Longitudinal Incidence of Adverse Outcomes of Age-Related Macular Degeneration FREE

Of conditions affecting function in elderly individuals, vision loss is surpassed only by arthritis and heart disease.^{1 - 2} Age-related macular degeneration (AMD) is the leading cause of severe visual impairment and blindness in the United States.^{3 - 6} Its onset increases with age, with an estimated 5-year incidence rate of up to 33% among Medicare beneficiaries aged 65 years or older.^{3 ,7}

Age-related macular degeneration has been linked to increased risk of depression, falls, fractures, and disability.^{8 - 14} In 1 study of patients with advanced AMD, 79% of individuals had at least 1 comorbid condition, most frequently hypertension and depression.⁸ Elderly persons with neovascular AMD had increased odds of hypertension, hypercholesterolemia, emphysema, chronic obstructive pulmonary disease, atherosclerosis, arthritis, and coronary heart disease.¹⁵

Our study evaluated cumulative incidence of visual and functional outcomes for 10 years (1994-2004) in a cohort of Medicare beneficiaries aged 68 years or older with AMD diagnosed in 1994 relative to persons without this diagnosis during the same period. We also compared prevalence of common general health conditions for persons with and without diagnosed AMD.

We used the 5% sample of Medicare claims data for information on diagnoses, types, and dates of service. These data were merged with Medicare denominator files for information on enrollment dates in Medicare fee-for-service insurance, death, and beneficiary demographic characteristics. Data were linked by a unique identifier, allowing longitudinal, person-specific analysis from 1991 to 2004.

Individuals were classified as having AMD if they had 1 or more Medicare claims with an AMD diagnosis (eTable, available at http://archophthalmol.com). We identified all individuals whose AMD diagnosis occurred in 1994 (N = 37 838) (Figure 1). To ensure that this was likely the first AMD diagnosis, we examined all claims for 3 years before the 1994 diagnosis and excluded beneficiaries with an AMD diagnosis during this period. Because persons who were younger than 68 years in 1994 did not have 3 years of prior claims data to examine, we excluded them (n = 4624) as well as individuals enrolled in a Medicare risk plan (health maintenance organization [HMO]) or who resided outside the United States for 12 months or longer in the 3 years before the 1994 AMD diagnosis (n = 508). These exclusions yielded an AMD analysis sample of 32 702 persons (AMD group).

We also selected a control group of individuals without AMD from 3.6 million beneficiaries aged 68 years or older. To ensure routine eye surveillance in the control group, only individuals with 1 or more visits to an ophthalmologist or optometrist or with a 65091-68899 Current Procedural Terminology code on a Medicare claim in 1994 and 1 or more such encounters during the 10-year follow-up were included in the control group. After the HMO and US residence restrictions were applied and participants were matched to the AMD group on age, sex, and race/ethnicity, the control group consisted of 32 702 persons.

Baseline was defined as the 1994 date on which AMD was first diagnosed in the beneficiary or as July 1, 1994, in the control group. Once included, beneficiaries were followed up until December 31, 2004, or until censoring, which occurred at the beneficiary's death, when the person withdrew from Medicare fee-for-service benefits for an HMO for 6 months or longer, or when the person resided outside of the United States for 12 months or longer.

Using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis and place of service codes from claims data, we identified outcome measures for vision loss and blindness, hip fracture, depression, and admission to a nursing home. We used vision loss, blindness, and hip fracture ICD-9-CM codes from a study by Javitt et al.¹⁶ Our categories of severe vision loss and blindness captured beneficiaries with visual acuities of 20/200 or less and 20/400 or less, respectively. Admission to a nursing home was ascertained from the place of service code (code 31) in the claims data. We assumed that persons with an encounter covered by Medicare Part B in a nursing home resided in such facilities at the time.

Because depression, hip fracture, and nursing home stays often reflect conditions other than AMD, we used the Charlson Comorbidity Index (CCI), which is primarily based on ICD-9-CM codes,¹⁷ to control for general health at baseline (eTable). Each category has an associated weight used to calculate the CCI score. In addition to using the total score, we separately evaluated each morbidity category used to construct the CCI, except for human immunodeficiency virus, to observe general health patterns in the 2 groups.

Using SAS 9.0 (SAS Institute Inc, Cary, North Carolina), we calculated cumulative incidence of primary outcomes using time-to-event analysis and compared prevalence of common comorbid conditions at years 1, 5, and 10 in persons with and without an AMD diagnosis during 1994-2004. Cumulative incidence was calculated for persons without the outcome at baseline. All analyses controlled for age, sex, and race according to the study design.

