Hearing-Impaired Children in the United Kingdom, I: Auditory Performance, Communication Skills, Educational Achievements, Quality of Life, and Cochlear Implantation Paula C. Stacey, Heather M. Fortnum, Garry R. Barton, and A. Quentin Summerfield Objectives: The objectives of this study were to identify variables that are associated with differ- ences in outcome among hearing-impaired children and to control those variables while assessing the impact of cochlear implantation. Study Design: In a cross-sectional study, the parents and teachers of a representative sample of hearing- impaired children were invited to complete ques- tionnaires about children���s auditory performance, spoken communication skills, educational achieve- ments, and quality of life. Multiple regression was used to measure the strength of association be- tween these outcomes and variables related to the child (average hearing level, age at onset of hearing impairment, age, gender, number of additional dis- abilities), the family (parental occupational skill level, ethnicity, and parental hearing status), and cochlear implantation. Results: Questionnaires were returned by the par- ents of 2858 children, 468 of whom had received a cochlear implant, and by the teachers of 2241 chil- dren, 383 of whom had received an implant. Across all domains, reported outcomes were better for children with fewer disabilities in addition to im- paired hearing. Across most domains, reported out- comes were better for children who were older, female, with a more favorable average hearing level, with a higher parental occupational skill level, and with an onset of hearing-impairment af- ter 3 years. When these variables were controlled, cochlear implantation was consistently associated with advantages in auditory performance and spo- ken communication skills, but less consistently as- sociated with advantages in educational achieve- ments and quality of life. Significant associations were found most commonly for children who were younger than 5 years when implanted, and had used implants for more than 4 years. These children, whose mean (preoperative, unaided) average hear- ing level was 118 dB, were reported to perform at the same level as nonimplanted children with aver- age hearing levels in the range from 80 dB to 104 dB, depending on the outcome measure. Conclusions: When rigorous statistical control is exercised in comparing implanted and nonim- planted children, pediatric cochlear implantation is associated with reported improvements both in spo- ken communication skills and in some aspects of educational achievements and quality of life, pro- vided that children receive implants before 5 years of age. (Ear & Hearing 2006 27 161���186) INTRODUCTION Summerfield & Marshall (1999) argued that co- chlear implants are provided to young children in the expectation that immediate benefits in the domain of auditory receptive capabilities will trig- ger a cascade of further benefits in spoken com- munication skills, educational achievements, so- cial independence, and quality of life. The cascade is reproduced in Figure 1. The greater the range of domains over which implantation is effective, the stronger the justification for implantation in chil- dren. In principle, the effectiveness of implanta- tion could be tested at any of the points in the cascade. In practice, too few implanted children have reached adulthood to permit tests across the entire range with adequate statistical power. Ac- cordingly, the present study tested the restricted hypothesis that implantation is associated with enhanced short- and medium-term outcomes that are expected to emerge within a few years of implantation. Outcomes in the following four do- mains were selected on grounds of feasibility and informativeness. CHOICE OF OUTCOME DOMAINS Auditory Performance Implantation has been reported to be strongly associated with improvements in auditory perfor- MRC Institute of Hearing Research, University Park, Notting- ham United Kingsom [P.C.S., G.R.B., H.M.F., A.Q.S.] Now at the Division of Primary Care, School of Community Health Sciences, University of Nottingham, Nottingham, United Kingdom [G.R.B.] Now at Department of Psychology, University of York, Heslington, York, United Kingdom [P.C.S., A.W.S.] Now at the Trent Research and Development Support Unit, Nottingham University Medical School, Queen���s Medical Centre, Nottingham, United Kingdom [H.M.F.]. 0196/0202/06/2702-0161/0 ��� Ear & Hearing ��� Copyright �� 2006 by Lippincott Williams & Wilkins ��� Printed in the U.S.A. 161

