Teletherapy in Children Who Are Deaf and Hard of Hearing
a study on Hearing Loss
This study seeks to determine the effectiveness of speech/language teletherapy to address disparities in speech and language outcomes in children who are deaf or hard-of-hearing (D/HH). The investigators will enroll D/HH children aged 0-27 months. 140 children who are publicly insured will be randomized to receive usual clinical care or to be given access to an 18-month course of speech-language teletherapy program. 70 children who are privately insured will also be enrolled and will receive usual care. Children will undergo, at baseline and every 9 months thereafter to a study endpoint of 18 months, for a total of 3 timepoints, a battery of in-person and parent-report assessments designed to provide a comprehensive measurement of the child's auditory function, speech, verbal- and non-verbal communication, spoken language, and quality of life.
Teletherapy to Address Language Disparities in Deaf and Hard-of-hearing Children
210 children aged 0-27 months with confirmed permanent hearing loss will be recruited from Otolaryngology and Audiology clinics at multiple pediatric hospitals. All provide care for a broadly diverse D/HH population with respect to the disparities of interest. Children who come into Otolaryngology/Audiology clinics will be screened for eligibility. Once eligibility is confirmed, enrollment will be offered. Once enrolled, all groups will undergo comprehensive speech, language, and quality-of-life assessments at baseline and every 9 months thereafter at each site. Assessments will include measures of language, speech, vocabulary, hearing-related quality of life, parental self-efficacy, and early intervention benefit. The investigators will additionally measure therapy utilization and baseline demographic and clinical characteristics. Children will be allocated to one of the following three study arms: 1) Low-UC (low-income children, randomized to receiving Usual Care); 2) Low-TT (low-income children, randomized to receive Usual Care plus Access to Supplemental Teletherapy), and; 3) Higher-UC (higher-income children, receiving Usual Care). Low-TT group will receive an 18-month course of teletherapy at their home. The overall goal of this study is to learn whether improving access to teletherapy for children who are D/HH can reduce disparities in language outcomes. Ongoing engagement with Parent and Stakeholder advisors will occur throughout the study to ensure patient-centeredness and dissemination potential. 1. Specific Aim 1: Primary Analysis I. Demographic and other baseline data including family characteristics data will be listed and summarized descriptively by the treatment arm. Categorical data will be presented as frequencies and percentages. For continuous data, mean, standard deviation, median, interquartile range, minimum, and maximum will be presented. The Full Analysis Set (FAS) comprises all children to whom trial treatment has been assigned by randomization. According to the intent-to-treat principle, analysis will be completed based on the treatment arm and strata to which children are assigned through the randomization procedure. The primary objective of this trial is to compare auditory comprehension (AC) at 18 months for low-income children receiving TT v. UC. The following statistical hypotheses will be tested to address the primary objective: Ho: θ1 = 0 vs. HA: θ1 ≠ 0 where θ1 is the difference between Low-TT and Low-UC at 18 months. The primary analysis to test this hypothesis and compare the two treatment arms will consist of a t-test generated from a linear regression model of the primary endpoint, AC, adjusted for stratification factors. The difference between treatments will be calculated, along with its 95% CI. A total of 140 low-income children will need to be randomized 1:1 (70 per arm) to TT or UC to achieve 90% statistical power to detect an effect size of 0.75 (estimated from Table 1 using public insurance as a proxy for income assuming a common standard deviation of 21.3 and difference of 16: 87 vs. 71, assuming a smaller difference) of the primary endpoint: PLS-5 AC at 18 months. This sample size was adjusted for the drop-out of 20% and variance inflation factor of 0.1 (50/(1-.1) to account for factors adjusted for in the model (squared multiple correlation coefficient of 0.1). This calculation was based on a two-sample t-test assuming equal variances (assuming the larger of the two standard deviations of 21.3 in Table 1) and a two-sided alpha level of 0.017 (Bonferroni adjusted: 0.05/3 co-primary hypotheses). 2. Specific Aim 1: Primary Analysis II & III. While the primary question of this study is to address whether teletherapy improves language outcomes and access to specialty services for disadvantaged families, the investigators also aim to study whether teletherapy can close the language outcomes gap between low and higher-income families. A non-randomized cohort study of families with higher income (but retaining them as an additional comparison) is justified given that the families with higher income likely have access to supplemental services if desired whether children are in the study or not. The investigators will be accruing higher-income patients in parallel to the RCT of low-income children. Therefore, it is important to test the statistical hypotheses: Ho: θ2 = 0 vs. HA: θ2 ≠ 0 where θ2 is the difference at 18 months in AC between low-income children receiving TT and higher-income children receiving UC; and Ho: θ3 = 0 vs. HA: θ3 ≠ 0 where θ3 is the difference at 18 months in AC between low-income children receiving UC and higher-income children receiving UC. Higher-income children will be matched on hearing-level and enrollment sites. For these latter comparisons, control variables will also consist of baseline characteristics known to be associated with AC, which includes clinical attributes and demographic disparities, and potential differences will be assessed using paired tests. To assess whether outcome trajectories differ by study group, the investigators will use mixed-effects linear (for continuous outcomes) and logistic (for dichotomous secondary outcomes) regression analyses. The investigators will flexibly model trajectories by testing whether including quadratic or cubic terms for time (up to 3 visits: baseline, 9 months, and 18 months) or random slopes for individuals improve the model fit and include them if indicated by a significant (P<0.05) likelihood ratio test. The overall difference will be assessed using an F-test and post-estimation t-test in SAS v 9.4. The investigators will also explore the rate of change of scores over time. All of the secondary outcomes will be assessed using this mixed modeling approach, including QOL outcomes. The investigators will also assess the model fit (e.g., residuals) and assess whether transforming the outcomes (e.g., log transformations) provides the best fit. To account for matching, the investigators will also include a random effect. The investigators will also assess whether the baseline values are subject to confounding by isolating within-person changes. The benefit of mixed-effects models is that such models produce unbiased estimates even when some individuals have missing observations, adjust for differential loss to follow-up, accommodate irregular time measurements, and account for clustering of individuals, as required in this study. A two-sided p-value less than 0.017 will be considered statistically significant (Bonferroni corrected for three hypotheses (0.05/3). Estimates and associated 95% confidence intervals, including corrected intervals for multiple testing, will be reported. 210 children will be recruited: 70 higher-income; and 140 low-income, with 70 receiving TT and 70 receiving UC, as described above. Estimated statistical power to compare the higher-income children to low-income children receiving TT and UC will be assessed via a repeated measures design with 3 visits. It assumes an ANOVA F-test with these three groups. The investigators assumed a correlation of 0.9, between-subject variance at 0.22, error variance of 2, and Bonferroni corrected alpha level 0.017 that resulted in 92% statistical power. The investigators expect that the pair-wise t-test comparison at 18 months between Low-UC and Higher-UC would achieve at least 90% statistical power to detect a difference of 27.5 (Table 1). For pair-wise t-test of Low-TT and Higher-UC, the investigators estimate from Table 1 a detectable difference of 10.1 resulting in 81% statistical power. All of the above calculations assume a 20% drop-out, collinearity adjustment of 0.1 (and for pair-wise comparisons, an intraclass correlation coefficient of 0.4). These sample size calculations were performed using Stata 15.1. 3 Specific Aim 2: Retrospective secondary analyses. Demographic and other baseline data will be listed and summarized descriptively by utilization group, TT or UC. Categorical and continuous data will be summarized as described in Aim 1. A linear model (similar to Aim 1) will be used to assess the association between AC, primary outcome, and each of the six disparities at 18 months. Heterogeneous treatment effects (HTEs) will be assessed using the standard HTE approach, an interaction between the utilization group and each disparity. Exploration of whether disparities act alone or in combination (≥2 disparities, such as those who are Spanish speaking and receive public insurance) will be assessed. Potential confounders (described above) and other demographic disparities that may influence the model results will be assessed by performing sensitivity analyses. The investigators will also explore whether the interaction is time-varying by fitting an interaction of time by treatment by disparity using the mixed model approach described in Aim 1. The investigators performed simulations to explore the potential statistical power to detect HTE. Each simulation was repeated 1000 times to provide precise estimates and the investigators included a total of six disparities, which is consistent with the analysis plan. The investigators varied the number of disparities that had known differential HTE (between 2 to 5 disparities) and the disparities that did not (i.e., no differential HTE), as well as the effect size (0.4 to 1.0). A total sample size of 210 children was assumed. The investigators expect approximately 78 children to utilize TT and 132 children to utilize UC. The investigators also assumed a Bonferroni correction, (0.008=0.05/6 