- Email: [email protected]
Validity of the language development survey in infants born preterm
Early Human Development 98 (2016) 11–16
Contents lists available at ScienceDirect
Early Human Development journal homepage: www.elsevier.com/locate/earlhumdev
Validity of the language development survey in infants born preterm Camille Beaulieu-Poulin a, Marie-Noëlle Simard b, Hélène Babakissa a, Francine Lefebvre a, Thuy Mai Luu a,⁎ a b
Department of Pediatrics, CHU Sainte-Justine, 3175 Chemin de la Côte-Ste-Catherine, Montreal, Quebec H3T 1C5, Canada School of Rehabilitation, Faculty of Medicine, University of Montreal, C.P. 6128, succursale Centre-Ville, Montreal, Quebec H3C 3J7, Canada
a r t i c l e
i n f o
Article history: Received 7 January 2016 Received in revised form 26 May 2016 Accepted 4 June 2016 Keywords: Preterm infants Language delay Developmental screening Language development survey Bayley scales of infant development
a b s t r a c t Background: Preterm infants are at greater risk of language delay. Early identification of language delay is essential to improve functional outcome in these children. Aim: To examine the concurrent validity of Rescorla's Language Development Survey and the Bayley Scales of Infant and Toddler Development (Bayley-III) at 18 months corrected age in preterm infants. Study design: Test accuracy study. Participants: 189 preterm infants born b29 weeks were assessed at 18 months. Outcome measures: The Language Development Survey, a parent-reported screening instrument, was administered in French concurrently with the Language Scales of the Bayley-III. Receiver-Operating-Characteristics curves were used to determine optimal cut-off score on the Language Development Survey to identify Bayley-III score b 85. Sensitivity, specificity, positive and negative predictive values, and κ coefficient were calculated. Results: Using Rescorla's original cut-off scores of ≤10 words for boys and ≤24 for girls, sensitivity was 76% and 88% for boys and girls, respectively, and specificity was 73% and 52% for boys and girls, respectively, in identifying language delay as per the Bayley-III. The optimal threshold was ≤10 words for both boys and girls. In girls, lowering the cut-off score decreased sensitivity (79%), but improved specificity (82%), thus lowering the number of false-positives. Conclusion: Our findings support using the Language Development Survey as an expressive language screener in preterm infants. © 2016 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Communication and language problems are frequently encountered in infants born prematurely [1]. More specifically, preterm children exhibit difficulties across all language sub-domains, including phonological awareness, semantics, grammar, discourse, and pragmatics [2,3]. Language difficulties persist throughout the school years with 15% of preterm children still requiring speech therapy at 12 years [4]. The first cues indicating language problems can be identified as early as during infancy. Indeed, among extremely preterm infants, 20–35% have language delay according to the Bayley Scales of Infant and Toddler Development 3rd edition (Bayley-III) at 18–24 months corrected age (CA) [5]. Language impairment has a significant long-term impact and may affect social relationships, literacy and educational attainment [6]. Therefore, early recognition of language delay is crucial to improve functional
Abbreviations: LDS, Language Development Survey; CA, corrected age. ⁎ Corresponding author. E-mail addresses: [email protected] (C. Beaulieu-Poulin), [email protected] (M.-N. Simard), [email protected] (H. Babakissa), [email protected] (F. Lefebvre), [email protected] (T.M. Luu).
http://dx.doi.org/10.1016/j.earlhumdev.2016.06.003 0378-3782/© 2016 Elsevier Ireland Ltd. All rights reserved.
and academic outcomes of at-risk children. As such, the American Academy of Pediatrics has been advocating for standardized developmental screening at specific ages using validated tools including parentreported questionnaires [7]. Prompt identification of language developmental delay allows the child to benefit from early speech and language intervention which has been shown to enhance language abilities and to reduce the burden of a persistent disability on child's function [8]. Standardized language-specific screening tools for infants that are parent-administered and quick to complete and score are scarce. One such parental questionnaire, the MacArthur-Bates Communicative Development Inventory (CDI), has been adapted in multiple languages and widely used for research with preterm populations in several countries [9–14]. At 12–24 months corrected ages, preterm infants display lower gesture/word production, as measured by the CDI, as well as smaller vocabulary size and shorter length of utterances compared to term controls, with lower gestational ages associated with more impairment [9,15,16]. Though the CDI has proven useful to assess language development in preterm infants, time required to complete the form (30–40 min) and to score it (20 min) is rather long and could hinder uptake by busy general health care providers. To overcome these barriers, short-form versions of the CDI were created that consist of an 89-word checklist for
12
C. Beaulieu-Poulin et al. / Early Human Development 98 (2016) 11–16
vocabulary comprehension and production at 8–18 months and of a 100-word checklist for vocabulary production at 16–30 months and a question on word combinations [17]. These have been adapted in several languages [18–21], but not validated in preterm populations yet. The Language Development Survey (LDS), another vocabulary production checklist filled by parents, is also a shorter alternative (b10 min to complete and b 5 min to score) that may be easy to use in practice [22]. The LDS has been recommended as a valuable tool to screen specifically for language delay in the general population [23]. In addition, it has been validated in bilingual homes in the United States using the English version of the questionnaire [24]. However, data in preterm infants remain limited. Only Mossabeb et al. compared results obtained from the LDS with the Bayley-III language composite score in preterm infants aged 22 to 26 months and showed a sensitivity of 61% and a specificity of 82% [25]. Psychometric properties of the LDS have not yet been determined at an earlier age in preterm infants. Therefore, this study aimed to examine the concurrent validity of the LDS compared to the language composite score of the Bayley-III in preterm infants born b 29 weeks of gestation at 18 months CA. This cohort consisted of French-speaking families. The 18-month visit is a key encounter to conduct developmental screening as recommended both by the American Academy of Pediatrics and the Canadian Pediatric Society [7,26]. We postulated that the LDS would be a reliable screening tool that is able to discriminate with an 80% sensitivity and specificity preterm born infants with language developmental delay. 2. Material and methods 2.1. Participants This was a retrospective cross-sectional analysis of a cohort of infants prospectively followed from neonatal discharge until 18 months corrected age. All infant survivors born b29 weeks of gestation and admitted to Centre Hospitalier Universitaire Sainte-Justine neonatal intensive care unit were eligible to be seen at the neonatal follow-up clinic as part of routine developmental surveillance and screening. There were no exclusion criteria for follow-up. For the purpose of this current study, we included all infants born between April 2009 and December 2012, aged 17 to 19 months corrected age at time of follow-up visit, who were able to complete the Bayley Scales of Infant and Toddler Development third edition (Bayley-III) and whose parents had filled the Language Development Survey (LDS). The local institutional ethics board approved of the study.
