A Comparison of different scoring terminations rules for visual acuity testing: from a computer simulation to a clinical study

Purpose: To compare four visual acuity (VA) scoring termination rules. Methods: A computer simulation generated 30,000 virtual patients who underwent 10 repetitions for each of four termination rules, on both the Snellen and ETDRS charts (2.4 million tests performed in total). Three termination rules focused on the smallest character row: all characters were correctly identified (100%), one character was incorrectly identified (one miss) and 50% or more of the characters were correctly identified (50%). The forth termination rule used a calculation in which each character, when correctly recognized, contributed a proportional increment (per-letter). Accuracy, test-retest variability (TRV) and test duration were measured. Next, a clinical study was conducted in which 254 subjects underwent three repetitions of the ETDRS VA test from 4 m, and VA scores for each of the four scoring termination rules were calculated. Results: In the Snellen simulation, the mean accuracy of the 100%, one miss, 50% and per-letter termination rules in decimal was 0.23 (-0.16 logMAR), 0.11 (-0.09 logMAR), 0.10 (-0.08 logMAR), and -0.08 (0.08 logMAR) respectively; while with the ETDRS simulation, the mean accuracy in decimal was 0.34 (-0.22 logMAR), 0.14 (-0.11 logMAR), 0.07 (-0.06 logMAR), and 0.07 (-0.05 logMAR), respectively. For the ETDRS simulation, the per-letter had the lowest TRV values and the longest test duration. In the clinical study (n = 254), the reproducibility of the 100%, one miss, 50% and per-letter was 0.50, 0.53, 0.17, 0.14, respectively. Conclusions: Clinical study and simulation data both suggest that the 100% and one-miss termination rules have higher TRVs, while the 50% and per-letter demonstrated much tighter, and rather close, TRV values.