Two-Year Outcomes of a Population-Based Intervention for Preschool Language Delay: An RCT
OBJECTIVE: We have previously shown short-term benefits to phonology, letter knowledge, and possibly expressive language from systematically ascertaining language delay at age 4 years followed by the Language for Learning intervention. Here, we report the trial’s definitive 6-year outcomes.
METHODS: Randomized trial nested in a population-based ascertainment. Children with language scores >1.25 SD below the mean at age 4 were randomized, with intervention children receiving 18 1-hour home-based therapy sessions. Primary outcome was receptive/expressive language. Secondary outcomes were phonological, receptive vocabulary, literacy, and narrative skills; parent-reported pragmatic language, behavior, and health-related quality of life; costs of intervention; and health service use. For intention-to-treat analyses, trial arms were compared using linear regression models.
RESULTS: Of 1464 children assessed at age 4, 266 were eligible and 200 randomized; 90% and 82% of intervention and control children were retained respectively. By age 6, mean language scores had normalized, but there was little evidence of a treatment effect for receptive (adjusted mean difference 2.3; 95% confidence interval [CI] –1.2 to 5.7; P = .20) or expressive (0.8; 95% CI –1.6 to 3.2; P = .49) language. Of the secondary outcomes, only phonological awareness skills (effect size 0.36; 95% CI 0.08–0.65; P = .01) showed benefit. Costs were higher for intervention families (mean difference AU$4276; 95% CI: $3424 to $5128).
CONCLUSIONS: Population-based intervention targeting 4-year-old language delay was feasible but did not have lasting impacts on language, possibly reflecting resolution in both groups. Long-term literacy benefits remain possible but must be weighed against its cost.
- CELF-P2 —
- Clinical Evaluation of Language Fundamentals, Preschool Edition
- CI —
- confidence interval
What’s Known on This Subject:
Preschool language delay predicts poorer academic performance, employment opportunities, and social relationships. Language for Learning, a systematic, population-based intervention for 4-year-olds with low language, is feasible, acceptable and has short-term benefits, but long-term benefits are unknown.
What This Study Adds:
Population ascertainment at age 4 followed by a yearlong, one-on-one home program benefited phonological skills (an important literacy determinant) at age 6, but not the primary language outcomes. To be cost-effective, future follow-up would need to demonstrate lasting academic benefits.
Language delay is linked to educational difficulties and well-being, and potentially has lasting implications through to adulthood.1,2 Indeed, language and cognitive development have been picked out as key “life-chance” indicators at 3 years of age.3
Developing effective interventions is therefore a priority. Systematic reviews support intervening for expressive language and phonological development for children with pronounced, clinical levels of difficulty.4,5 Unfortunately, benefits to receptive language difficulties are less clear, and some interventions may even be contraindicated.6 A related area of development that clearly shares common factors, is phonics, and specifically the role that this plays in the development of preliteracy skills.7,8 Oral language and literacy are closely linked from a developmental perspective, and it is a reasonable assumption that intervening for the former may also benefit the latter. This has led to interventions that target both oral language and literacy. One recent study showed that phonic intervention improved decoding, and working on oral language had a significant effect on reading comprehension.9
This rapid growth in higher-quality trials has undoubtedly enhanced knowledge of how to intervene clinically for language delay. Despite this, policy-makers still lack useful guidance on how to intervene at the community or societal level. This reflects a number of weaknesses in the literature. Heterogeneity is high, most trials are underpowered, optimal timing for intervention remains unknown, and few studies have addressed flow-on benefits to naturalistic use of language (eg, narrative skills) or child well-being and behavior, both associated with language difficulties during the school years. The paucity of economic analyses and unclear ascertainment strategies is also problematic because systematic population ascertainment (regardless of whether a screening or more definitive measure is used) would be expected to yield more false-positives and mild cases than clinical presentation, which could profoundly alter cost-versus-benefit equations.
