additional contribution from visual impairment to mortality. Another explanation is “selective survival,” that is, persons genetically predisposed to die of causes related to visual impairment will do so at a relatively younger age, leaving those without this predisposition to survive into very old age.

We found that CVI also increased mortality risk via both direct and indirect pathways. Although the direct effect of CVI on mortality risk seems counterintuitive, the association may occur in both directions. The SEM provides information on the magnitude of the association but not its direction. There are several mechanisms that could explain the associations we found between visual impairment and mortality. Persons with various disabilities in walking may be less likely to see a doctor regularly or to have prescriptions for critical medications filled. They may be more socially isolated, have a poorer and relatively unvaried diet, and less able to seek urgent help when needed. They may also be less likely to exercise regularly, leading to lower cardiorespiratory reserve and greater risk of death during the stress of illness. Disabilities in walking are also associated with increasing risk of falls and fractures (eg, hip fractures),32 which may lead to an increased risk of death.

For the direct pathway, there are probably many unidentified covariates that were not accounted for. One example is poor diet, which is associated with age-related macular degeneration,3335 cataract,36,37 cancer,38,39 diabetes,40,41 and cardiovascular disease.38,39,42 We found some evidence of an indirect association through lower BMI to mortality for NCVI, which may support this speculation. The direct pathway may also be explained by some systemic processes that are common to both visual impairment and mortality, such as chronic inflammation.

To our knowledge, only 1 other study has used SEM to examine the pathways from visual impairment to mortality. In agreement with their findings, our study reported a small but significant indirect effect through disability.24 In contrast to our results, however, these authors reported that self-rated health was a significant indirect pathway that was independent of disability in walking. Other notable differences from this earlier study were that the indirect pathways had a lower magnitude than the direct pathway and the HRs were much lower in comparison with those estimated from our study sample. There were 2 main differences between this earlier study and ours that may explain this disagreement. First, while the earlier study was comparable with ours in many ways, there was no objective measure of VA. Visual function was assessed by self-reporting and proxy reporting of visual impairment by asking whether each person was blind or had difficulty seeing from one or both eyes. This could have introduced bias because persons with lower self-rated health and/or disability might have been more likely to overestimate visual impairment and vice versa. Second, the earlier study included persons 18 years and older in comparison with 49 years and older in our study.

The strengths of our study include its large population-based data set, with high participation and long follow-up period, standardized VA assessment, use of Australian National Death Index mortality and causes of death data, and detailed data on the health and functional status of participants. Limitations include the possibility that not all potential mortality markers were included in the model, such as exercise, diet, and nutrition variables. Also, the relatively low number of persons with visual impairment after age stratification limits our ability to detect weak associations that could be significant. Limitations of SEM include the assumption that relationships between variables in the model are linear and that the directions of the arrows in the path model are assumed but cannot be proven.

In conclusion, this study reaffirms that visual impairment is associated with an increased risk of all-cause mortality. Analysis using SEM suggests both direct and indirect pathways for this relationship. Disability in walking may represent an important indirect pathway to mortality for persons with visual impairment, and adjusting for this factor in statistical analysis may overadjust for the indirect effect of visual impairment on mortality risk. The impact of visual impairment on mortality may in fact be greater than that reported from previous studies that have used traditional statistical models.

Back to top
Article Information

Correspondence: Jie Jin Wang, MMed, PhD, Centre for Vision Research, Department of Ophthalmology, University of Sydney, Westmead Hospital, Hawkesbury Road, Westmead, New South Wales, Australia 2145 (jiejin_wang@wmi.usyd.edu.au).

Submitted for Publication: January 12, 2009; final revision received June 5, 2009; accepted June 14, 2009.

Financial Disclosure: None reported.

Funding/Support: The Blue Mountains Eye Study was supported by National Health and Medical Research Council (Australia) grants 974159, 991407, 211069, and 262120.

