Developmental language disorders (DLDs) affect about

Original Article Early Identification of Language Delay by Direct Language Assessment or Parent Report? Steffi Sachse, PhD, Waldemar Von Suchodoletz, MD ABSTRACT: Objective: The goal of the study was to compare the accuracy and diagnostic power of a parent report measure and direct language assessment for early identification of children with language delay. Method: The parent language report and direct language measures were compared for 47 typically developing toddlers and 70 late-talking toddlers aged 24 to 26 months. One year later, language abilities of 102 of the 117 children were reassessed. Results: The concurrent validity of the parent report was high both for judging language skills and for identifying language delay. No evidence was found of differences in the rating accuracy of mothers with different educational levels. Language abilities 1 year later were predicted better with direct language measurement than with the parent report. However, there were no differences between the accuracy of the parent report and individual language assessment concerning the prediction of language delay at age 3. Conclusion: The results suggest that the parent language report is a valid and efficient tool for assessing productive language abilities and judging expressive language delay in 2-year-old toddlers. The measurement characteristics of the parent report are comparable with those of direct language measures. (J Dev Behav Pediatr 29:34 41, 2008) Index Terms: late talkers, early identification, parent report, concurrent and predictive validity From the Department for Child and Adolescent Psychiatry, Psychosomatic and Psychotherapy, Ludwig-Maximilians-Universität, Waltherstr. 23, D-80337 München, Germany. Received February 2007; accepted June 2007. Address correspondence to: Waldemar v. Suchodoletz, MD, Department of Child and Adolescent Psychiatry, Psychosomatic and Psychotherapy, Ludwig- Maximilians-Universität, Waltherstr. 23, D-80337 München, Germany; E-mail: Suchodoletz@lrz.uni-muenchen.de Copyright 2008 Lippincott Williams & Wilkins Developmental language disorders (DLDs) affect about 5% to 10% of children (for a review, see Law et al 1 ). They are characterized by serious difficulties in language acquisition without an identifiable cause such as general mental retardation, physical disability, hearing impairment, or overall communicative disorder. Children with DLDs are assumed to be at high risk of later academic, behavioral, and social difficulties. 2 6 Early identification and early intervention may improve the outcome of affected children. 7 9 Language delay can usually be identified at the age of 24 months. 10 12 Children with a productive vocabulary of fewer than 50 words or no multiple-word utterances at that age are referred to as late talkers (LTs). 13 Previous studies indicate that 10% to 20% of toddlers meet this criterion. 10,12 Other criteria often used to define language delay are a productive vocabulary below the 5th or the 10th percentile. 7,14,15 Up to now, there have been no clinically relevant instruments for identifying language retardation earlier. Those instruments used so far to try to identify language retardation during the preverbal or very early verbal period have a very poor predictive validity. 16 18 During pediatric examinations, it is often difficult to judge language abilities via conversation in young children. In the unfamiliar clinical setting, toddlers tend to refuse contact or they produce too few utterances to obtain a large enough language sample. Therefore, other techniques for language assessment need to be employed. As an alternative to direct language tests, parent reports of language and communication skills are being used increasingly. Direct language assessment has the advantage that observable behavior is evaluated. However, toddlers often show low compliance in situations decontextualized from the child s everyday communicative activities. A period of adaptation to the laboratory setting is necessary for the child to overcome refusal of compliance with the examiner s requests. Furthermore, the attention span of toddlers is short and therefore an assessment must also be short and cannot use a large number of language items. Hence, a clinical language assessment is time-consuming and expensive. Parent reports are based on a representative sample of the child s utterances from a wide variety of naturalistic situations. They are quick, easy to use, and less costintensive than direct language assessments. Therefore, such language questionnaires are being used more and more frequently in clinical practice and research. Concerns have been raised, however, about the validity and reliability of parent questionnaires. 16,19,20 Parent reports reflect an adult s perception of the child s speech and are therefore considered to be not very objective. Furthermore, parents interest in children s activities and parents attention to the child s verbal productions differ, and there might be further distinctions related to a family s socioeconomic situation. Moreover, Pine et al 21 have reported differences between predominant word types in parent reports and spontaneous language samples. The parent questionnaires most commonly used for 34 Early Identification of Language Delay by Parent Report Journal of Developmental & Behavioral Pediatrics

