A Response to: United States Preventative Services Task Force (USPSTF) Report Regarding Universal Newborn Hearing Screening

Christine Yoshinaga-Itano, Ph.D.
Professor
University of Colorado, Boulder
Department of Speech, Language & Hearing Sciences
Boulder, Colorado 80309-0409
Phone: 303 492-3050
email: Christie.Yoshi@colorado.edu'

This paper is written as a response to the USPSTF paper titled "Recommendations and Rationale Newborn Hearing Screening." The USPSTF paper was published in November 2001, the reader is referred to the original (www.ahcpr.gov/news/press/pr2001/newbornpr.htm and https://www.ahcpr.gov/clinic/3rduspstf/newhearrr.htm).

The American Speech-Language-Hearing Association has responded to the USPSTF paper ( see /news/displaynews.asp?id=342) as has the American Academy of Audiology ( see /news/displaynews.asp?id=341).

Nonetheless, I would like to address this same topic, in more detail, on an issue-by-issue basis. The bold print is essentially the point made by the USPSTF, the standard print is my response.

Issue 1 -- The USPSTF stated ... "There are no prospective, controlled studies that directly examine whether newborn hearing screening and earlier intervention result in improved speech, language or educational development."

There is a new population of early-identified children with significant hearing loss, whose language levels are within the low average range. There has never before been a "population" of children with significant hearing loss who have achieved average age-appropriate language skills in the history of the research literature. These children have done well, and their hearing losses were all identified early. If their language levels are not a result of early identification programs, what explains their status? These children with early-identified hearing loss, and no secondary disabilities, demonstrate this level of language development in their first five years of life (Yoshinaga-Itano, Coulter & Thomson, 2001).

Therefore, although not a prospective study from screening or controlled random assignment to screening study, the conclusion is clear; hearing impaired children, when identified early and when treated with appropriate intervention have the ability to develop approximately normal language skills.

A follow-through study, such as that recommended by the USPSTF would cost millions of dollars and would likely require several generations of children. Statistics on children with significant hearing loss indicate that five hundred to six hundred must be screened to find a single child with hearing loss. Simply, the costs of the recommended study would be prohibitive. Longitudinal follow-through of children past the age of five years has never been done with any population of children with significant hearing loss with the exception of very small sample sizes, making generalizability to the population at-large difficult. The study the USPSTF proposed would be fantastic, but improbable. I think we do not have to wait until the study is funded and completed to draw reasonable and obvious conclusions.

Issue 2 -- The US Preventative Task Force questions whether there were other factors that might contribute to language development. "Other factors, such as family involvement (an important contributor to language development), the degree of other disabilities, or the quality of pediatric care, might have influenced the time of identification and the language outcome."

In educational literature the ability to account for 50 to 70% of the variance in development is rarely if ever seen, predominantly because there are so many variables that usually contribute to developmental outcome and not all of them can be included in a single research design. Researchers in language development and significant hearing loss are continually searching for variables that will account for additional variance in outcome. However, when non-verbal cognitive development and age of identification of the hearing loss account for over 50% of the variance, and family involvement, family interaction variables account for additional variance, the other variables such as method of communication, degree of hearing loss, and socio-economic variables, either add no additional variance or only a very small amount.

The question of whether children who are not screened at birth would be identified earlier because of family involvement is an interesting one raised by the Task Force. Family involvement, as described in the literature, is measured after diagnosis of hearing loss and when the family is involved in intervention services. Hearing loss is an "invisible disability" and is typically not identified by the most caring and involved parents until language and speech fail to develop. Children with hearing loss can fool their parents simply by being very visually aware of their environment. To imply that a child would be identified later because parents are not "involved" is as inappropriate as implying that a child is identified late because the managing physician is "unconcerned" or has not provided appropriate care. Testimony from physicians, who are involved parents, whose own children were late identified with hearing loss can attest to the fact that hearing loss is an "invisible disability."

