NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Berkman ND, Wallace I, Watson L, et al. Screening for Speech and Language Delays and Disorders in Children Age 5 Years or Younger: A Systematic Review for the U.S. Preventive Services Task Force [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2015 Jul. (Evidence Syntheses, No. 120.)
Screening for Speech and Language Delays and Disorders in Children Age 5 Years or Younger: A Systematic Review for the U.S. Preventive Services Task Force [Internet].
Show detailsIn this chapter, we summarize the findings of the 2006 report64 about screening preschool children for speech and language delays. We note the 2006 USPSTF recommendations and comment on the implications of this new synthesis for previous conclusions. Then we discuss the context for these updated results, applicability, limitations of the review and the literature, research gaps, and conclusions.
Overall, the USPSTF issued an I recommendation following the 2006 review, concluding that “the evidence is insufficient to recommend for or against routine use of brief, formal screening instruments in primary care to detect speech and language delay in children up to 5 years of age.”
Speech and language delays affect 5 to 8 percent of preschool children, often persists into the school years, and may be associated with lowered school performance and psychosocial problems. The USPSTF found insufficient evidence that brief, formal screening instruments that are suitable for use in primary care for assessing speech and language development can accurately identify children who would benefit from further evaluation and intervention. Fair evidence suggested that interventions can improve the results of short-term assessments of speech and language skills; however, no studies assessed long-term outcomes. Furthermore, no studies assessed any additional benefits that may be gained by treating children identified through brief, formal screening who would not be identified by addressing clinical or parental concerns. No studies addressed the potential harms of screening or interventions for speech and language delays, such as labeling, parental anxiety, or unnecessary evaluation and intervention. Thus, the USPSTF could not determine the balance of benefits and harms of using brief, formal screening instruments to screen for speech and language delay in the primary care setting.
Summary of Review Findings
Key Question 1
The 2006 report found no studies that met the inclusion criteria to answer the question: “Does screening for speech and language delays result in improved speech and language, as well as improved other nonspeech and language outcomes?” The update changed the wording of the question to specify disorders as well as delays: “Does screening for speech and language delays or disorders lead to improved speech and language outcomes, as well as improved outcomes in domains other than speech and language?” Although one new RCT met our inclusion criteria70,71 by randomizing a large national sample of children who received regularly scheduled care at child health centers to early screening and measuring outcomes in both groups at age 8 years, the study was not included in our analysis because it was rated as poor quality due to various flaws. The most serious flaw is the large attrition, with less than 60 percent of the intervention group being fully screened and only about half of the fully screened group contributing outcomes. The study obtained outcomes on an even smaller percentage of children in the control group. Other flaws included not using a standard instrument for measuring speech and language at the endpoint but rather using a more indirect measure based on self-report, and then not conducting an analysis that considered other possible diagnoses that may have occurred unevenly in the two groups over the long followup period and influenced the findings, including autism spectrum disorder and other developmental or emotional delays or disorders. Nevertheless, we acknowledge the difficulty in conducting and maintaining a study of this kind (Table 9).
Table 9
Summary of Evidence.
Key Question 2
Prior Review Findings on Screening
The 2006 review examined several aspects of the question of whether screening evaluations in the primary care setting accurately identify children for diagnostic evaluations and interventions.
The first was whether identification of risk factors improves screening. The 2006 review found 16 studies that met its inclusion criteria and concluded that a small number of characteristics, such as male sex, family history, and parental education, were linked to language delay. We discuss these and other risk factors as part of Contextual Question 2. However, we found no studies that used these risk factors to improve accuracy, nor did we find any studies that examined the role of child race and ethnicity on the accuracy of speech-language screening results.
The second and third subquestions addressed screening techniques, how screening differed by age, screening accuracy, and how accuracy differed by age. The 2006 review evaluated the performance characteristics of instruments to screen for speech and language delay. It included studies of instruments that took 30 minutes or less to administer. The included studies were generally focused on children age 5 years or younger who did not have a previously diagnosed condition such as autism, mental retardation, or orofacial malformations.
The 2006 review included a total of 43 studies that described 32 instruments taking no more than 30 minutes to administer. (Note: The 2006 review counted a study90 with two substudies as two separate studies; we count it as one study.) In the good- or fair-quality studies of instruments, sensitivity for detecting a speech or language delay ranged from 17 to 100 percent, and specificity for detecting typical language ranged from 45 to 100 percent. The previous review further identified the Early Language Milestone Scale, Clinical Linguistic and Auditory Milestone Scale, LDS, SKOLD, and Levett-Muir Language Screening Test as the five instruments with the highest sensitivity and specificity. However, the reviewers concluded that the best methods and ages for screening could not be determined from the studies included in the review because of a number of factors (e.g., instruments were not designed for screening or lacked comparisons across populations, venues were outside the primary care setting, speech and language delays have no gold standard reference). The fourth subquestion examined the optimal ages and frequency for screening. No studies addressed this question.
