Health care utilization and adults who are deaf: Relationship with age at onset of deafness

Health Services Research, Feb, 2002 by Steven Barnett, Peter Franks

Limitations

The categories used in this study are likely to result in some misclassification bias. For example, not all people deafened prelingually (before age 3 years) use American Sign Language; thus, their sociocultural ties to the deaf community may be weaker than those of ASL users (Ebert and Heckerling 1995a). The analysis also does not examine the possibility that additional categories based on the age at onset of hearing loss might reveal insight into significant differences among postlingually deafened adults. However, the categories of people with hearing loss used in these analyses yield a more socially valid way to study the deaf population than used in many previous studies of health care services utilization. Categories of people with hearing loss based on hearing loss severity and age at onset of hearing loss have been used in health care-related analyses elsewhere (Cooper 1976; Ebert and Heckerling 1995b; Jones and White 1990; Lass et al. 1978; McEwen and Anton-Culver 1988), including studies with NHIS d ata (Gentile 1975; Ries 1994). Diversity in the sociocultural identity of people with hearing loss has also been discussed previously (Padden and Humphries 1988). Age at onset of hearing loss is a good surrogate measure of the preferred communication mode of deaf people. Preferred mode of communication might be a more valid way to categorize deaf people for the purpose of these analyses, but those data are not available. The findings of this study suggest that more direct questions about modes of communication of deaf persons are warranted in future surveys. It is unlikely that survey nonresponse or the use of self-report produce significant bias. The overall nonresponse rate of the NHIS is 4.4 percent; 2.7 percent is a result of respondent refusal, and the remainder is a result of failure to locate the respondent (Ries 1994). Self-reported health status (Stewart, Hays, and Ware 1988) and self-reported hearing status on the NHIS Hearing Supplement (Schein, Gentile, and Haase 1970) have been shown to be valid measures.

The NHIS is a cross-sectional study; causality cannot be assessed. It is unlikely, however, that the health characteristics analyzed caused the different categories of hearing loss. It is possible that unmeasured variables explain the findings. For example, some comorbid conditions may cause deafness and affect health care utilization. The adjustments used for health status as well as sociodemographics should limit biases due to confounding.

Some NHIS questions regarding health characteristics were asked only of a subset of households in the survey. The resulting small samples for deaf population categories, particularly in the prelingually deafened respondents, limit the power to detect differences among the deaf and general populations. Conclusions about analyses finding no differences among the general and deaf populations may be misleading.

Conclusions from statistically significant results of analyses may be questioned because of the small unweighted sample size for the prelingually deafened study population. Because these data are from a nationally representative sample, these analyses are valuable despite the sample size limitations. Future surveys could oversample for people with deafness.