References
1.
Cugati  SCumming  RGSmith  WBurlutsky  GMitchell  PWang  JJ Visual impairment, age-related macular degeneration, cataract, and long-term mortality: the Blue Mountains Eye Study. Arch Ophthalmol 2007;125 (7) 917- 924
PubMedArticle
2.
Freeman  EEEgleston  BLWest  SKBandeen-Roche  KRubin  G Visual acuity change and mortality in older adults. Invest Ophthalmol Vis Sci 2005;46 (11) 4040- 4045
PubMedArticle
3.
Lee  DJGomez-Marin  OLam  BLZheng  DD Visual acuity impairment and mortality in US adults. Arch Ophthalmol 2002;120 (11) 1544- 1550
PubMedArticle
4.
Klein  RKlein  BEMoss  SE Age-related eye disease and survival: the Beaver Dam Eye Study. Arch Ophthalmol 1995;113 (3) 333- 339
PubMedArticle
5.
McCarty  CANanjan  MBTaylor  HR Vision impairment predicts 5 year mortality. Br J Ophthalmol 2001;85 (3) 322- 326
PubMedArticle
6.
Clemons  TEKurinij  NSperduto  RDAREDS Research Group, Associations of mortality with ocular disorders and an intervention of high-dose antioxidants and zinc in the Age-Related Eye Disease Study: AREDS Report No. 13. Arch Ophthalmol 2004;122 (5) 716- 726
PubMedArticle
7.
Wang  JJMitchell  PSimpson  JMCumming  RGSmith  W Visual impairment, age-related cataract, and mortality. Arch Ophthalmol 2001;119 (8) 1186- 1190
PubMedArticle
8.
Knudtson  MDKlein  BEKlein  R Age-related eye disease, visual impairment, and survival: the Beaver Dam Eye Study. Arch Ophthalmol 2006;124 (2) 243- 249
PubMedArticle
9.
Lee  DJGomez-Marin  OLam  BLZheng  DD Visual impairment and unintentional injury mortality: the National Health Interview Survey 1986-1994. Am J Ophthalmol 2003;136 (6) 1152- 1154
PubMedArticle
10.
Kulmala  JEra  PParssinen  O  et al.  Lowered vision as a risk factor for injurious accidents in older people. Aging Clin Exp Res 2008;20 (1) 25- 30
PubMedArticle
11.
Kulmala  JEra  PTormakangas  TParssinen  ORantanen  THeikkinen  E Visual acuity and mortality in older people and factors on the pathway. Ophthalmic Epidemiol 2008;15 (2) 128- 134
PubMedArticle
12.
Evans  JRFletcher  AEWormald  RP Depression and anxiety in visually impaired older people. Ophthalmology 2007;114 (2) 283- 288
PubMedArticle
13.
Thiagarajan  MEvans  JRSmeeth  LWormald  RPFletcher  AE Cause-specific visual impairment and mortality: results from a population-based study of older people in the United Kingdom. Arch Ophthalmol 2005;123 (10) 1397- 1403
PubMedArticle
14.
Smith  WMitchell  PLeeder  SRWang  JJ Plasma fibrinogen levels, other cardiovascular risk factors, and age-related maculopathy: the Blue Mountains Eye Study. Arch Ophthalmol 1998;116 (5) 583- 587
PubMedArticle
15.
Klein  BEKlein  RLee  KECruickshanks  KJ Performance-based and self-assessed measures of visual function as related to history of falls, hip fractures, and measured gait time: the Beaver Dam Eye Study. Ophthalmology 1998;105 (1) 160- 164
PubMedArticle
16.
Wang  JJMitchell  PSmith  W Vision and low self-rated health: the Blue Mountains Eye Study. Invest Ophthalmol Vis Sci 2000;41 (1) 49- 54
PubMed
17.
Thompson  JRGibson  JMJagger  C The association between visual impairment and mortality in elderly people. Age Ageing 1989;18 (2) 83- 88
PubMedArticle
18.
Mooijaart  SPKoeijvoets  KMSijbrands  EJDaha  MRWestendorp  RG Complement Factor H polymorphism Y402H associates with inflammation, visual acuity, and cardiovascular mortality in the elderly population at large. Exp Gerontol 2007;42 (11) 1116- 1122
PubMedArticle
19.
Hennis  AWu  SYLi  XNemesure  BLeske  MCBarbados Eye Study Group, Lens opacities and mortality: the Barbados Eye Studies. Ophthalmology 2001;108 (3) 498- 504
PubMedArticle
20.
West  SKMunoz  BIstre  J  et al.  Mixed lens opacities and subsequent mortality. Arch Ophthalmol 2000;118 (3) 393- 397
PubMedArticle
21.