The USPSTF found insufficient evidence that brief, formal screening instruments that are suitable for use in primary care for assessing speech and language skills can accurately identify children who would benefit from further evaluation and intervention.
Implications of the New Synthesis on Prior Conclusions on Screening
Of the 42 studies (43 articles) identified in the 2006 review, 16 studies (17 articles) continued to meet the inclusion criteria for this update and were determined to be of good or fair quality.54,91–95,97–100,102–106,108,122 In the current review, we included only studies that provided accuracy statistics or data that allowed us to calculate accuracy. We were stricter in determining whether a study met the population inclusion criteria. We excluded studies if some children in the population were outside the age range or had a previously diagnosed condition and the study did not stratify for age and condition. We used the most recent USPSTF criteria for determining quality of studies. To these previously included 16 studies, we added an additional eight newly found studies (nine articles). In doing so, we were able to address one identified limitation of the previous review—namely, the lack of studies comparing the same instrument in different populations. We also examined the studies in our review by considering parent-rated screening instruments separately from those that are administered by trained examiners, including those in primary care. The current review also included one study that examined screening for preverbal language88 (Table 10).
Table 10
Risk Factors: Earlier Speech and Language Concerns Through Parental Education.
Altogether, 14 studies (in 16 articles)54,72–78,88,95,98,102–104,108,122 examined the accuracy of screening instruments in which parents rated their young children’s speech and language skills. Sensitivity ranged from 50 to 94 percent, and specificity ranged from 45 to 96 percent.
Nine of the parent-rated screening studies (11 articles) examined three instruments widely used in the United States—the ASQ Communication Domain, CDI, and LDS. Several of these instruments exhibited LRs suggesting that there is a moderate to large increase in the likelihood of language delays in children who screen positive or a moderate to large decrease in the likelihood of language delays in children who screen negative. Although the LRs tended not to be consistent across all studies that included a particular instrument, both the positive and negative LRs were moderate to large in two studies examining the CDI.74,76,77 Another parent-reported instrument, the ITC, is also used in the United States but in a somewhat younger population (i.e., ages 6 to 24 months).
Because the ASQ and CDI have versions for infants, toddlers, and preschool-age children, we were able to examine the accuracy of the instruments at different ages. Studies examining the ASQ in children ages 2 years,74 3.5 years,73 and 4.5 years72 reported comparably low sensitivity at all three ages (ranging from 50% to 59%), and better specificity for the 2- and 3.5-year-old samples (95% and 92%) than for the older sample (79% and 83%). These results suggest that use of the ASQ for screening for language delays, especially in preschool-age children, may result in many false-negatives. Comparisons indicated that sensitivity and specificity of the toddler and preschool CDI versions were fairly close, suggesting that the CDI is robust in its ability to detect a language delay across the toddler and preschool years. The study examining the ITC subdivided the sample into two age groups—younger and older toddlers. Sensitivity, specificity, and LRs were nearly identical in the two age groups.
However, studies indicate that the screening accuracy of parent-reported instruments diminishes somewhat over time. Two studies (in four articles)76,77,98,122 examined the accuracy of screening instruments completed by parents of children age 2 years in relation to language assessments administered to the children at both age 2 and 3 years. In one study122 that evaluated the LDS, sensitivity was reduced after a year, but specificity remained the same. In the second study,76,77 which evaluated the German version of the CDI, sensitivity was about the same at ages 2 and 3 years, but specificity was reduced at 3 years. Forty-four percent of the children who had been classified as having a language delay at age 2 years had typical language at age 3 years. The reduction in specificity over time illustrates the finding that some children with language delays will “catch up” and display more typical language skills with development.3
We reviewed 12 studies72,86,91–94,97,99,100,103,105,106 that examined the accuracy of screening tests administered by trained examiners, all but two of which require direct testing of the child. The variability in accuracy across these instruments was greater than the variability across the parent-rated instruments; sensitivity ranged from 17 to 100 percent and specificity ranged from 48 to 100 percent. Several of the trained examiner instruments also had moderate or large LRs, indicating an increase in the likelihood of a language delay for children who screened positive or a decrease in the likelihood of a language delay for those who screened negative. Many of the screening instruments performed as well as the parent-rated instruments, but aside from the DDST (now known as the Denver II), most are not used in primary care offices and would require a dedicated, trained professional or paraprofessional to directly test the child. The study of the original version of the DDST found excellent specificity (100%) but poor sensitivity (46%); no studies provided information on the accuracy of the language component of the current version.