2.3. Index test Rescorla's LDS is a 310-word checklist used to identify language delay in 18–35 month-old children [22]. Parents are asked to mark all words spontaneously said by the child and to describe word combinations. Additional questions inquire on number of otitis media, number of languages spoken at home, family history of language delay, and parental concerns regarding child's language development. The total number of single words represents the infant's score. Above 24 months, length of word combinations is also taken into account. For infants between 18 and 23 months, a score ≤ 10 words for boys and ≤24 words for girls (≤15th percentile) suggests expressive language delay. Strong reliability has been demonstrated with test-retest reliability and Cronbach's alpha internal consistency of both 0.99 [22,24]. Correlation between 0.72 and 0.87 with other vocabulary tests indicates good validity [24,27]. Sensitivity is ≥ 80% and specificity ≥ 85% [24]. Completion takes b 10 min and requires a fifth-grade level reading skill. 2.4. Reference test The Bayley-III is a widely used and validated measure of child's developmental status which yields three main scales: cognitive, language and motor [28]. The language composite score combines results obtained from the expressive and receptive scalded scales. Mean is set at 100 with a standard deviation of 15. Language delay is defined as a language composite score b 85. Reliability is above 0.80. Validity measures range between 71% and 83% for the language scale when compared to the Wechsler Preschool and Primary Scale of Intelligence third edition (WPPSI-III) [28]. 2.5. Statistical analysis Descriptive statistics were computed in terms of means, standard deviations, and frequencies. Receiver-operating-characteristic (ROC) curves were constructed for boys and girls separately to find the optimal cut-off value on the LDS to identify a Bayley-III language composite score b 85. We calculated Cohen's kappa to determine chance independent agreement between classifications (i.e. normal language vs. language delay). Concurrent validity of the LDS in identifying children with Bayley-III language composite score b 85 was further assessed with sensitivity, specificity, positive predictive value, negative predictive value and 95% confidence interval. Statistical analyses were performed using SPSS 21.0. 3. Results
2.2. Procedure 3.1. Descriptive analyses As part of the 18-month CA visit routinely offered to preterm infants born b 29 weeks of gestation, all children were administered the Bayley-III which was performed in French by a trained psychologist who was blind to prematurity-related complications and results of the LDS (see below for details on the Bayley-III and LDS). Indeed, the psychologist, who was meeting with the family for the first time, only reviewed medical charts after completion of testing. Afterwards, regardless of performance on developmental testing, all parents were asked to fill the French version of the LDS. If the child said words in another language, parents were told to report them as well as described by Rescorla et al. [24]. Although they were present during Bayley-III administration, they were unaware of Bayley-III scores while completing the LDS. Birth history and perinatal data were obtained from the infant's medical record. Parents were also asked about their educational level, living arrangements, and child language exposure. Finally, a pediatrician completed the physical and neurological examination. Due to the retrospective nature of the study, we could not record reasons why children could not be tested on the Bayley-III or why parents did not fill the LDS.
Of 350 survivors born b29 weeks gestation, 289 (83%) attended their routine follow-up visit at 18 months CA, but 18 were outside the age range (N 19 months CA) and therefore excluded leaving 271 eligible infants (Fig. 1). Complete dataset for both the Bayley-III and the French version of the Language Development Survey (LDS) was available for 189 of them. Participants did not differ from non-participants in terms of gestational age, birth weight, sex, oxygen use at 36 weeks CA, severe retinopathy of prematurity, severe brain lesions, duration of neonatal hospitalization, household arrangement, maternal age and maternal ethnicity. However, participants were more likely to be singletons (81% vs. 55%) with mothers with an educational level above high school (67% vs. 50%). Developmental assessments were conducted at a mean corrected age (CA) of 18.2 months (range: 17 to 19 months). Sixty-nine percent of mothers described themselves as Caucasians. For the majority of families, primary language spoken at home was French (89%) with one third of children exposed to a bilingual environment. A high proportion of parents (43%) expressed concerns regarding their child's language
C. Beaulieu-Poulin et al. / Early Human Development 98 (2016) 11–16
13
Fig. 1. Flow diagram.
development. Table 1 shows detailed characteristics of the study population. The Bayley-III yielded a mean language composite score of 90 for the entire group (boys: 88 ± 15; girls: 93 ± 16). Mean expressive and receptive language scores were 8 (boys: 8 ± 3; girls: 9 ± 3). Language delay, defined by a score b 85, was found in 38/105 (36%) boys and in 24/84 (29%) girls. On the LDS, median number of words spontaneously spoken was 15 for the entire group. Median scores were 13.75 (range 0–152) for boys and 16.8 (range 0–163) for girls. Twelve children overall (6%) had no spoken word. Using Rescorla's cut-off for expressive language delay, 47/105 (45%) of preterm boys said ≤ 10 words and 50/84 (60%) of preterm girls had a vocabulary of ≤24 words. 3.2. Comparison of results on the Bayley-III and the language development survey Tables 2 and 3 display psychometric properties of the LDS when using the Bayley-III as a reference standard. On one hand, 29/38 (76%) of boys and 21/24 (88%) of girls with language delay as ascertained by the Bayley-III were correctly identified as such by the LDS when using Rescorla's cut-off scores, suggesting fairly good sensitivity. On the other hand, 49/67 (73%) of boys, but only 31/60 (52%) of girls with normal language development according to the Bayley-III scored in the normal range on the LDS, which demonstrated poor specificity for girls. 3.3. Missed referrals We looked more specifically at children with language delay as per the Bayley-III which were missed by the LDS. Nine boys out of 105
Table 1 Characteristics of participants.
Characteristics Maternal education, n (%) Below high school High school Above high school Paternal education, n (%) Below high school High school Above high school Caucasian, n (%) Single parent household, n (%) Primary language spoken at home, n (%) French Other Bilingual family, n (%) Gestational age, mean ± SD (range), weeks Birth weight, mean ± SD (range), g Male sex, n (%) Postnatal steroids, n (%) Severe retinopathy of prematuritya, n (%) Severe brain lesionb, n (%) Neurodevelopmental disabilities at 18 months, n (%) Cerebral palsy Neurosensory deafness Blindness Bayley scores at 18 months Cognitive composite, mean ± SD (range) Motor composite, mean ± SD (range)
Preterm infants n = 189 27/189 (14%) 35/189 (19%) 127/189 (67%) 34/177 (19%) 51/177 (29%) 92/177 (52%) 130/189 (69%) 10/189 (5%) 168/189 (89%) 21/189 (11%) 60/189 (32%) 26.7 ± 1.3 (23.0–28.9) 959 ± 224 (480–1580) 105/189 (56%) 48/187 (26%) 13/183 (7%) 14/188 (7%) 16/189 (9%) 4/187 (2%) 0 (0%) 95 ± 13 (49–130) 86 ± 14 (21–115)
a Severe retinopathy of prematurity defined as PLUS disease (most severe vascular changes observed on the retina) or requiring intervention. b Severe brain lesion defined as grade 3–4 intraventricular hemorrhage, periventricular leukomalacia or hydrocephalus.
14
C. Beaulieu-Poulin et al. / Early Human Development 98 (2016) 11–16
Table 2 Test properties of the LDS at 18 months CA for predicting a Bayley-III language composite score of b85 using different cut-off values in preterm boys (n = 105).
Table 4 Characteristics of infants according to test classification.
LDS cut-off score for delay
Sensitivity (95% CI) Specificity (95% CI) PPV (95% CI)b NPV (95% CI)c Over-referrals Under-referrals Cohen's κ (standard error) a b c
≤5
≤10a
≤15
≤20
55% (40–70) 91% (83–96) 78% (60–90) 78% (68–86) 5% 17% 0.38 (0.09)
76% (61–88) 73% (62–83) 62% (48–75) 85% (74–92) 17% 9% 0.44 (0.09)
84% (71–93) 63% (51–74) 56% (43–69) 88% (76–95) 24% 6% 0.44 (0.08)
92% (81–98) 52% (40–64) 52% (40–64) 92% (81–98) 30% 3% 0.44 (0.07)
Cut-off determined by Rescorla. Positive predictive value. Negative predictive value.
