JSLHR
Research Article
Stability of Language and Literacy Profiles of Children With Language Impairment in the Public Schools Sherine R. Tambyraja,a Mary Beth Schmitt,a Kelly Farquharson,a and Laura M. Justicea
Purpose: The present study focused on the identification and stability of language and literacy profiles of primary school children receiving school-based language therapy over the course of one academic year. Method: Participants included 272 early elementary schoolage children (144 boys, 128 girls) who had been clinically identified as having a language impairment. A latent profile analysis was used to identify distinct profiles on the basis of a battery of language and literacy assessments in the fall and spring of the academic year. Results: Four profiles were identified in both fall and spring that could be best described as representing high, average, and low overall abilities. Two average groups were identified
that differentiated according to phonological awareness abilities. Children’s profile membership was variable from fall to spring with nearly 60% of children shifting into a higher profile. The results of t tests comparing children who shifted into higher profiles from those who remained stable in profile membership revealed group differences regarding language severity, socio-economic status, and proportion of therapy sessions received in the classroom. Conclusion: These results provide further evidence regarding the heterogeneity of children with language impairment served in the public schools, indicating that differences may be best conceptualized along a continuum of severity.
T
of work focused on identifying clinically relevant language and literacy profiles among children with LI, focusing specifically on those receiving school-based language services in U.S. public schools. In addition, it examines the extent to which children’s profiles remain stable over the course of an academic year during receipt of language services and explores child- and therapy-level factors that alter profile stability.
here is little debate in the research literature that children with language impairment (LI) constitute a heterogeneous group with respect to the domains of language affected and levels of severity, with some work showing that these reflect clinically distinct LI profiles (e.g., Beitchman et al., 1989; Conti-Ramsden, Crutchley, & Botting, 1997). Although there has been some consistency with respect to the nature of identifiable profiles of LI, few studies have examined the extent to which language therapy may affect the nature and stability of such profiles. Most importantly, little work has focused on examining the heterogeneity among clinical populations of children, such as those being treated for LI in public schools (but see Conti-Ramsden & Botting, 1999; Conti-Ramsden et al., 1997). The present study contributes to the relatively modest body
a
The Ohio State University, Crane Center for Early Childhood Research and Policy, Columbus. Mary Beth Schmitt is now at Texas Tech University Health Sciences Center, Lubbock, and Kelly Farquharson is now at Emerson College, Boston, MA Correspondence to Sherine R. Tambyraja: [email protected] Editor: Rhea Paul Associate Editor: Ann Tyler Received July 18, 2014 Revision received December 11, 2014 Accepted April 19, 2015 DOI: 10.1044/2015_JSLHR-L-14-0197
Approaches to Understanding LI subgroups Some previous research aimed at understanding the heterogeneity of children with LI has suggested that there are divergent theoretical frameworks to understanding how LI is best conceptualized. One such perspective is based on the notion that language is multidimensional, or comprised of separable, yet related linguistic domains (e.g., grammar, phonology, vocabulary); consequently, children with LI might be best subgrouped according to domain-specific deficits. For example, work by van der Lely, Rosen, and McClelland (1998) showed that children with specific LI (SLI) demonstrated particular deficits on grammar-based tasks, but relatively superior performance on nongrammatical language tests. A contrasting theoretical perspective is Disclosure: The authors have declared that no competing interests existed at the time of publication.
Journal of Speech, Language, and Hearing Research • Vol. 58 • 1167–1181 • August 2015 • Copyright © 2015 American Speech-Language-Hearing Association
Downloaded From: http://jslhr.pubs.asha.org/ by a New York University User on 02/14/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1167
offered by unidimensional accounts of language (e.g., Bishop & Edmundson, 1987; Dollaghan, 2011; Tomblin & Zhang, 2006; Tomblin, Zhang, Buckwalter, & O’Brien, 2003). Within this framework, children with LI are thought to have similar patterns of language development to typically developing children, but differ primarily with respect to the severity of their language deficit. Indeed, such theoretical perspectives may be useful in guiding the interpretation of the observed patterns of ability among children with LI (Beitchman et al., 1989; ContiRamsden et al., 1997). However, studies utilizing personcentered approaches to understanding within-group variability have actually lent support to both organizing frameworks. In terms of methodology, examination of patterns that identify groups of persons on the basis of sets of variables is called person-centered research. As applied to children with LI, person-centered approaches, including cluster analysis and latent profile analysis (LPA), can determine whether groups of children with LI can be classified according to linguistic domains (i.e., poor grammatical skills but average vocabulary), overall levels of language ability (i.e., low, average, or high ability across language measures), or both.