language screening are the MacArthur Communicative Development Inventories (CDIs), 22 which enable the measurement of language skills in infants (8 16 months) and toddlers (16 30 months). The Infant Form evaluates expressive and receptive language abilities and communication behavior, but cannot predict future language impairment 16,17 (for the German adaptation, see Sachse et al 18 ). The Toddler Form assesses language production only. There is a strong correspondence between CDI, Toddler Form results and findings on other parent questionnaires, e.g., the Language Development Survey. 13,23 The CDIs have been translated into about 40 languages. A German version of the Toddler Form is the ELFRA-2 (parent questionnaire for 2-year-old children: language and communication), 25 which was used in the present study. The concurrent validity of the CDI was evaluated by comparison with direct language assessment measures in typically developing toddlers and toddlers at risk of language disorders (for a review, see Heilmann et al 15 ). A moderate to strong correlation between language test results and CDI parameters (productive vocabulary, three longest utterances, complexity) was found (correlations ranged from 0.60 to 0.80). Slightly weaker associations, ranging from 0.40 to 0.70, have been reported between CDI scores and spontaneous language sample parameters. 15 No differences in the strength of the correlations have been found between children with and without developmental impairments. 24 The concurrent validity of the Language Development Survey is reported to be similar to that of the CDI. 10,12 The correlative predictive validity of parent language reports has been evaluated for typically developing toddlers and LTs. Strong and significant correlations have been reported between the CDI vocabulary score at the age of 2 years (CDI, Toddler Form) and the productive (CDI-III: r Sp.58) 25 and receptive vocabulary (Peabody Picture Vocabulary Test: r Sp.50) at the age of 3 years. 19 In clinical practice, language judgment with parent reports is used predominantly for identifying children with language delay. However, few studies have dealt with the accuracy of classifying toddlers into typically developing and late-talking children. In a study of 30- month-old toddlers, Heilmann et al 15 calculated the likelihood ratio for CDI parameters to estimate the concurrent validity for sorting children into groups with and without language delay. With a cutoff at the 19th percentile, the specificity (0.81) and sensitivity (0.79) of the CDI vocabulary score were modest to good. Feldman et al 25 investigated the predictive validity for different cutoffs of productive vocabulary scores regarding the language status at age 3. They reported high specificity but only modest sensitivity. The usefulness of the CDIs for identification of language delay in individual children has been questioned, 16 however, because of the high variability in language skills during the first 2 years, low association between the results of the Infant and Toddler Forms of the CDIs, and significant effects of sociodemographic background data. Owing to the ongoing controversy about the measurement properties of parent language reports, the great clinical relevance of language disorders and the wide use of parent questionnaires for early identification of language delay, further studies of the accuracy of parent reports are necessary. The goal of the present study was to determine the diagnostic power of the parent report for early identification of language delay compared with that of a direct assessment of language abilities. So far the predictive validity of parent reports used to identify language delays in early childhood has not been directly compared with the predictive validity of a language test in the same sample of children. To this end, concurrent and predictive validity were estimated for concordance between language scores and for classification of toddlers into LTs and children with typical language development (TLDs). In addition, the relevance of the mother s educational level for the accuracy of parent rating was examined. METHODS Sample The sample was recruited via birth announcements in a newspaper that reports the birth of a child unless the parents submit a formal objection. The German version of the MacArthur Communicative Development Inventories (CDIs), Toddler Form (ELFRA-2) 26 was sent to 1490 parents 1 or 2 weeks before the child s second birthday. The return rate was 71% (n 1056). Only children from a monolingual German-speaking home were included in the study (n 932). The parents of all children who were classified as late talkers (LTs) based on the ELFRA-2 (n 154) were invited to have their child s language skills assessed with a standardized language test. Additionally, children with a vocabulary on the ELFRA-2 50 words were selected randomly for language assessment, but with a preference given to boys (n 109). The reason for selecting more boys than girls was that there were more boys in the LT group. Similar gender ratios have been reported in other studies. 