Test-Retest Reliability of the Coordinate Response Measure in Adults with Normal Hearing or Cochlear Implants.

Abstract

Background: Speech perception in noise is especially challenging for cochlear implant (CI) recipients; thus, speech in noise tests are used to clinically evaluate functional hearing with CIs. The coordinate response measure (CRM) corpus can be utilized in an adaptive speech perception test with competing speakers as the masker. Determining the critical difference for CRM thresholds can enable it to be used to evaluate changes in CI outcomes for clinical and research purposes. If a change in CRM exceeds the critical difference, then this would indicate significant improvement or decrement in speech perception. Additionally, this information provides figures for power calculations that could be used for planning studies and clinical trials [Bland JM: An Introduction to Medical Statistics, 2000].

Objectives: This study determined the test-retest reliability of the CRM for adults with normal hearing (NH) and adults with CIs. The replicability, variability, and repeatability of the CRM were evaluated for the two groups separately.

Method: Thirty-three NH adults and thirteen adult CI recipients were recruited and tested with the CRM twice, 1 month apart. The CI group was tested with two talkers only, while the NH group was tested with seven talkers as well as two talkers.

Results: CRM had better replicability, repeatability and lower variability for the CI adults compared to NH adults. The critical difference (at p < 0.05) in the two-talker CRM speech reception thresholds (SRTs) among CI users was >5.2 dB, and it was >6.2 dB for the NH if an individual were to be tested under two different conditions. The critical difference (at p < 0.05) in the seven-talker CRM SRT was >6.49. The Mann-Whitney U test showed that CI recipients' CRM scores' variance (Mdn = -0.94) was significantly less than the NH group's (Mdn = 2.2) (U = 54, p < 0.0001). Although the NH had significantly better SRTs in the two-talker condition than in the seven-talker condition (t = -20.29, df = 65, p < 0.0001), the Wilcoxon signed ranks test showed no significant difference between the CRM scores' variance in the two conditions (Z = -1, N = 33, p = 0.08).

Conclusions: NH adults had significantly lower CRM SRTs than the CI recipients; t (31.16) = -23.91, p < 0.001. CRM had better replicability, stability and lower variability for the CI adults compared to NH adults.