(9%) and 3/84 girls (4%) had a reported vocabulary above Rescorla's cutoff levels and yet, scored b85 on the Bayley-III. Among these twelve children, six parents had expressed concerns regarding language development. LDS scores for boys ranged between 13 and 73, whereas girls had scores between 32 and 43. 3.4. Determination of optimal cut-off points Receiver-operating-characteristic (ROC) curves were built using word count to identify Bayley-III language composite score b 85. Area under the curve (AUC) value was 0.85 (95% CI 0.77–0.93) for boys and 0.85 (95% CI 0.76–0.95) for girls indicating the good discriminative property of the LDS. A cut-off value of 10 for both boys (Table 2) and girls (Table 3) provided the optimal balance between sensitivity and specificity while minimizing under-referral rates. When combining data for all infants, a cut-off value of 10 had a sensitivity of 77% (95% CI 66–87) and a specificity of 77% (95% CI 69–84) to detect language delay as per the Bayley-III. We performed additional descriptive analyses a posteriori to characterize the four different classification subgroups (Table 4). Infants with true positive screening test (n = 48), compared to those with false positive (n = 29) and true negative tests (n = 98), had longer neonatal hospitalization (ANOVA with post-hoc tests, p b 0.05). Their parents were also more likely to raise concerns with language development (p b 0.05). They were also of lower gestational age than infants with true negative results (p b 0.01). There was no difference in language environment between the four groups. The 14 infants with false negative screening did not differ from the 98 with true negative screening in terms of neonatal characteristics and parental education. However, their parents were more likely to have concerns regarding their language development (50% vs. 17%, p-value = 0.03), even though the reported number of words spoken was higher than the cut-off value for a
Table 3 Test properties of the LDS at 18 months CA for predicting a Bayley-III language composite score of b85 using different cut-off values in preterm girls (n = 84). LDS cut-off score for delay
Sensitivity (95% CI) Specificity (95% CI) PPV (95% CI)b NPV (95% CI)c Over-referrals Under-referrals Cohen's κ (standard error) a b c
≤10
≤15
≤20
≤24a
79% (60–92) 82% (71–90) 63% (46–79) 91% (81–97) 13% 6% 0.44 (0.10)
83% (65–95) 68% (56–79) 51% (36–67) 91% (81–97) 23% 5% 0.41 (0.09)
88% (71–97) 52% (39–64) 42% (29–56) 91% (79–98) 35% 4% 0.35 (0.09)
88% (71–97) 52% (39–64) 42% (29–56) 91% (79–98) 35% 4% 0.35 (0.09)
Cut-off determined by Rescorla. Positive predictive value. Negative predictive value.
Neonatal characteristics Mean gestational age ± SD, weeks Mean birthweight ± SD, g Median days of hospitalization (range) Family characteristics Maternal education ≤high school, % Primarily French speaking, % Exposure to more than one language at home, % Parental concerns regarding child language development, % Family history of language delay, % Test scores Median number of words (range) Mean Bayley scores ± SD Composite language score Expressive language score Receptive language score Cognitive score Composite motor score
True positive N = 48
False positive N = 29
False negative N = 14
True negative N = 98
26 ± 1 888 ± 231 110 (53–658)
27 ± 1 966 ± 216 98 (52–203)
27 ± 1 996 ± 213 98 (53–169)
27 ± 1 986 ± 219 92 (39–189)
33% 79% 50%
41% 97% 31%
36% 86% 50%
30% 92% 34%
83%
68%
50%
17%
22%
33%
33%
31%
3 (0−10)
7 (2−10)
19 (13–73)
31 (11–163)
72 ± 10 5±2 5±2 86 ± 12 77 ± 15
91 ± 6 8±1 9±2 95 ± 13 88 ± 12
78 ± 7 7±2 7±1 93 ± 12 78 ± 21
100 ± 10 10 ± 2 10 ± 2 101 ± 9 92 ± 9
positive screen. Finally, the 29 infants with false positive screening scored significantly lower on Bayley motor and language scales compared to those with true negative screening (p ≤ 0.001). 4. Discussion This study aimed to determine the concurrent validity of the Language Development Survey (LDS) and the language composite score of the Bayley-III in preterm infants b29 weeks of gestation at 18 months corrected age (CA). In practice, developmental screening tests should accurately discriminate children with and without developmental delay with a 70 to 80% sensitivity and a specificity of around 80% [7]. Our results showed that using Rescorla's recommended cutoff points, the LDS had fairly good sensitivity in all preterm infants, fair specificity in preterm boys, but low specificity in preterm girls. In clinical terms, high sensitivity means that the majority of infants with language delay would have been identified with the LDS thus allowing proper referral to a speech therapist. The low specificity of the LDS in preterm girls would result in several infants with normal language development being labelled at risk of language delay and possibly referred unnecessarily for speech and language assessment. However, Glascoe [29] demonstrated that children with a false-positive screening test performed less well than those who were considered ‘normal’ on both the screening and assessment measures. Therefore, children who fail on a language screening test, but have normal language development as determined by a complete language assessment, may still require close developmental monitoring. Indeed, our results were consistent with Glascoe's findings: girls with false-positive screens using a cut-off ≤24 words had a median language composite score which was lower than those with true-negative screens (94 vs. 106). The same results were observed in boys (89 vs. 97). Nevertheless, we considered a specificity value of 52% with a positive predictive value of only 42% to be unacceptably low, especially in a context of limited health resources. Using a cut-off of ≤ 10 words in girls, the under-referral rate was 6% (vs. 4% using a cut-off of ≤24) and the over-referral rate 13% (vs. 35%), which is thought to be more manageable in a real-life setting. To our knowledge, only another study [25] tested the validity of the LDS in preterm born infants at 22 to 26 months CA (n = 178). Our results revealed a slightly higher sensitivity and equivalent specificity (61% and 82%, respectively, in Mossabeb's study). This can be explained
C. Beaulieu-Poulin et al. / Early Human Development 98 (2016) 11–16
by the fact that study populations were different: participants in Mossabeb's study were less premature, heavier at birth and were older at time of assessment. Furthermore, Mossabeb et al. used a Bayley-III language composite score above 90 (as opposed to above 85) to define normal language. Rescorla and Alley [27] have conducted a series of studies in the general population and found sensitivity values ranging from 73 to 94% and specificity values of 67–95% when using either the Reynell expressive language scale, the Bayley Scales of Infant Development or items from the Standford-Binet as reference standards in older [30]. When using the MacArthur–Bates CDI: Words and Sentences for comparison, correlation was around 90% in a sample of 23–25 month-old children [31]. Overall, these studies including ours, which took place in a French-speaking setting, support the use of the LDS to identify children at risk of language delays. The psychometric properties of the LDS compares favorably with other language development screening tests validated in preterm born infants and toddlers. Boyer et al. [32] found that the Ages & Stages Questionnaire communication scale had a sensitivity of 79% and a specificity of 63% to identify later language difficulties in preterm infants at 24 months CA. Recently, Blaggan et al. showed that the Parent Report of Children's Abilities-Revised (PARCA-R) had a sensitivity of 90% and a specificity of 76% to identify moderate to late preterm infants with a combined Bayley-III language or cognitive score b 80 [33]. The LDS has the advantage of being quick and easy to administer, but only assesses expressive language development. Results of this study must be interpreted in light of the following limitations. Due to the retrospective nature of the study, several participants had missing values: 70% of the eligible population completed both screening questionnaire and Bayley-III, leading to potential for selection bias. Indeed, included infants were less likely to be multiples, who usually have less favorable developmental prognosis than singleton. This could bias our results given that parents of healthier infants tend to be less concerned with their development [34] and could therefore be less accurate when reporting on their child's difficulties; however this would mean that the sensitivity and specificity of the LDS are underestimated. Reporting bias also needs to be considered since parents were present during the psychologist's assessment. Although they were not aware of the test results while filling the questionnaire, their child's performance during the evaluation may have influenced the way they completed the screening test, thus increasing correlation between the two tests in our study. Moreover, the LDS only screens for expressive language delay as opposed to other tests like the MacArthur-Bates CDI: Words and Gestures form for 8–18 month-old infants that also assesses gesture production and word comprehension. However, Sansavini et al. have shown that word production was strongly correlated with gesture/action production as well as receptive vocabulary at 18 months corrected age in preterm infants, highlighting the interconnectedness of these competencies and suggesting that delay in expressive language is possibly associated with delay in receptive language [16]. Finally, the Bayley-III is not an evaluation of language per se. As such, it cannot be inferred that in our study the LDS was able to accurately identify true language delay as a full assessment by a certified speech therapist was not done on all infants. Nevertheless, the Bayley-III is widely used and accepted in the research setting as a measure of language development. 