Purchase this article with an account.
Susan Kelly, Yi Pang, Dana Richter, Calvin Vance, David McIntosh, Becky Yeung; Improving the Reliability of the CSV-1000 Test. Invest. Ophthalmol. Vis. Sci. 2012;53(14):4804.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Chart-based contrast sensitivity tests are easy to administer, take little time and are cost-effective. However, their test-retest reliability is generally low. We recently reported that the Vector Vision CSV-1000 has low test-re-test reliability as measured by the coefficient of repeatability (COR), Bland-Altman limits of agreement and the intraclass correlation coefficient (ICC). The high variability apparent in the test-retest scores resulted in part from the subjects’ willingness to continue to guess the location of the test patch after they no longer detected it. A procedure that reduced the guess rate would likely increase the reproducibility of the test. In the procedure described below we made minor changes in both the procedure as well as the scoring of the chart which produced significant improvements in the test’s reliability.
The CSV-1000 measures contrast required to detect 4 different spatial frequencies. For a given frequency the subject is asked to indicate the location of the test patch (top or bottom) or respond "blank" if no grating patch is detected. The contrast for each frequency varies between 20% and 0.8% in seven 0.15 log contrast steps and one 0.3 log contrast step. Subjects are tested in a descending order and the contrast threshold is defined as the lowest contrast patch the subject detects correctly.In the revised protocol we asked 78 subjects to indicate the location of the test patch (top or bottom) for each of the 8 contrast levels at each frequency (average age = 25.2 yr; visual acuity 20/40 or better). Subjects were not allowed to respond "blank". Subjects’ were tested on all 8 contrast levels for each frequency even if one of the target locations was judged incorrectly. After all 4 frequencies were tested in the standard descending order of contrast we went back to the 1st frequency and asked subjects to again indicate the location of the test patch but this time the task was ascending so that contrast varied from low to high. The contrast threshold was defined as the lowest contrast level that was detected correctly on both ascending and descending trials. This shifted the % correct needed for threshold from 50% up to 75%.
The reliability of the test was defined by the ICC. These coefficients were compared with those obtained from the standard protocol. The ICC values for the latter were 0.24, 0.57*, 0.07, and 0.31. The ICC values for the new protocol were: 0.38*, 0.52*, 0.62* and 0.51*. The asterisk indicates the coefficients are significant.
These results indicate that the ICC can be improved by changing the test protocol and scoring procedure. The new protocol increased test time by a full minute (from 2 to 3 min) but we suggest additional changes that will decrease testing time while preserving the improved ICC values.
This PDF is available to Subscribers Only