Early-identified children have families who are rated higher in family involvement than families with later-identified children. There are probably a whole host of variables that contribute to this effect. Some researchers have also found a very high relationship between family involvement and socio-economic status. One of the striking findings of the early-identified children is that, at least in the early years of life, the impact of socio-economic status is either eliminated or compensated for, by early intervention services, such that socio-economic status has not been found to be a primary predictor variable of language outcome in the first five years of life.

The degree of other disabilities does factor into the time of identification. Children being treated for other medical conditions are often tested for hearing loss, particularly if the medical conditions are typically associated with hearing loss. Early-identified children with milder hearing losses are only identified early with universal newborn hearing screening programs (UNHS) if they have other disabilities, since these children would be observed in the home to respond to sounds and even speech and delays in expressive language would not be evident until the child is two to three years of age at the earliest, but most typically after the age of five years.

The quality of pediatric care is an interesting question. Since the average age of identification of congenital hearing loss in the United States has been very late, typically after two years of age, this question seems to have been answered. Without UNHS programs, even the highest quality of pediatric care appears unable to result in the diagnosis hearing loss in the newborn period, or even in the first twelve months of life. The US Preventative Task Force fails to identify one study that shows that identification of hearing loss for the average child can occur within the first six months or twelve months of life through "usual care."

Issue 3 - Does universal newborn hearing screening result in better outcomes than targeted screening or typical care without screening?

If we were to answer this question with all due consideration for scientific methodology, all babies would have to be randomly assigned to one of three groups. 1- Traditional/typical care (no screen), 2- Targeted screen (perhaps based on high-risk factors) or 3- UNHS screen. In traditional medical and other clinical studies, patients can be randomly assigned to either a control "no treatment" group or an experimental "treatment" group. The statistical results of these assignments would be telling and self-evident.

However, babies assigned to these groups would have to be followed prospectively at least into elementary school to determine whether they would eventually be identified with hearing loss. Because of the low incidence of hearing loss and the exclusion of specific groups of children with hearing loss from consideration, I estimate that 20,000 children in each group would have to be followed to get even a small number of children in each of the three groups.

Importantly, federal law prohibits random assignment to treatment, in the timing of the initiation of treatment and in the type of treatment. Additionally, families have the right to choose whether they desire intervention, and if so, the intervention they wish. Thus, random assignment to one of the three groups is arguably not legally possible. Further, identification of hearing loss with concomitant failure to disclose this information to the parents, in order to randomly assign their child to the "no treatment"group, would also be prohibited by law. The USPSTF does acknowledge that such a design could probably not be accomplished. Further, a "no treatment" group that naturally emerges in the rare instance that a family chooses "no treatment" would be characteristically different from the "treatment" group resulting in unequal comparison groups by characteristics of the family and child and also in number of families.

Further exacerbating the problem is the fact that the characteristics of the family and child may influence the intervention choices. For example, families with children who have mild hearing loss tend to choose more auditory/oral interventions because their children have significantly more benefit from amplification. Many families with children who have multiple disabilities choose a simultaneous communication mode. The impact of commitment of the family to the intervention of choice and the skill of the family in the intervention of choice are variables that have previously been found to affect developmental outcome. Random assignment to treatment would ignore these variables.

The USPSTF suggests that an inception cohort of newborns offered UNHS with infants managed by usual care (including selective screening) should be compared. In order for these comparisons to be made in a meaningful manner, a sufficient number of infants must be identified both in the traditional care versus the UNHS group.

My best estimate is that the minimum number of infants required would approach about 120,000 infants, to obtain a cohort of 30 to 50 eligible participants (with no secondary disabilities and with bilateral moderate to profound permanent hearing loss) in each group. These infants would have to be followed for their entire educational career to demonstrate "better function later in life." Of course if the study was initiated this year (2002), and if we accepted the termination of their college BA/BS degrees as the endpoint of the study, in 2024 we can expect to analyze the resultant data.

Importantly, random assignment would in no way guarantee equal groups by socio-economic status, cognitive quotients, degree of hearing loss, gender, ethnicity, or presence of an additional disability etc. Random assignment results in equal comparison groups only when the population sampled is normally distributed, and hearing loss is not. The variability of the population would potentially necessitate doubling the number of infants in the study from 120,000 to 240,000 to guarantee sufficient numbers for statistical analysis.