Person-Centered Studies of Children with LI Until recent years, cluster analysis was the primary method used for conducting person-centered research on children with LI (e.g., Beitchman et al., 1989; Conti-Ramsden et al., 1997; van Weerdenburg, Verhoeven, & van Balkom, 2006). Conti-Ramsden et al. (1997) used cluster analysis to identify subgroups of children with SLI among 242 seven-year-old British children. The children had been previously identified by clinicians as requiring language therapy or “language unit” placement in their schools, and were administered a broad battery of standardized tests measuring grammar, vocabulary, word reading, articulation, narrative ability, and number skills. Further, these researchers examined the extent to which their findings would be consistent with that of Rapin and Allen’s (1987) classification studies of children with LI that yielded six separate subgroups. Conti-Ramsden et al. (1997) utilized a cluster analysis to identify six separate subgroups, five of which exhibited comparable characteristics to those described by Rapin and Allen. Of note, however, is that the clusters that resembled subgroups identified by Rapin and Allen were most readily discerned by measures of articulation and expressive speech, and one cluster distinguished by pragmatic language abilities. Overall, the three largest groups (comprising 68% of the full sample) all performed poorly on five out of the six measures, whereas the two smallest groups (about 20% of the sample) exhibited varying patterns of ability across measures. Study findings therefore showed that a majority of children with SLI were subgrouped according to overall ability, although some groups of children did appear to demonstrate differences specific to linguistic domains. Results from a cluster analysis of 347 five-year-old Canadian children with speech difficulties and LI (Beitchman
1168
et al., 1989) similarly showed that the four identified clusters were best described as representing “high,” “low,” and two “average” groups (which differed according to articulation and auditory processing abilities). A more recent cluster analysis of 147 six-year-old and 136 eight-year-old Dutch-speaking children with SLI also resulted in four subgroups at each age group exhibiting similar characteristics (van Weerdenburg et al., 2006). A battery of language skills was administered, including speech output, auditory perception, morpho-syntactic abilities, vocabulary, narrative recall, and several memory tasks. A factor analysis determined that the test scores could be conceptualized by four distinct factors representing lexical-semantic, speech production, syntactic-sequential, or auditory processing abilities. Using these four factors, the authors then found that children’s performances on these assessments could be represented by four distinct subgroups at each age group. In accordance with previous cluster analysis research, some clusters represented domain-specific differences (e.g., Cluster 3 children had relatively low scores on lexical-semantic tests but higher scores on auditory processing) and other clusters represented varying levels of severity across most administered measures (e.g., Cluster 1 children had relatively high scores on all factors, and Cluster 2 had relatively low scores on all factors). As in previous cluster analyses, subgroups demonstrating domain-specific discrepancies differed on measures of speech output and articulation. Taken together, data from large-scale, cluster-analytic studies of children with LI suggest that although some subgroups are best identified according to language severity, others indeed demonstrate domain-specific differences, although the nature of such differences has not always been consistent. Furthermore, previous work has only examined non-U.S. populations of young children. The present study aims to build upon this research by examining the extent to which language and literacy profiles are identifiable among children with LI receiving treatment for language disorders within the public schools of the United States—an important, yet understudied population. The present work differs from previous similarly focused research in two additional ways. First, we utilized a relatively small set of assessments that each tapped a distinctive linguistic or literacy skill. Specifically, we examined the constructs of grammar, vocabulary, listening comprehension, PA, and letter/word identification. This set of five tasks included one receptive language test (listening comprehension) and four expressive language tests. Previous work has been broader in scope, including measures of articulation and number skill (ContiRamsden et al., 1997), working memory, and rapid automated naming (van Weerdenburg et al., 2006). Second, we utilized LPA as the methodology for examining profiles of children with LI. LPA, to our knowledge, has not been used to distinguish reliable profiles (or subgroups) of school-age children with LI. LPA confers some important advantages over cluster analysis, including tests of statistical significance and goodness-of-fit statistics. Additionally, LPA allows for identification of an optimal number of profiles as informed by the data; some methods
Journal of Speech, Language, and Hearing Research • Vol. 58 • 1167–1181 • August 2015
Downloaded From: http://jslhr.pubs.asha.org/ by a New York University User on 02/14/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
of cluster analysis (e.g., K-means) require researchers to predetermine a specific number of groupings, which may be arbitrary or unnecessarily restrict the data (Boscardin, 2012).