5. Conclusion Preterm children are at greater risk of long-lasting language difficulties and need to be identified promptly to improve functional outcomes. The Language Development Survey is a rapid screening instrument that performs well in detecting 18 month-old preterm infants at risk of language delays. The optimal cut-off set at a word count of ≤10 makes the Language Development Survey very simple to use and score by community pediatricians or neonatal follow-up programs that cannot afford direct assessment of language development, thus ensuring that all
15
preterm infants, despite limited resources, can be properly screened using a validated tool. Acknowledgement This study was supported by grants from the Fonds de Recherche en Santé du Québec and by the Centre de recherche du CHU Sainte-Justine. References [1] N. Barre, A. Morgan, L.W. Doyle, P.J. Anderson, Language abilities in children who were very preterm and/or very low birth weight: a meta-analysis, J. Pediatr. 158 (5) (2011) 766–774 (e1). [2] N. Reidy, A. Morgan, D.K. Thompson, T.E. Inder, L.W. Doyle, P.J. Anderson, Impaired language abilities and white matter abnormalities in children born very preterm and/or very low birth weight, J. Pediatr. 162 (4) (2013) 719–724. [3] I.L. van Noort-van der Spek, M.C. Franken, N. Weisglas-Kuperus, Language functions in preterm-born children: a systematic review and meta-analysis, Pediatrics 129 (4) (2012) 745–754. [4] T.M. Luu, L.R. Ment, K.C. Schneider, K.H. Katz, W.C. Allan, B.R. Vohr, Lasting effects of preterm birth and neonatal brain hemorrhage at 12 years of age, Pediatrics 123 (3) (2009) 1037–1044. [5] P.J. Anderson, C.R. De Luca, E. Hutchinson, G. Roberts, L.W. Doyle, Victorian infant collaborative G. Underestimation of developmental delay by the new Bayley-III scale, Arch. Pediatr. Adolesc. Med. 164 (4) (2010) 352–356. [6] K. Durkin, G. Conti-Ramsden, Language, social behavior, and the quality of friendships in adolescents with and without a history of specific language impairment, Child Dev. 78 (5) (2007) 1441–1457. [7] L. Nadeau, M. Boivin, R. Tessier, F. Lefebvre, P. Robaey, Mediators of behavioral problems in 7-year-old children born after 24 to 28 weeks of gestation, J. Dev. Behav. Pediatr. 22 (1) (2001) 1–10. [8] I.F. Wallace, N.D. Berkman, L.R. Watson, T. Coyne-Beasley, C.T. Wood, K. Cullen, et al., Screening for speech and language delay in children 5 years old and younger: a systematic review, Pediatrics 136 (2) (2015) (e448-e62). [9] S. Foster-Cohen, J.O. Edgin, P.R. Champion, L.J. Woodward, Early delayed language development in very preterm infants: evidence from the MacArthur-bates CDI, J. Child Lang. 34 (3) (2007) 655–675. [10] L. Marston, J.L. Peacock, S.A. Calvert, A. Greenough, N. Marlow, Factors affecting vocabulary acquisition at age 2 in children born between 23 and 28 weeks' gestation, Dev. Med. Child Neurol. 49 (8) (2007) 591–596. [11] M. Perez-Pereira, P. Fernandez, M.L. Gomez-Taibo, M. Resches, Language development of low risk preterm infants up to the age of 30 months, Early Hum. Dev. 90 (10) (2014) 649–656. [12] A. Sansavini, A. Bello, A. Guarini, S. Savini, R. Alessandroni, G. Faldella, et al., Noun and predicate comprehension/production and gestures in extremely preterm children at two years of age: are they delayed? J. Commun. Disord. 58 (2015) 126–142. [13] A. Schults, T. Tulviste, E. Haan, Early vocabulary in full term and preterm Estonian children, Early Hum. Dev. 89 (9) (2013) 721–726. [14] L. Fenson, P.S. Dale, J.S. Reznick, D. Thal, E. Bates, J.P. Hartung, et al., The MacArthur Communicative Development Inventories: User's guide and technical manual, Singular Publishing Group, San Diego, 1993. [15] S. Kern, F. Gayraud, Influence of preterm birth on early lexical and grammatical acquisition, First Lang. 27 (2) (2007) 159–173. [16] A. Sansavini, A. Guarini, S. Savini, S. Broccoli, L. Justice, R. Alessandroni, et al., Longitudinal trajectories of gestural and linguistic abilities in very preterm infants in the second year of life, Neuropsychologia 49 (13) (2011) 3677–3688. [17] L. Fenson, S.J. Pethick, C. Renda, J.L. Cox, P.S. Dale, J.S. Reznick, Short-form versions of the MacArthur communicative development inventories, Appl. Psycholinguist. 21 (01) (2000) 95–116. [18] S.W. Kim, H.R. Jeon, E.J. Park, H.I. Kim, W. Jung da, M.R. Woo, The usefulness of M-B CDI-K short form as screening test in children with language developmental delay, Ann. Rehabil. Med. 38 (3) (2014) 376–380. [19] W. Vach, D. Bleses, R. Jorgensen, Construction of a Danish CDI short form for language screening at the age of 36 months: methodological considerations and results, Clin. Linguist. Phon. 24 (8) (2010) 602–621. [20] M. Westerlund, E. Berglund, M. Eriksson, Can severely language delayed 3-year-olds be identified at 18 months? Evaluation of a screening version of the MacArthurbates communicative development inventories, J. Speech Lang. Hear. Res. 49 (2) (2006) 237–247. [21] D. Jackson-Maldonado, V.A. Marchman, L.C.H. Fernald, Short-form versions of the Spanish MacArthur–Bates communicative development inventories, Appl. Psycholinguist. 34 (04) (2013) 837–868. [22] L. Rescorla, The language development survey: a screening tool for delayed language in toddlers, J. Speech Hear. Disord. 54 (4) (1989) 587–599. [23] D. Drotar, T. Stancin, P.H. Dworkin, L. Sices, S. Wood, Selecting developmental surveillance and screening tools, Pediatrics in review / American Academy of Pediatrics. 29 (10) (2008) e52–e58. [24] L. Rescorla, T.M. Achenbach, Use of the language development survey (LDS) in a national probability sample of children 18 to 35 months old, J. Speech Lang. Hear. Res. 45 (4) (2002) 733–743. [25] R. Mossabeb, K.C. Wade, K. Finnegan, E. Sivieri, S. Abbasi, Language development survey provides a useful screening tool for language delay in preterm infants, Clin. Pediatr. 51 (7) (2012) 638–644.
16
C. Beaulieu-Poulin et al. / Early Human Development 98 (2016) 11–16
[26] R. Williams, J. Clinton, Canadian Paediatric society early years task force. Getting it right at 18 months: in support of an enhanced well-baby visit, Paediatr. Child Health 16 (10) (2011) 647–654. [27] L. Rescorla, A. Alley, Validation of the language development survey (LDS): a parent report tool for identifying language delay in toddlers, J. Speech Lang. Hear. Res. 44 (2) (2001) 434–445. [28] C.A. Albers, A.J. Grieve, Test review: Bayley, N. (2006). Bayley scales of infant and toddler development - third edition. San Antonio, TX: Harcourt assessment, J. Psychoeduc. Assess. 25 (2) (2007) 18. [29] F.P. Glascoe, Are overreferrals on developmental screening tests really a problem? Archives of pediatrics & adolescent medicine. 155 (1) (2001) 54–59. [30] L. Rescorla, M. Hadicke-Wiley, E. Escarce, Epidemiological investigation of expressive language delay at age two, First Lang. J. 13 (5) (1993).
[31] L. Rescorla, N.B. Ratner, P. Jusczyk, A.M. Jusczyk, Concurrent validity of the language development survey: associations with the MacArthur-Bates communicative development inventories: words and sentences, American journal of speech-language pathology / American Speech-Language-Hearing Association. 14 (2) (2005) 156–163. [32] J. Boyer, C. Flamant, G. Boussicault, I. Berlie, G. Gascoin, B. Branger, et al., Characterizing early detection of language difficulties in children born preterm, Early Hum. Dev. 90 (6) (2014) 281–286. [33] S. Blaggan, A. Guy, E.M. Boyle, E. Spata, B.N. Manktelow, D. Wolke, et al., A parent questionnaire for developmental screening in infants born late and moderately preterm, Pediatrics 134 (1) (2014) e55–e62. [34] A. Heiser, O. Curcin, C. Luhr, I. Grimmer, B. Metze, M. Obladen, Parental and professional agreement in developmental assessment of very-low-birthweight and term infants, Dev. Med. Child Neurol. 42 (1) (2000) 21–24.
Contents lists available at ScienceDirect
Early Human Development journal homepage: www.elsevier.com/locate/earlhumdev
Validity of the language development survey in infants born preterm Camille Beaulieu-Poulin a, Marie-Noëlle Simard b, Hélène Babakissa a, Francine Lefebvre a, Thuy Mai Luu a,⁎ a b
Department of Pediatrics, CHU Sainte-Justine, 3175 Chemin de la Côte-Ste-Catherine, Montreal, Quebec H3T 1C5, Canada School of Rehabilitation, Faculty of Medicine, University of Montreal, C.P. 6128, succursale Centre-Ville, Montreal, Quebec H3C 3J7, Canada
a r t i c l e
i n f o
Article history: Received 7 January 2016 Received in revised form 26 May 2016 Accepted 4 June 2016 Keywords: Preterm infants Language delay Developmental screening Language development survey Bayley scales of infant development
a b s t r a c t Background: Preterm infants are at greater risk of language delay. Early identification of language delay is essential to improve functional outcome in these children. Aim: To examine the concurrent validity of Rescorla's Language Development Survey and the Bayley Scales of Infant and Toddler Development (Bayley-III) at 18 months corrected age in preterm infants. Study design: Test accuracy study. Participants: 189 preterm infants born b29 weeks were assessed at 18 months. Outcome measures: The Language Development Survey, a parent-reported screening instrument, was administered in French concurrently with the Language Scales of the Bayley-III. Receiver-Operating-Characteristics curves were used to determine optimal cut-off score on the Language Development Survey to identify Bayley-III score b 85. Sensitivity, specificity, positive and negative predictive values, and κ coefficient were calculated. Results: Using Rescorla's original cut-off scores of ≤10 words for boys and ≤24 for girls, sensitivity was 76% and 88% for boys and girls, respectively, and specificity was 73% and 52% for boys and girls, respectively, in identifying language delay as per the Bayley-III. The optimal threshold was ≤10 words for both boys and girls. In girls, lowering the cut-off score decreased sensitivity (79%), but improved specificity (82%), thus lowering the number of false-positives. Conclusion: Our findings support using the Language Development Survey as an expressive language screener in preterm infants. © 2016 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Communication and language problems are frequently encountered in infants born prematurely [1]. More specifically, preterm children exhibit difficulties across all language sub-domains, including phonological awareness, semantics, grammar, discourse, and pragmatics [2,3]. Language difficulties persist throughout the school years with 15% of preterm children still requiring speech therapy at 12 years [4]. The first cues indicating language problems can be identified as early as during infancy. Indeed, among extremely preterm infants, 20–35% have language delay according to the Bayley Scales of Infant and Toddler Development 3rd edition (Bayley-III) at 18–24 months corrected age (CA) [5]. Language impairment has a significant long-term impact and may affect social relationships, literacy and educational attainment [6]. Therefore, early recognition of language delay is crucial to improve functional
Abbreviations: LDS, Language Development Survey; CA, corrected age. ⁎ Corresponding author. E-mail addresses: [email protected] (C. Beaulieu-Poulin), [email protected] (M.-N. Simard), [email protected] (H. Babakissa), [email protected] (F. Lefebvre), [email protected] (T.M. Luu).
http://dx.doi.org/10.1016/j.earlhumdev.2016.06.003 0378-3782/© 2016 Elsevier Ireland Ltd. All rights reserved.
and academic outcomes of at-risk children. As such, the American Academy of Pediatrics has been advocating for standardized developmental screening at specific ages using validated tools including parentreported questionnaires [7]. Prompt identification of language developmental delay allows the child to benefit from early speech and language intervention which has been shown to enhance language abilities and to reduce the burden of a persistent disability on child's function [8]. Standardized language-specific screening tools for infants that are parent-administered and quick to complete and score are scarce. One such parental questionnaire, the MacArthur-Bates Communicative Development Inventory (CDI), has been adapted in multiple languages and widely used for research with preterm populations in several countries [9–14]. At 12–24 months corrected ages, preterm infants display lower gesture/word production, as measured by the CDI, as well as smaller vocabulary size and shorter length of utterances compared to term controls, with lower gestational ages associated with more impairment [9,15,16]. Though the CDI has proven useful to assess language development in preterm infants, time required to complete the form (30–40 min) and to score it (20 min) is rather long and could hinder uptake by busy general health care providers. To overcome these barriers, short-form versions of the CDI were created that consist of an 89-word checklist for
12
C. Beaulieu-Poulin et al. / Early Human Development 98 (2016) 11–16
vocabulary comprehension and production at 8–18 months and of a 100-word checklist for vocabulary production at 16–30 months and a question on word combinations [17]. These have been adapted in several languages [18–21], but not validated in preterm populations yet. The Language Development Survey (LDS), another vocabulary production checklist filled by parents, is also a shorter alternative (b10 min to complete and b 5 min to score) that may be easy to use in practice [22]. The LDS has been recommended as a valuable tool to screen specifically for language delay in the general population [23]. In addition, it has been validated in bilingual homes in the United States using the English version of the questionnaire [24]. However, data in preterm infants remain limited. Only Mossabeb et al. compared results obtained from the LDS with the Bayley-III language composite score in preterm infants aged 22 to 26 months and showed a sensitivity of 61% and a specificity of 82% [25]. Psychometric properties of the LDS have not yet been determined at an earlier age in preterm infants. Therefore, this study aimed to examine the concurrent validity of the LDS compared to the language composite score of the Bayley-III in preterm infants born b 29 weeks of gestation at 18 months CA. This cohort consisted of French-speaking families. The 18-month visit is a key encounter to conduct developmental screening as recommended both by the American Academy of Pediatrics and the Canadian Pediatric Society [7,26]. We postulated that the LDS would be a reliable screening tool that is able to discriminate with an 80% sensitivity and specificity preterm born infants with language developmental delay. 2. Material and methods 2.1. Participants This was a retrospective cross-sectional analysis of a cohort of infants prospectively followed from neonatal discharge until 18 months corrected age. All infant survivors born b29 weeks of gestation and admitted to Centre Hospitalier Universitaire Sainte-Justine neonatal intensive care unit were eligible to be seen at the neonatal follow-up clinic as part of routine developmental surveillance and screening. There were no exclusion criteria for follow-up. For the purpose of this current study, we included all infants born between April 2009 and December 2012, aged 17 to 19 months corrected age at time of follow-up visit, who were able to complete the Bayley Scales of Infant and Toddler Development third edition (Bayley-III) and whose parents had filled the Language Development Survey (LDS). The local institutional ethics board approved of the study.
2.3. Index test Rescorla's LDS is a 310-word checklist used to identify language delay in 18–35 month-old children [22]. Parents are asked to mark all words spontaneously said by the child and to describe word combinations. Additional questions inquire on number of otitis media, number of languages spoken at home, family history of language delay, and parental concerns regarding child's language development. The total number of single words represents the infant's score. Above 24 months, length of word combinations is also taken into account. For infants between 18 and 23 months, a score ≤ 10 words for boys and ≤24 words for girls (≤15th percentile) suggests expressive language delay. Strong reliability has been demonstrated with test-retest reliability and Cronbach's alpha internal consistency of both 0.99 [22,24]. Correlation between 0.72 and 0.87 with other vocabulary tests indicates good validity [24,27]. Sensitivity is ≥ 80% and specificity ≥ 85% [24]. Completion takes b 10 min and requires a fifth-grade level reading skill. 2.4. Reference test The Bayley-III is a widely used and validated measure of child's developmental status which yields three main scales: cognitive, language and motor [28]. The language composite score combines results obtained from the expressive and receptive scalded scales. Mean is set at 100 with a standard deviation of 15. Language delay is defined as a language composite score b 85. Reliability is above 0.80. Validity measures range between 71% and 83% for the language scale when compared to the Wechsler Preschool and Primary Scale of Intelligence third edition (WPPSI-III) [28]. 2.5. Statistical analysis Descriptive statistics were computed in terms of means, standard deviations, and frequencies. Receiver-operating-characteristic (ROC) curves were constructed for boys and girls separately to find the optimal cut-off value on the LDS to identify a Bayley-III language composite score b 85. We calculated Cohen's kappa to determine chance independent agreement between classifications (i.e. normal language vs. language delay). Concurrent validity of the LDS in identifying children with Bayley-III language composite score b 85 was further assessed with sensitivity, specificity, positive predictive value, negative predictive value and 95% confidence interval. Statistical analyses were performed using SPSS 21.0. 3. Results
2.2. Procedure 3.1. Descriptive analyses As part of the 18-month CA visit routinely offered to preterm infants born b 29 weeks of gestation, all children were administered the Bayley-III which was performed in French by a trained psychologist who was blind to prematurity-related complications and results of the LDS (see below for details on the Bayley-III and LDS). Indeed, the psychologist, who was meeting with the family for the first time, only reviewed medical charts after completion of testing. Afterwards, regardless of performance on developmental testing, all parents were asked to fill the French version of the LDS. If the child said words in another language, parents were told to report them as well as described by Rescorla et al. [24]. Although they were present during Bayley-III administration, they were unaware of Bayley-III scores while completing the LDS. Birth history and perinatal data were obtained from the infant's medical record. Parents were also asked about their educational level, living arrangements, and child language exposure. Finally, a pediatrician completed the physical and neurological examination. Due to the retrospective nature of the study, we could not record reasons why children could not be tested on the Bayley-III or why parents did not fill the LDS.
Of 350 survivors born b29 weeks gestation, 289 (83%) attended their routine follow-up visit at 18 months CA, but 18 were outside the age range (N 19 months CA) and therefore excluded leaving 271 eligible infants (Fig. 1). Complete dataset for both the Bayley-III and the French version of the Language Development Survey (LDS) was available for 189 of them. Participants did not differ from non-participants in terms of gestational age, birth weight, sex, oxygen use at 36 weeks CA, severe retinopathy of prematurity, severe brain lesions, duration of neonatal hospitalization, household arrangement, maternal age and maternal ethnicity. However, participants were more likely to be singletons (81% vs. 55%) with mothers with an educational level above high school (67% vs. 50%). Developmental assessments were conducted at a mean corrected age (CA) of 18.2 months (range: 17 to 19 months). Sixty-nine percent of mothers described themselves as Caucasians. For the majority of families, primary language spoken at home was French (89%) with one third of children exposed to a bilingual environment. A high proportion of parents (43%) expressed concerns regarding their child's language
C. Beaulieu-Poulin et al. / Early Human Development 98 (2016) 11–16
13
Fig. 1. Flow diagram.
development. Table 1 shows detailed characteristics of the study population. The Bayley-III yielded a mean language composite score of 90 for the entire group (boys: 88 ± 15; girls: 93 ± 16). Mean expressive and receptive language scores were 8 (boys: 8 ± 3; girls: 9 ± 3). Language delay, defined by a score b 85, was found in 38/105 (36%) boys and in 24/84 (29%) girls. On the LDS, median number of words spontaneously spoken was 15 for the entire group. Median scores were 13.75 (range 0–152) for boys and 16.8 (range 0–163) for girls. Twelve children overall (6%) had no spoken word. Using Rescorla's cut-off for expressive language delay, 47/105 (45%) of preterm boys said ≤ 10 words and 50/84 (60%) of preterm girls had a vocabulary of ≤24 words. 3.2. Comparison of results on the Bayley-III and the language development survey Tables 2 and 3 display psychometric properties of the LDS when using the Bayley-III as a reference standard. On one hand, 29/38 (76%) of boys and 21/24 (88%) of girls with language delay as ascertained by the Bayley-III were correctly identified as such by the LDS when using Rescorla's cut-off scores, suggesting fairly good sensitivity. On the other hand, 49/67 (73%) of boys, but only 31/60 (52%) of girls with normal language development according to the Bayley-III scored in the normal range on the LDS, which demonstrated poor specificity for girls. 3.3. Missed referrals We looked more specifically at children with language delay as per the Bayley-III which were missed by the LDS. Nine boys out of 105
Table 1 Characteristics of participants.
Characteristics Maternal education, n (%) Below high school High school Above high school Paternal education, n (%) Below high school High school Above high school Caucasian, n (%) Single parent household, n (%) Primary language spoken at home, n (%) French Other Bilingual family, n (%) Gestational age, mean ± SD (range), weeks Birth weight, mean ± SD (range), g Male sex, n (%) Postnatal steroids, n (%) Severe retinopathy of prematuritya, n (%) Severe brain lesionb, n (%) Neurodevelopmental disabilities at 18 months, n (%) Cerebral palsy Neurosensory deafness Blindness Bayley scores at 18 months Cognitive composite, mean ± SD (range) Motor composite, mean ± SD (range)
Preterm infants n = 189 27/189 (14%) 35/189 (19%) 127/189 (67%) 34/177 (19%) 51/177 (29%) 92/177 (52%) 130/189 (69%) 10/189 (5%) 168/189 (89%) 21/189 (11%) 60/189 (32%) 26.7 ± 1.3 (23.0–28.9) 959 ± 224 (480–1580) 105/189 (56%) 48/187 (26%) 13/183 (7%) 14/188 (7%) 16/189 (9%) 4/187 (2%) 0 (0%) 95 ± 13 (49–130) 86 ± 14 (21–115)
a Severe retinopathy of prematurity defined as PLUS disease (most severe vascular changes observed on the retina) or requiring intervention. b Severe brain lesion defined as grade 3–4 intraventricular hemorrhage, periventricular leukomalacia or hydrocephalus.
14
C. Beaulieu-Poulin et al. / Early Human Development 98 (2016) 11–16
Table 2 Test properties of the LDS at 18 months CA for predicting a Bayley-III language composite score of b85 using different cut-off values in preterm boys (n = 105).
Table 4 Characteristics of infants according to test classification.
LDS cut-off score for delay
Sensitivity (95% CI) Specificity (95% CI) PPV (95% CI)b NPV (95% CI)c Over-referrals Under-referrals Cohen's κ (standard error) a b c
≤5
≤10a
≤15
≤20
55% (40–70) 91% (83–96) 78% (60–90) 78% (68–86) 5% 17% 0.38 (0.09)
76% (61–88) 73% (62–83) 62% (48–75) 85% (74–92) 17% 9% 0.44 (0.09)
84% (71–93) 63% (51–74) 56% (43–69) 88% (76–95) 24% 6% 0.44 (0.08)
92% (81–98) 52% (40–64) 52% (40–64) 92% (81–98) 30% 3% 0.44 (0.07)
Cut-off determined by Rescorla. Positive predictive value. Negative predictive value.
(9%) and 3/84 girls (4%) had a reported vocabulary above Rescorla's cutoff levels and yet, scored b85 on the Bayley-III. Among these twelve children, six parents had expressed concerns regarding language development. LDS scores for boys ranged between 13 and 73, whereas girls had scores between 32 and 43. 3.4. Determination of optimal cut-off points Receiver-operating-characteristic (ROC) curves were built using word count to identify Bayley-III language composite score b 85. Area under the curve (AUC) value was 0.85 (95% CI 0.77–0.93) for boys and 0.85 (95% CI 0.76–0.95) for girls indicating the good discriminative property of the LDS. A cut-off value of 10 for both boys (Table 2) and girls (Table 3) provided the optimal balance between sensitivity and specificity while minimizing under-referral rates. When combining data for all infants, a cut-off value of 10 had a sensitivity of 77% (95% CI 66–87) and a specificity of 77% (95% CI 69–84) to detect language delay as per the Bayley-III. We performed additional descriptive analyses a posteriori to characterize the four different classification subgroups (Table 4). Infants with true positive screening test (n = 48), compared to those with false positive (n = 29) and true negative tests (n = 98), had longer neonatal hospitalization (ANOVA with post-hoc tests, p b 0.05). Their parents were also more likely to raise concerns with language development (p b 0.05). They were also of lower gestational age than infants with true negative results (p b 0.01). There was no difference in language environment between the four groups. The 14 infants with false negative screening did not differ from the 98 with true negative screening in terms of neonatal characteristics and parental education. However, their parents were more likely to have concerns regarding their language development (50% vs. 17%, p-value = 0.03), even though the reported number of words spoken was higher than the cut-off value for a
Table 3 Test properties of the LDS at 18 months CA for predicting a Bayley-III language composite score of b85 using different cut-off values in preterm girls (n = 84). LDS cut-off score for delay
Sensitivity (95% CI) Specificity (95% CI) PPV (95% CI)b NPV (95% CI)c Over-referrals Under-referrals Cohen's κ (standard error) a b c
≤10
≤15
≤20
≤24a
79% (60–92) 82% (71–90) 63% (46–79) 91% (81–97) 13% 6% 0.44 (0.10)
83% (65–95) 68% (56–79) 51% (36–67) 91% (81–97) 23% 5% 0.41 (0.09)
88% (71–97) 52% (39–64) 42% (29–56) 91% (79–98) 35% 4% 0.35 (0.09)
88% (71–97) 52% (39–64) 42% (29–56) 91% (79–98) 35% 4% 0.35 (0.09)
Cut-off determined by Rescorla. Positive predictive value. Negative predictive value.
Neonatal characteristics Mean gestational age ± SD, weeks Mean birthweight ± SD, g Median days of hospitalization (range) Family characteristics Maternal education ≤high school, % Primarily French speaking, % Exposure to more than one language at home, % Parental concerns regarding child language development, % Family history of language delay, % Test scores Median number of words (range) Mean Bayley scores ± SD Composite language score Expressive language score Receptive language score Cognitive score Composite motor score
True positive N = 48
False positive N = 29
False negative N = 14
True negative N = 98
26 ± 1 888 ± 231 110 (53–658)
27 ± 1 966 ± 216 98 (52–203)
27 ± 1 996 ± 213 98 (53–169)
27 ± 1 986 ± 219 92 (39–189)
33% 79% 50%
41% 97% 31%
36% 86% 50%
30% 92% 34%
83%
68%
50%
17%
22%
33%
33%
31%
3 (0−10)
7 (2−10)
19 (13–73)
31 (11–163)
72 ± 10 5±2 5±2 86 ± 12 77 ± 15
91 ± 6 8±1 9±2 95 ± 13 88 ± 12
78 ± 7 7±2 7±1 93 ± 12 78 ± 21
100 ± 10 10 ± 2 10 ± 2 101 ± 9 92 ± 9
positive screen. Finally, the 29 infants with false positive screening scored significantly lower on Bayley motor and language scales compared to those with true negative screening (p ≤ 0.001). 4. Discussion This study aimed to determine the concurrent validity of the Language Development Survey (LDS) and the language composite score of the Bayley-III in preterm infants b29 weeks of gestation at 18 months corrected age (CA). In practice, developmental screening tests should accurately discriminate children with and without developmental delay with a 70 to 80% sensitivity and a specificity of around 80% [7]. Our results showed that using Rescorla's recommended cutoff points, the LDS had fairly good sensitivity in all preterm infants, fair specificity in preterm boys, but low specificity in preterm girls. In clinical terms, high sensitivity means that the majority of infants with language delay would have been identified with the LDS thus allowing proper referral to a speech therapist. The low specificity of the LDS in preterm girls would result in several infants with normal language development being labelled at risk of language delay and possibly referred unnecessarily for speech and language assessment. However, Glascoe [29] demonstrated that children with a false-positive screening test performed less well than those who were considered ‘normal’ on both the screening and assessment measures. Therefore, children who fail on a language screening test, but have normal language development as determined by a complete language assessment, may still require close developmental monitoring. Indeed, our results were consistent with Glascoe's findings: girls with false-positive screens using a cut-off ≤24 words had a median language composite score which was lower than those with true-negative screens (94 vs. 106). The same results were observed in boys (89 vs. 97). Nevertheless, we considered a specificity value of 52% with a positive predictive value of only 42% to be unacceptably low, especially in a context of limited health resources. Using a cut-off of ≤ 10 words in girls, the under-referral rate was 6% (vs. 4% using a cut-off of ≤24) and the over-referral rate 13% (vs. 35%), which is thought to be more manageable in a real-life setting. To our knowledge, only another study [25] tested the validity of the LDS in preterm born infants at 22 to 26 months CA (n = 178). Our results revealed a slightly higher sensitivity and equivalent specificity (61% and 82%, respectively, in Mossabeb's study). This can be explained
C. Beaulieu-Poulin et al. / Early Human Development 98 (2016) 11–16
by the fact that study populations were different: participants in Mossabeb's study were less premature, heavier at birth and were older at time of assessment. Furthermore, Mossabeb et al. used a Bayley-III language composite score above 90 (as opposed to above 85) to define normal language. Rescorla and Alley [27] have conducted a series of studies in the general population and found sensitivity values ranging from 73 to 94% and specificity values of 67–95% when using either the Reynell expressive language scale, the Bayley Scales of Infant Development or items from the Standford-Binet as reference standards in older [30]. When using the MacArthur–Bates CDI: Words and Sentences for comparison, correlation was around 90% in a sample of 23–25 month-old children [31]. Overall, these studies including ours, which took place in a French-speaking setting, support the use of the LDS to identify children at risk of language delays. The psychometric properties of the LDS compares favorably with other language development screening tests validated in preterm born infants and toddlers. Boyer et al. [32] found that the Ages & Stages Questionnaire communication scale had a sensitivity of 79% and a specificity of 63% to identify later language difficulties in preterm infants at 24 months CA. Recently, Blaggan et al. showed that the Parent Report of Children's Abilities-Revised (PARCA-R) had a sensitivity of 90% and a specificity of 76% to identify moderate to late preterm infants with a combined Bayley-III language or cognitive score b 80 [33]. The LDS has the advantage of being quick and easy to administer, but only assesses expressive language development. Results of this study must be interpreted in light of the following limitations. Due to the retrospective nature of the study, several participants had missing values: 70% of the eligible population completed both screening questionnaire and Bayley-III, leading to potential for selection bias. Indeed, included infants were less likely to be multiples, who usually have less favorable developmental prognosis than singleton. This could bias our results given that parents of healthier infants tend to be less concerned with their development [34] and could therefore be less accurate when reporting on their child's difficulties; however this would mean that the sensitivity and specificity of the LDS are underestimated. Reporting bias also needs to be considered since parents were present during the psychologist's assessment. Although they were not aware of the test results while filling the questionnaire, their child's performance during the evaluation may have influenced the way they completed the screening test, thus increasing correlation between the two tests in our study. Moreover, the LDS only screens for expressive language delay as opposed to other tests like the MacArthur-Bates CDI: Words and Gestures form for 8–18 month-old infants that also assesses gesture production and word comprehension. However, Sansavini et al. have shown that word production was strongly correlated with gesture/action production as well as receptive vocabulary at 18 months corrected age in preterm infants, highlighting the interconnectedness of these competencies and suggesting that delay in expressive language is possibly associated with delay in receptive language [16]. Finally, the Bayley-III is not an evaluation of language per se. As such, it cannot be inferred that in our study the LDS was able to accurately identify true language delay as a full assessment by a certified speech therapist was not done on all infants. Nevertheless, the Bayley-III is widely used and accepted in the research setting as a measure of language development. 5. Conclusion Preterm children are at greater risk of long-lasting language difficulties and need to be identified promptly to improve functional outcomes. The Language Development Survey is a rapid screening instrument that performs well in detecting 18 month-old preterm infants at risk of language delays. The optimal cut-off set at a word count of ≤10 makes the Language Development Survey very simple to use and score by community pediatricians or neonatal follow-up programs that cannot afford direct assessment of language development, thus ensuring that all