15,27 Of the parents selected, 57% agreed to a detailed assessment of their child. Children with poor vision (n 1), a hearing impairment (n 2), or an abnormal result on a hearing screening test (transient evoked otoacoustic emissions, for details, see Measures section) (n 8) as well as children with missing subtests due to only brief cooperation (LTs: n 11; children with typical language development [TLDs]: n 9) were excluded from the study. The final sample consisted of 117 children, 70 LTs and 47 TLDs. The children were 24, 25, or 26 months old (n 3, 95, and 19, respectively). A second assessment, involving 102 of the original 117 children was done 1 year later. Of the 70 LTs, 11 (16%) and of the 47 TLDs, four (9%) could not be re-examined. The difference in the dropout rate between LTs and TLDs was not significant ( 2 1.306; p.253). Information about developmental milestones, medical history (complications during pregnancy or birth, prematurity, chronic disorders, history of otitis media or other ear disorders), and socioeconomic characteristics was obtained by having the parents complete a questionnaire. Vol. 29, No. 1, February 2008 2008 Lippincott Williams & Wilkins 35

Table 1 gives a summary of important characteristics of the sample. Except for the differences in language skills, there were no significant differences between the LT and TLD groups on any of the socioeconomic or developmental parameters (t-test or 2 test). According to the parent report and the clinical and psychological examinations, there was no evidence of autism or any general medical disorders. As a measure of nonverbal cognitive abilities, two subtests of the Münchener Funktionelle Entwicklungsdiagnostik 28 Table 1. Characteristics of the Sample According to Parent Report Total (n 117) TLDs (n 47) LTs (n 70) Variables n % n % n % Child s variables Gender Male 79 68 27 57 52 74 Female 38 33 20 43 18 26 Ear problem in past Otitis media 25 22 11 23 14 20 Ear operation 2 2 0 0 2 3 Complications during Pregnancy 31 27 9 19 22 31 Birth 45 39 21 46 24 35 Experience in a group of children Play group 70 64 29 64 41 63 Day care center 21 19 11 27 10 15 Family variables No. of siblings No siblings 45 39 22 47 23 33 1 52 44 21 45 31 44 1 20 17 4 9 16 23 Birth order 1st child 53 45 27 57 26 37 2nd child 49 42 18 38 31 44 3rd child 15 13 2 4 13 19 Parent marital status Married 105 90 41 87 64 91 Unmarried 12 10 6 13 6 9 Sociodemographic variables Mother s education 8 9 yr 15 13 5 11 10 14 10 yr 28 24 10 22 18 26 13 yr 73 63 31 67 42 60 Father s education 8 9 yr 17 15 5 11 12 17 10 yr 16 14 6 13 10 15 13 yr 82 71 35 76 47 68 Mother s employment Part time 36 31 15 33 21 30 Full time 3 3 1 2 2 3 Father s employment Part time 6 5 5 11 1 1 Full time 108 93 41 87 67 97 TLDs, children with typical language development; LTs, late talkers (no significant differences between TLDs and LTs, 2 test). Not all variables include 117 children because there are a few missing. Educational level: 8 years is the lowest level of a school leaving certificate; 10 years (mainstream school) secondary school level I certificate (O level); 13 years corresponds to a general qualification for university entrance. 36 Early Identification of Language Delay by Parent Report Journal of Developmental & Behavioral Pediatrics

were used. The mean developmental age of the LTs was about 1 month lower than that of the TLDs (perception: mean 25.0, SD 2.5 months vs mean 26.6, SD 2.0 months; motor skills of the hand: mean 25.4, SD 2.8 months vs mean 26.5, SD 2.6 months). One late-talking child had a subnormal score on the perception scale. All other children had scores in the normal range on both nonverbal subtests. When the children were 3 years old, the parents were asked whether there had been any speech-language intervention for their child in the past year. Except for two mothers of LTs who participated in a parent-focused language intervention program to learn language-facilitating behavior, all parents reported that there had not been any special interventions. Because of this marginal rate of intervention, the development in the sample can be regarded as a spontaneous course of language development. Each child was assigned an identification number and the parents