Extending Profiles to Include Literacy In large part, previous work aimed at identifying LI subgroups has relied upon varying domain-specific language measures. For school-age children with LI, however, including literacy skills in the profile analysis may also relay clinically relevant information. It is well established that children with LI are at heightened risk for poor literacy skills (e.g., Botting, Simkin, & Conti-Ramsden, 2006; Catts, 1993; Catts, Fey, Tomblin, & Zhang, 2002), but several studies have also shown that some children with LI exhibit normally developing literacy skills, particularly with respect to code-based literacy skills (e.g., Botting et al., 2006), such as letter recognition and phonological awareness (PA). In fact, some research has shown that performance on code-based literacy skills may differentiate subgroups of children with LI (e.g., Catts, Adlof, Hogan, & Weismer, 2005; Kelso, Fletcher, & Lee, 2007). However, the cluster analysis of school-age children, by Conti-Ramsden et al. (1997), also included a word reading measure as part of the assessment battery, and for the most part, scores on that test were not a distinguishing factor. In other words, children with overall low language scores had similarly poor word reading skills. However, in the present study we include two fundamental code-based literacy skills (i.e., letter/ word recognition, PA) to more thoroughly examine the extent to which these skills inform profiles of school-age children with LI. In addition to understanding the ways in which literacy skills may differ among children with LI, considering code-based literacy skills is especially relevant within the context of educational clinical services. In recent years, speech-language pathologists (SLPs) have been charged with addressing the literacy needs of children on their caseload in addition to, or in conjunction with, traditional language therapy (American Speech-Language-Hearing Association [ASHA], 2001). However, recent evidence reveals that many SLPs do not regularly focus on literacy in therapy (Fallon & Katz, 2011; Tambyraja, Schmitt, Justice, Logan, & Schwarz, 2014) and that many SLPs received very limited training to do so (Blood, Mamett, Gordon, & Blood, 2010). Therefore, it is critical that research continues to investigate the extent to which certain subgroups of children with LI may be more likely to experience code-based literacy difficulties than others, in order to enhance and support school-based language therapy provision.
Stability of LI One point of convergence among studies of children with LI is that language deficits are potentially long-term in nature, particularly for those with severe language difficulties (Beitchman, Wilson, Brownlie, Walters, & Lancee, 1996; Durkin & Conti-Ramsden, 2007; Johnson et al., 1999;
Stothard, Snowling, Bishop, Chipchase, & Kaplan, 1998; Tomblin et al., 2003). However, some longitudinal research also suggests that the direction and quality of language change exhibited by children with LI is inconsistent over time (e.g., Beitchman et al., 1996; Conti-Ramsden & Botting, 1999), with some subgroups of children demonstrating more language growth in specific domains than others. In a follow up to the 1997 study, Conti-Ramsden and Botting (1999) repeated the cluster analysis procedure with their participants to evaluate the stability of the previously determined subgroups. They administered the same language and academic measures 1 year later; results revealed similar characteristics for five out of the six clusters. Cluster membership, however, was variable to the extent that 45% of the participating children shifted into a different group at Time 2. For example, a child who was previously in the lexical-syntactic deficit group at age 7 years had skills more consistent with the phonological-syntactic deficit at age 8 years. Thus, although the structure for most of the subgroups remained similar over time, children’s group membership was not always consistent. The notion that the characterization of subgroups remains stable, even though the subgroup membership changes over time, was also observed in a recent LPA study of emergent literacy (e.g., Cabell, Justice, Logan, & Konold, 2013). Cabell et al. (2013) examined the emergent literacy skills of 484 preschoolers from low socio-economic status (SES) backgrounds in the fall and spring of an academic year and determined that five distinct emergent literacy subgroups, or profiles, were evident within their group of participants. Moreover, they found that the primary features of the identified profiles were similar from fall to spring, although fewer distinct profiles were determined on the basis of spring scores. Additionally, 35% of the participants shifted into a different profile by the spring. Stability was more evident for the profiles representing either “high” or “low” emergent literacy skill level, suggesting that movement between profiles may be more likely for children with relatively average levels of ability. Cabell et al. further found that child-specific and demographic variables (i.e., initial skill level, SES, ethnicity) differentiated children who remained in the same emergent literacy profile from those who shifted into a higher performing group. Previous work investigating subgroups of children with LI (e.g., Beitchman et al., 1996; Conti-Ramsden & Botting, 1999; van Weerdenburg et al., 2006) has not sought to identify whether these same child-specific and demographic variables might distinguish children whose profile membership changes over time. Following procedures in Cabell et al. (2013), the present study examines the extent to which children’s initial language skills and demographic/background factors (i.e., SES) might associate with subgroup membership changes over time. Additionally, because the present study concerns language profiles among children receiving language therapy in U.S. public schools, we also examined the extent to which the total number of therapy sessions received as well as service delivery models (classroom or pullout) would differentiate children whose profile membership
