Mean total and component annual health care costs per patient after index by vision loss cohort (n = 181 922). Total costs are not always equal to the sum of component costs because of rounding. Each component cost is rounded to the nearest integer, and the total cost is rounded to the nearest integer separately.
Frequency of events by vision loss cohort.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Bramley T, Peeples P, Walt JG, Juhasz M, Hansen JE. Impact of Vision Loss on Costs and Outcomes in Medicare Beneficiaries With Glaucoma. Arch Ophthalmol. 2008;126(6):849–856. doi:10.1001/archopht.126.6.849
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
To assess the impact of vision loss severity on costs and health outcomes among Medicare beneficiaries with glaucoma.
A retrospective cohort analysis was conducted using Medicare claims. Patients were stratified into 4 categories: no vision loss, moderate vision loss, severe vision loss, and blindness. Outcomes of interest were mean annual medical costs by category, component costs, and frequency of depression, falls and/or accidents, injury, femur fracture, and nursing home placement.
Multivariate regression analysis showed that patients with any degree of vision loss had 46.7% higher total costs compared with patients without vision loss. Mean total and component costs increased with onset and severity ($8157 for no vision loss to $18 670 for blindness). Patients with vision loss were significantly more likely to be placed in a nursing home (odds ratio = 2.18; 95% confidence interval, 2.06-2.31), develop depression (odds ratio = 1.63; 95% confidence interval, 1.54-1.73), fracture a femur (odds ratio = 1.67; 95% confidence interval, 1.53-2.83), or experience a fall or accident (odds ratio = 1.59; 95% confidence interval, 1.50-1.68) vs patients without vision loss.
Vision loss in glaucoma is costly, and costs increase with severity. There is significantly increased risk of nursing home admission, depression, falls and/or accidents, injury, or femur fracture with vision loss compared with no vision loss.
More than 3.3 million Americans had visual impairment or blindness in 2004; glaucoma, a chronic optic neuropathy, is a leading cause of vision impairment.1Loss of vision is feared more than human immunodeficiency virus and AIDS, cancer, stroke,2or death.3Visual impairment increases risks of falls and accidents,4,5hip fractures,6-13injury,4,14-17motor vehicle crashes,4,18,19nursing home placement,5,8,20and mortality.21,22Moreover, inadequate vision is related to functional status decline23-27and depression.28-32
The economic burden of all visual disorders to the United States was estimated at $35.4 billion, with a governmental budgetary impact of $13.7 billion (2005 US dollars).33These estimates do not include costs of associated comorbidities (depression, traumatic injury) or eye-related well care and were not stratified by vision loss cause.
Because glaucoma is responsible for approximately 3 of every 4 cases of visual impairment and is predominantly a disease of the elderly,34its cost impact is highly important to Medicare. Numerous glaucoma-specific burden of illness studies have been conducted.5,35-38However, some are limited by the patient population studied (eg, end-stage glaucoma only,37non-Medicare patients36,37), method of glaucoma severity staging,35and reliance on chart review methods35,37that measure prescribed—not consumed—resource use. Another study lacked visual loss severity stratification,37and there is substantial evidence that costs and adverse consequences increase with visual impairment severity.5,35,36,39-46Each of these considerations limits the applicability and generalizability of results.
A retrospective claims analysis conducted by Javitt et al5described an epidemiologically validated vision loss severity categorization using International Classification of Diseases, Ninth Revision, Clinical Modification(ICD-9-CM)47codes for vision loss. An estimated $2.14 billion in non–eye-related medical care costs was attributed to blindness and vision loss in 2003, and 27% to 41% of excess vision loss costs were due to increased risk of depression, injury, skilled nursing facility use, and long-term care facility admission. The study, although important and methodologically sound, included Medicare beneficiaries aged 69 years or older with vision loss from any cause, used data spanning 1999 to 2003, and did not show costs by resource use category.
A more current estimate of total and component costs is needed to better understand the full economic implications of glaucoma-specific vision loss to Medicare and to determine the impact of these costs by severity. With the recent emphasis by Medicare on glaucoma awareness and disease progression,48-50it is imperative to document the risk of associated adverse consequences (depression, injury, femur fracture, and nursing home admission) stratified by vision loss in a glaucoma population. Therefore, the objective of this research was to assess the impact of vision loss severity on costs and health outcomes among Medicare beneficiaries with glaucoma.
This retrospective cohort analysis was conducted with the frequently used51,52and nationally representative Medicare 5% random sample, which contains covered inpatient and outpatient fee-for-service claims for beneficiaries aged 65 years or older. Data elements are derived from discharge records and contain unique identifier-linked medical claims and charges for inpatient hospitalization, physician visits, injectable medications, and services related to laboratory, skilled nursing facility, ambulatory surgery, home health, and hospice care in addition to mortality records. The study database was developed using a limited data set with masked patient identifiers, and materials were handled in compliance with the Health Insurance Portability and Accountability Act of 1996.
