This study is subject to a number of limitations and assumptions. Results are subject to uncertainty due to data limitations; we estimate a 95% credible interval for the economic burden to be $111 billion to $174 billion. We do not include a monetary value for health utility or disability loss in our primary results due to controversy over their inclusion in economic burden estimates, whether they include productivity losses, and the valuation of quality or disability adjusted life years.
We believe the two greatest limitations to this analysis are uncertainty in the prevalence of visual impairment and blindness, and the quality of self-reported diagnosis data. The sensitivity analysis identified the prevalence of vision loss as the primary cause of uncertainty in results, due almost entirely to its impact on productivity losses, and to a lesser extent due to its impact on long-term care cost and informal care cost estimates. The prevalence of vision loss has no effect on our estimation of medical costs or other direct and indirect costs which are based on other data sources. Epidemiological evidence of prevalence is available only for the population aged 40 and older, but due to sample size limitations is still subject to substantial uncertainty.[2, 3]
The use of NHANES data to provide prevalence estimates for the population aged 12-39 may introduce bias as this data relies on autorefractors to provide estimates of best-corrected near distance acuity, and contrast sensitivity and visual field are not assessed among participants in this age group. In addition, NHANES does not assess acuity among participants younger than age 12. We imputed prevalence in this age group based on the incidence of blindness reported in the United Kingdom and the prevalence among older children in NHANES data; which may introduce bias and we expect this may underestimate the prevalence of visual impairment at very young ages. Using these disparate sources of data for prevalence for different age groups leads to a decrease in estimated visual impairment prevalence from ages 18-39 to ages 40-64. This counter-intuitive finding appears to be a failing of this approach; however this age pattern can be seen in NHANES data alone when comparing the prevalence from 18-39 to 40-64.
The use of MEPS data for the estimation of medical costs has advantages and disadvantages versus an alternative of claims costs. Claims costs can provide much larger samples of diagnosed patients providing narrower confidence intervals of results. However, claims costs will not capture many costs attributable to eye disorders and vision loss due to poor diagnostic coding, indirect links between visual status and other health conditions, and the fact that available claims databases do not capture significant sources of payments for eye care services. The structure of MEPS relies on individuals to report health conditions, past diagnoses, and medical utilization. MEPS then confirms utilization and expenditures by surveying the medical providers of the individual respondents, and assigns 3-digit ICD-9 diagnosis codes to respondents on the basis of available information. This approach will underreport the true prevalence of disorders. The use of 3-digit ICD-9 codes precludes identification of disorders defined at the 4th or 5th digit, which include important diseases of diabetic retinopathy, age related macular degeneration and amblyopia, inhibiting the estimation of cost for these specific disorders. Due to the structure of MEPS, the costs for optometry visits and medical vision aids are self-reported by MEPS respondents and are not verified by providers. This structure eases the estimation of these costs, but the failure to verify these expenditures may lead to underreporting of costs due to poor recall or patients possibly reporting copayments as the total cost.
DALY losses are similarly sensitive to the prevalence of vision loss. The methodology of assessing quality of life losses based on self-reported quality of life among respondents reporting low vision in MEPS data used by the previous estimate cannot be replicated in the younger population. We used an alternative approach where we applied published disability loss weights to the prevalent populations mildly impaired, moderately impaired and blind to estimate disability adjusted life year (DALY) costs of low vision.
We found no data on the relative demand for assistive living devices or informal care for children due to vision loss in the United States. We assume the relative impact on demand due to vision loss in the United States is identical to rates observed in Europe, which might introduce bias. We do not include the cost of vision screening other than school and preschool screening and eye examinations, such as acuity chart screening in annual physicals or child well-checks.
Finally, we do not include the monetized value of quality of life or disability losses in our primary results because of limitations and uncertainty in the utility loss associated with vision loss, the monetary value of a QALY or DALY, and controversy over their inclusion in economic burden studies.