Because few studies of screening instruments administered by trained examiners examined the same instrument with different populations or with different ages, it is unclear how instruments for multiple ages fare more broadly or whether there is an optimal age for screening. We were able to examine cross-age accuracy with only two instruments that are published and used in the United States. Two studies92,105 examined the FPSLST in children age 3 years and in children ages 4 to 5 years. Specificity was greater in the study105 in older children (ranging from 85% to 95% in two samples with two reference measures each) than the study of younger children92 (81%), but sensitivity was generally low in both studies (17% to 74% in the older cohort and 60% in the younger group).
A second study,93 which reported on an instrument (the SKOLD) designed for children ages 2 to 5 years, provided separate accuracy statistics for each of three age groups (ages 2.5 to 3, 3 to 3.5, and 3.5 to 4 years) and for speakers of African American dialect and SE, and generally found excellent sensitivity and specificity for each age and linguistic group. Although the accuracy of this instrument suggests that it is a good candidate for screening children ages 2.5 to 4 years, particularly speakers of African American dialect, its widespread utility may be limited by the necessary training. The developers of the instrument93 caution, “For successful administration and scoring, screeners need an understanding of normal and impaired language development, black English, and familiarity with administration and scoring procedures of SKOLD. Ideally, paraprofessionals should be trained by speech-language pathologists in the above areas.”
The 2006 review concluded that, despite the availability of brief screening instruments, screening for speech and language delays has serious limitations. For example, optimal screening methods had not been established, an accepted gold reference standard was lacking, data comparing an instrument across different populations and different ages were limited, and sensitivity and specificity varied. With the addition of eight newly identified studies and the exclusion of 14 of the 35 studies from the 2006 review, the evidence in this review differs somewhat. We identified several studies that speak to the accuracy of the CDI and LDS in multiple populations and multiple ages. Although there is no gold standard for speech and language assessment, the reference standards used in these studies are well-regarded instruments that speech-language pathologists routinely use. The sensitivity and specificity of these instruments are acceptable,163 and because parents complete them, adopting them in a screening program should not burden a primary care practice with training someone in test administration. The findings related to the CDI and LDS point to the importance of involving parents in identifying young children with speech and language delays and disorders. In addition, each of these instrument focuses on language, and the more extensive information that parents provide specifically related to their children’s language skills may help explain the fact that the CDI and LDS are more accurate in identifying children with speech and language delays or disorders than broad-based instruments that include fewer items to screen for speech and language problems. In summary, this synthesis yields evidence that two parent-rated instruments, the CDI and LDS, would likely be interpreted with little difficulty in the primary care setting and can accurately identify children for diagnostic evaluations and interventions. Our findings do not address the importance of speech and language assessments for children with intellectual disabilities, sensory or motor impairments, or structural abnormalities of the head or neck.
Key Question 3
The 2006 review found that no studies addressed the question of adverse effects of screening. The authors suggested potential adverse effects, such as false-positive and false-negative results, which would have deleterious consequences, such as erroneously labeling a child with typical speech and language as having a delay or disorder, or missing a child with a true speech and language impairment who then fails to benefit from timely intervention services. We also found no evidence to address this question.
Key Question 4
The 2006 study found no studies examining the role of enhanced surveillance by a primary care clinician once a child is identified as possibly having a speech and language delay. We asked a related question: “Does surveillance (active monitoring) by primary care clinicians play a role in accurately identifying children for diagnostic evaluations and interventions?” We found no evidence to answer this question.
Key Question 5
Prior Review Findings on Speech and Language Outcomes of Treatment
In the 2006 review, studies evaluated the effects of individual or group interventions directed by clinicians or parents that focused on specific speech and language domains. These domains included expressive and receptive language, articulation, phonology, and syntax. Interventions were short term, commonly lasting from 3 to 6 months, and took place in speech and language specialty clinics, community clinics, homes, and schools. Outcomes were measured by subjective reports from parents and by scores on standardized instruments.