were assured of the confidentiality of the results. All children s parents gave informed consent on both examination dates. All parents agreed at the first assessment to be contacted again when their child turned 3. The general design of the study was approved by the ethics commission of the medical faculty. Measures The parents rated the expressive language abilities of their child on the German version of the CDI, Toddler Form (ELFRA-2). The ELFRA-2 consists of three scales: Productive Vocabulary (260 items), Syntax (25 items), and Morphology (11 items). Productive vocabulary represents the total number of words that the parents indicated their child produces. Syntactic and morphologic abilities were scored by adding ratings from 0 to 2 for word combinations of differing complexity. According to the test instructions, LTs are defined as children with a productive vocabulary 50 words or with a productive vocabulary of 50 to 80 words and grammatical scores below the cutoff points (syntax score 7 and morphology score 2). In the ELFRA-2 norming sample, 14% of the toddlers met one of these criteria. This is similar to the results of Horwitz et al 29 in a community sample where 13% of 18- to 23-month-old toddlers and 15% of those aged 24 to 29 months were found to have a language delay. The children were tested in a quiet room in the outpatient department of our hospital during two 1-hour sessions. The sessions included a standardized language test (SETK-2, details later), nonverbal subtests of a developmental scale (Münchener Funktionelle Entwicklungsdiagnostik), and a hearing screening (measurement of transient evoked otoacoustic emissions [TEOAEs] by ECHO- SCREEN Plus T, Mack Medizintechnik, Pfaffenhofen, Germany). The examiner was unaware of the children s language scores on the parent report. While the child was being evaluated, the mother sat in the same room and completed questionnaires. The direct language assessment was performed with the SETK-2 (language test for 2-year-old children 30 ). This instrument consists of four subtests (Word Production, Sentence Production, Word Comprehension, Sentence Comprehension). The SETK-2 is standardized for two groups of children, aged 24 to 29 months and 30 to 36 months. The language assessment was audiotaped for later scoring. In accordance with the test instructions, a language delay was assumed if a child scored below normal on one of the four subtests. A comparison with the ELFRA-2 judgment was achieved by choosing a cutoff of 1 SD below the mean for classification of a child as having below-normal language skills. One year later, the language skills of 102 of the 117 children were again assessed with a standardized language test (SETK-3/5) (mean age, 37.0 months; SD 0.7 months) in an identical setting. The SETK-3/5 (Language test for 3- to 5-year-old children) 30 consists of four subtests (Picture Description, Marking of Plurals, Sentence Comprehension, Repetition of Pseudo-Words). As a judgment of the responses on the subtest Repetition of Pseudo-Words was impossible for children with phonological impairments, and such impairments were common in the children classified earlier as LTs, this subtest was excluded from the analysis. If a child scored 1 SD below the mean in one of the three subtests included, a language delay was assumed. RESULTS Descriptive Statistics Descriptive statistics for the language abilities of the two groups of children are summarized in Table 2. There were highly significant differences between the two groups not only in the parent report but also in the direct language measures. In contrast to the ELFRA-2, the SETK-2 judges not only productive but also receptive language skills. Of the 70 LTs, 47 met the criteria for a pure expressive language delay (SETK-2: T value 40 on one of the production subtests, T value 40 on the comprehension subtests) and 23 for a mixed receptive-expressive language delay (SETK-2: T value 40 on both one of the production subtests and one of the comprehension subtests). None of the children had a receptive language delay without an expressive delay. Concurrent Validity The correlation between the parent report and the direct language assessment was computed for each scale of the ELFRA-2 and the SETK-2 (Table 3). As the scales of the ELFRA-2 estimate several aspects of language production without considering comprehension skills, strong correlations with SETK-2 production scales and weaker correlations with SETK-2 comprehension scales were anticipated. As expected, the strongest correlation was found for the vocabulary scale of the parent report and the word production scale of the language test. The correlations for the grammar scales of the ELFRA-2 and the production scales of the SETK-2 were modest to high. Finally, the correlations for the ELFRA-2 scales and the comprehension subscales of the SETK-2, although only modest, were also significant. Vol. 29, No. 1, February 2008 2008 Lippincott Williams & Wilkins 37