Issue 4 - Regarding, the USPSTF assertion that previous studies do not establish that "screening low-risk newborns" is an important factor because there is not a control group of newborns with usual care.

The fact that no cohort of early-identified children (first six months of life) has existed in the literature and they do indeed now exist, addresses this issue and should suffice. Research indicates that fifty percent of the infants born with hearing loss are missed through high-risk register screenings. In countries that have used a universal "distraction testing" (a visual reinforcement head-turning screen) at 9 months of age, they (see data from the UK, Australia, Scandinavia, Germany, and Holland) report an average age of hearing loss identification between 2 and 2.5 years.

Issue 5 -- The USPSTF asserts that samples of participants in the studies of early-identified children could represent a biased sample. The USPSTF stated "None compared an inception cohort of low-risk newborns identified by screening with those identified in usual care, making it impossible to exclude selection bias as an explanation for the results."

I believe there was no selection bias. The Colorado Home Intervention Program outcomes are prospective studies of the language outcomes of the population not a sample of children in Colorado followed from identification of hearing loss. Today, in Colorado, the identification of hearing loss is a direct follow-through from universal newborn hearing screening. The data management in the early years of UNHS in Colorado did not include child identification information, but included aggregate data, numbers born, numbers screened, numbers passed and referred. Parents, therefore, reported whether their infants had participated in UNHS. A few articles include longitudinal data analysis although the vast majority of the data published so far are cross-sectional.

In the Colorado studies, the later-identified groups were children identified through typical care or the high-risk register. Since screening began in 1992 in the state of Colorado, all children born before 1992 were identified through either typical care, high-risk register, or a targeted screen in a newborn intensive care unit.

The Yoshinaga-Itano, Coulter & Thomson (2000) study compared 1) children identified through universal newborn hearing screening, 2) children unscreened between 1992 and 1999, and 3) children unscreened through UNHS because they were born prior to the initiation of screening in Colorado. It should be noted that the actual average age of diagnosis of the unscreened children is probably much higher because only children enrolled in the early intervention program, who were identified with hearing loss in the first three years of life were included. Children identified through preschool screening or school screening programs for the first time would not be included in this data. Thus, children with mild and moderate hearing losses who were identified later in life are not represented in these statistics, making the average age of identification for this population lower than the actual state statistics.

An early-identified child from selective screening, i.e. newborn intensive care screening, distraction testing, or high risk register screening, has the same outcomes as those identified through UNHS (Yoshinaga-Itano, Coulter & Thomson, 2000, Yoshinaga-Itano, Coulter & Thomson, 2002). The early-identified children in the Apuzzo & Yoshinaga-Itano (1995) study were children identified through high risk register. Their outcomes were virtually identical to those from both the 1998 (Yoshinaga-Itano, Sedey, Coulter & Mehl, 1998) and 2000 studies (Mayne, Yoshinaga-Itano, Sedey, 2000; Yoshinaga-Itano, Coulter & Thomson, 2000). When considering only children with cognitive quotients of at least 80, 31.6% of the children in the "not screened" group, 62.1% in the "probably not screened" group (parent report of "no UNHS" but no state records), and 44.3% of the children born before 1992 had language quotients that were 80 or above. This compares with the 82.4% in the "screened" group and 81.0% in the "probably screened" group (parent report of UNHS but no state records) who had language quotients that were 80 or above.

Issue 6 - The USPSTF queried .... Did attrition of subjects and failure to follow-up bias the samples of the existing literature? They reported, "Studies had unclear criteria for selection of subjects. Early-identified children who remained in the program may have had better results than early-identified children who were not available for follow-up. None of the studies provide information on attrition or follow-up rates."

However, in the Colorado cohort over 90% of the children for whom early intervention is recommended, enroll in the Colorado Home Intervention Program. Most of the remaining 10% enroll in other intervention services. Some of these children are children with multiple disabilities and hearing loss is not the primary disability. Others enroll in private intervention programs or school-based programs. However, many of these providers also participate in the outcome data collection. Over the last twenty years of data collection, there has only been a 5% attrition rate. Attrition typically occurs because the family moves out of state or because of a death. Since the program is a publicly funded program, and the program is state-wide very few children are lost to follow-up. Early identified children who do not continue follow-up in the program are typically those who are no longer receiving special educational services because their developmental levels are so high that they no longer qualify for services. Thus, if we are missing any early-identified children, they are most likely the ones with the highest functioning levels. All children in the state who had been early-identified were included as potential subjects for these studies. Seventy percent of the available participants, children identified with hearing loss, as a result of UNHS, were included in the study.

Issue 7 -- THE USPSTF report states: "The diagnosis of congenital hearing loss is often delayed. In one survey conducted before hearing screening was common, the median age at diagnosis was 13 months for infants with severe to profound bilateral SNHL and 17 months for those with mild-to-moderate hearing losses. (Harrison & Roush, 1996)"

Studies using surveys of early intervention programs throughout the United States cannot be used as an index of the average age of identification of hearing loss. While surveys provide valuable information for the profession, they were not designed as an accurate measure of the average age of diagnosis of the hearing loss in the United States. The Harrison & Roush (1996) study analyzed surveys sent to intervention programs throughout the United States. As in most survey studies, a small percentage of those contacted responded to the surveys. Early intervention providers were asked to fill out a response indicating the age at which children in their program were typically diagnosed. No audiogram data was submitted to verify the information provided. Probably the most significant issue is that the programs only enrolled children between the ages of birth and three years of age, therefore, any child identified after the age of three would not be included in the average age of identification. The characteristics of the intervention programs (public versus private) responding may influence the data obtained. The actual number of children in each program who were identified at specific age ranges was not reported by the respondants.

Yoshinaga et al. (2002) reported that 84% of the children in the "screen" group were identified by 6 months of age as compared to 8% in the "no screen" group. Sixteen percent in the "screen" group were identified after 6 months of age, while 92% in the "no screen" group were identified after 6 months of age.

Issue 8 -- The USPSTF stated that "Studies of statewide universal newborn hearing screening programs in the United States have found that the mean age of identification of hearing impairment has decreased from 12 to 13 months before screening programs were introduced to 3 to 6 months since their introduction."

The data from Rhode Island actually reported that the average age of diagnosis of hearing loss in their state was 20 months prior to the introduction of screening, not 12 to 13 months. No public or private program has ever published data indicating that the average age of identification of hearing loss prior to universal newborn hearing screening was 12 to 13 months.

Issue 9 -- The USPSTF queries....Does UNHS result in earlier treatment?

One important issue to be addressed is the definition of treatment. Does treatment mean medical management or audiologic management (identification/diagnosis, amplification fitting, and therapeutic intervention for speech, language and social-emotional development)? The treatment after identification of permanent sensori-neural hearing loss is typically not medical and does not involve medication or surgical procedures. For almost all children, initiation of intervention begins with hearing aid fitting.

Issue 10 -- Does universal newborn hearing screening reduce the age at which hearing aids are fit? The USPSTF report states: "The mean age at which hearing aids are fit has been reduced from 13 to 16 months before screening programs were introduced to 5 to 7 months."

The studies do not indicate that the mean age at which hearing aids are fit has been reduced from 13 to 16 months "before UNHS programs were introduced" to 5 to 7 months "after UNHS programs were introduced". The only way an accurate indication of mean age at hearing aid fitting is to look at least 7 years of data in order to include the many children who are diagnosed later in life. The state surveys conducted by the Marion Downs National Center indicate that age of identification of hearing loss and age at amplification fitting were much more delayed than reported by the USPTF.

Issue 11 -- The USPSTF concludes that the cost of screening children in the well-baby nursery has too low of a yield to be cost-effective. Current procedures miss fifty percent of the children with congenital hearing loss and the group of children missed, have the highest potential for the greatest benefit from early-identification. The USPSTF states that "To use a hypothetical example, if 50% of the newborns with PHL would have poor language ability if diagnosed after age 10 months, and early intervention reduces this by 50%, then the NNS to prevent 1 additional case of delayed language acquisition would be 6771."

This statistic only includes moderate to profound bilateral hearing loss. The argument here is flawed because there is no research evidence that 50% of the newborns with PHL would have poor language if diagnosed after age 10 months. In fact, the statistic is more likely greater than 65% of the population and may approach 80%. The comparison groups in all of the studies compared children who were enrolled in early intervention programs and therefore, identified by three years of age. They excluded children who were identified after three years of age. Only a small percentage of the children with significant hearing loss have age appropriate language levels as they enter adulthood. This statistic is further flawed because the population statistics include 40% of the population with additional disabilities, and if children with hearing loss only were separated from these statistics, the impact would be even greater.

Rather than looking at whether the cost is worth 1 case in 6771 screened, the emphasis should be placed upon the majority of children with significant hearing loss having the opportunity for normal language development versus what exists today, when the majority of children have language development between 50 and 60% of chronological age and these language quotients either remain or worsen as the child ages.

The cost of later-identification of hearing loss goes well beyond the initial birth period. The cost to the individual child and family, the educational system and the aggregate cost to society is too high, when only about 16 to 20% of the later-identified children achieve and maintain age-appropriate language levels.

Issue 12 -- The USPTF found good evidence that the prevalence of hearing loss in infants in the newborn intensive care unit (NICU) and those with other specific risk factors is 10 to 20 times higher than the prevalence of hearing loss in the general population of newborns. Both the yield of screening and the proportion of true positive results will be substantially higher when screening is targeted at these high-risk infants, but selective screening programs typically do not identify all infants with risk factors. Evidence that early identification and intervention for hearing loss improves speech, language, or auditory outcomes in high-risk populations is also limited.

Although yield is higher in the NICU, children with risk factors only account for 50% of the population of children with hearing loss. Thus, targeted screening would have a high yield of identification of hearing loss per child screened, but would miss 50% of the children. High risk screening has existed in the United States for decades. It has failed to identify a large cohort of children in the first year of life. Further, those children identified in the first year of life have a significantly higher incidence of secondary disabilities (approximately 66%) than the children identified through UNHS and the well-baby nurseries (about 30%). Thus, the children with the highest potential for success are the most likely to be missed.

Issue 13 -- The USPSTF stated that "Although early identification and intervention may improve the quality of life for the infant and family during the first year of life, and prevent regret by the family over delayed diagnosis of hearing loss, the USPTF found few data addressing these benefits. The USPTF could not determine from existing studies whether these potential benefits outweigh the potential harms of false-positive tests that many low-risk infants would experience following universal screening in both high- and low-risk groups."

The literature does not indicate that there are potential harms of false-positive tests. There has been no evidence that newborn hearing screening causes parental harm since parental stress levels of the "false positive" referral group are similar to those present in the general population and to those families whose babies have not been screened. There is a body of evidence that screening does not result in parental harm when measured by traditional instruments of parental stress and depression. A summary of the literature indicates the following findings. Families of children who are screened for hearing and those whose children are not screened do not differ significantly in their stress levels. Families of children who are screened and pass and families of children who are screened and refer do not differ significantly in their stress levels. Approximately 6-13% have high levels of stress in both groups. This level of high stress is the level of stress found in the general public. Families of children with hearing loss identified through UNHS and those identified through other means do not differ in their stress levels. UNHS does not result in increased family stress.

Studies demonstrating a better quality of life as a result of UNHS, early identification and early intervention have not yet been conducted. It is a reasonable assumption that as earlier identification and intervention have a positive impact upon the developmental outcomes of children with significant hearing loss, the quality of life of the family should be significantly improved.