mance (e.g., Boothroyd & Eran, 1994 Nikolopoulos, Archbold, & O���Donoghue, 1999). We obtained esti- mates of the ability of children to detect and inter- pret environmental sounds and to respond to simple speech sounds as indexes of auditory performance. Spoken Communication Skills Implantation has been reported to be associated with improvements in speech perception (e.g., Blamey et al., 2001 Meyer, Svirsky, Kirk, & Miy- amoto, 1998 Osberger et al. 1991 Staller, Beiter, Brimacombe, Mecklenburg, & Arndt, 1991a Tyler, Fryauf���Bertschy, Gantz, Kelsay, & Woodworth, 1997a) and speech production (Tobey, Geers, Bren- ner, Altuna, & Gabbert, 2003). We obtained esti- mates of the ability to use, to understand, and to be understood when using spoken language as indexes of spoken communication skills. Educational Achievements In the 1970s, profoundly hearing-impaired school- leavers lagged their hearing peers by up to 7 years in reading and up to 4 years in mathematics (Conrad, 1979 Quigley & Kretschmer, 1982). The gap has closed since that time, probably as a result of earlier identification of deafness (e.g., Dalzell et al., 2000) and more effective early intervention (Moeller, 2000). Evi- dence has now been reported (e.g., Spencer, Barker, & Tomblin, 2003 Tomblin, Spencer, & Gantz, 2000) that implantation is associated with a further closing of the gap. We obtained three types of estimate of educa- tional achievement: informal assessments of attain- ments in reading, writing, number work, and other quantitative skills, plus an estimate of reading age formal assessments of progress through the National Curriculum obtained from Standard Assessment Tests and assessments of participation and engage- ment in the process of education. Quality of Life Studies of changes in quality of life associated with pediatric cochlear implantation have been of two sorts. First, parents have reported the relation- ship between their own expectations and outcomes (e.g., Kelsay & Tyler, 1996 Nikolopoulos, Lloyd, Archbold, & O���Donoghue, 2001). These studies em- phasize the importance to parents of the emergence of skills in listening and communicating. Second, measures of children���s health status (health-related quality of life) have been obtained by proxy from parents (Cheng et al., 2000 Barton, Stacey, Fort- num & Summerfield, submitted[b]). We obtained estimates of two aspects of quality of life: ���child- centered��� quality of life, estimated by parents and teachers in terms of a child���s happiness, capacity to form friendships, and independence of function and ���family-centered��� quality of life, estimated by par- ents in relation to their own level of anxiety about their child and about the degree of disruption to family life that they attributed to their child���s hear- ing impairment. CHOICE OF EXPLANATORY VARIABLES The study had a cross-sectional design in which outcomes in the four domains identified above were obtained for a representative sample of hearing- impaired children, including children with implants and children without implants. A set of explanatory variables was also obtained for each child. Statisti- cal analyses determined whether there was a signif- icant association between implantation and each outcome measure, while controlling the influence of the other explanatory variables. The choice of ex- Fig. 1. A cascade of benefits that can be hypothesized to flow from the provision of cochlear implants to young deaf chil- dren. [Reproduced from Summerfield & Marshall (1999), with permission.] 162 EAR & HEARING / APRIL 2006

planatory variables was guided by three consider- ations. First, classes of variable were identified that have been shown to be independent predictors of outcomes within the general population of children, the hearing-impaired population, or the implanted population. Second, variables were identified that differ between the implanted and nonimplanted populations of children in the United Kingdom (Fortnum, Marshall, & Summerfield, 2002b). Third, additional variables were identified that were shown to differ between the implanted and nonim- planted participants in the present study (Fortnum, Stacey, & Summerfield, submitted). Nine explana- tory variables were identified: average (preopera- tive, unaided) hearing level, age at the onset of hearing impairment, age, gender, number of addi- tional disabilities, socioeconomic status, ethnicity, parental hearing status, and mode of communica- tion used in teaching. Average Hearing Level Among children without implants, many out- comes are poorer, the greater the severity of the hearing loss (Bamford, Wilson, Atkinson, & Bench, 1981 Blamey et al., 2001 Conrad, 1979 Svirsky, Sloan, Caldwell, & Miyamoto, 2000b). Among hear- ing-impaired children in the United Kingdom, the average (preoperative, unaided) hearing level (AHL) is significantly greater for children with implants than for hearing-impaired children without im- plants (Fortnum et al., 2002b). Thus, failure to control AHL when comparing implanted and nonim- planted children might lead to an underestimate of the effectiveness of implantation. AHL was calcu- lated as the average of hearing levels at 0.5 kHz, 1 kHz, 2 kHz, and 4 kHz.