Because the analytic file provided dates of service by quarter rather than specific day, inclusion criteria reflected quarter date stamping. Patients aged 65 years or older with a new diagnosis of glaucoma (index diagnosis, ICD-9-CMcode 365.x) between the third quarter in 1999 and the fourth quarter in 2004 and with continuous Medicare enrollment in the 2 quarters prior to and 5 quarters after index diagnosis were selected, as this approximated 1 year of follow-up. Individuals with health maintenance organization enrollment or who died prior to the end of the minimally required 5-quarter follow-up period were excluded.
Patients were stratified into 4 mutually exclusive categories based on their worst degree of vision loss during follow-up using 57 ICD-9-CMcodes previously described and validated by Javitt et al5: no vision loss, moderate vision loss (ICD-9-CMcodes 369.6-369.9), severe vision loss (ICD-9-CMcodes 369.1-369.4), and blindness (ICD-9-CMcodes 369.0-369.08).
The direct medical cost analysis was from the Medicare perspective and included the amount reimbursed to the provider, excluding premiums, deductibles, and co-pays. Inpatient, outpatient, rehabilitation, physician visit, office-administered pharmaceutical, laboratory, skilled nursing facility care (≤ 100 days), hospice, and home health care costs were included. Because Medicare did not cover outpatient prescription drugs at the time of study conduct, these costs were excluded. Costs were adjusted for inflation using the medical care Consumer Price Index and are presented in 2007 US dollars.53
The primary cost outcome variable was the adjusted mean total annual health care costs in the postindex period by vision loss category. Component costs—inpatient, outpatient, emergency, physician, and other—were compared between cohorts. Other outcomes included incident depression, falls and/or accidents, injury, femur fracture, and nursing home placement (ie, admission to a skilled nursing facility or long-term care facility) and were based on ICD-9-CMcodes (Table 1).
Risk assessments for nursing home placement and depression were conducted on the population at risk at the time of index vision loss diagnosis; therefore, only patients who did not have the event of interest at that point were included. For falls or accidents, injury, and femur fracture, patients with prior events were not excluded.
Comorbidities were assessed using the valid and reliable54-57Charlson comorbidity score with Deyo modification.58,59The calculated Deyo-Charlson Index score ranges from 0 to 29 with higher scores representing a higher burden of comorbidity, defined as the cumulative increased likelihood of 1-year mortality.
Univariate analyses of frequencies, medians, and means with standard deviations were performed to describe the study population. Statistical differences between vision loss cohorts were assessed using χ2tests of proportionality for categorical variables and analysis of variance for continuous variables.
Ordinary least squares regression models estimated and compared mean annual total and component health care costs between cohorts during the 5 quarters of follow-up from the index date. Five quarters approximated 1 year owing to the varying dates of initiation within the entry quarter. To produce the best fit, each model was adjusted for a subset of covariates found to be significantly associated with the costs. Covariates considered for analysis were age, sex, race, geographic region, presence of neurological disorders or inpatient claims in the preindex period, Deyo-Charlson Index score, sum of all existing diagnoses in the preindex period, and total costs in the preindex period. For patients with some degree of vision loss, the index date was reassigned as the date of first diagnostic evidence of the most severe vision loss. For those with no vision loss, the index date remained the first medical claim for glaucoma diagnosis. In the cost analyses, the mean overall cost per patient was first calculated for each group. When using statistical inference through modeling, cost variables were log transformed to adjust for the highly right-skewed distribution typical of cost data.60The reference group for statistical comparison was beneficiaries with no evidence of vision loss during follow-up.
Separate stepwise parsimonious binary logistic regression models were derived to assess differences in the risk of experiencing each secondary outcome event, controlling for baseline covariates that may have impacted risk between those with vision loss and those with no vision loss. Odds ratios (ORs) of their occurrence and 95% confidence intervals (CIs) were derived. Baseline covariates considered were age, sex, race, geographic region, preindex comorbid diseases (dementia, Alzheimer disease, Parkinson disease, depression), Deyo-Charlson Index score, unique count of disease states beyond those considered in the Deyo-Charlson Index, and number of preindex inpatient claims. Variables that were significant at P < .05 were retained in the models. Pvalues were considered significant when the 2-sided value was less than .05. All of the analyses were carried out using SAS version 9.1 statistical software (SAS Institute, Inc, Cary, North Carolina).
