[Test-retest reliability of a web-based colour vision test]

Screening tests should be simple, with high acceptance, low costs and offer a high sensitivity and specificity as well as high reliability. The test-retest reliability of a web-based colour vision test was evaluated.

The results of 386 subjects, who had performed the web-based colour vision test with 25 colour plates twice (available at http://www.farbsehtest.de), were analysed from an mySQL data base. Retests were performed spontaneously, normally on the same day, or through an email newsletter with a request to repeat the test. Test-retest reliability was assessed by calculating the kappa coefficient and the corresponding confidence interval. The McNemar test was used as a complementary procedure to detect the presence of systematic differences between test results.

The kappa coefficient rises with increasing time interval between the first test performance and the test repetition. Thus it was 0.24 when the test was repeated within one hour (confidence interval 0.07 - 0.41, n = 63), it increases to 0.83 (confidence interval 0.75 - 0.92, n = 240) in the subjects who repeated the test at the earliest after two years. While the proportion of failed tests (< 3 mistakes) in the second test performance markedly decreases when the test is repeated in the short term within one hour (p < 0.001, McNemar), we were unable to identify any systematic differences (p = 0.602) for a time interval of more than one day.

The web-based colour vision test is a reliable screening test with good reproducible results as well under uncontrolled examination conditions. Subjects must perform the test with serious intent and should not aim at a positive test result by training or manipulation because of self-delusion. Test results are reliable if the time interval between test and retest is more than one day.