disparities) to help protect against family-wise error rate (FWER) of 0.05. The investigators included the main effects in the linear regression model (i.e., utilization TT vs. UC group), disparity (dichotomized), and the interaction effect between the disparity and utilization group. The results from this simulation indicated that there was over 80% statistical power to detect known HTE for 2 disparities regardless of the effect size, ES = 0.4 or ES = 1.0, respectively, while controlling the FWER at 5% for the remaining 4 disparities without HTE present. To detect 5 known disparities, there was 68% and 83% statistical power when the ES = 0.4 and ES=1.0, respectively. The investigators used SAS v.9.4 to estimate the statistical power. The 20% dropout rate is based on the historical 18-month follow-up rate (80%) in the UCSF D/HH clinic for low-income children who are D/HH. Travel and time will be significant for the 3 required study assessments, but will be compensated; information acquired from these assessments will also be a valuable contributor to clinical care and will be shared with the clinical care and Early Intervention teams. The assessments will also be aligned with the child's standard clinical care, which requires in-person audiology visits every 6-9 months. As a comparative effectiveness study, the investigators accept contamination between groups and an unblinded design: families may make different decisions regarding what exact services to pursue, depending on whether were assigned to the teletherapy group or not. Families may be influenced by their Early Intervention centers and community support groups; contamination via these forums may affect service choices external to this study and diminish effect size relative to our retrospective data. Services will be carefully measured for all groups. The investigators will explore whether therapy utilization is linked to the outcomes, possibly as a mediator. To show that therapy utilization is a mediator of the intervention at 18 months, the investigators will assess whether it has a main or interaction effect on the primary outcome. The investigators will estimate the direct and indirect effects of our regression model when exposure-mediator interaction is present. To account for possible effects of crossover and contamination, the investigators will consider performing marginal structural modeling as a sensitivity analysis.
Hearing Loss Speech and Language Development Delay Due to Hearing Loss Hearing Loss in Children Congenital Hearing Loss Speech-Language Teletherapy Deafness
You can join if…
Open to people ages up to 27 months
- Age 0-27 months;
- Hearing loss, as determined by auditory brainstem response (ABR) or behavioral audiometry (average of pure-tone air-conduction hearing thresholds (0.5-4 kHz; PTA) calculated from at least 2 frequencies from ABR (dB eHL) or behavioral audiometry (dB
- bilateral sensorineural, mixed, or permanent conductive hearing loss with better-ear PTA > 20 dB.
- single-sided deafness (unilateral SNHL with PTA > 70 dB);
- unilateral complete aural atresia; or
- bilateral auditory-neuropathy spectrum disorder, as determined by ABR.
- Primary home language is English or Spanish, determined by electronic medical record or direct parent report.
- For children with PTA > 20 dB, either:
- Currently fit with hearing aid or using a cochlear implant; OR
- Date identified for hearing-aid fitting or cochlear-implant activation within 3 months of enrollment.
You CAN'T join if...
- Family does NOT have the intention to pursue listening and spoken language for their child, based on parent report;
- Moderate to severe global developmental delay, as determined by managing audiologist and/or otolaryngologist, based on:
- ICD-10 diagnosis code or chart review of medical progress notes indicating global developmental delay;
- presence of syndrome known to be associated with this delay (such as Trisomy 21, 22q11 syndrome, or CHARGE syndrome); OR
- parent report.
- Speech-language teletherapy received through a clinical provider outside of this study at the time of enrollment, based on parent report.
- No prognosis for access to sound as determined by managing audiologist and/or otolaryngologist. All children with bilateral severe-to-profound SNHL must have imaging to confirm this criterion prior to enrollment in the study. Children will be excluded if they have bilateral severe-to-profound sensorineural hearing loss and either:
- Contraindication to cochlear implantation, OR
- Temporal bone abnormalities that lead to great concern for poor cochlear-implant outcomes, including common cavity and/or cochlear nerve deficiency on imaging.
- UCSF Benioff Children's Hospital - Oakland
Oakland California 94609 United States
- The University of California - San Francisco
San Francisco California 94158 United States
Lead Scientist at UCSF
- Dylan K Chan, MD, PhD
Dylan K. Chan, MD, PhD, is an Associate Professor in Residence and he is the Director of the Children’s Communication Center (CCC) in the Department of Otolaryngology – Head and Neck Surgery (ONHS) at the University of California, San Francisco.
- not yet accepting patients
- Start Date
- Completion Date
- University of California, San Francisco
- Study Type
- Last Updated