Eight fair- or good-quality studies focusing on the treatment of children age 3 years or younger found mixed results, with five studies reporting improvement on a variety of speech and language domains, including clinician-directed treatment to improve expressive and receptive language delay, parent-directed therapy to improve expressive delay, and clinician-directed therapy to improve receptive auditory comprehension. Results were also mixed for seven fair-quality studies focusing on children ages 3 to 5 years; five found significant improvement and two reported no differences.
Implications of the New Synthesis on Prior Conclusions on Speech and Language Outcomes of Treatment
The previous review reported significant effects of treatment on speech and language outcomes across the age range of 2 to 5 years, although significant findings were not universal across included trials. We did not include six of the previously included studies in our review because we considered them to be comparative, examining the relative merit of a new intervention compared with treatment as usual. One newly identified trial was unique in examining the treatment of children who were all younger than age 2 years;128 no significant effects on language outcomes were detected, but it is not possible to evaluate whether this finding was related to the young age of the children or other factors.
The evidence of maintained benefits of a school-based language treatment program for preschoolers with low language scores125 is an important addition to this update. The current review also adds evidence from two small trials for the potential effectiveness of treating preschool children who stutter, with both trials testing the same treatment, the Lidcombe Program of Early Stuttering Intervention.126,127 Thus, there is some cumulative evidence for benefits of targeting outcomes in the areas of language (six of 11 trials reporting significant positive results), speech sounds (six of eight trials reporting significant positive results), and fluency (two of two trials reporting significant positive results) among toddlers and preschoolers with speech and language delays or disorders.
The addition of new evidence related to the treatment of speech and language problems in children does little to clarify the characteristics of effective treatments. Two of the three largest trials included in this review, and the only two of good quality, reported limited to no benefits associated with treatment.128,135 A potential explanation for these results is that the trials examined the lowest-intensity treatments evaluated in any of the studies included in our review (about 6 hours of individual speech and language therapy in one case and 12 hours of parent group meetings in the other). The addition of findings from a second trial of low-intensity treatment provides more reason to question the benefits of such low-intensity treatments for young children with speech and language delays. However, because the heterogeneity across the included studies related to many factors in addition to intensity, it is not possible to be certain that treatment intensity explains the null findings. In fact, one trial entailing only 13 to 16 hours of parent group meetings produced large effects on language outcomes,136 and another study that provided the second most intensive treatment of any of these trials (individual treatment) found no main effects on child language or intelligibility.129 Thus, we are unable to comment on the minimum amount of intensity required for a treatment to likely be effective, and intensity alone cannot account for either positive or null findings among these trials.
This review also includes a study that identified an interaction between a child’s baseline characteristic and the response to a particular treatment strategy (recasting).129 Although the generalizability of this specific finding is limited, it is the only evidence related to the benefits of matching treatments to individual child characteristics. Given the improbability that a treatment for any condition will benefit everyone with the condition, there is a need for such evidence.
We grouped outcome measures into the broad categories of language, speech sounds, and fluency. We generally would anticipate correlations of at least moderate size among different measures within one of these broad categories. For example, children who are slow to acquire vocabulary generally will also have relatively short mean lengths of utterances; children who make many errors on consonant sounds generally will be less intelligible than children who make few errors. However, the strength of these correlations for any given subpopulation of children with speech or language disorders is an empirical question—that is, we cannot assume that one measure within a category such as language will be equivalent to another measure in that category, or that the effects of a treatment on one measure will be generalizable to other measures within the outcome category. Across the trials that report outcomes within the categories of language and speech sounds, diverse outcome measures were used, with no single measure used in a majority of trials. For example, among the trials that report on language outcomes, the most often used measure was mean length of utterance, an index of expressive language structural complexity; however, this measure was used in only four of 10 trials reporting language outcomes.
Trials also varied in the way outcomes were reported. For example, most trials omitted information about effect sizes, and some did not report the statistics needed to compute effect sizes. In a few cases, outcome measures were reported that speak directly to clinical significance, such as the relative number of children in treatment versus control who reduced stuttering to less than 1 percent of syllables126 or improvement on the clinical criteria used for study entry.135 In most cases, however, outcomes were not reported in terms that are easily interpreted with respect to their clinical or functional effects.