Table 2. Descriptive Statistics for the Results on the Language Test (SETK-2) and the Parent Report (ELFRA-2) Group TLDs (n 47) LTs (n 70) Significance Level TLD/LT Mean SD Mean SD t p SETK-2 scales Production Words 57.5 9.9 36.5 7.0 12.7.001 Sentences 50.0 6.7 35.5 4.6 13.0.001 Comprehension Words 56.4 6.9 48.3 8.7 5.6.001 Sentences 55.5 7.2 43.6 10.9 7.2.001 ELFRA-2 scales Vocabulary 148.8 35.2 33.8 17.9 20.7.001 Syntax 20.3 8.9 2.6 4.3 12.7.001 Morphology 6.9 3.7 0.4 1.0 11.8.001 TLDs, children with typical language development; LTs, late talkers; SETK-2, T-values; ELFRA-2, raw scores; two-tailed t-test. Table 3. Correlation Between the Language Test (SETK-2) and the Parent Report (ELFRA-2) Production SETK-2 Scales Comprehension ELFRA-2 scales Words Sentences Words Sentences Vocabulary 0.87 0.83 0.40 0.44 Syntax 0.67 0.75 0.45 0.48 Morphology 0.70 0.76 0.45 0.53 n 117; Spearman rank correlation coefficients; significance level of all correlations: p.001. The question of whether the accuracy of the parent report depends on the mother s educational level was addressed by calculating the language measures separately for three educational levels. The scores on the Vocabulary subscale of the parent report and on the Word Production subscale of the language test tended to be lower in toddlers with less educated mothers, but the differences were not significant. The Spearman rank correlation coefficients for the two productive vocabulary scores were similar in all groups (mother s education 8 9 years: r Sp.83; 10 years: r Sp.92; 13 years: r Sp.85). No marked association between educational level and accuracy of judgment was found. In clinical practice, it is less important to determine the exact vocabulary of children than to judge whether a language delay exists. The classification of the toddlers as LTs and TLDs was the same for 106 (91%) of the 117 children and different for only 11 (9%). Five children were classified as LTs with the language test but not with the parent report and six with the parent report but not with the language test. With the language test results (SETK-2) serving as the gold standard, the ELFRA-2 measurement properties (sensitivity, specificity, positive predictive value, and negative predictive value) are excellent (Table 4). Predictive Validity Predictive validity is an especially important type of validity for an instrument used for early identification. A parent report can be recommended for clinical practice only if it predicts later language skills and enables early detection of language-impaired children that is comparable with detection by the direct measurement of language by a professional examiner. Predictive validity was evaluated by reassessing the language skills of 102 of the original 117 children when the children were 3 years old. The correlations between language measures at ages 2 and 3 are modest, and somewhat weaker for parent report scores (r Sp from.34 to.52) than for direct language measures (r Sp from.47 to.62), but all correlations are highly significant. At age 3 years, 67 of the 102 children had normal results and 35 scored 1 SD or more below the mean on at least one of the SETK-3/5 subtests. Except for two children, all children with below-normal language abilities at the age of 3 had been labeled as LTs on both the language test and the parent report 1 year earlier. However, not all Table 4. Validity Scores for the Parent Report (ELFRA-2) and the Language Test (SETK-2) at Age 2 Sensitivity Specificity PPV NPV Gold standard: language test classification at age 2 (SETK-2 result) ELFRA-2 0.93 0.88 0.91 0.89 Gold standard: language test classification at age 3 (SETK-3/5 result) ELFRA-2 0.61 0.94 0.56 0.95 SETK-2 0.64 0.94 0.58 0.96 PPV, positive predictive value; NPV, negative predictive value. 38 Early Identification of Language Delay by Parent Report Journal of Developmental & Behavioral Pediatrics

toddlers who had been classified as LTs at age 2 showed a language delay 1 year later: 24 (42%) of the children classified as LTs on the SETK-2 and 26 (44%) classified as LTs on the ELFRA-2 had results on the language test at age 3 that were in the normal range (Table 5). Therefore, the specificity and the positive predictive value were in a moderate range and were lower than the sensitivity and the negative predictive value. As Table 4 shows, the measurement properties for the direct language assessment with the SETK-2 and for the parent language report with the ELFRA-2 are very similar. DISCUSSION In pediatric outpatient departments, the possibilities for direct language assessment to enable early identification of language delay are limited. Language tests are time-consuming, and administering them requires specialized linguistic knowledge. Parent questionnaires appear to be a reasonable alternative. The use of such questionnaires is easy to implement, and the parent report appears to be an especially useful tool for language evaluation in young children. Parent reports are increasingly used in practice for identification of several types of developmental impairments. However, the accuracy of parent language reports has been questioned. 