Ditlevsen  SChristensen  ULynch  JDamsgaard  MTKeiding  N The mediation proportion: a structural equation approach for estimating the proportion of exposure effect on outcome explained by an intermediate variable. Epidemiology 2005;16 (1) 114- 120
PubMedArticle
22.
Rothman  KJGreenland  S Modern Epidemiology. 2nd ed. Philadelphia, PA Lippincott-Raven1998;
23.
Bentler  PMStein  JA Structural equation models in medical research. Stat Methods Med Res 1992;1 (2) 159- 181
PubMedArticle
24.
Christ  SLLee  DJLam  BLZheng  DDArheart  KL Assessment of the effect of visual impairment on mortality through multiple health pathways: structural equation modeling. Invest Ophthalmol Vis Sci 2008;49 (8) 3318- 3323
PubMedArticle
25.
Attebo  KMitchell  PSmith  W Visual acuity and the causes of visual loss in Australia: the Blue Mountains Eye Study. Ophthalmology 1996;103 (3) 357- 364
PubMedArticle
26.
Powers  JBall  JAdamson  LDobson  A Effectiveness of the National Death Index for establishing the vital status of older women in the Australian Longitudinal Study on Women's Health. Aust N Z J Public Health 2000;24 (5) 526- 528
PubMedArticle
27.
Whitworth  JAWorld Health Organization, International Society of Hypertension Writing Group, 2003 World Health Organization (WHO)/International Society of Hypertension (ISH) statement on management of hypertension. J Hypertens 2003;21 (11) 1983- 1992
PubMedArticle
28.
Muthen  LKMuthen  BO Mplus User's Guide. 5th ed. Los Angeles, CA Muthen & Muthen2007;
29.
Weverling-Rijnsburger  AWBlauw  GJLagaay  AMKnook  DLMeinders  AEWestendorp  RG Total cholesterol and risk of mortality in the oldest old. Lancet 1997;350 (9085) 1119- 1123
PubMedArticle
30.
Langer  RDGaniats  TGBarrett-Connor  E Paradoxical survival of elderly men with high blood pressure. BMJ 1989;298 (6684) 1356- 1357
PubMedArticle
31.
van Bemmel  TGussekloo  JWestendorp  RGBlauw  GJ In a population-based prospective study, no association between high blood pressure and mortality after age 85 years. J Hypertens 2006;24 (2) 287- 292
PubMedArticle
32.
Ivers  RQCumming  RGMitchell  PSimpson  JMPeduto  AJ Visual risk factors for hip fracture in older people. J Am Geriatr Soc 2003;51 (3) 356- 363
PubMedArticle
33.
Chong  EWSinclair  AJGuymer  RH Facts on fats. Clin Experiment Ophthalmol 2006;34 (5) 464- 471
PubMedArticle
34.
Bone  RALandrum  JTGuerra  LHRuiz  CA Lutein and zeaxanthin dietary supplements raise macular pigment density and serum concentrations of these carotenoids in humans. J Nutr 2003;133 (4) 992- 998
PubMed
35.
Seddon  JMCote  JRosner  B Progression of age-related macular degeneration: association with dietary fat, transunsaturated fat, nuts, and fish intake. Arch Ophthalmol 2003;121 (12) 1728- 1737
PubMedArticle
36.
Chiu  CJTaylor  A Nutritional antioxidants and age-related cataract and maculopathy. Exp Eye Res 2007;84 (2) 229- 245
PubMedArticle
37.
Gale  CRHall  NFPhillips  DIMartyn  CN Plasma antioxidant vitamins and carotenoids and age-related cataract. Ophthalmology 2001;108 (11) 1992- 1998
PubMedArticle
38.
Ames  BNShigenaga  MKHagen  TM Oxidants, antioxidants, and the degenerative diseases of aging. Proc Natl Acad Sci U S A 1993;90 (17) 7915- 7922
PubMedArticle
39.
Heidemann  CSchulze  MBFranco  OHvan Dam  RMMantzoros  CSHu  FB Dietary patterns and risk of mortality from cardiovascular disease, cancer, and all causes in a prospective cohort of women. Circulation 2008;118 (3) 230- 237
PubMedArticle
40.
van Dam  RMWillett  WCRimm  EBStampfer  MJHu  FB Dietary fat and meat intake in relation to risk of type 2 diabetes in men. Diabetes Care 2002;25 (3) 417- 424
PubMedArticle
41.
van Dam  RMRimm  EBWillett  WCStampfer  MJHu  FB Dietary patterns and risk for type 2 diabetes mellitus in U.S. men. Ann Intern Med 2002;136 (3) 201- 209
PubMedArticle
42.
Giugliano  DCeriello  AEsposito  K The effects of diet on inflammation: emphasis on the metabolic syndrome. J Am Coll Cardiol 2006;48 (4) 677- 685
PubMedArticle
×