On the statistical reliability of letter-chart visual acuity measurements

Abstract

Purpose: To determine boundary values of test-retest reliability and sensitivity to acuity change that are unlikely to be exceeded in any clinical situation, for a popular visual acuity chart (Lighthouse/ETDRS) with three scoring methods; to discuss general methodological issues associated with statistical accuracy of optotype chart testing; and to link measures of test reliability to measures of sensitivity to change.

Methods: Five highly practiced subjects were tested using a computer-controlled acuity testing system in a procedure designed to reduce measurement error. Subjects read the computerized chart 156 times, yielding a sample of 78 test-retest comparisons.

Results: Under conditions likely to minimize variability, visual acuity may, with 95% confidence, be ascertained only within +/- 0.1 log units, using this chart with the recommended letter-by-letter scoring. Detecting a significant change in visual acuity requires about +/- 0.14 log units for the same degree of confidence.

Conclusions: These measurements may be viewed as approaching the upper limit of reliability of this letter chart. Reliability probably is considerably less in typical usage.