16,20 Nevertheless, across several studies, the correlation between parent report and language sample measures has been found to be moderate to strong. 15 However, correlation analysis does not provide insight into the test s ability to identify language delay in individual children. The most important measurement property in the clinical context, the accuracy of a judgment about language delay, has been evaluated only rarely. In particular, there is a lack of information about the diagnostic power of the parent language report in direct comparison with that of individual language assessment and also about the validity of language ratings by parents with different educational levels. The present study focuses on these issues. Language delay can be identified at the age of 2 years. In younger children, the validity and reliability of language assessment are limited. In several studies, the low Table 5. Predictive Validity of Parent Report Measures (ELFRA-2) and Language Test Scores (SETK-2) for Classification into Language Ability Groups ELFRA-2 Language Skills at Age 3 Normal Below normal Total TLDs 41 (95%) 2 (5%) 43 (100%) LTs 26 (44%) 33 (56%) 59 (100%) 67 (66%) 35 (33%) 102 (100%) SETK-2 TLDs 43 (96%) 2 (4%) 45 (100%) LTs 24 (42%) 33 (58%) 57 (100%) Total 67 (66%) 35 (35%) 102 (100%) n 102, gold standard: language skills at age 3. Bold type represents concordant classification. predictive validity of the Infant Form of the MacArthur Communicative Development Inventories (CDIs) has been established 16,17 (for the German adaptation, see Sachse et al 18 ). Therefore, at present, language screening for early identification of language impairment cannot be recommended for children younger than 18 months. Our results provide evidence that the parent language report is an effective tool for assessment of language skills at the age of 24 months. Even though the basis of information about the child s language skills in a parent report is different from that in direct language assessment measures, the two approaches yield very similar results in toddlers. Our findings using standardized language tests as the gold standard are consistent with those of previous studies in typically developing children and children with Down syndrome that assessed the concurrent validity of the CDI by comparison with spontaneous language sample parameters and a language score calculated from language items of the Bayley Scales of Infant Development. 24,31 The strongest correlation between parent report and direct language measures was found for language production scores, in particular productive vocabulary. In contrast, the correlation between parent report scores and the comprehensive scales of the language test was relatively weak. This may be the result of differences in the development of expressive and receptive language skills in individual children and the relatively poor reliability of language comprehension measures. 16,32,33 Heilmann et al 15 reported that the CDI was very accurate in classification of typically developing and late-talking toddlers into language ability groups at the age of 30 months. Our findings demonstrate that a parent report yields comparable results in children aged 24 months. If parent questionnaires are to be used in pediatric practice, their applicability in families from different socioeconomic backgrounds and with different educational levels must be ensured. Pan et al 19 reported lower validity scores in their sample of low-income families than other authors found for middle-class monolingual families. This raises the question of whether the accuracy of the parent report depends on socioeconomic variables. Moreover, across several studies, significant correlations between vocabulary scores and socioeconomic status have been found. 34 36 However, the correlations have generally been weak. For example, when Rescorla and Achenbach 11 collected normative data for the Language Development Survey in children with highly diverse socioeconomic status the correlation was 0.14. Furthermore, Feldman et al 25 reported a significant dependence of the CDI vocabulary score of 24-month-old toddlers on the mother s educational level. Because of a lack of studies about the reliability of parent reports by parents of different socioeconomic or educational levels, several authors have expressed concerns about the applicability of parent language reports in families with different backgrounds. 16,20 In line with the cited studies, we found a tendency to a smaller productive vocabulary on the parent report for children whose mothers had less education. Additionally, Vol. 29, No. 1, February 2008 2008 Lippincott Williams & Wilkins 39