In summary, the majority of the 13 trials that met inclusion criteria for this review offer evidence supporting the effectiveness of treating speech and language delays and disorders in young children. Positive findings have emerged from studies examining various service delivery models, including individual and small-group treatment, and various intervention agents, including parents supported/trained by professionals, speech-language pathologists, and trained teaching or therapy assistants. Some included trials reported null findings for language and speech sound outcomes. Confident interpretation of this body of evidence on the treatment of speech and language delays is limited by multiple factors, including 1) the small size of many of the trials, which constrains the examination of moderators and mediators of treatment effectiveness; 2) the lack of replicated positive findings for any treatment approach except the Lidcombe program for stuttering; 3) the wide variability across trials in the age of children treated and intervention agents (e.g., speech-language pathologists, teaching assistants, parents, research staff), intensity, content, and strategies; 4) the relatively small number of trials that have examined manualized treatments or otherwise provide enough details of the treatment approach to permit replication; 5) a corresponding lack of reporting of treatment fidelity in many trials; and 6) the lack of common outcome measures and the inconsistency in how results are reported across trials. Because of these constraints, the current body of evidence does not lend itself to meta-analysis and offers little guidance on the specific factors associated with effective treatments for young children with speech and language delays or disorders.
Key Question 6
Prior Review Findings on Other Nonspeech and Language Outcomes of Treatment
In the 2006 review, four good- or fair-quality studies included functional outcomes other than speech or language. However, the interventions and outcomes varied across the studies and lacked appropriate comparison cohorts. The 2006 review also examined “additional” outcomes and cost-effectiveness issues but did not find any studies that addressed these questions.
Implications of the New Synthesis on Prior Conclusions on Other Nonspeech and Language Outcomes of Treatment
As in the previous review, few trials examined other outcomes of speech and language treatment in children (i.e., outcomes beyond speech and language). One new trial provided evidence supporting the contributions of oral language to proficiency in early reading comprehension.125 Although this is widely assumed to be the case based on prior longitudinal correlational research, the trial provides better evidence for a causal relationship. The other outcomes measured in the four trials that included nonspeech and nonlanguage outcomes are disparate and thus allow no synthesis of findings across studies.
Key Question 7
The 2006 review found no studies that addressed this question. The update found insufficient evidence to address this question (one outcome in each of two studies).
Applicability of Findings
The included studies have mixed applicability for primary care settings. In a few studies, screening occurred in primary care settings,78,86,99,100,103 and in two cases, primary care providers administered the screening to the children.86,103 It should be noted that none of these studies occurred in the United States, and the extent to which conclusions reached from screening in primary care settings in Sweden, Australia, and the United Kingdom are transferable to primary care settings in the United States is not known.
Other settings for screening included early childhood care centers, preschools, and elementary schools; developmental evaluation centers; university research laboratories; and hospitals. Whether it is realistic for screening to occur in another setting and to have the results sent to a primary care provider is not known, although with training and supervision, a staff member in the primary care setting could administer some of the screening instruments. In some studies, parent-rated instruments were completed at home and mailed or brought to the investigator, and in other cases, they were completed when the child was seen for the administration of the reference test. Either of these settings appears to be applicable to the primary care setting. However, aside from the ASQ, which is used in the primary care setting, parent-rated instruments have not been widely adopted in the United States.
Most of the intervention trials (eight of 12) were conducted in countries other than the United States; three in the United Kingdom,125,135,136 two in Australia,127,128,134,137,138 and one in New Zealand.126 As with the screening studies, whether conclusions reached from trials conducted in countries with different medical, health insurance, and school systems are applicable to the United States is debatable.
Many screening studies examined accuracy in only a subset of children—those who scored positive on the instrument and either a random selection of children who scored negative or a separate cohort of children with typical language. The applicability to an unselected group of children in a primary care setting is not known. However, it is highly likely that the positive and negative predictive values that we calculated are inaccurate because of an incorrect prevalence estimate. An important next step is to conduct screening studies in a general population of preschoolers, in which the prevalence of language delay is closer to the 8 percent found in prevalence studies.
There is also mixed applicability for the interventions in community settings. One study explicitly tested the effectiveness of immediately referring young children identified with speech or language delays/disorders to usual community speech-language therapy services compared with a control condition (watchful waiting).135 This test is valuable in providing information on whether it is helpful for a primary care provider to refer children with speech and language delays or disorders for speech-language pathology treatment. However, the question this study answers is similar to asking, “Is it effective for a person with symptoms of illness to go to a physician?” Speech-language pathology services entail a diversity of treatments that are individualized to a child’s symptoms and ability to participate in different types of interventions, and will also be influenced by the training, experiences, and preferences of the speech-language pathologist serving the child. The rigor of an RCT is unlikely to be relevant to clinical treatment, where it is important to recognize the individual’s needs.