Adverse outcomes were identified at the first service date on which the study diagnosis or place of service first appeared and remained until death or the end of the observational period. Cox proportional hazards models were used for unadjusted and adjusted time-to-event (survival) analyses of the primary outcomes. A negative log survivor or cumulative hazard function was used to estimate cumulative incidence and hazard ratios.¹⁸ Time-to-event analysis allowed us to compare the AMD and control groups while adjusting for censored data and covariates over time. Numbers of events, censored observations, cumulative event frequencies, and number of subjects at risk were calculated per year.

Analyses of blindness and severe and moderate vision loss included covariates for glaucoma, cataracts, background and proliferative diabetic retinopathy, macular edema, and CCI score. The eye diseases were based on diagnoses recorded on claims in the 3 years before AMD diagnosis and the CCI score on diagnoses used to construct the CCI for the year prior to the baseline date (a day in 1994 for the AMD group and July 1, 1994, for the control group). Covariates in the analysis of hip fracture and depression were CCI score, blindness, severe and moderate vision loss, and osteoporosis (only for hip fracture), based on diagnostic information during the 3 years before AMD diagnosis for all variables except CCI score. The analysis of nursing home placement only controlled for baseline CCI score.

To gauge the contribution of AMD to population health, we calculated cumulative incidence. This is the difference in the number of individuals with a particular diagnosed complication, eg, hip fracture, between (1) all Medicare beneficiaries aged 68 years and older with fee-for-service Medicare who had AMD first diagnosed in 1994 and were enrolled in Medicare fee-for-service benefits in 2004, and (2) individuals with the same characteristics except diagnoses of AMD from 1994 to 2004. Ten-year population-level cumulative incidence estimates were calculated as the number of individuals who survived our 10-year analysis without being censored multiplied by the 10-year cumulative incidence rates, which was then multiplied by 20, given our use of the 5% Medicare sample.

For the analysis of common comorbid conditions, annual prevalence rates were calculated for 1994-2004. Logistic regression was used to calculate odds ratios (ORs) and 95% confidence intervals (CIs) for differences in prevalence at the 1-, 5-, and 10-year follow-ups. This analysis did not contain any covariates except a binary variable for AMD vs control group.

By design, demographic characteristics were the same for the 2 groups (Table 1). Individuals in the AMD group were followed up for an average of 7.48 years compared with 7.74 years in the control group (P < .001). Persons in the AMD group developed all study visual outcomes at higher rates than those without AMD (Figure 2). Cumulative incidence of blindness (3.2% vs 1.2%) and severe (5.4% vs 1.3%) and moderate (6.0% vs 3.6%) vision loss was higher in the AMD group than in the control group after accounting for preexisting eye diagnoses and CCI score at baseline. Adjusted hazard ratios (HRs) were 2.34 for developing blindness (95% CI, 2.05-2.67), 3.70 for severe vision loss (95% CI, 3.28-4.19), and 1.69 for moderate vision loss (95% CI, 1.55-1.85) in individuals with AMD compared with controls (Figure 3). In multivariate analysis, preexisting diagnoses of glaucoma (HR, 2.37; 95% CI, 2.08-2.70) and proliferative diabetic retinopathy (HR, 2.56; 95% CI, 1.64-3.99) were most strongly associated with onset of blindness; having cataracts was protective (HR, 0.84; 95% CI, 0.73-0.95). Among all covariate eye conditions, AMD had the strongest association with blindness and onset of severe and moderate vision loss. Within the control group, preexisting diagnosis of glaucoma was most strongly associated with blindness onset (HR, 3.60; 95% CI, 2.90-4.47).

Persons with AMD experienced higher rates of all study functional outcomes following the AMD diagnosis (Figure 2). Significantly more persons with AMD experienced a hip fracture than controls (P < .001). Cumulative incidence of hip fracture in the AMD group was 22.0% compared with 19.9% in the control group. The 10-year population-level cumulative incidence for hip fractures was 5230 (58 665 persons with AMD who were aged 68 years or older vs 53 435 controls). Controlling for baseline values of CCI score, blindness, severe and moderate vision loss, and osteoporosis, we found that the adjusted HR was 1.09 (95% CI, 1.04-1.14). The 10-year cumulative incidence rate of depression was 40.3% in those with AMD vs 38.2% in controls (P < .001). This corresponds to a 10-year population-level cumulative incidence of 4843 (AMD group, n = 107 469; control group, n = 102 626). Unadjusted and adjusted HRs were 1.06 (95% CI, 1.02-1.09) and 1.06 (95% CI, 1.03-1.10), respectively (Figure 3). The probability of being admitted to a nursing home was higher in the AMD group (58.2% vs 55.5%; P < .001), with a 10-year population-level cumulative incidence of 6167 cases in the AMD and control groups (AMD group, n = 155 196; control group, n = 149 029). Adjusting for the CCI score, the HR of being admitted to a nursing home was 1.06 (95% CI, 1.03-1.09) in the AMD group compared with the control group.

