The functional hearing inventory: criterion-related validity and interrater reliability

The Functional Hearing Inventory (FHI), an observational instrument for functional hearing, provides information about how a deafblind child uses his/her residual hearing within his/her natural environment. This study obtained evidence of the validity and reliability of the FHI.

In particular, criterion-related validity for the FHI was investigated by correlating it with teachers' and parents' ratings of functional hearing, and the traditional measure of hearing, the audiogram. Interrater reliability for the FHI was studied through correlating the FHI ratings of deafblind subjects by two trained evaluators using point-by-point and consensus methods. The two raters included the researcher and one other rater who was trained by the researcher. The raters observed students in their natural settings and recorded the information on the FHI observation form. The subjects for this study were a purposeful sample of students between the ages of three and twenty-one who were reported on the Federal Deafblind Census.

There were 14 participants for whom there was complete information, comprising 6 females (43%) and 8 males (57%). The demographic section indicated that 57% of the participants were male, and 57% were Caucasian. The majority of the participants are in their teens with 21% being in the 7* grade. Over 42% of the participants had a primary disability of deafblindness and four of the participants had a secondary disability of either hearing or visual impairment.

Cohen' s kappa was used to measure agreement for criterion validity as well as to determine interrater reliability. Null hypotheses of no relationship between the FHI and teachers' ratings, and between the FHI and parents' ratings were rejected, with a moderate relationship in the former case (K = 0.46, p = 0.0043), and with a somewhat weaker relationship in the latter case (K = 0.22, p = 0.01 ). The null hypothesis between the FHI and the audiogram could not be rejected (K = 0.13, p = 0.26).

The null hypothesis for interrater reliability was rejected for environmental conditions/background noise, signal, and response levels. The respective kappas were 0.96 (p < 0.0000001), 0.85 (p < 0.0000001), and 0.81 (p < 0.0000001), all considered to be high levels of association.