Tambyraja et al.: LI Profiles
Downloaded From: http://jslhr.pubs.asha.org/ by a New York University User on 02/14/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1169
remained consistent from those who shifted into a different spring profile.
Study Aims The current study contributes to the line of personcentered research seeking to identify clinically meaningful and distinct profiles of school-age children with LI on the basis of measures of language and literacy skill. Efforts to understand the variability of children with LI may be particularly prudent for children receiving therapy in U.S. public schools, as these children constitute a considerable proportion of school-based SLP caseloads (ASHA, 2012). Similar to previous work (e.g., Conti-Ramsden et al., 1997), we included measures of code-based literacy, as studies show that such skills are variably affected in children with LI, and addressing literacy is considered within the scope of practice for SLPs (ASHA, 2001). If distinct language and literacy profiles can be discerned among children who receive school-based services, such information could greatly affect and streamline the provision of language intervention. Additionally, as this work extends previous person-centered studies of children with LI by utilizing LPA, a data-driven analytical approach, it may serve to confirm some of the characteristics and descriptions of SLI profiles that have been previously identified and described (e.g., Conti-Ramsden et al., 1997; van Weerdenburg et al., 2006). Thus, the current study aims to (a) identify the language and literacy profiles within a group of clinically identified children with LI at the beginning of an academic year, (b) determine the stability of these profiles, as well as membership stability, at the end of the academic year, and (c) identify child- and therapy-specific factors that differentiate children who move across profiles from those whose profile membership remains stable.
Method Participants Participants in the present study were drawn from a larger, multicohort study, titled Speech Therapy Experiences in Primary Schools (STEPS), involving children who were receiving speech-language therapy in rural and urban public schools and their SLPs. In STEPS, school-based SLPs were asked to distribute recruitment materials to children on their caseload who met the following criteria: (a) had an Individualized Education Plan for speechlanguage services, (b) were in kindergarten, first, or second grade, and (c) did not exhibit severe cognitive impairment. Families who provided informed consent for their child’s participation were then contacted by research staff. The final set of participants included 293 children, enrolled across three separate academic years (2009–2010, 2010–2011, 2011–2012) and their 75 SLPs. The primary purpose of the larger study was to explore relations among key aspects of therapy with children’s outcomes in a business-as-usual framework; therefore, the participants in the current and larger study may be considered a clinically identified sample
1170
of children with LI in that their primary diagnosis was originally ascertained by a school-based SLP. Additional assessments administered by study personnel served to further describe the language and literacy abilities of the enrolled participants. The present study was concerned with whether children with LI who were in receipt of school-based language therapy could be profiled according to performance on specific measures of language and literacy (i.e., grammar, listening comprehension, vocabulary, letter/word recognition). Of the 293 children in the larger study, a small percentage achieved scores within normal range on a researcheradministered standardized language measure (i.e., above −1 SD on at least one measure of language and literacy; n = 15). Additionally, six children scored more than 2 SD below the mean (i.e.,
Research Article
Stability of Language and Literacy Profiles of Children With Language Impairment in the Public Schools Sherine R. Tambyraja,a Mary Beth Schmitt,a Kelly Farquharson,a and Laura M. Justicea
Purpose: The present study focused on the identification and stability of language and literacy profiles of primary school children receiving school-based language therapy over the course of one academic year. Method: Participants included 272 early elementary schoolage children (144 boys, 128 girls) who had been clinically identified as having a language impairment. A latent profile analysis was used to identify distinct profiles on the basis of a battery of language and literacy assessments in the fall and spring of the academic year. Results: Four profiles were identified in both fall and spring that could be best described as representing high, average, and low overall abilities. Two average groups were identified
that differentiated according to phonological awareness abilities. Children’s profile membership was variable from fall to spring with nearly 60% of children shifting into a higher profile. The results of t tests comparing children who shifted into higher profiles from those who remained stable in profile membership revealed group differences regarding language severity, socio-economic status, and proportion of therapy sessions received in the classroom. Conclusion: These results provide further evidence regarding the heterogeneity of children with language impairment served in the public schools, indicating that differences may be best conceptualized along a continuum of severity.
T
of work focused on identifying clinically relevant language and literacy profiles among children with LI, focusing specifically on those receiving school-based language services in U.S. public schools. In addition, it examines the extent to which children’s profiles remain stable over the course of an academic year during receipt of language services and explores child- and therapy-level factors that alter profile stability.
here is little debate in the research literature that children with language impairment (LI) constitute a heterogeneous group with respect to the domains of language affected and levels of severity, with some work showing that these reflect clinically distinct LI profiles (e.g., Beitchman et al., 1989; Conti-Ramsden, Crutchley, & Botting, 1997). Although there has been some consistency with respect to the nature of identifiable profiles of LI, few studies have examined the extent to which language therapy may affect the nature and stability of such profiles. Most importantly, little work has focused on examining the heterogeneity among clinical populations of children, such as those being treated for LI in public schools (but see Conti-Ramsden & Botting, 1999; Conti-Ramsden et al., 1997). The present study contributes to the relatively modest body
a
The Ohio State University, Crane Center for Early Childhood Research and Policy, Columbus. Mary Beth Schmitt is now at Texas Tech University Health Sciences Center, Lubbock, and Kelly Farquharson is now at Emerson College, Boston, MA Correspondence to Sherine R. Tambyraja: [email protected] Editor: Rhea Paul Associate Editor: Ann Tyler Received July 18, 2014 Revision received December 11, 2014 Accepted April 19, 2015 DOI: 10.1044/2015_JSLHR-L-14-0197
Approaches to Understanding LI subgroups Some previous research aimed at understanding the heterogeneity of children with LI has suggested that there are divergent theoretical frameworks to understanding how LI is best conceptualized. One such perspective is based on the notion that language is multidimensional, or comprised of separable, yet related linguistic domains (e.g., grammar, phonology, vocabulary); consequently, children with LI might be best subgrouped according to domain-specific deficits. For example, work by van der Lely, Rosen, and McClelland (1998) showed that children with specific LI (SLI) demonstrated particular deficits on grammar-based tasks, but relatively superior performance on nongrammatical language tests. A contrasting theoretical perspective is Disclosure: The authors have declared that no competing interests existed at the time of publication.
Journal of Speech, Language, and Hearing Research • Vol. 58 • 1167–1181 • August 2015 • Copyright © 2015 American Speech-Language-Hearing Association
Downloaded From: http://jslhr.pubs.asha.org/ by a New York University User on 02/14/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1167
offered by unidimensional accounts of language (e.g., Bishop & Edmundson, 1987; Dollaghan, 2011; Tomblin & Zhang, 2006; Tomblin, Zhang, Buckwalter, & O’Brien, 2003). Within this framework, children with LI are thought to have similar patterns of language development to typically developing children, but differ primarily with respect to the severity of their language deficit. Indeed, such theoretical perspectives may be useful in guiding the interpretation of the observed patterns of ability among children with LI (Beitchman et al., 1989; ContiRamsden et al., 1997). However, studies utilizing personcentered approaches to understanding within-group variability have actually lent support to both organizing frameworks. In terms of methodology, examination of patterns that identify groups of persons on the basis of sets of variables is called person-centered research. As applied to children with LI, person-centered approaches, including cluster analysis and latent profile analysis (LPA), can determine whether groups of children with LI can be classified according to linguistic domains (i.e., poor grammatical skills but average vocabulary), overall levels of language ability (i.e., low, average, or high ability across language measures), or both.
Person-Centered Studies of Children with LI Until recent years, cluster analysis was the primary method used for conducting person-centered research on children with LI (e.g., Beitchman et al., 1989; Conti-Ramsden et al., 1997; van Weerdenburg, Verhoeven, & van Balkom, 2006). Conti-Ramsden et al. (1997) used cluster analysis to identify subgroups of children with SLI among 242 seven-year-old British children. The children had been previously identified by clinicians as requiring language therapy or “language unit” placement in their schools, and were administered a broad battery of standardized tests measuring grammar, vocabulary, word reading, articulation, narrative ability, and number skills. Further, these researchers examined the extent to which their findings would be consistent with that of Rapin and Allen’s (1987) classification studies of children with LI that yielded six separate subgroups. Conti-Ramsden et al. (1997) utilized a cluster analysis to identify six separate subgroups, five of which exhibited comparable characteristics to those described by Rapin and Allen. Of note, however, is that the clusters that resembled subgroups identified by Rapin and Allen were most readily discerned by measures of articulation and expressive speech, and one cluster distinguished by pragmatic language abilities. Overall, the three largest groups (comprising 68% of the full sample) all performed poorly on five out of the six measures, whereas the two smallest groups (about 20% of the sample) exhibited varying patterns of ability across measures. Study findings therefore showed that a majority of children with SLI were subgrouped according to overall ability, although some groups of children did appear to demonstrate differences specific to linguistic domains. Results from a cluster analysis of 347 five-year-old Canadian children with speech difficulties and LI (Beitchman
1168
et al., 1989) similarly showed that the four identified clusters were best described as representing “high,” “low,” and two “average” groups (which differed according to articulation and auditory processing abilities). A more recent cluster analysis of 147 six-year-old and 136 eight-year-old Dutch-speaking children with SLI also resulted in four subgroups at each age group exhibiting similar characteristics (van Weerdenburg et al., 2006). A battery of language skills was administered, including speech output, auditory perception, morpho-syntactic abilities, vocabulary, narrative recall, and several memory tasks. A factor analysis determined that the test scores could be conceptualized by four distinct factors representing lexical-semantic, speech production, syntactic-sequential, or auditory processing abilities. Using these four factors, the authors then found that children’s performances on these assessments could be represented by four distinct subgroups at each age group. In accordance with previous cluster analysis research, some clusters represented domain-specific differences (e.g., Cluster 3 children had relatively low scores on lexical-semantic tests but higher scores on auditory processing) and other clusters represented varying levels of severity across most administered measures (e.g., Cluster 1 children had relatively high scores on all factors, and Cluster 2 had relatively low scores on all factors). As in previous cluster analyses, subgroups demonstrating domain-specific discrepancies differed on measures of speech output and articulation. Taken together, data from large-scale, cluster-analytic studies of children with LI suggest that although some subgroups are best identified according to language severity, others indeed demonstrate domain-specific differences, although the nature of such differences has not always been consistent. Furthermore, previous work has only examined non-U.S. populations of young children. The present study aims to build upon this research by examining the extent to which language and literacy profiles are identifiable among children with LI receiving treatment for language disorders within the public schools of the United States—an important, yet understudied population. The present work differs from previous similarly focused research in two additional ways. First, we utilized a relatively small set of assessments that each tapped a distinctive linguistic or literacy skill. Specifically, we examined the constructs of grammar, vocabulary, listening comprehension, PA, and letter/word identification. This set of five tasks included one receptive language test (listening comprehension) and four expressive language tests. Previous work has been broader in scope, including measures of articulation and number skill (ContiRamsden et al., 1997), working memory, and rapid automated naming (van Weerdenburg et al., 2006). Second, we utilized LPA as the methodology for examining profiles of children with LI. LPA, to our knowledge, has not been used to distinguish reliable profiles (or subgroups) of school-age children with LI. LPA confers some important advantages over cluster analysis, including tests of statistical significance and goodness-of-fit statistics. Additionally, LPA allows for identification of an optimal number of profiles as informed by the data; some methods
Journal of Speech, Language, and Hearing Research • Vol. 58 • 1167–1181 • August 2015
Downloaded From: http://jslhr.pubs.asha.org/ by a New York University User on 02/14/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
of cluster analysis (e.g., K-means) require researchers to predetermine a specific number of groupings, which may be arbitrary or unnecessarily restrict the data (Boscardin, 2012).
Extending Profiles to Include Literacy In large part, previous work aimed at identifying LI subgroups has relied upon varying domain-specific language measures. For school-age children with LI, however, including literacy skills in the profile analysis may also relay clinically relevant information. It is well established that children with LI are at heightened risk for poor literacy skills (e.g., Botting, Simkin, & Conti-Ramsden, 2006; Catts, 1993; Catts, Fey, Tomblin, & Zhang, 2002), but several studies have also shown that some children with LI exhibit normally developing literacy skills, particularly with respect to code-based literacy skills (e.g., Botting et al., 2006), such as letter recognition and phonological awareness (PA). In fact, some research has shown that performance on code-based literacy skills may differentiate subgroups of children with LI (e.g., Catts, Adlof, Hogan, & Weismer, 2005; Kelso, Fletcher, & Lee, 2007). However, the cluster analysis of school-age children, by Conti-Ramsden et al. (1997), also included a word reading measure as part of the assessment battery, and for the most part, scores on that test were not a distinguishing factor. In other words, children with overall low language scores had similarly poor word reading skills. However, in the present study we include two fundamental code-based literacy skills (i.e., letter/ word recognition, PA) to more thoroughly examine the extent to which these skills inform profiles of school-age children with LI. In addition to understanding the ways in which literacy skills may differ among children with LI, considering code-based literacy skills is especially relevant within the context of educational clinical services. In recent years, speech-language pathologists (SLPs) have been charged with addressing the literacy needs of children on their caseload in addition to, or in conjunction with, traditional language therapy (American Speech-Language-Hearing Association [ASHA], 2001). However, recent evidence reveals that many SLPs do not regularly focus on literacy in therapy (Fallon & Katz, 2011; Tambyraja, Schmitt, Justice, Logan, & Schwarz, 2014) and that many SLPs received very limited training to do so (Blood, Mamett, Gordon, & Blood, 2010). Therefore, it is critical that research continues to investigate the extent to which certain subgroups of children with LI may be more likely to experience code-based literacy difficulties than others, in order to enhance and support school-based language therapy provision.
Stability of LI One point of convergence among studies of children with LI is that language deficits are potentially long-term in nature, particularly for those with severe language difficulties (Beitchman, Wilson, Brownlie, Walters, & Lancee, 1996; Durkin & Conti-Ramsden, 2007; Johnson et al., 1999;
Stothard, Snowling, Bishop, Chipchase, & Kaplan, 1998; Tomblin et al., 2003). However, some longitudinal research also suggests that the direction and quality of language change exhibited by children with LI is inconsistent over time (e.g., Beitchman et al., 1996; Conti-Ramsden & Botting, 1999), with some subgroups of children demonstrating more language growth in specific domains than others. In a follow up to the 1997 study, Conti-Ramsden and Botting (1999) repeated the cluster analysis procedure with their participants to evaluate the stability of the previously determined subgroups. They administered the same language and academic measures 1 year later; results revealed similar characteristics for five out of the six clusters. Cluster membership, however, was variable to the extent that 45% of the participating children shifted into a different group at Time 2. For example, a child who was previously in the lexical-syntactic deficit group at age 7 years had skills more consistent with the phonological-syntactic deficit at age 8 years. Thus, although the structure for most of the subgroups remained similar over time, children’s group membership was not always consistent. The notion that the characterization of subgroups remains stable, even though the subgroup membership changes over time, was also observed in a recent LPA study of emergent literacy (e.g., Cabell, Justice, Logan, & Konold, 2013). Cabell et al. (2013) examined the emergent literacy skills of 484 preschoolers from low socio-economic status (SES) backgrounds in the fall and spring of an academic year and determined that five distinct emergent literacy subgroups, or profiles, were evident within their group of participants. Moreover, they found that the primary features of the identified profiles were similar from fall to spring, although fewer distinct profiles were determined on the basis of spring scores. Additionally, 35% of the participants shifted into a different profile by the spring. Stability was more evident for the profiles representing either “high” or “low” emergent literacy skill level, suggesting that movement between profiles may be more likely for children with relatively average levels of ability. Cabell et al. further found that child-specific and demographic variables (i.e., initial skill level, SES, ethnicity) differentiated children who remained in the same emergent literacy profile from those who shifted into a higher performing group. Previous work investigating subgroups of children with LI (e.g., Beitchman et al., 1996; Conti-Ramsden & Botting, 1999; van Weerdenburg et al., 2006) has not sought to identify whether these same child-specific and demographic variables might distinguish children whose profile membership changes over time. Following procedures in Cabell et al. (2013), the present study examines the extent to which children’s initial language skills and demographic/background factors (i.e., SES) might associate with subgroup membership changes over time. Additionally, because the present study concerns language profiles among children receiving language therapy in U.S. public schools, we also examined the extent to which the total number of therapy sessions received as well as service delivery models (classroom or pullout) would differentiate children whose profile membership
Tambyraja et al.: LI Profiles
Downloaded From: http://jslhr.pubs.asha.org/ by a New York University User on 02/14/2016 Terms of Use: http://pubs.asha.org/ss/rights_and_permissions.aspx
1169
remained consistent from those who shifted into a different spring profile.
Study Aims The current study contributes to the line of personcentered research seeking to identify clinically meaningful and distinct profiles of school-age children with LI on the basis of measures of language and literacy skill. Efforts to understand the variability of children with LI may be particularly prudent for children receiving therapy in U.S. public schools, as these children constitute a considerable proportion of school-based SLP caseloads (ASHA, 2012). Similar to previous work (e.g., Conti-Ramsden et al., 1997), we included measures of code-based literacy, as studies show that such skills are variably affected in children with LI, and addressing literacy is considered within the scope of practice for SLPs (ASHA, 2001). If distinct language and literacy profiles can be discerned among children who receive school-based services, such information could greatly affect and streamline the provision of language intervention. Additionally, as this work extends previous person-centered studies of children with LI by utilizing LPA, a data-driven analytical approach, it may serve to confirm some of the characteristics and descriptions of SLI profiles that have been previously identified and described (e.g., Conti-Ramsden et al., 1997; van Weerdenburg et al., 2006). Thus, the current study aims to (a) identify the language and literacy profiles within a group of clinically identified children with LI at the beginning of an academic year, (b) determine the stability of these profiles, as well as membership stability, at the end of the academic year, and (c) identify child- and therapy-specific factors that differentiate children who move across profiles from those whose profile membership remains stable.
Method Participants Participants in the present study were drawn from a larger, multicohort study, titled Speech Therapy Experiences in Primary Schools (STEPS), involving children who were receiving speech-language therapy in rural and urban public schools and their SLPs. In STEPS, school-based SLPs were asked to distribute recruitment materials to children on their caseload who met the following criteria: (a) had an Individualized Education Plan for speechlanguage services, (b) were in kindergarten, first, or second grade, and (c) did not exhibit severe cognitive impairment. Families who provided informed consent for their child’s participation were then contacted by research staff. The final set of participants included 293 children, enrolled across three separate academic years (2009–2010, 2010–2011, 2011–2012) and their 75 SLPs. The primary purpose of the larger study was to explore relations among key aspects of therapy with children’s outcomes in a business-as-usual framework; therefore, the participants in the current and larger study may be considered a clinically identified sample
1170
of children with LI in that their primary diagnosis was originally ascertained by a school-based SLP. Additional assessments administered by study personnel served to further describe the language and literacy abilities of the enrolled participants. The present study was concerned with whether children with LI who were in receipt of school-based language therapy could be profiled according to performance on specific measures of language and literacy (i.e., grammar, listening comprehension, vocabulary, letter/word recognition). Of the 293 children in the larger study, a small percentage achieved scores within normal range on a researcheradministered standardized language measure (i.e., above −1 SD on at least one measure of language and literacy; n = 15). Additionally, six children scored more than 2 SD below the mean (i.e.,