We report the final outcomes of the Language for Learning study, a novel population-based trial targeting 4-year-olds and designed so that the intervention could feasibly be rolled out in the community and address identified evidence gaps.10 Although there is no consensus as to the optimal age for population-based intervention, we chose 4 years because at earlier ages, language delay is both nonspecific and insensitive in predicting later language delay11–13 (ie, low sensitivity), whereas school-age intervention may be too late.14 The program was designed to be sufficiently standardized for population-based delivery by nonspecialists, yet flexible enough to meet the varying needs of children at this age. We aimed to determine whether a 10-month intervention targeting 4-year-olds ascertained with language delay would improve language and associated outcomes at age 6 years. Interim results at age 5 years (immediately postintervention) suggested improved phonological awareness and letter knowledge, as well as weak evidence of benefit to expressive, but not receptive, language.15
We hypothesized a priori that at 6 years (well over a year after the program ended) lasting benefits to the intervention group would include better mean scores on the following:
standardized measures of expressive and receptive language (primary outcomes) and
related secondary outcomes of phonological skills, receptive vocabulary, literacy skills, narrative skills, pragmatic language skills, behavior, and health-related quality of life.
We also report program costs and health service use to determine the cost comparison between the intervention program and usual care.
The Language for Learning (ISRCTN03981121) randomized controlled trial was nested within a cross-sectional population-based ascertainment of language delay at age 4 years in 8 of the 31 local government areas comprising greater Melbourne (population 4.0 million in 201116), Australia. Participants were drawn from 2 earlier low-intensity language and literacy promotion trials with null findings. The Let’s Learn Language13 trial was conducted within 3 local government areas selected to represent lower, middle, and upper socioeconomic areas according to Australian census-derived disadvantage indicators.17 The 5 local government areas selected for the Let’s Read18 trial comprised a convenience sample selected from those in the lowest tertile for disadvantage. Methods are detailed in the published protocol10 and are briefly outlined here.
As children reached their fourth birthday (Let’s Read, March–July 2010; Let’s Learn Language, May–December 2010), all 1661 parents were sent a questionnaire and reconsent form. Research assistants conducted a formal language assessment with children of consenting parents either in the home or at a convenient local venue (Fig 1).
The eligibility criterion was expressive and/or receptive language scores >1.25 SD below the normative mean of 100 on the Clinical Evaluation of Language Fundamentals, Preschool Edition (CELF-P2).19 Exclusion criteria were known intellectual disability, major medical condition, hearing loss >40 dB in the better ear, autism spectrum disorder, or parents with insufficient English to complete written questionnaires without an interpreter.
Randomization and Blinding
Participants were randomized in a 1:1 ratio to the intervention or usual care (control) arms using a computer-generated random number sequence provided by an independent statistician. Randomization was stratified by previous trial (Let’s Read or Let’s Learn Language) and nature of language problem (receptive, expressive, or both receptive and expressive) and blocked within each stratum using randomly permuted block sizes in a nonsystematic sequence. Allocation was concealed using sealed opaque envelopes held by an independent researcher. Outcome assessors were blind to group allocation, but once allocated, participants could not be blinded.
Figure 1 summarizes the intervention’s timeline and content (for more details, see the trial’s published protocol10 and interim results15). The replicable program was manualized (modeled on that of Boyle14) and sufficiently standardized for population-based delivery by nonspecialists, yet flexible enough to meet the varying needs of children at this age. The 10 trained “language assistants” who delivered the intervention were university graduates with psychology and sociology qualifications who were supervised throughout by a speech pathologist (N.Z.).
Briefly, both oral language (narrative skills, vocabulary and grammar) and preliteracy skills (phonological awareness and letter knowledge) were targeted because they are clearly related20 and important to developing language,21 academic,22,23 and social skills. We chose to deliver the one-on-one, 18-session program (three, 6-week blocks of weekly sessions, starting every 3 months) in the home so that parents as well as children could be targeted. It also allowed for the fact that Australian preschool-age children attend a variety of service settings within and across states, so that a center-based program might not be feasible in a local-government led public health framework.