we found similar differences on direct language measures. However, the correlations between parent report and language test scores, i.e., the concurrent validity, were similar in groups of children from different educational backgrounds. These findings provide evidence that the differences in the productive vocabulary of toddlers on parent reports from parents with different educational levels is not caused by differences in the mothers language-rating ability but rather by differences in the children s language skills. Hence, there are no indications that the validity of the language report is poorer if the parents have less education. In the present study, we looked at two types of predictive validity: prediction of language skills and prediction of language delay at age 3. The correspondence between the measures of age-appropriate language abilities at ages 2 and 3 was moderate overall. In previous studies, similar correlations have been reported regarding predictive properties of parent reports. 15,25 Until now the predictive validity of the parent report has not been compared with the predictive diagnostic power of direct language measurement. Our results provide evidence that for prediction of language skills, the predictive validity of the language test is higher than that of the parent report. The correlations between the language test scores at ages 2 and 3 were somewhat higher than the correlations between parent report ratings at age 2 and language measures at age 3. As the most important part of our assessment of predictive validity, we looked at prediction of language delay at age 3. The parent report had an excellent specificity (0.94) and negative predictive value (0.95), and a moderate sensitivity (0.61) and positive predictive value (0.56). In light of the numerous influences on child development during the 1-year interval from age 2 to age 3, the predictive power of 0.61 can be considered to be quite high. The findings are similar to those of Feldman et al 25 in a sample of children with persistent middle-ear effusion with a threshold of the 15th percentile on the vocabulary scale of the CDI at age 2 and 1 SD below the mean on the vocabulary scale of the CDI-III at age 3 as the gold standard (sensitivity: 0.57; positive predictive value: 0.57; specificity: 0.85; negative predictive value: 0.85). In clinical practice, the instruments with the best diagnostic accuracy should be used. However, until now there has been a lack of information about the accuracy of parent report in comparison with direct language measures. In the present study, the prognostic validity of parent report and individual language measures was assessed in the same sample. Table 4 shows that there are no marked differences between the two instruments regarding the accuracy of prediction of language delay. Consequently, it appears that the diagnostic power of parent report is as high as that of language tests and the relatively low sensitivity is not an indication of inadequate parent reports, but rather an expression of low stability of language delay during the first few years of life. About 50% of children who were late talkers (LTs) at the age of 24 months have language abilities in the normal range 1 year later. 37 39 Such children are referred to as late bloomers. These results suggest that the accuracy of parent rating in toddlers is comparable with that of direct language assessment regarding identification of developmental language delay. In summary, our findings show that parents are able to judge their toddler s expressive language development with reasonable accuracy and that the parent language report is a valid tool for assessing the productive language abilities of children at the age of 24 months. Parent report and direct language measures are highly correlated regarding both productive language abilities and identification of children with a language delay. In mothers who have had at least 8 years of schooling, the accuracy of the parent report seems to be independent of the mother s education. The predictive validity of the parent report at age 2, as measured by rank correlation coefficients in relation to individual language measures 1 year later, is slightly lower than that of direct language assessment. However, the diagnostic power of the two assessment methods for prediction of a language delay at age 3 appears to be similar. The present study extends previous findings by directly comparing the validity of parent report and individual language measures in the same children. Furthermore, in contrast to other studies, 19,25 standardized language tests were used as the gold standard at both age 2 and age 3, and not only language production skills but also language comprehension skills were assessed. Moreover, our sample was selected from a general population, whereas in other studies that addressed a similar issue specific types of children were selected, such as children from low-income families, including bilingual children, 19 and those with persistent middle-ear effusion. 