Some trials evaluated manualized programs for which resources and training are available (e.g., Lidcombe Program of Early Stuttering Intervention, Hanen Parent Program, the cycles approach to phonological therapy). Using the Hanen program requires certification, which is relatively expensive (Appendix F). The treatments used in most trials would be difficult to replicate in the community because of insufficient published information on the program, as well as the difficulty that community practitioners have in accessing information in many peer-reviewed journals.
Context for Findings
Contextual Question 1. Techniques for Screening for Speech and Language Delays or Disorders and Differences by Age and Cultural Background
In the 2006 review, the question concerning techniques for speech and language screening was examined as part of addressing accuracy in KQ 2. The 2006 review, which considered all techniques taking 30 minutes or less to complete as having potential for screening, found 43 articles describing the characteristics of 51 speech and language screening instruments. (Note: The 2006 review counted a study164 with two substudies as two separate studies; we count it as one study.) The conclusion was that there was no gold standard and that studies using these instruments provided limited details about the participants.
In the current review, we limited our focus to instruments that either take no more than 10 minutes to administer in the primary care facility or, if administered outside the primary care practice, could be interpreted in 10 minutes or less. We also limited it to instruments that we used to address KQ 2. We found 20 studies that described instruments that met criteria for addressing KQ 2. Descriptions of the screening instruments are in Table 4. Both parent-rated and trained examiner tools are included, with the latter appropriate for children who are somewhat older.
Contextual Question 2. Risk Factors Associated With Speech and Language Delays or Disorders
We searched the evidence for consistent, reliable, and valid risk factors that clinicians could use to identify children at highest risk for speech and language delays (Tables 10 and 11). The ability to reliably stratify children by risk could promote efficiencies in screening activities, ideally assisting in earlier identification of children with speech and language disorders that would translate into earlier intervention and improved speech and language outcomes. Predicting which children are at high risk for speech and language disorders is complicated, however, by the many types of speech and language disorders, heterogeneity in populations across studies, inconsistent identification of potential risk factors across studies, and inconsistent adjustment for potential confounders (i.e., other characteristics that may simultaneously be related to a child’s risk for a speech and language problem). To adjust for confounders, all but six studies165–170 included multivariate analyses of cohorts or a case-control design. We limit our report of cohort studies to their multivariate findings, where available.
Table 11
Risk Factors: Low Birth Weight Through Other Associations.
Evidence for valid risk factors is also limited by lack of discussion of causal links describing how an associated risk factor may lead to a speech or language delay. For example, male sex is listed as a risk factor for speech and language delays in a number of studies, but it is unclear how and why male sex may contribute to speech and language delay. We aimed to update the evidence on risk stratification.
Our review includes 38 studies conducted in 28 cohorts and one review of studies on characteristics of late-talking toddlers.171 Twenty-one of the cohorts were English-speaking and seven were non-English speaking.
Among studies in English-speaking populations, sample sizes ranged from 60165 to 11,383172 subjects. Most studies evaluated outcomes measuring language delay with or without speech delay. Speech and language outcome domains included expressive and receptive language and vocabulary, number of words, early language and communication difficulty, stuttering, and parental report of speech and language impairment. Male sex was a significant risk factor in 11 of the 14 studies examining it.83,167,173–181 Only one large cohort study of children age 5 years in Britain reported that male sex decreased the probability of both SLI and nonspecific language impairment.46 In these multivariate analyses, proximal factors such as family overcrowding, the child being in preschool, and the parent being a poor reader were found to be significant risk factors for poorer outcomes. Family history of speech and language impairment was also a consistent risk factor, significantly associated with delay in seven of nine studies.166,168,173,175,176, 178,182 However, family history was generally measured by self-report and described nonspecifically (i.e., family members who were late talkers or had language disorders and speech problems). Family history was not found to be a risk factor for stuttering onset in one cohort that measured outcomes at ages 3 and 4 years.177,181 Parental education had an inconsistent association with speech and language delay. Nine of 15 studies reported a significant association between lower parental education level (either mother or father) and speech and/or language delay.46,167,168,172,173, 179,182,183 The study of risk factors related to stuttering onset found that stuttering was associated with the mother having a higher level of education.177,181 Other risk factors identified in two or more studies among English-speaking populations included lower socioeconomic status, earlier identified speech and language delays, poorer parenting practices, greater parental stress, and poorer maternal mental health. Minority race was significant in two of the five studies that examined it.180,184
Four studies examined speech and language delays in preterm birth cohorts, measured at ages 18 months to 4 years; studies mostly examined nonoverlapping sets of risk factors.83,180,184,185 However, two of the studies found that males were at higher risk for poor outcomes.83,180 Perinatal risk factors were inconsistently measured across other cohorts and included prematurity, low birth weight, being born late in the family birth order, less breastfeeding, maternal alcohol consumption during pregnancy, and younger maternal age at birth. Perinatal factors determined to be risk factors in at least one study that measured them were maternal binge drinking, prematurity, low birth weight, and younger maternal age.