Each session comprised 3 components: (1) phonological awareness tasks; (2) an oral language activity specifically selected and tailored to each child’s profile, severity, and progress; and (3) shared story-book reading. Session format was standardized to cover a brief review of the previous week, then activities (1) directed at the child; (2) for parent and child together, supported by the language assistant; and (3) for home practice. Each activity had standardized supporting materials and manual instructions. Language assistants were supplied with an intervention kit, which included A4-size letter cards (a–z plus “ch,” “sh,” and “th”) and picture cards for the phonological awareness and letter knowledge activities; a range of generic toys such as a farmhouse and animals, cash register, and groceries for the oral language activity; and a range of age-appropriate books for the shared book reading activity. The 19th session was a detailed 5-year-old language assessment, followed by a final 20th feedback and recommendations session.
Outcome measures were collected after the child’s sixth birthday (Table 1). The primary outcomes were directly assessed receptive and expressive language. Secondary outcomes spanned directly assessed phonological, receptive vocabulary, literacy and narrative skills, and parent-reported pragmatic language, behavior, and health-related quality of life. Trained research assistants blinded to trial arm status conducted all direct assessments in participants’ homes (May 2012– January 2013).
Resources used were costed in 2012 Australian dollars (AU$), using intervention provider records and parent-reported health service use.15 Travel was valued at a standard unit cost of AU$0.80 per kilometer, materials were valued at market price, and parent time was valued at the Australian average wage of AU$35.60 per hour.16 Children’s time was assigned a value of $0. Health service use unit costs were obtained from government standard sources.24
We anticipated that 1500 children would be assessed at 4 years, allowing for 20% loss to follow-up. Assuming 240 (16%) would be eligible, 10% would decline participation, and 1% would be excluded, we estimated 210 children would enter the trial (105 in each arm). Ten percent attrition would provide 94 children in each trial arm at outcome, giving 80% power to detect a difference of 0.41 SD at the 5% level of significance.
Trial arms were analyzed using the intention-to-treat principle. Mean outcomes were compared using linear regression in unadjusted analyses and analyses adjusted for the following prognostic factors: child gender, mother’s education level, recruitment from Let’s Read or Let’s Learn Language, expressive and receptive language scores at baseline, and baseline measure of the outcome being considered when available.
Figure 2 summarizes the participant flow. Of the 1661 children eligible, 1464 (88.1%) completed the baseline assessment and, of the 266 children who had low language, 200 entered the trial. Retention rates were high, with 89 intervention (89.9% of 99) and 83 control (82.2% of 101) children retained at 6 years of age; 91% of families received at least 2 blocks of therapy (mean 17 of the 18 intervention sessions). Table 2 shows that the arms were well balanced at entry for expressive (mean 80.8 vs 79.8) and receptive (mean 76.2 vs 75.5) language scores.
Table 3 shows that, in both arms, expressive language scores improved by ∼10 standard score points (equivalent to two-thirds of a SD in the normative population) and receptive language by around 14 points (nearly 1 SD) over the 2 years from baseline, with receptive and expressive mean scores for both arms close to 90, that is, well into the normal range, by age 6. There was, however, little evidence of a treatment effect on either of these primary language outcomes. Of the secondary outcomes, benefits were maintained for phonological awareness skills (effect size 0.36; 95% confidence interval [CI]: 0.08–0.65; P = .01), a direct program target, with possible flow-on benefit to reading (P = .10). The 2 arms were similar on all other secondary outcomes, including health-related quality of life.
Post hoc interaction analyses did not suggest differential benefit to children with less versus more educated mothers or with lower versus higher baseline language scores.
Program Costs and Health Service Use
Table 4 shows the costs of all service use related to child speech and language concerns between child age 4 and 6 years. The intervention costs were AU$3231 to government15 plus AU$694 cost of parent time devoted to the intervention. Across the trial, 29% of families reported using additional health services for their child’s speech and language concerns (outside the Language for Learning program) between 4 and 5 years and 24% between 5 and 6 years, most commonly speech pathologists, pediatricians, general practitioners, and hearing specialists. However, the Language for Learning program had little impact on these additional care episodes (mean difference, intervention vs control arms, 2.6 episodes; 95% CI –3.5 to 8.7, P = .39). Thus, compared with controls, costs in the intervention group were on average AU$3427 (CI $3145 to $3708) higher to government and a further $849 (CI $185 to $1512) higher to families, which is essentially the cost of the Language for Learning program itself.