15
preterm infants, despite limited resources, can be properly screened using a validated tool. Acknowledgement This study was supported by grants from the Fonds de Recherche en Santé du Québec and by the Centre de recherche du CHU Sainte-Justine. References [1] N. Barre, A. Morgan, L.W. Doyle, P.J. Anderson, Language abilities in children who were very preterm and/or very low birth weight: a meta-analysis, J. Pediatr. 158 (5) (2011) 766–774 (e1). [2] N. Reidy, A. Morgan, D.K. Thompson, T.E. Inder, L.W. Doyle, P.J. Anderson, Impaired language abilities and white matter abnormalities in children born very preterm and/or very low birth weight, J. Pediatr. 162 (4) (2013) 719–724. [3] I.L. van Noort-van der Spek, M.C. Franken, N. Weisglas-Kuperus, Language functions in preterm-born children: a systematic review and meta-analysis, Pediatrics 129 (4) (2012) 745–754. [4] T.M. Luu, L.R. Ment, K.C. Schneider, K.H. Katz, W.C. Allan, B.R. Vohr, Lasting effects of preterm birth and neonatal brain hemorrhage at 12 years of age, Pediatrics 123 (3) (2009) 1037–1044. [5] P.J. Anderson, C.R. De Luca, E. Hutchinson, G. Roberts, L.W. Doyle, Victorian infant collaborative G. Underestimation of developmental delay by the new Bayley-III scale, Arch. Pediatr. Adolesc. Med. 164 (4) (2010) 352–356. [6] K. Durkin, G. Conti-Ramsden, Language, social behavior, and the quality of friendships in adolescents with and without a history of specific language impairment, Child Dev. 78 (5) (2007) 1441–1457. [7] L. Nadeau, M. Boivin, R. Tessier, F. Lefebvre, P. Robaey, Mediators of behavioral problems in 7-year-old children born after 24 to 28 weeks of gestation, J. Dev. Behav. Pediatr. 22 (1) (2001) 1–10. [8] I.F. Wallace, N.D. Berkman, L.R. Watson, T. Coyne-Beasley, C.T. Wood, K. Cullen, et al., Screening for speech and language delay in children 5 years old and younger: a systematic review, Pediatrics 136 (2) (2015) (e448-e62). [9] S. Foster-Cohen, J.O. Edgin, P.R. Champion, L.J. Woodward, Early delayed language development in very preterm infants: evidence from the MacArthur-bates CDI, J. Child Lang. 34 (3) (2007) 655–675. [10] L. Marston, J.L. Peacock, S.A. Calvert, A. Greenough, N. Marlow, Factors affecting vocabulary acquisition at age 2 in children born between 23 and 28 weeks' gestation, Dev. Med. Child Neurol. 49 (8) (2007) 591–596. [11] M. Perez-Pereira, P. Fernandez, M.L. Gomez-Taibo, M. Resches, Language development of low risk preterm infants up to the age of 30 months, Early Hum. Dev. 90 (10) (2014) 649–656. [12] A. Sansavini, A. Bello, A. Guarini, S. Savini, R. Alessandroni, G. Faldella, et al., Noun and predicate comprehension/production and gestures in extremely preterm children at two years of age: are they delayed? J. Commun. Disord. 58 (2015) 126–142. [13] A. Schults, T. Tulviste, E. Haan, Early vocabulary in full term and preterm Estonian children, Early Hum. Dev. 89 (9) (2013) 721–726. [14] L. Fenson, P.S. Dale, J.S. Reznick, D. Thal, E. Bates, J.P. Hartung, et al., The MacArthur Communicative Development Inventories: User's guide and technical manual, Singular Publishing Group, San Diego, 1993. [15] S. Kern, F. Gayraud, Influence of preterm birth on early lexical and grammatical acquisition, First Lang. 27 (2) (2007) 159–173. [16] A. Sansavini, A. Guarini, S. Savini, S. Broccoli, L. Justice, R. Alessandroni, et al., Longitudinal trajectories of gestural and linguistic abilities in very preterm infants in the second year of life, Neuropsychologia 49 (13) (2011) 3677–3688. [17] L. Fenson, S.J. Pethick, C. Renda, J.L. Cox, P.S. Dale, J.S. Reznick, Short-form versions of the MacArthur communicative development inventories, Appl. Psycholinguist. 21 (01) (2000) 95–116. [18] S.W. Kim, H.R. Jeon, E.J. Park, H.I. Kim, W. Jung da, M.R. Woo, The usefulness of M-B CDI-K short form as screening test in children with language developmental delay, Ann. Rehabil. Med. 38 (3) (2014) 376–380. [19] W. Vach, D. Bleses, R. Jorgensen, Construction of a Danish CDI short form for language screening at the age of 36 months: methodological considerations and results, Clin. Linguist. Phon. 24 (8) (2010) 602–621. [20] M. Westerlund, E. Berglund, M. Eriksson, Can severely language delayed 3-year-olds be identified at 18 months? Evaluation of a screening version of the MacArthurbates communicative development inventories, J. Speech Lang. Hear. Res. 49 (2) (2006) 237–247. [21] D. Jackson-Maldonado, V.A. Marchman, L.C.H. Fernald, Short-form versions of the Spanish MacArthur–Bates communicative development inventories, Appl. Psycholinguist. 34 (04) (2013) 837–868. [22] L. Rescorla, The language development survey: a screening tool for delayed language in toddlers, J. Speech Hear. Disord. 54 (4) (1989) 587–599. [23] D. Drotar, T. Stancin, P.H. Dworkin, L. Sices, S. Wood, Selecting developmental surveillance and screening tools, Pediatrics in review / American Academy of Pediatrics. 29 (10) (2008) e52–e58. [24] L. Rescorla, T.M. Achenbach, Use of the language development survey (LDS) in a national probability sample of children 18 to 35 months old, J. Speech Lang. Hear. Res. 45 (4) (2002) 733–743. [25] R. Mossabeb, K.C. Wade, K. Finnegan, E. Sivieri, S. Abbasi, Language development survey provides a useful screening tool for language delay in preterm infants, Clin. Pediatr. 51 (7) (2012) 638–644.
16
C. Beaulieu-Poulin et al. / Early Human Development 98 (2016) 11–16
[26] R. Williams, J. Clinton, Canadian Paediatric society early years task force. Getting it right at 18 months: in support of an enhanced well-baby visit, Paediatr. Child Health 16 (10) (2011) 647–654. [27] L. Rescorla, A. Alley, Validation of the language development survey (LDS): a parent report tool for identifying language delay in toddlers, J. Speech Lang. Hear. Res. 44 (2) (2001) 434–445. [28] C.A. Albers, A.J. Grieve, Test review: Bayley, N. (2006). Bayley scales of infant and toddler development - third edition. San Antonio, TX: Harcourt assessment, J. Psychoeduc. Assess. 25 (2) (2007) 18. [29] F.P. Glascoe, Are overreferrals on developmental screening tests really a problem? Archives of pediatrics & adolescent medicine. 155 (1) (2001) 54–59. [30] L. Rescorla, M. Hadicke-Wiley, E. Escarce, Epidemiological investigation of expressive language delay at age two, First Lang. J. 13 (5) (1993).
[31] L. Rescorla, N.B. Ratner, P. Jusczyk, A.M. Jusczyk, Concurrent validity of the language development survey: associations with the MacArthur-Bates communicative development inventories: words and sentences, American journal of speech-language pathology / American Speech-Language-Hearing Association. 14 (2) (2005) 156–163. [32] J. Boyer, C. Flamant, G. Boussicault, I. Berlie, G. Gascoin, B. Branger, et al., Characterizing early detection of language difficulties in children born preterm, Early Hum. Dev. 90 (6) (2014) 281–286. [33] S. Blaggan, A. Guy, E.M. Boyle, E. Spata, B.N. Manktelow, D. Wolke, et al., A parent questionnaire for developmental screening in infants born late and moderately preterm, Pediatrics 134 (1) (2014) e55–e62. [34] A. Heiser, O. Curcin, C. Luhr, I. Grimmer, B. Metze, M. Obladen, Parental and professional agreement in developmental assessment of very-low-birthweight and term infants, Dev. Med. Child Neurol. 42 (1) (2000) 21–24.