Sources of parameter uncertainty in the projections of the cost of vision problems include uncertainty in the population projections and uncertainty in the cost estimates derived from the Cost of Vision report. The Cost of Vision report included a probabilistic sensitivity analysis based on simultaneously sampling over 70 underlying parameters from their prior distributions and recording the resulting cost estimates, and then repeating this process 10,000 times. The distribution of cost estimates over these 10,000 replications equates to the full spectrum of possible results, with the 2.5th percentile and 97.5th percentile cost values representing the 95 percent credible interval of costs. A credible interval is essentially the Bayesian statistical equivalent of a confidence interval.

The projected credible intervals of total costs by year are shown in **Figure 5.2** and **Table 5.2** below. As with the prevalence projections, the increased range of uncertainty is associated with both higher costs and later years of projection. In 2014, the baseline real cost projection is $145 billion, while the lower bound of the 95% credible interval is $117 billion, the upper bound is $182 billion, equating to a range of $65 billion. By 2032, the lower and upper bound cost projections are $199 billion and $309 billion, yielding a range of $110 billion. By 2050, the lower and upper bound cost projections reach $300 billion and 467 billion; a range of $167 billion.

#### Figure 5.2. 95% Confidence Intervals of Cost Projections