This trial addressed the important question of whether it is cost-effective to deliver systematic, population-based intervention for low language in 4-year-old children. Although it proved feasible and acceptable for non–speech pathologists to implement a relatively intensive home-based language program, this did not result in better language scores. Notably, by age 6, both groups showed improvements in mean expressive and receptive (approximately two-thirds and 1 SD scores, respectively). This suggests that language, at least as measured by standardized tests such as the CELF, remains fluid between these ages and that much of the observed changes we typically ascribe to treatment may actually be driven by child development.
Although the program costs could be considered modest given its intensity and duration, its cost-effectiveness must be judged in terms of the outcomes achieved. The lasting, sizeable treatment benefit to the secondary outcome of phonological awareness skills (among the strongest precursors of literacy skills) could perhaps be achieved by a simpler, cheaper intervention solely targeting phonological awareness skills. We are now instigating an additional 9-year-old follow-up, which will include results on national standardized academic outcomes in middle childhood.
Strengths and Weaknesses of the Study
The program was positively rated by parents, achieved high engagement and completion, and included a population sampling frame, random allocation, quality control, and blinding of outcomes. The manualized yet tailored program could be implemented by non–speech pathologists with a view to practicability of future rollout if effective.
There are of, course, limitations. Like all null trials, a program of greater intensity or different content might have had greater effects. Regarding generalizability, we excluded families with little or no English. Despite oversampling disadvantaged areas, the trial ultimately included few severely disadvantaged families, and the children were clustered toward the mild end of the low language spectrum. Recent data suggest that children from the most disadvantaged families25 and/or with more severe difficulties may be more responsive to such early interventions,26,27 albeit with much higher levels of intervention than in the present trial. However, it is difficult to envisage a systematic sampling frame that would reach all but only all of these children, and this would exclude the many children who are less disadvantaged and/or have milder but still debilitating language delays. Nor did our post hoc interaction analyses support such a strategy.
Rather than a screening test, baseline eligibility was determined by the full CELF-P2 because our main goal was to test the intervention itself. This gave greater certainty about the children’s entry language status, but it would not be feasible to extend this to whole populations, and it did not protect against substantial natural resolution by age 6. Finally, the single substantive benefit to phonological skills could still have been a chance finding, even though it was robust across both the interim (5 years) and final (6 years) outcomes.
Interpretation in Light of Other Study Findings
This report adds to the few relatively large randomized trials testing systematized interventions after large-scale ascertainment of measured language delays. Boyle’s school-based trial determined which children had specific language impairments and then provided a slightly lower dose of intervention than the present trial; like Language for Learning, no benefits were seen to language, but phonological awareness outcomes were not studied.14 Targeting English children screened as having low language at approximately their fourth birthday, Fricke et al22 showed that an intensive 30-week intervention, delivered within the education setting several times weekly, can improve oral language, vocabulary, and narrative skills for at least 6 months postintervention. Notably, vocabulary change occurred not with the nursery but with the later school component; this may suggest that, beyond the toddler years, intensity, frequency, and timing are more important than parent engagement.
Most positive trials have targeted children with more pronounced problems and/or disadvantage. For example, Gallagher and Chiat showed positive effects on all outcomes in 8 children on a specialist waiting list randomized to intensive speech-language therapy for 4 hours a week over 24 weeks.28 A trial of the Nuffield Language Program, providing alternating individual and group school-based sessions over 20 weeks, demonstrated gains in instructed vocabulary, grammar, and narrative ability in the Oral Language group compared with the Phonology with Reading group (effect size = 1.02, 0.33, and 0.15, respectively).9 Furthermore, the interventions targeting oral language skills benefited language, whereas those targeting phonological and early reading skills had a differential effect in that area. However, these children were older than those in the present trial, had more severe difficulties at baseline, and 30% were in receipt of free school meals, the measure for social disadvantage used within the UK school system.
