Table 1 Key efficacy variables used in the model.
| Variables | Point estimate (95% CI) |
|---|---|
| Prevalence variables | |
| Prevalence (normal cases) | 0.681 (0.670–0.692) |
| Prevalence (mild retinopathy) | 0.266 (0.255–0.276) |
| Prevalence (observable retinopathy/maculopathy) | 0.013 (0.011–0.016) |
| Prevalence (referable retinopathy/maculopathy) | 0.040 (0.035–0.045) |
| Technical failure rate | 0.082 (0.076–0.089) |
| Efficacy of level 1 manual grading | |
| Detection rate for technical failures | 0.969 (0.951–0.981) |
| Proportion of normal cases appropriately recalled | 0.920 (0.911–0.927) |
| Detection rate for mild retinopathy | 0.819 (0.800–0.837) |
| Detection rate for observable retinopathy/maculopathy | 1.000 (0.956–1.000) |
| Detection rate for referable retinopathy/maculopathy | 0.992 (0.971–0.998) |
| Efficacy of level 1 automated grading | |
| Detection rate for technical failures | 0.998 (0.990–1.000) |
| Proportion of normal cases appropriately recalled | 0.674 (0.660–0.688) |
| Detection rate for mild retinopathy | 0.859 (0.841–0.875) |
| Detection rate for observable retinopathy/maculopathy | 0.976 (0.917–0.993) |
| Detection rate for referable retinopathy/maculopathy | 0.980 (0.953–0.991) |
| Efficacy of level 2 manual grading | |
| Technical failures | |
| Detection rate for technical failures* | 0.819 (0.786–0.852) |
| Proportion of technical failures incorrectly referred to level 3 | 0.103 (0.077–0.128) |
| Proportion of technical failures incorrectly recalled at 12 months | 0.076 (0.054–0.099) |
| Proportion of technical failures incorrectly recalled at 6 months | 0.002 (0.000–0.010) |
| No retinopathy | |
| Proportion of normal cases appropriately recalled* | 0.938 (0.912–0.964) |
| Proportion of normal cases incorrectly referred to slit‐lamp | 0.056 (0.032–0.081) |
| Proportion of normal cases incorrectly referred to level 3 | 0.006 (0.002–0.021) |
| Proportion of normal cases incorrectly recalled at 6 months | 0.000 (0.000–0.011) |
| Mild retinopathy | |
| Detection rate for mild cases* | 0.955 (0.944–0.966) |
| Proportion of mild cases incorrectly referred to slit‐lamp | 0.007 (0.004–0.014) |
| Proportion of mild cases incorrectly referred to level 3 | 0.025 (0.016–0.033) |
| Proportion of mild cases incorrectly recalled at 6 months | 0.013 (0.007–0.019) |
| Observable retinopathy | |
| Detection rate for observable cases* | 0.607 (0.503–0.712) |
| Proportion of observable cases incorrectly referred to slit‐lamp | 0.000 (0.000–0.044) |
| Proportion of observable cases incorrectly referred to level 3 | 0.143 (0.068–0.218) |
| Proportion of observable cases incorrectly recalled at 12 months | 0.250 (0.157–0.343) |
| Referable retinopathy | |
| Detection rate for referable cases* | 0.885 (0.845–0.925) |
| Proportion of referable cases incorrectly referred to slit‐lamp | 0.000 (0.000–0.016) |
| Proportion of referable cases incorrectly recalled at 6 months | 0.045 (0.019–0.071) |
| Proportion of referable cases incorrectly recalled at 12 months | 0.070 (0.038–0.102) |
*Denotes appropriate level 2 grading outcomes for each category of retinopathy (beta distributions were applied to all efficacy parameters in probabilistic sensitivity analysis).