Individuals with newly diagnosed AMD in 1994 had a higher prevalence (11 of 16) of comorbid general health conditions measured at 1, 5, and 10 years compared with controls (Table 2). Persons in the AMD group had a statistically significantly increased likelihood of having the following comorbid conditions diagnosed in year 1 (in descending order): moderate to severe liver disease (OR, 2.16; 95% CI, 1.35-3.43), renal disease (OR, 1.45; 95% CI, 1.27-1.65), congestive heart failure (OR, 1.29; 95% CI, 1.21-1.38), paraplegia and hemiplegia (OR, 1.29; 95% CI, 1.13-1.47), cerebrovascular disease (OR, 1.22; 95% CI, 1.14-1.30), dementia (OR, 1.21; 95% CI, 1.09-1.33), myocardial infarction (OR, 1.21; 95% CI, 1.11-1.32), cancer (OR, 1.20; 95% CI, 1.11-1.30), peptic ulcer disease (OR, 1.17; 95% CI, 1.05-1.30), metastatic cancer (OR, 1.14; 95% CI, 1.01-1.29), diabetes with complications (OR, 1.13; 95% CI, 1.00-1.26), rheumatic disease (OR, 1.13; 95% CI, 1.00-1.28), and peripheral vascular disease (OR, 1.10; 95% CI, 1.03-1.17). Incidences of these conditions tended to be higher in the first year following AMD diagnosis; by year 10, only congestive heart failure (OR, 1.13; 95% CI, 1.01-1.27) and paraplegia and hemiplegia (OR, 1.40; 95% CI, 1.10-1.78) were more likely to occur among persons with AMD, while those with AMD were less likely to have complications of diabetes mellitus diagnosed by year 10 (OR, 0.72; 95% CI, 0.58-0.90). At year 5, persons with AMD were less likely to have been diagnosed with mild liver disease (OR, 0.82; 95% CI, 0.70-0.95). There were no significant differences in the likelihood of chronic obstructive pulmonary disease or diabetes mellitus between the 2 groups.

This study provides a nationally representative, decade-long examination of visual and functional health complications in persons with AMD newly diagnosed in 1994. No previous epidemiological study of AMD has been based on a national longitudinal database with an observational period as long as a decade. Our analysis has confirmed earlier studies' findings that elderly individuals with newly diagnosed AMD experience higher cumulative incidence rates of blindness, severe and moderate vision loss, depression, hip fracture, and nursing home placement than those without AMD. Differences persisted even after controlling for relevant covariates present in the baseline year. Individuals with AMD had a higher prevalence of 11 of 16 general health conditions.

That AMD is the most common cause of poor vision and blindness in the United States is well known.^{3 - 6} We found that the risk of developing blindness was 2 times higher in individuals with AMD than controls after 10 years of follow-up, accounting for other comorbid eye conditions. Individuals with AMD also developed severe and moderate vision loss at rates 3.5 and 1.5 times higher than controls, respectively, in both unadjusted and adjusted analyses. Rates of blindness and vision loss in the control group were similar to those estimated for the US population.⁴

Klein et al¹⁹ reported vision loss in 8% of study subjects during 15 years in the Beaver Dam Eye Study. In a cohort aged 65 years or older at baseline, not limited to persons with AMD, 2% experienced severe vision loss after 15 years. We reported moderate vision loss of 6.0% and severe vision loss in 5.4% in persons with AMD during 10 years.

Previous studies have reported increased prevalence of depression in individuals with AMD (around 30%).^{8 - 9 ,20 - 22} We observed a cumulative incidence rate of 40.3% for depression during 10 years following AMD diagnosis. Because persons with a claim documenting depression during the look-back period were excluded from our analysis of cumulative incidence of depression, the 40.3% value applies to incident cases. The corresponding estimate for controls was 38.2%. Our multivariate survival analysis using a Cox proportional hazards model revealed a 6% higher probability of depression diagnosis among persons with AMD. Rovner et al^{23 - 24} attributed depression among individuals with AMD to the relinquishment of their normal or valued activities, but given nearly as high rates in the non-AMD population, other factors affecting depression also apply. For ophthalmologists, these findings highlight the AMD population's vulnerability to depression and the importance of ongoing depression screening.