* Age at Onset of Hearing Impairment An older age at the onset of hearing impairment is associated with better outcomes in speech percep- tion among nonimplanted children (Blamey et al., 2001) and implanted children (Blamey et al., 2001 Osberger et al., 1991 Staller et al., 1991a) and also with better outcomes in speech production and lan- guage among implanted and nonimplanted children (Blamey et al., 2001). Among hearing-impaired chil- dren in the United Kingdom, the average age at the onset of hearing impairment is greater for children with implants than for children without implants (Fortnum et al., 2002b). Therefore, failure to control age at the onset of hearing impairment when com- paring implanted and nonimplanted children might lead to an overestimate of the effectiveness of im- plantation. Age Communication skills improve with increasing age in the general population (Fenson et al., 2000 Mogford & Bishop, 1993), the hearing-impaired pop- ulation (Blamey et al., 2001 Meyer et al., 1998), and the implanted population (Blamey et al., 2001 Svir- sky, Robbins, Kirk, Pisoni, & Miyamoto, 2000a). The sample of children in the present study was drawn from the population of hearing-impaired children in the United Kingdom in 1998. At that time, children with implants were younger than the average of the population (Fortnum et al., 2002b). Thus, failure to control age when comparing implanted and nonim- planted children might lead to an underestimate of the effectiveness of implantation. Gender In the United Kingdom and internationally, there are more hearing-impaired boys than girls, with an overall ratio of about 1.16 boys for every girl (Fort- num, 2003). However, in the United Kingdom, more girls than boys have cochlear implants (Fortnum et al., 2002b). In the general population, girls display better early communication skills (Fenson et al., 2000) and higher educational achievements than boys (Gillborn & Mirza, 2000). A similar pattern has been documented in the hearing-impaired popula- tion (Allen, 1986 Easterbrooks & O���Rourke, 2001) and in the implanted population (Geers, 2003 Tobey et al., 2003). Therefore, failure to control for gender when comparing implanted and nonimplanted chil- dren might lead to an overestimate of the effective- ness of implantation. Number of Additional Disabilities Children with cochlear implants in the United Kingdom have fewer additional disabilities than other hearing-impaired children (Fortnum et al., 2002b). The US Department of Education Office of Special Education and Rehabilitative Services (2002) reported that pupils with disabilities dropped *In the United Kingdom, it has been traditional to summarize hearing sensitivity as the average of the hearing levels measured at 500 Hz, 1 kHz, 2 kHz, and 4 kHz. We refer to this four- frequency average as AHL. In North America, hearing sensitivity is often summarized as a three-frequency pure-tone average (PTA) computed at 500 Hz, 1 kHz, and 2 kHz. Values of AHL are expected to be numerically greater than values of PTA because AHL includes higher frequencies, where sensitivity is likely to be poorer. We assessed the consequences of using one metric rather than the other by comparing PTA with AHL for 2378 children for whom both values could be calculated. The linear regression equation is PTA 0.916 0.987(AHL), with an adjusted r2 of 0.97. Thus, for example, in the present sample of children, an AHL of 95 dB corresponds to a PTA of 93 dB. EAR & HEARING, VOL. 27 NO. 2 163

out of school at twice the rate of their peers and that their enrolment in higher education was 50% lower than in the general population. Within the hearing- impaired population, negative effects of disabilities additional to hearing impairment have also been reported. For example, compared with hearing-im- paired children with no additional disabilities, hear- ing-impaired children with additional disabilities display lower nonverbal IQ (Conrad, 1979) and achieve poorer educational qualifications as mea- sured by national examinations at school-leaving age (Powers, 2003). Although children with addi- tional disabilities display improvements in the abil- ity to perceive speech after cochlear implantation, the improvements emerge at a slower rate compared with children