Division of Psychology, Communication and Human Neuroscience, The University of Manchester, United Kingdom.
Department of Language and Communication Science, City University of London, United Kingdom.
J Speech Lang Hear Res. 2024 Jul 9;67(7):2191-2221. doi: 10.1044/2024_JSLHR-23-00490. Epub 2024 May 24.
This systematic review and multilevel meta-analysis examines the accuracy of sentence repetition (SR) tasks in distinguishing between typically developing (TD) children and children with developmental language disorder (DLD). It explores variation in the way that SR tasks are administered and/or evaluated and examines whether variability in the reported ability of SR to detect DLD is related to these differences.
Four databases were searched to identify studies that had used an SR task on groups of monolingual children with DLD and TD children. Searches produced 3,459 articles, of which, after screening, 66 were included in the systematic review. A multilevel meta-analysis was then conducted using 46 of these studies. Multiple preregistered subgroup analyses were conducted in order to explore the sources of heterogeneity.
The systematic review found a great deal of methodological variation, with studies spanning 19 languages, 39 SR tasks, and four main methods of production scoring. There was also variation in study design, with different sampling (clinical and population sampling) and matching (age and language matching) methods. The overall meta-analysis found that, on average, TD children outperformed children with DLD on the SR tasks by 2.08 s. Subgroup analyses found that effect size only varied as a function of the matching method and language of the task.
Our results indicate that SR tasks can distinguish children with DLD from both age- and language-matched samples of TD children. The usefulness of SR appears robust to most kinds of task and study variation.
本系统评价和多层次元分析旨在检查句子重复（SR）任务在区分典型发育（TD）儿童和发育性语言障碍（DLD）儿童方面的准确性。它探讨了 SR 任务实施和/或评估方式的差异，并检查 SR 检测 DLD 的能力的变异性是否与这些差异有关。
通过搜索四个数据库，确定了对患有 DLD 和 TD 儿童的单语儿童群体使用 SR 任务的研究。搜索产生了 3459 篇文章，经过筛选，有 66 篇文章被纳入系统评价。然后使用其中的 46 项研究进行了多层次元分析。进行了多个预先注册的子组分析，以探讨异质性的来源。
系统评价发现存在大量的方法学差异，研究涵盖了 19 种语言、39 项 SR 任务和 4 种主要的产生评分方法。研究设计也存在差异，采用了不同的抽样（临床和人口抽样）和匹配（年龄和语言匹配）方法。总体元分析发现，TD 儿童在 SR 任务中的表现平均比 DLD 儿童好 2.08 秒。子组分析发现，效应大小仅随匹配方法和任务语言的变化而变化。
我们的结果表明，SR 任务可以区分 DLD 儿童和年龄和语言匹配的 TD 儿童样本。SR 的有用性似乎对大多数任务和研究变化都具有稳健性。
