Twenty seven studies that reported 32 comparisons met inclusion criteria (n=28,032, range 92 to 11,734); two were longitudinal and 25 cross-sectional in design. Of the 32 comparisons: 19 used a representative patient sample; eight reported using a reliable reference standard; 29 avoided partial and/or differential verification bias; 21 avoided incorporation bias; 15 blinded interpreters of index test and/or reference standard results; 20 avoided clinical review bias; 16 reported uninterpretable results; and 17 reported on withdrawals.
When subsequent measures were used as a reference standard (one study), sensitivity was 86% (95% CI 64 to 97) and specificity was 99% (95% CI 98 to 99). When intensive screening was compared to a reference standard of less intensive screening (one study that evaluated several different tests), sensitivity ranged from 0% (95% CI 0 to 20) to 64% (95% CI 44 to 81) and specificity from 84% (95% CI 81 to 86) to 100% (95% CI 99 to 100).
When tests conducted by a nurse, screening aid or lay person were compared to those conducted by an ophthalmologist or orthoptologist (six studies), sensitivity ranged from 9% (95% CI 4 to 18) to 83% (95% CI 65 to 94) and specificity from 68% (95% CI 52 to 82) to 100% (95% CI 99 to 100).
When compared to a reference standard test: visual acuity tests (eight studies) had sensitivities from 9% (95% CI 1 to 28) to 100% (95% CI 3 to 100) and specificities from 8% (95% CI 4 to 15) to 100% (95% CI 97 to 100); photoscreeners (nine studies) had sensitivities from 46% (95% CI 30 to 63) to 91% (95% CI 82 to 96) and specificities from 53% (95% CI 42 to 65) to 100% (95% CI 99 to 100); and stereo acuity tests (seven studies) had sensitivities from 14% (95% CI 0 to 58) to 100% (95% CI 3 to 100) and specificities from 76% (95% CI 68 to 82) to 99% (95% CI 97 to 100).