A total of 365 345 Medicare beneficiaries with glaucoma met the inclusion criteria. Excluded were 50.2% (n = 183 423) owing to the following: death prior to the end of the 5-quarter follow-up period (n = 15 846); health maintenance organization enrollment (n = 22 904); or lack of continuous eligibility or complete data availability during the preindex and postindex periods (n = 144 673). Thus, 181 922 patients with glaucoma were categorized into vision loss cohorts: the majority (92.1%) as having no vision loss, 4.3% as having moderate vision loss, 1.9% as having severe vision loss, and 1.7% as blind. Table 2shows patient characteristics of the study population.
The simple mean total and component health care costs increased with onset and degree of vision loss, ranging from $8157 for no vision loss to $18 670 for blindness (Figure 1). Onset of moderate vision loss was associated with a $5005 increase in costs compared with those with glaucoma but no vision loss ($13 162 vs $8157, respectively). The cost for patients with severe vision loss was $7155 more than that for patients without any degree of vision loss, whereas the cost for blind patients was $10 513 higher than that for patients without vision loss.
Results from the covariate-adjusted multivariate regression analysis showed that patients with any degree of vision loss had 46.7% higher total costs compared with patients without vision loss (P < .001) (Table 3).
Frequency of events invariably increased with greater vision loss for all of the investigated events (Figure 2). In blind patients, 25.3% were placed in nursing homes, 17.0% had incident depression, 15.5% experienced a fall and/or accident, 16.9% sustained injury, and 7.0% fractured a femur. Conversely, in patients with no vision loss, no single event exceeded 7.8% in the total population.
Patients with vision loss were significantly more likely to be placed in a nursing home (OR = 2.18; 95% CI, 2.06-2.31) and to develop depression (OR = 1.63; 95% CI, 1.54-1.73) compared with patients with no vision loss. Additionally, they were 67.4% more likely to fracture a femur (OR = 1.67; 95% CI, 1.53-1.83) and 58.6% more likely to experience a fall or accident (OR = 1.59; 95% CI, 1.50-1.68) vs the reference group. In addition, there was a 33.4% higher risk of injury associated with vision loss (OR = 1.33; 95% CI, 1.27-1.41) compared with no vision loss. The odds of an individual with vision loss experiencing selected adverse events during the postindex period compared with patients with no vision loss, adjusted for all of the significant predictor variables, are shown in Table 4.
The progressive nature of glaucoma and permanence of vision loss48make it an important disease to properly manage with early intervention. Results from this study showed that as visual field limitations increased, resource consumption, adverse consequences, and intensity of care also increased. Medicare beneficiaries with glaucoma who experienced any degree of ICD-9-CM–coded vision loss had significantly higher covariate-adjusted total health care costs than patients without vision loss. Both total and component costs progressively increased with escalating severity of vision loss.
Most economic analyses within glaucoma evaluated costs associated with glaucoma severity rather than vision loss or patient-rated impairment. Javitt et al5were the first to describe the vision loss categories used in our study, and their research estimated overall average annual non–eye-related costs in 2003 US dollars as $5954 to $7527 for no vision loss, $10 050 to $14 876 for moderate vision loss, $14 311 to $15 250 for severe vision loss, and $14 905 to $19 366 for blindness. In general, our average annual total (including eye-related care) costs were similar: $8157 for no vision loss; $13 162 for moderate vision loss; $15 312 for severe vision loss; and $18 670 for blindness (2007 US dollars). There are some important methodological differences between our study and the research by Javitt and colleagues. Of the patients in their study sample, 37% to 43% had a glaucoma diagnosis and 38% to 66% had macular degeneration compared with 100% of the patients with glaucoma in our study. Our analysis included all health care costs, whereas Javitt and colleagues segmented costs by eye-related and non–eye-related costs and also investigated incremental vision loss costs. We show mean costs from 1999 to 2005 for year 1 after diagnosis of initial vision loss or glaucoma (for the cohort with no vision loss); Javitt and colleagues derived annualized data during 4 years of follow-up with either stable or progressive vision loss (in 1999-2003).
Similar to our economic burden findings with vision loss severity, numerous studies have shown that glaucoma severity is associated with increased treatment costs.35,36,39-46Lee et al35found that patients with ocular hypertension or earliest-phase glaucoma had mean annual eye-related direct medical costs of $623 (2001 US dollars), rising to $2511 for those with end-stage disease (typically blindness), which were slightly higher than the eye-related costs in Medicare beneficiaries described by Javitt et al5that excluded prescription drugs. Kobelt-Nguyen et al44found that mean 2-year costs for glaucoma were $2188 (1995 US dollars) and increased with severity.