who have no additional disabilities (Holt & Kirk, 2005 Pyman, Blamey, Lacy, Clark, & Dowell, 2000 Rajput, Brown, & Bamiou, 2003 Waltzman, Scalchunes, & Cohen, 2000). Therefore, failure to control for other disabilities when compar- ing implanted and nonimplanted children might lead to an overestimate of the effectiveness of im- plantation. Socioeconomic Status Children with cochlear implants in the United Kingdom tend to come from more affluent families than children without implants (Fortnum et al., 2002b). Children from more affluent families tend to attain better communication skills and better edu- cational attainments. These effects are found for children in the general population (Arriaga, Fenson, Cronan, & Pethick, 1998 Gillborn & Mirza, 2000), the hearing-impaired population (Kluwin, 1994 Powers, 2003 Van Den Horst, & Kamstra, 1979), and the implanted population (Geers, 2003 Tobey et al., 2003). Therefore, failure to control for socioeco- nomic status when comparing implanted with non- implanted children could lead to an overestimate of the effectiveness of implantation. Ethnicity Compared with the white majority in the United Kingdom, members of ethnic minorities are more likely to be unemployed, to have lower levels of household income, and to live in low-income house- holds (White, 2002). Additionally, the prevalence profile of disease and illness differs between ethnic minorities and the white majority (Donaldson & Clayton, 1984), including the prevalence of child- hood hearing impairment (Fortnum et al., 2002b). Children from some ethnic minorities attain lower educational qualifications (Gillborn & Mirza, 2000 Powers, 2003), and hearing-impaired children from some ethnic minorities demonstrate lower levels of achievement than their white peers (Allen, 1986 Cohen, Fischgrund, & Redding, 1990 Kluwin, 1994). The relationships among ethnicity, socioeco- nomic status, and outcomes in health and education are complex and not fully understood. Accordingly, epidemiologists are encouraged to report the ethnic affiliations of study samples and to control ethnicity alongside socioeconomic status. Parental Hearing Status Parental hearing status was included in the list of explanatory variables because some studies have concluded that hearing-impaired children born to hearing-impaired parents achieve better educa- tional outcomes than hearing-impaired children who are born to hearing parents (Powers, 2003). Communication Mode Children with cochlear implants in the United Kingdom are more likely than profoundly hearing- impaired nonimplanted children to be taught using oral communication only (Fortnum, Marshall, Bam- ford, & Summerfield, 2002a). Additionally, the per- centage of implanted children who use oral commu- nication increases with time after implantation (Archbold et al., 2000). Although it had been con- cluded that there is no difference in aural-oral skills between children using spoken language alone and children using a mixture of speaking and signing (Wilbur, 1979), more recent evidence suggests that speech perception, speech production, and oral com- munication skills are better among children who rely solely on an oral approach (e.g., Geers & Moog, 1992). Advantages associated with oral communica- tion have also been documented among children with implants (Archbold et al., 2000 Geers, 2000), where the amount of time spent speaking and lis- tening is positively associated with better speech perception (Geers, Brenner, & Davidson, 2003), speech production (Tobey et al., 2003), and use of language (Geers, Nicholas, & Sedey, 2003). Al- though children may have performed better because they used oral approaches, it is also possible that children used oral-only approaches because they performed better or were judged to have the capacity to perform better. The causal relationship between communication mode and outcomes is unclear (Dow- ell, Blamey, & Clark, 1995). Nonetheless, in some analyses, we controlled the mode of communication to determine whether it explained an independent component of the variance in outcomes. Previous studies of the effectiveness of pediatric cochlear implantation have adopted one of five ex- perimental designs, as reviewed by Meyer et al. (1998) and by Summerfield (2002): prospective un- 164 EAR & HEARING / APRIL 2006