Implications for Clinicians and Policy Makers
The evidence suggests that screening unselected populations for early language delay and rolling out targeted home-based interventions should not be recommended at this stage. Nonetheless, as a group, such children remain at risk for subsequent difficulties and generate parent, practitioner, and policy concern; therefore, how risk is assessed, monitored, and managed needs to be further explored.
We suggest that trajectories of language development in typically developing children have yet to be mapped out sufficiently well to differentiate improvements due to intervention and normal developmental processes. In this trial, only ∼30% of the cohort still had receptive and/or expressive language scores outside 1.25 SD below the mean by age 6 years. Although patterns may be relatively stable by middle childhood,29 there appears to be far greater variability in the early years, the period covered in this study.30,31 Thus, the window when early intervention is most commonly recommended is also the most problematic for targeting. In a recent review of specific language impairment, we suggested that eligibility should only be determined once a child’s performance has been monitored over time, possibly for a full 12 months.32 Many if not most of the children who participated in this study would have been ineligible under this stricter criterion.
Population-based interventions for low oral language skills in young children are feasible but their individual and societal benefit remains uncertain. This trial highlights some key evidence gaps that we believe are impeding progress. First, documenting reproducible pathways to normal and impaired language by middle childhood would help to better understand intervention impact over and above natural history. Second, predictive tools are needed that quantify an individual child’s absolute risk of lasting language impairment. These would help researchers to focus on the children who are unlikely to resolve spontaneously, providing better “signal to noise” for determining true intervention impacts, help governments avoid unnecessary costs, and reassure parents and practitioners when appropriate. Finally, gauging societal benefit requires both long-term follow-up and knowledge of the burdens and costs of impaired language pathways including key health, educational, and health economic outcomes.
We are indebted to Professor Luigi Girolametto for his central role in the preceding Let’s Learn Language trial and to both Professor Girolametto and Dr Jemma Skeat for their early advice on the structure and development of the Language for Learning intervention. We especially acknowledge the outstanding contributions of Sherryn Tobin, Natasha Napiza, and Ruth Nicholls as Project Managers of the earlier Let’s Learn Language and Let’s Read trials, without whom our population sample would not have existed. We thank all the children, parents, maternal and child health nurses, and research assistants (including Penny Levickis, Jon Quach, Hannah Bryson, Lisa Belford, Jane Sheehan, Elizabeth Varelias, Liz Varley, Rebecca Nadalin, Laura Punaro, Leah Zelencich, and Peta Newell) who took part in the trial.
- Accepted July 27, 2015.
- Address correspondence to Melissa Wake, MBChB, MD, FRACP, GDip, Epi Biostats, Centre for Community Child Health, Royal Children’s Hospital, Flemington Rd, Parkville, VIC 3052, Australia. E-mail:
Dr Wake conceptualized and designed the trial, she takes overall responsibility for all aspects of the trial, and contributed to and critically reviewed the final manuscript; Dr Tobin was the project manager, she coordinated and supervised data collection, and reviewed and revised the manuscript; Dr Levickis assisted with the management of the project, prepared the data for analysis, and reviewed and revised the manuscript; Dr Gold conceptualized and designed the trial, provided advice on the economic evaluation, carried out the analysis of the health service use data, and reviewed and revised the manuscript; Dr Ukoumunne conceptualized and designed the trial, advised on statistical issues, carried out the statistical analyses, and reviewed and revised the manuscript; Dr Zens designed the intervention with advice from Dr Jemma Skeat, assisted with the management of the project, and reviewed and revised the manuscript; Dr Goldfeld advised on the translational aspects of the trial and reviewed and revised the manuscript; Ms Le provided advice on the economic evaluation, carried out the analysis of the health service use data, and reviewed and revised the manuscript; Dr Law conceptualized and designed the trial, designed the intervention with advice from Dr Jemma Skeat, advised on the measures to be used in the study, and reviewed and revised the manuscript; Dr Reilly conceptualized and designed the trial, assisted in the development of the intervention, provided guidance regarding the data collection instruments, and reviewed and revised the manuscript; and all authors approved the final manuscript as submitted.
