Using scalar blur analyses to examine the effect of test criteria on performance of vision screening tests

Abstract

Purpose : School vision screening programmes play a key role in the early detection of uncorrected refractive error and other ocular conditions in children. The efficacy of vision screening depends on the specific test protocol and criteria used. Here, we used scalar blur values (SBV; representing blur attributed to refractive error; Blendowske, 2015), derived from childhood refractive error data, to simulate the effect of modifying cut-off criteria on the accuracy of visual acuity (VA) in vision screening tests

Methods : Calculations were based on eye examination data from 255 primary and high school students aged 4-18 years in rural Queensland, Australia. Data included uncorrected monocular and binocular logMAR distance and near VA, distance VA with +1.50DS blur, and with cycloplegic autorefraction. The sphero-cylindrical autorefractor results (range of spherical equivalents -3.63D to +8.13D) were converted to SBV and used to classify the presence of refractive error. ROC curves were generated for a range of SBV and screening test performance simulated by calculating the effect of different VA cut-off values. The area under the curve (AUC) provided an overall diagnostic accuracy measure, a conservative 0.9 criterion was applied.

Results : Sensitivity and specificity depended on both VA cut-off values, and the magnitude of refractive error (SBV) considered significant. The 0.9 AUC criterion was achieved for myopes at SBV values of 1.25D and 1.50D for monocular and binocular measures respectively. Applying a cut-off value of 0.3 logMAR with these SBV yielded a sensitivity of 100% and specificity of 85% for myopes tested monocularly. For hyperopes, the same criterion was achieved at SBV values of 3.50D (monocular and binocular). Unlike near VA which did not improve SBV, adding a +1.50 blur test for hyperopes substantially reduced these values: 2.00D (monocular) and 2.25D (binocular).

Conclusions : The SBV approach allows the effect of different screening tests and criteria on sensitivity and specificity to be simulated and quantified. Distance VA measured in isolation was effective at detecting approximately 1D of myopia, compared to 3.5D of hyperopia. The inclusion of a +1.50DS blur test in children with good distance acuity substantially improved hyperopia detection. Using the SBV approach on larger datasets will help determine the most effective test protocol for school vision screening.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.