controlled (e.g., Staller, Dowell, Beiter, & Brima- combe, 1991b Tyler, Fryauf-Bertschy et al., 1997b) historically controlled (e.g., Svirsky et al., 2000a Tait & Lutman, 1997) cross-sectional with cross- sectional controls (e.g., Boothroyd & Eran, 1994 Nakisa et al., 2001) prospective with cross-sectional controls (e.g., Meyer et al., 1998 Osberger et al., 1991 Snik, Vermeulen, Brokx, Beijk, & van den Broek, 1997 Svirsky & Meyer, 1999) and prospec- tive with prospective controls (e.g., Blamey et al., 2001 Geers & Moog, 1994). With the exception of the prospective case-control study by Geers & Moog (1994), none of these studies exercised control over as wide a range of variables as were included in this study, although a similar degree of control was exercised in explorations of the variables associated with different levels of outcome among implanted children (Geers & Brenner, 2003). MAIN HYPOTHESES The study tested two hypotheses: (1) Implanta- tion is effective both in the short-term domains of auditory receptive capabilities and spoken commu- nication skills and in the medium-term domains of educational achievements and quality of life. (2) However, benefits of implantation are more robust and emerge sooner after implantation, in short-term compared with medium-term domains. To address these hypotheses, we compared measures of the effectiveness of implantation between domains and, within domains, between groups of children who had used implants for different lengths of time. We made these comparisons with three measures of effective- ness: estimates of the strength of association be- tween implantation and outcomes, estimates of the importance of implantation as an explanatory vari- able for outcomes compared with other variables, and estimates of functionally equivalent, equiva- lent, and significant hearing levels. Functionally Equivalent, Equivalent, and Significant Hearing Levels Several authors have summarized the results of comparisons between implanted and nonimplanted children as a functionally equivalent hearing level (Blamey et al., 2001 Boothroyd & Eran, 1994 Geers & Moog, 1994 Meyer et al., 1998 Nakisa et al., 2001 Osberger et al., 1991 Osberger, Maso, & Sam, 1993 Snik et al., 1997 Svirsky & Meyer, 1999). This is the unaided AHL of nonimplanted children who achieve the same level of outcome as a child with an implant. Boothroyd & Eran (1994) argued that a functionally equivalent hearing level can be more informative than the outcome score itself that is, it can be more informative to learn that an implanted child whose unaided preoperative AHL was, for example, 115 dB now functions like nonimplanted children with an AHL of 85 dB, than to be told that the implanted child achieves a score of 45% correct. A further advantage is that AHL is a common metric in which the effectiveness of implantation can be compared between domains. We estimated functionally equivalent hearing levels along with two other values: the ���equivalent��� hearing level is the AHL at which the outcome score for implanted children is the same as the outcome score for nonimplanted children the ���significant��� hearing level is the lowest (most favorable) AHL at which the level of outcome for implanted children differs significantly from the level of outcome for nonimplanted children. We used the methods de- scribed in the Appendix for estimating functionally equivalent, equivalent, and significant hearing lev- els while controlling the effects of potentially con- founding variables. SUBSIDIARY QUESTIONS Mode of Communication Used in Teaching Geers (Geers, 2003 Geers, Nicholas, & Sedey, 2003) identified explanatory variables for outcomes in the domains of language and reading in im- planted children. Better outcomes were associated with variables in the child (e.g., higher IQ, female gender), the child���s family (e.g., fewer siblings, high socioeconomic status), and the implant system (e.g., modern processor, larger dynamic range). After ac- counting for variance in outcomes with these vari- ables, the mode of communication used in teaching was significantly associated with the residual vari- ance in each domain, such that better outcomes were associated with oral education. We first determined whether this result can be generalized by asking whether variables in the child and the child���s family are associated with differences in outcome among hearing-impaired children generally, not just im- planted children. We then examined whether resid- ual differences in academic achievements can be explained by the mode of communication used in teaching. Consequences of Controlling Different Numbers of Explanatory Variables In observational studies (i.e., studies like the present one in which subjects are assigned to treat- ments by methods other than random assignment), variables that differ between cases and control sub- jects must be controlled if an accurate estimate of the effect of the treatment is to be obtained. In practice, it may not be practical to measure more EAR & HEARING, VOL. 27 NO. 2 165