Summary:

The Colorado studies investigated the relationship of universal newborn hearing screening through logistical regression statistical techniques on a population of 294 children with significant hearing loss. No children were excluded from this study. All children meeting the selection criteria were included in the study. The study included children with hearing loss only and those with multiple disabilities. In addition, the Colorado studies investigated the question of early-identification and early intervention for children with hearing loss and developmental outcomes through studies that included matched designs, multi-variance analysis of covariance statistical designs, and multiple regression techniques using both step-wise and block designs. All studies, regardless of the number of participants resulted in the same results, a robust and repeatable impact of early identification and initiation of intervention in the first six months of life on developmental outcomes. Although the USPSTF believes that statistical analysis and experimental group statistical control are not sufficient, if the phenomena were not real, the number of replications of the effect are well beyond chance findings. UNHS programs result in earlier identification of the hearing loss, earlier initiation of intervention including fitting of amplification, and better outcomes for the child and the family in language, speech and social-emotional development.

References:

Apuzzo, M., & Yoshinaga-Itano, C. (1995). Early identification of infants with significant hearing loss and the Minnesota Child Development Inventory. Seminars in Hearing, 16, 124-139.

Calderon, R. & Naidu, S. (2000). Further support of the benefits of early identification and intervention with children with hearing loss. In C. Yoshinaga-Itano & A.L. Sedey (Eds.), Language, Speech and Social-Emotional Development of Children Who Are Deaf and Hard-of Hearing: The Early Years, The Volta Review, 100, 53-84.

Mayne, A., Yoshinaga-Itano, C., & Sedey, A. (2000). Receptive vocabulary development of infants and toddlers who are deaf or hard of hearing. In C. Yoshinaga-Itano & A.L. Sedey (Eds.), Language, Speech and Social-Emotional Development of Children Who Are Deaf and Hard-of Hearing: The Early Years, The Volta Review, 100, 29-52.

Mayne, A., Yoshinaga-Itano, C., Sedey, A.L., & Carey, A. (2000). Expressive vocabulary development of infants and toddlers who are deaf or hard of hearing. In C. Yoshinaga-Itano & A.L. Sedey (Eds.), Language, Speech and Social-Emotional Development of Children Who Are Deaf and Hard-of-Hearing: The Early Years, The Volta Review, 100, 1-28.

Moeller, M.P. (2000). Early intervention and language development in children who are deaf and hard of hearing. Pediatrics, 106, E43.

Pipp-Siegel, S., Sedey, A.L., & Yoshinaga-Itano, C. (2001). Predictors of parental stress of mothers of young children with hearing loss. Journal of Deaf Studies and Deaf Education. 7(1), 1-17.

Pressman, L., Pipp-Siegel, S., Yoshinaga-Itano, C., & Deas, A. (1999). The relation of sensitivity to child expressive language gain in deaf and hard-of-hearing children whose caregivers are hearing. Journal of Deaf Studies and Deaf Education, 4, 294-304.

Yoshinaga-Itano, C., Coulter, D., & Thomson, V. (2000). The Colorado Newborn Hearing Screening Project: Effects on speech and language development for children with hearing loss. In Philbin, M.K., Graven, S.N., Robertson, A. (Eds.) The influence of auditory experience on the fetus, newborn, and preterm infant: report of the sound study group of the national resource center: The physical and developmental environment of the high risk infant. Journal of Perinatology, 20(8:2), S132-S137.

Yoshinaga-Itano, C., Coulter, D., & Thomson, V. (2002). Developmental outcomes of children born in Colorado hospitals with universal newborn hearing screening programs. Seminars in Neonatology.

Yoshinaga-Itano, C., & Sedey, A. (2000). Early speech development in children who are deaf or hard-of-hearing: Interrelationships with language and hearing. In C. Yoshinaga-Itano & A.L. Sedey (Eds.), Language, Speech and Social-Emotional Development of Children Who Are Deaf and Hard-of-Hearing: The Early Years, The Volta Review, 100, 181-211.

Yoshinaga-Itano, C., Sedey, A.L., Coulter, D.K., & Mehl, A.L. (1998). The language of early-and later-identified children with hearing loss, Pediatrics, 102, 1161-1171.