The primary driver of total cost was inpatient visits, increasing by severity (52.5%-69.4% of total costs). Physician visit costs ranged from 9.0% (blindness) to 23.1% (no vision loss) of total costs. Others42have found that component costs for patients with open-angle glaucoma were dominated by procedures (45% of total costs) and physician visits (43%). As severity increased, physician visit costs decreased to 38% and procedure costs increased to 50%. Outpatient prescription medications were not included in our study because they were not covered by Medicare at the time of the study. The advent of Medicare Part D prescription drug coverage will likely increase the total cost estimate, and other studies36,39,40have estimated medication component costs to be dependent on disease severity41and highly variable, ranging from 12%42to 61%.35
The adjusted risk of nursing home placement was 2.18 times higher for patients with any vision loss compared with individuals with no vision loss, confirming results from other studies.5,8,20Our results of a 33.4% increased risk of injury, 58.6% increased risk of falls and accidents, and 67.4% increased risk of femur fracture are also within the parameter risk estimates by other researchers.5,8Haymes et al4found that compared with control subjects, patients with glaucoma had more than a 3-fold higher risk of falls in the previous year. Colón-Emeric et al13found that glaucoma was significantly associated with hip fracture risk in ambulatory men (OR = 2.6; 95% CI, 1.0-6.2).
Studies show that approximately one-third of visually impaired older individuals report clinically significant depressive symptoms,28-30,61-63and rates are significantly higher than those in the healthy population of similar age and are similar to or greater than those in other medically ill elderly individuals.28-32Our analysis found a 63.0% increased risk of depression diagnosis, similar to the depression risk increases of 40% to 80% found by Javitt et al.5Numerous other studies found links between vision loss and depression, family dysfunction, and marital problems28-32and between depression and physical functioning.64
Several limitations should be considered when interpreting the results of this study. The data source was administrative claims and the design was retrospective; thus, causality cannot be ascribed. Although numerous other studies have shown similar trends with costs and adverse consequences, the possibility cannot be ruled out that factors other than visual loss severity were responsible for differences in outcomes.
The Medicare 5% analytic sample did not include outpatient medications or long-term care and may limit generalizability of the results to the non-Medicare population, to elderly individuals who do not receive their primary health care through Medicare, or to the post–Medicare Part D period. However, the Medicare 5% analytic database is a representative reflection of those individuals aged 65 years or older with glaucoma who receive medical care through Medicare coverage. Moreover, median glaucoma-associated costs have been shown to be similar for patients younger than 65 years compared with patients aged 65 years or older.36
Derived costs represented first-year postdiagnosis total health care cost, reflecting a snapshot in time rather than longitudinal costs, and were not specifically attributed to vision loss. Year 1 costs have been shown to be both lower35and higher36,37,39in patients with glaucoma compared with post–year 1 costs. Severity-associated costs may have been due to other diseases coexisting with visual loss deterioration, and differences in outcomes may have been affected if cohorts were not equally balanced with respect to these factors. To moderate this effect, the multivariate cost and risk statistical analyses controlled for age, sex, preindex health care utilization, and other confounders.
The ICD-9-CMcodes were used for both glaucoma diagnosis and vision loss, and they are subject to the limitations in coding variance and projection of administrative claims data to larger populations.65,66The ICD-9-CM code 365.0 may be used for patients with ocular hypertension or glaucoma suspects, and these patients typically have a much lower risk of current or future vision loss. As such, the cohort with no vision loss may not exclusively contain patients with confirmed glaucoma who have no vision loss, thus the estimated costs and risks for this category may be conservative. Although the vision loss categories used in our study have previously been described and validated against epidemiologic studies34by Javitt et al,5the possibility remains that the codes were not accurate reflections of either the presence of vision loss or vision loss gradations. Further, the severity categorizations are susceptible to individual clinician judgment, which introduces a degree of interrater reliability and variance.
The risk of blindness or significant, permanent loss of visual field in glaucoma is costly, and the total direct medical and component costs increase with vision loss severity. There is a significant increase in the risk of major adverse consequences such as nursing home admission, depression, falls and accidents, injury, or femur fracture with any degree of vision loss compared with no vision loss.
Correspondence:Thomas Bramley, PhD, Xcenda, 1528 Preston St, Salt Lake City, UT 84108 (firstname.lastname@example.org).
Submitted for Publication:August 23, 2007; final revision received December 20, 2007; accepted January 3, 2008.
Financial Disclosure:None reported.
Author Contributions:Dr Bramley 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.
Funding/Support:This work was supported by Allergan, Inc.
Additional Contributions:Elizabeth Davis, PhD, contributed to the development of the manuscript, and Laurie Kozbelt, MA, contributed to the preparation of the manuscript.
Create a personal account or sign in to: