Age at Intervention for Permanent Hearing Loss and 5-Year Language Outcomes

The participants were 350 children with PCHL and a comparison group of 120 children with normal hearing that was similar to the group of children with PCHL with respect to age, sex, and socioeconomic status (SES). In this study, we focused on the impact of intervention within the age range for which there is uncertainty about the effect of age of intervention, and we therefore excluded children whose device was fitted later than 3 years of age and those whose long-term audiograms revealed a progressive hearing loss. The participants included children from the population-based cohort from 3 states in Australia who participated in the LOCHI study (see Introduction and Fig 1) and who had received by 3 years of age a device of the same type (ie, HA or CI) that they were using at age 5 years. This study was approved by the AH Human Research Ethics Committee.

The AH national service network served as the sampling frame. All families of children who met inclusion criteria were presented with an information statement and consent form at a time as close to initial fitting of HAs as possible, in person or by post. Once written informed consent was obtained from parents, their child was assessed at several intervals.

Postdiagnostic follow-up for all children with PCHL, including this cohort, is performed by AH audiologists until the children reach 26 years of age. The service includes audiological assessment and fitting of HAs and referrals for CI, medical, and communication intervention, in accordance with the AH national protocol.²¹ 

Main outcome measures^22,–32 are summarized in Table 1. Standardized measures of language, vocabulary, letter knowledge, speech production, and oral reading were directly administered to children at age 5 to 6 years by speech pathologists who were blinded, as much as possible, to age at intervention and hearing loss severity. Parents reported on their child’s functional auditory performance in real life, language, and social skills. All measures are widely used for assessing development of children with normal hearing or PCHL, with known validity and reliability. The same test instruments and protocols were administered across all sites.

Information about age at HA fitting, age at CI activation, audiometric thresholds, and hearing device were obtained from clinical records. The degree of PCHL at the time of evaluation was characterized in terms of better-ear 4-frequency average (BE4FA) hearing loss (average of hearing thresholds at 0.5, 1, 2, and 4 kHz) in dB hearing level (HL). Nonverbal IQ was assessed by a psychologist. Parents completed custom-designed questionnaires on demographic characteristics.

Potential a priori confounders were determined on the basis of current literature, including earlier findings at 3 years of age.¹⁶ The variables included birth weight, sex, degree of hearing loss, nonverbal IQ, additional disabilities, maternal education (university degree, vocational training certificate, or ≤12 years of formal schooling), SES (census-based Index of Relative Socioeconomic Advantage and Disadvantage for home postcode³³; higher scores reflect greater advantage), and communication mode during educational intervention (oral versus combined [speech and sign]). These were used as predictors in regression analyses to maximally account for variance in scores so as to increase the sensitivity with which we can assess the impact of intervention timing or access to UNHS.

The statistical analysis was done by using R, version 3.0.2,³⁴ and Statistica, version 10.³⁵ All analyses used 2-tailed tests unless otherwise stated, with statistical significance set at P < .05.

Children unable to be directly assessed because of parent-reported or apparent disabilities or not using spoken English were assigned the poorest score observed for that test (receptive and expressive language: n = 33, 9.4%; vocabulary: n = 62, 17.7%; speech production: n = 47, 13.4%; reading: n = 66, 18.9%; letter name: n = 63, 18.0%; mathematics: n = 69, 19.7%). A multiple imputation method³⁶ was applied to handle missing values in variables, assuming that data were missing at random. The analysis was averaged over 10 imputations.

Scores from 20 measures were aggregated by using factor analysis to form a global language score. This approach was justified by an exploratory factor analysis showing that such measures all loaded onto a single underlying factor. The global score was scaled so that a normal population would have a mean of 100 and an SD of 15. The use of a global score eliminates the disadvantages associated with performing multiple hypothesis testing on correlated data and increases reliability by reducing random measurement error through averaging across measures.