Age-related macular degeneration has been associated with an increased risk of hip fracture.²⁵ Anastasopoulos et al²⁵ found an 11% increased risk of hip fractures in persons with atrophic AMD compared with those without AMD during a 4-year follow-up. The Rotterdam Study reported an increased risk of falling, with no change in risk of incident hip fracture in elderly individuals with visual field loss.²⁶ Javitt et al¹⁶ observed a greater than 1.5-fold increased odds of injury for persons with progressive vision loss to blindness compared with others. Using the same definition of hip fracture as Javitt et al, we found statistically significant increases in the probability of injury for individuals with AMD, even after controlling for CCI scores, diagnosis of osteoporosis, and vision loss at baseline.

Residence in a nursing home has been linked to visual impairment.^{27 - 28} Javitt and coauthors found that progression from normal vision to blindness was associated with a 2.5- to 3.5-fold increased odds of using such facilities. The Blue Mountains Eye Study found a 5-fold increase of bilateral and unilateral blindness in nursing home residents compared with local community residents. Age-related macular degeneration was the second most common cause of blindness, affecting 12% of the residents.²⁹ A follow-up study found a relative risk of nursing home admission of 1.8 among persons with visual impairment compared with controls.²⁷ Our results confirmed these findings of increased use of nursing home facilities,³⁰ but they are specific to AMD.

A cross-sectional analysis of Medicare beneficiaries found that 82% of beneficiaries had 1 or more chronic conditions and that 65% had 2 or more.³¹ Our results documented a significantly higher prevalence (11 of 16) of general health conditions in individuals with AMD than the non-AMD cohort. Previous studies have presented conflicting evidence of associations between AMD and several comorbid conditions, including diabetes, hypercholesterolemia, and acute myocardial infarction.³² It has been hypothesized that risk factors for development of AMD, including smoking,³³ age, obesity,^{34 - 35} inflammation, and nutrition, may be associated with increased prevalence of many of these conditions.

Strengths of our study include use of a large, longitudinal, nationally representative sample of elderly individuals in the United States. A 10-year follow-up of 2 demographically matched cohorts of individuals with and without a diagnosis of AMD allowed us to monitor the development of time-dependent changes in visual, functional, and general health conditions more reliably than with cross-sectional data, cross-study comparisons, or even with data from randomized, controlled trials, which typically have a shorter follow-up period. With our data, we could assess how care is actually delivered. We used survival or time-to-event analysis to evaluate the primary outcomes, which allowed us to compare time-to-event data between the AMD and control groups while adjusting for censoring and covariates.

We acknowledge several study limitations. First, diagnosis of AMD, primary outcomes, and general health conditions were based on Medicare claims data. While highly specific, claims data, designed for administrative use, are often insensitive, failing to identify all persons with the outcome of interest.^{36 - 37} While this deficiency mainly applies to blindness and vision loss, codes for blindness/low vision have been used as the best available measures of these outcomes as in a population in a previous article.¹⁶ To the extent that underascertainment is greater in the AMD than in the control group, our estimates of vision loss that was attributable to AMD would be understated and hence conservative. We have mitigated this effect by only including persons with at least 1 ophthalmologist or optometrist visit in 1994 and at least 1 more in the 10-year follow-up period. Since our estimates are based on AMD incidence, differences in rates of visual and functional outcomes may be expected to be lower than in a prevalence-based study for this reason as well. Second, we followed up individuals with AMD from first diagnosis as ascertained in claims, which typically differs from the date of the actual onset of a disease. Third, diagnostic information from claims data, which are designed for billing purposes, may be less accurate than that from clinical studies. Given their use in billing, one may assume a tendency for overdiagnosis; underdiagnosis may actually be more common.³⁸ Finally, persons enrolled in HMOs for extended periods were eliminated from our analysis. Individuals enrolled in HMOs are typically healthier than those in fee-for-service programs.³⁹

In summary, this nationally longitudinal study documents increased rates of visual and functional health complications that occurred within 10 years of an AMD diagnosis. These findings demonstrate that the health issues of the AMD population are multifaceted, especially when viewed in a 10-year period, highlighting the importance of a multidisciplinary, integrated approach to the care of elderly persons with an AMD diagnosis.

Correspondence: Frank A. Sloan, PhD, Duke University, Rubenstein Hall, Room 114, Box 90253, Durham, NC 27708 (fsloan@duke.edu).

Submitted for Publication: June 6, 2008; final revision received November 11, 2008; accepted November 24, 2008.

Author Contributions: Dr Sloan had full access to the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Financial Disclosure: None reported.

Funding/Support: This research was supported in part by grant 2R 37-AG-17473-05A1 from the National Institute on Aging.

Disclaimer: The sponsor had no role in the design or conduct of this study.

Previous Presentation: This study was presented at the Association for Research in Vision and Ophthalmology Annual Meeting; April 27, 2008-May 1, 2008; Ft Lauderdale, Florida.