The 13 studies assessing risk in non–English-speaking populations, conducted in eight cohorts, included sample sizes from 24 to 42,107 subjects and evaluated various types of delay, including vocabulary, communication, word production, speech, stuttering, and expressive and receptive language. Significant associations were reported in five studies in four cohorts evaluating risk associated with male sex57,58,186,187 and two studies evaluating family history of speech and language concerns.187,188 Perinatal risk factors were examined in a Netherlands study comparing a preterm and term cohort; the study found prematurity to be associated with communication delays at age 4 years.186,189–191 Several studies, including one based on a large Finnish cohort (n=8,276) found that low birth weight was also associated with poorer speech and language outcomes.58,192,193 Other associated risk factors reported less consistently include parental education level and family factors such as size and overcrowding. These studies did not find associations with the mother’s stuttering or speaking style or rate, mother’s age, or child’s temperament.
A review of late-talking toddlers ages 18 to 34 months found a statistically significant association with family history of language disorders, socioeconomic status, and parental stress but no association with parents’ education level.171 The review identified some of the challenges inherent in identifying risk factors for speech and language disorders. First, some studies are limited to children with an expressive vocabulary delay, excluding children with receptive language deficits, even though many children age 2 years have deficits of both comprehension and expression.171 Also, the instruments used to measure expressive vocabulary across studies are inconsistent. The review author concluded that future research should take into account the lack of homogeneity observed within the population of children with a vocabulary delay at age 2 years and consider a multifactorial perspective of child development to further understand this phenomenon.
Although more recent studies examine more proximal risk factors, such as social determinants of health, rather than distal risk factors such as race, speech and language studies continue to have dissimilar inclusion and exclusion criteria and assess dissimilarly measured risk factors and outcomes. Because of these dissimilarities, it is difficult to determine which of these more proximal factors may be the attributable factor for the speech and language disorder.
Contextual Question 3. Role of Primary Care Providers in Screening in Children Age 5 Years or Younger That Is Performed in Other Venues
The 2006 review did not address the role of primary care providers in screening in children age 5 years or younger that is performed in other venues (such as Head Start or preschool). We found two studies91,194 that examined screening in preschool venues; however, neither discussed the role of the medical provider. Thus, we have no evidence on the interface between this aspect of the screening process and primary care providers.
Limitations of the Review
The 2006 review identified a number of limitations of the literature base, including a lack of studies specific to screening; inconsistencies in terminology across studies; assessment instruments and interventions that address specific aspects of language development rather than a common, global indicator of speech or language; and difficulties evaluating the effects of complex interventions, especially those related to screening. Many of these issues continue to plague the field.
We found additional limitations. One difficulty in drawing conclusions about whether screening for speech and language delays or disorders leads to improved outcomes is the lack of well-designed studies that address this overarching question. The ideal study would randomize children to screening and no screening; follow up with those who are screened, both positive and negative results; and at some later point, assess all children, while collecting enough data to understand what occurred during the intervening time. Although de Koning et al70 designed a randomized trial of screening versus no screening, their study had a large attrition, and they did not use a uniform method of assessing language outcomes. Thus, this trial did not provide evidence about screening.
We are beginning to answer the question of whether screening can accurately identify children and have identified some candidate measures. Yet many studies included in the review are less than ideal because they include selected groups of children; that is, many studies include a sample of children with and without language delays. Use of such predetermined samples makes it difficult to examine whether screening is accurate in unselected samples, the likely target for such activities. In addition, because such studies tend to have a greater number of children with language delays or disorders, estimates of prevalence are skewed, leading to inaccurate estimates of positive and negative predictive values. Only a few studies examined how well screening instruments detect speech and language disorders over the long term. Such studies are critical in calculating the real benefit of early detection. Examining long-term outcomes may identify children with a language impairment rather than a transient language delay, enabling providers to target intervention resources to those who have a greater need.
We also encountered studies that purported to screen for speech and language delays but used instruments that were not specific to linguistic skills, instead screening for developmental delays. Other studies validated instruments by examining their accuracy in relation to other instruments, not to recognized reference standards. The issue is not that the instruments are deficient; rather, it is the study designs that are deficient.