We performed standard multiple linear regression analysis with the global score as the dependent variable. Predictors included age at initial HA fitting (log transformed), degree of PCHL, nonverbal IQ, birth weight, and an interaction between age at HA fitting and degree of PCHL as continuous variables and sex, additional disabilities, maternal education, SES, and communication mode as categorical variables.

We performed a multiple linear regression analysis using the global score as a dependent variable, with predictor variables as for the HA group, but we performed it with age at HA fitting replaced by age at activation of the first CI (log transformed) and without the interaction between age at HA fitting and degree of PCHL.

Analyses for subgroups of users of HAs and CIs were as described above but with age at intervention replaced by screening status (screened vs nonscreened).

Figure 1 summarizes participant enrollment. Of the 728 infants who met inclusion criteria for the LOCHI study, 545 (74.8%) families gave consent for participation. Participants were similar to nonparticipants with respect to age, sex, degree of PCHL, and SES. By 5 years of age, 472 were enrolled in the study, of whom 384 children had nonprogressive hearing loss. Of these, 350 completed the required evaluations. All except 3 completed at least 1 directly administered test. Demographic characteristics of the cohort are shown in Table 2.

Figure 2 shows the distribution of the global language scores of the cohort.

Children who received HAs earlier had better language outcomes. In addition, the regression model indicates that the effect of age at intervention increased as hearing loss increased (P = .035; 1-tailed). Table 3 gives the effect sizes associated with changes in age at amplification for different degrees of PCHL. Compared with amplification at 3 months, those who started amplification at 24 months had poorer language: −11.8 score points (95% CI: −18.7 to −4.8) for 70-dB HL and −6.8 score points (95% CI: −10.8 to −2.8) for 50-dB HL (see Fig 3). These effects were observed after adjustments for nonverbal IQ, birth weight, additional disabilities, maternal education, sex, SES, and communication mode in a model that accounted for 74% of total variance in global score.

Earlier age at CI activation was associated with better language at 5 years of age. Table 4 gives the effect sizes for variations in age at implantation. Compared with those who first received CIs at age 6 months, children who received a CI at 24 months had lower scores by 1.4 SD (−21.4 score points, 95% CI: −33.8 to −9.0). These effects were observed after adjustments for nonverbal IQ, additional disabilities, maternal education, and communication mode in a multilinear regression model that accounted for 70% of total variance in global score. Figure 4 shows the relationship between age at implantation and the adjusted global scores.

The effect size of hearing screening (screened vs nonscreened) was 0.3 score points (95% CI: −4.2 to 4.9; P = .88) for children using HAs and 6.4 score points (95% CI: −2.8 to 16.2; P = .19) for children using CIs.

In this prospective study, we demonstrated a strong positive effect of early intervention on 5-year language in children with PCHL. The younger a child received intervention, the better the language outcomes. In addition, more substantial benefits of earlier access to useful hearing via both HAs and CIs were obtained by those with worse hearing. Earlier intervention, rather than access to UNHS, improved outcomes.

This study has a sample size that was powered for investigating the effect of intervention timing on outcomes separately for children using HAs or CIs after allowing for the effects of multiple demographic characteristics. In this relatively large study, we include language data from 350 children fitted with HAs before 3 years of age, comprising 189 (54%) before age 6 months (at 5 years: 111 use HAs, 78 use CIs) and 161 between 6 months and 3 years (at 5 years: 125 use HAs, 36 use CIs). These data therefore constitute the best opportunity to examine how severity of PCHL influences the association between age at intervention and language outcomes. Because all children received uniform hearing services and high-level technology from a single government-funded agency, we controlled for variations in postdiagnostic services. We evaluated outcomes using direct assessments of children by researchers who were blinded to age at diagnosis and parent reports.

The median age at HA fitting for the entire screened group (3.9 months; interquartile range: 2.3–10.1) is earlier than that of the cohorts in the Wessex study (15 months; interquartile range: 10–40),¹¹ the Netherlands study (mean 15.7 months; SD, 14.0),¹⁵ and an earlier Australian study.³⁷ Unlike recent population-based studies in which researchers included few children with CIs (Wessex study: 16; Netherlands study: 32),^11,15 the present sample comprised 114 children who received CIs before 3 years of age, 42 (36.5%) of whom received them before 12 months of age. All children were assessed at 5 years of age, unlike previous studies that evaluated children at an older (Wessex study: 6–10 years) or a younger age (Netherlands study: 3–5 years).

Although our sample size was large, the 48.1% (350 out of 728) inclusion rate was lower than that in the Wessex study,¹¹ but it was comparable to the proportion with measured outcomes in the Netherlands study¹⁵ or a recent Australian study.³⁸ As age, sex, hearing loss and SES were similar in participants and nonparticipants, response bias seems unlikely. The participation rate may limit inferences on the merits of UNHS, but it does not affect conclusions on the impact of intervention timing on outcomes.

Although our regression models have accounted for a major proportion of variance in global scores, there remains about 30% of unexplained variance, so some combination of other factors and measurement error has also influenced outcomes. Future investigations of the effect of intervention timing on scores for specific assessments would also be worthwhile because age at intervention may affect some abilities more than others.¹⁸ 

Finally, we excluded children with progressive hearing loss (n = 88), so the results will likely not generalize to that population.

With our findings, we broadly agree with but also extend findings in recent reports. Earlier intervention improved language outcomes of children with PCHL.¹⁸ Unlike previous studies in which researchers used cutoff points for early identification of 6⁷ or 9 months,¹¹ suggesting a time window for language development to maximally benefit from intervention or a “sensitive period” that ends before 1 year,¹⁸ our data do not support this concept. Rather, the earlier intervention is provided (at least down to 3 months for HA and 6 months for CI) over the first 3 years of life, the better. Furthermore, the benefit increases as hearing loss increases (see Tables 3 and 4). The positive strong effect of early CI is consistent with results at age 3 years¹⁶ and previous literature.^39,40 

Although access to UNHS led to earlier intervention (72% of the screened group commenced amplification before 6 months, compared with 32% in the unscreened group), and earlier intervention led to better outcomes, there is insufficient evidence to conclude that UNHS is beneficial. The diminished effect size of UNHS compared with that of age at intervention is likely because although UNHS maximizes the opportunity of early intervention, not all screened children commenced amplification early and not all unscreened children commenced amplification late (see Table 2).

The present findings, consistent with a recent review of evidence,¹⁸ have important implications. First, timely device-fitting and educational support needs to follow UNHS to ensure earlier access to hearing, communication, and language development.⁴¹ Second, research needs to focus on postdiagnostic intervention to capitalize on early detection. Third, future follow-up of the existing cohort could evaluate whether the long-term language outcomes improve as the age at intervention decreases.⁴² 

In this study, we show that it is vital to implement a seamless clinical pathway from screening to diagnosis to intervention (fitting HA or CI) so that the opportunity offered by UNHS to improve language development can be captured. In addition, the significance of early implantation highlights the importance of vigilant monitoring of early outcomes with amplification so that children at risk for language delay can receive CIs at the earliest possible age.

 
• 95% CI

  95% confidence interval

 
• AH

  Australian Hearing

 
• BE4FA

  better-ear 4-frequency average

 
• CI

  cochlear implant

 
• HA

  hearing aid

 
• HL

  hearing level

 
• LOCHI

  Longitudinal Outcomes of Children with Hearing Impairment

 
• PCHL

  permanent childhood hearing loss

 
• SES

  socioeconomic status

 
• UNHS

  universal newborn hearing screening

We thank all the families and children for their participation. We also acknowledge the assistance of the many clinicians who provided audiological management of the children and the assistance of professionals at early intervention agencies and CI clinics in the collection of demographic and outcomes data. We also thank Julia Day, Kathryn Crowe, and Nicole Mahler for their assistance with data collection in the early phase of the study and Vicky W. Zhang and Sanna Hou for their help with preparation of the manuscript. Last but not least, we thank the anonymous reviewers for their detailed comments and suggestions for improving earlier versions of the manuscript.