This trial has been registered with Current Controlled Trials (identifier number ISRCTN03981121).
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Supported by the Australian National Health and Medical Research Council (NHMRC Project Grant 607407). The NHMRC played no role in the trial’s design; in the collection, analysis, and interpretation of data; in the writing of the manuscript; or in the decision to submit the manuscript for publication. Over the course of the trial, Dr Wake has been supported by NHMRC Population Health Career Development Grant 546405 and Senior Research Fellowship 1046518; Drs Goldfeld and Ukoumunne by NHMRC Capacity Building Grant 436914; Dr Reilly by NHMRC Practitioner Fellowships 1041892 and 491210; Dr Gold by NHMRC Capacity Building Grant 425855 and NHMRC Early Career Fellowship 1035100; and Dr Levickis and Ms Le by NHMRC Centre for Research Excellence in Child Language 1023493. Dr Ukoumunne is funded by the National Institute for Health Research (NIHR) Collaboration for Leadership in Applied Health Research and Care (CLAHRC) for the South West Peninsula at the Royal Devon and Exeter NHS Foundation Trust. Research at the Murdoch Childrens Research Institute is supported by the Victorian Government’s Operational Infrastructure Support Program. The views expressed in this publication are those of the authors and not necessarily those of the NHMRC, NHS, the NIHR, or the Department of Health in England.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Field F
- Brooks G,
- Torgerson CJ
- Reilly S, Wake M, Ukoumunne OC, Bavin E, Prior M, Cini E, et al. Predicting language outcomes at 4 years of age: findings from Early Language in Victoria Study. Pediatrics 2010;126(6). Available at: www.pediatrics.org/cgi/content/full/126/6/e1530
- Wake M,
- Tobin S,
- Girolametto L,
- et al
- Boyle J,
- McCartney E,
- Forbes J,
- O’Hare A
- ↵Australian Bureau of Statistics. Census data. Available at: http://www.abs.gov.au/websitedbs/censushome.nsf/home/data?opendocument#from-banner=LN. Accessed July 5 2012
- ↵Australian Bureau of Statistics. Socio-Economic Indexes for Areas. Canberra, Australia; 2006. Available at: http://www.ausstats.abs.gov.au/ausstats/subscriber.nsf/0/D729075E079F9FDECA2574170011B088/$File/20390_2006.pdf. Accessed July 5, 2012
- ↵Wiig E, Secord W, Semel E. Clinical Evaluation of Language Fundamentals Preschool. 2nd ed, Australian and New Zealand Standardised Edition. Sydney, Australia: Harcourt Assessment; 2006
- Gillon GT
- ↵Australian Government. MBS online. Department of Health. Available at: http://www.mbsonline.gov.au/internet/mbsonline/publishing.nsf/Content/Medicare-Benefits-Schedule-mbs-downloads: Accessed July 5th 2012
- Law J
- Whaley SE,
- Jiang L,
- Gomez J,
- Jenks E
- Gallagher AL,
- Chiat S
- Semel E, Wiig E, Secord W. The Clinical Evaluation of Language Fundamentals. 4th ed, Australian Standardised Edition (CELF-4 Australian). New South Wales, Australia: Harcourt Assessment Inc; 2006
- Wilkinson SG,
- Robertson JG
- Renfrew C
- Wagner RK,
- Torgesen JK,
- Rashotte CA
- Dunn LM, Dunn LM. Peabody Picture Vocabulary Test—Third Edition (PPVTIII). Circle Pines, MN: AGS Publishing; 1997
- Bishop D
- Feeny D
- Perera R,
- Heneghan C,
- Yudkin P
- Copyright © 2015 by the American Academy of Pediatrics