One limitation of the included intervention trials was that the studies often did not include information on whether the children were receiving community services for their speech and language symptoms outside of the study. Exceptions126,129,135 provided information about community speech and language services. Understanding what services children in both arms of the intervention study receive is critical to interpreting treatment effects, or lack thereof.
It is challenging, at least in the United States, to conduct an RCT comparing speech and language treatment versus no treatment in children with severe enough symptoms to be identified as having a speech and language delay or disorder. Under IDEA, children from birth to age 5 years with special needs are entitled to services through the early intervention programs in their resident State. States have some latitude in setting eligibility criteria for these services, and as funding has become tighter, the trend is to limit eligibility, requiring that the children served have more severe problems. The result of this law and the associated policies in the United States is that children with more significant problems will likely receive public early intervention/preschool services, making it unlikely that researchers could conduct an RCT that compares children receiving a speech-language treatment with children not receiving any speech-language treatment. Although it may be possible to conduct such trials in children who have milder symptoms that do not qualify them for public services, such trials would not be representative of the full population of children with speech and language delays and disorders, and would largely exclude the children with the greatest needs.
Across the included trials, the majority of control groups offered intervention to children on a delayed schedule. This condition would likely make parents more willing to consent to their child participating in an RCT, but constrains our ability to examine long-range outcomes for treated versus untreated children.
Future Research Needs
In order to sufficiently answer the question, “Does screening for speech and language delays or disorders lead to improved speech and language outcomes, as well as improved outcomes in domains other than speech and language?”, studies need to be specifically designed and executed for this purpose. Neither the current review nor the 2006 review could answer this question directly; rather, both addressed the question by considering subquestions. This research gap presents an opportunity for a large study to test the efficacy of systematic routine screening for speech and language delays and disorders compared with not implementing routine screening in primary care. In tandem with this, the field would benefit from a study to examine the feasibility of speech- and language-specific screening as part of the more general developmental screening that is already recommended.49 Better designed studies of risk factors, including children’s background characteristics, would also facilitate clinicians’ ability to identify children who are at highest risk for speech and language delays.
Only a few of the screening studies included children who were speakers of languages other than English. Future studies should include such children in studies of both language screening and language intervention.
Given Federal mandates under IDEA that all children with a documented speech or language delay receive early intervention, going forward, it may be difficult to conduct RCTs to examine the efficacy of interventions. Future research protocols may adopt quasiexperimental designs of sufficient rigor to answer intervention questions. For example, regression discontinuity designs seem applicable to addressing treatment efficacy because they can be used when there is a cutoff in a continuous measure that is used to identify children who are eligible for the treatment. The effect size is evaluated at the point of discontinuity, dividing those who met by those who did not meet eligibility criteria. Well-designed and implemented regression discontinuity designs can now meet standards for rigor without reservation for the U.S. Department of Education, whose Institute of Education Sciences sponsors evaluations of evidence.
We recommend that stakeholders with an interest in screening develop research agendas and funding targeted to answer the important questions that could not be addressed in this review. To build the necessary evidence that screening children for speech and language delays and disorders can lead to improved outcomes, it will be necessary to design and conduct studies that can specifically address that question.
Conclusion
We found no evidence to answer the overarching question of whether screening for speech and language delays or disorders leads to improved speech and language outcomes. However, this should not be interpreted to mean that screening for speech and language delays is not beneficial; rather, we do not know whether there is a benefit because of a lack of evidence to answer this question. The studies from the 2006 review, as well as the newly identified studies, suggest that some screening instruments for detecting speech and language delays and disorders are accurate. Although these parent-rated instruments require only that the primary care provider interpret the findings, studies have not examined how receptive providers are to doing so. As in the 2006 review, we found no studies that addressed the harms of screening for speech and language delays, nor did we find any evidence about the role of enhanced surveillance by a primary care clinician once a child is identified as possibly having a speech and language delay. Building on the studies identified in the 2006 review, we found evidence supporting the effectiveness of treating speech and language delays and disorders in children. However, the body of evidence does not provide guidance regarding the specific factors associated with effective treatments for young children with speech and language delays or disorders. Finally, this review found no evidence relating to the harms of treating speech and language delays or disorders.
- Discussion - Screening for Speech and Language Delays and Disorders in Children ...Discussion - Screening for Speech and Language Delays and Disorders in Children Age 5 Years or Younger
Your browsing activity is empty.
Activity recording is turned off.
See more...