Sources of parameter uncertainty in the projections of eye disease prevalence include uncertainty in the population projections and the current prevalence of eye disease. The US Census’ population projections are released as a middle series, as well as low and high series based on alternative assumptions. We assume the low and high series projections are analogous to the credible range of population projections.
The Vision Problems in the U.S. database does not report confidence intervals or standard errors for the prevalence estimates. However, this data is based on a series of studies released by the Eye Disease Prevalence Research Group (EDPRG) in 2004.[7-13] The EDPRG included most of the underlying studies included in the Vision Problems database, and reported 95% confidence intervals for prevalence estimates for each disease by age, sex and race. We applied the percent ranges associated with the EDPRG confidence intervals to the prevalence estimates reported in the Vision Problems database to generate approximations of confidence intervals for the current prevalence estimates. We believe that this is a conservative approach to estimating uncertainty in the current estimates because the Vision Problems database includes all of the data from the EDPRG plus additional studies, which results in larger sample sizes than were used to create the confidence intervals. Thus, we believe that it is likely that the Vision Problems prevalence estimates would have less uncertainty than we are applying based on the level of uncertainty observed in the earlier EDPRG estimates.
Figure 5.1 and Table 5.1 show the 95% confidence interval range of disease prevalence projections. Uncertainty is a function of both uncertainty in disease prevalence, which is static over time, and uncertainty in population projections, which increases over time. The impact of population uncertainty is much smaller than that of prevalence uncertainty, but nonetheless contributes to the range between the low and high prevalence projections increasing over time.
In terms of overall scale of uncertainty in the current prevalent population, cataract and diabetic retinopathy dwarf the other conditions due primarily to their high overall prevalence estimate. The 2014 cataract prevalence estimate ranges from 23.9 million to 27.4, a range of over 3.5 million. Diabetic retinopathy has a current range of uncertainty of over 2 million, while all other disorders are less than 1 million. However, in terms of percentage range of uncertainty, the prevalence of blindness is by far the most uncertain, with the difference between the high and low values equal to half the baseline prevalence estimate. Uncertainty in each disorder increases slightly over time.
Figure 5.1. 95% Confidence Intervals of Prevalence Projections