25 Finally, our findings demonstrate that the validity of parent report is comparable in mothers with different educational levels. General use of the parent language report in health care centers could contribute to early identification of children at risk of developmental language disorders. Children with a language delay at age 3 can already be identified by parent report at age 2. In 2-year-old children with parent report values in the normal range, unremarkable language development at 1-year follow-up is very likely. However, a prediction about the language development of 2-year-old late-talking toddlers is uncertain regardless of the assessment method. About half of the LTs catch up with typically developing peers within 1 year. REFERENCES 1. Law J, Boyle J, Harris F, et al. Prevalence and natural history of primary speech and language delay: findings from a systematic review of the literature. Int J Lang Commun Disord. 2000;35: 165 188. 2. Clegg J, Hollis C, Mawhood L, et al. Developmental language disorders a follow-up in later adult life. Cognitive, language and psychosocial outcomes. J Child Psychol Psychiatry. 2005;46:128 149. 3. Cohen NJ. Developmental language disorders. In: Howlin, P 40 Early Identification of Language Delay by Parent Report Journal of Developmental & Behavioral Pediatrics

Udwin O, eds. Outcomes in Neurodevelopmental and Genetic Disorders. Cambridge: Cambridge University Press; 2002:25 55. 4. Lyytinen P, Eklund K, Lyytinen H. Language development and literacy skills in late-talking toddlers with and without familial risk for dyslexia. Ann Dyslexia. 2005;55:166 192. 5. Rescorla L. Age 13 language and reading outcomes in late-talking toddlers. J Speech Lang Hear Res. 2005;48:459 472. 6. Snowling MJ, Adams JW, Bishop DVM, et al. Educational attainments of school leavers with a preschool history of speechlanguage impairments. Int J Lang Commun Disord. 2001;36:173 183. 7. Girolametto L, Pearce PS, Weitzman E. Interactive focused stimulation for toddlers with expressive vocabulary delays. J Speech Hear Res. 1996;39:1274 1283. 8. Robertson SB, Ellis Weismer S. Effects of treatment on linguistic and social skills in toddlers with delayed language development. J Speech Lang Hear Res. 1999;42:1234 1248. 9. Whitehurst GJ, Fischel JE, Lonigan CJ, et al. Treatment of early expressive language delay: If, when, and how. Top Lang Disord. 1991;11:55 68. 10. Klee T, Carson DK, Gavin WJ, et al. Concurrent and predictive validity of an early language screening program. J Speech Lang Hear Res. 1998;41:627 641. 11. Rescorla L, Achenbach TM. Use of the Language Development Survey (LDS) in a national probability sample of children 18 to 35 months old. J Speech Lang Hear Res. 2002;45:733 743. 12. Rescorla L, Alley A. Validation of the Language Development Survey (LDS): a parent report tool for identifying language delay in toddlers. J Speech Lang Hear Res. 2001;44:434 445. 13. Rescorla L, Ratner NB, Jusczyk P, et al. Concurrent validity of the Language Development Survey: associations with the MacArthur- Bates Communicative Development Inventories: Words and Sentences. Am J Speech Lang Pathol. 2005;14:156 163. 14. Girolametto L, Wiigs M, Smyth R, et al. Children with a history of expressive vocabulary delay: outcomes at 5 years of age. Am J Speech Lang Pathol. 2001;10:358 369. 15. Heilmann J, Weismer SE, Evans J, et al. Utility of the MacArthur- Bates Communicative Development Inventory in identifying language abilities of late-talking and typically developing toddlers. Am J Speech Lang Pathol. 2005;14:40 51. 16. Feldman HM, Dollaghan CA, Campbell TF, et al. Measurement properties of the MacArthur Communicative Development Inventories at ages one and two years. Child Dev. 2000;71:310 322. 17. Fenson L, Bates E, Dale P, et al. Measuring variability in early child language: don t shoot the messenger. Child Dev. 2000;71: 323 328. 18. Sachse S, Saracino M, von Suchodoletz W. Prognostische Validität des ELFRA-1 bei der Früherkennung von Sprachentwicklungsstörungen. Klin Pädiatr. 2007;219:17 22. 19. Pan BA, Rowe ML, Spier E, et al. Measuring productive vocabulary of toddlers in low-income families: concurrent and predictive validity of three sources of data. J Child Lang. 2004; 31:587 608. 20. Roberts JE, Burchinal M, Durham M. Parents report of vocabulary and grammatical development of African American preschoolers: child and environmental associations. Child Dev. 1999;70:92 106. 21. Pine JM, Lieve EV, Rowland C. Observational and checklist measures of vocabulary composition: what do they mean? J Child Lang. 1996;23:573 589. 22. Fenson L, Dale PS, Reznick JS, et al. MacArthur Communicative Development Inventories: User s Guide and Technical Manual. San Diego, CA: Singular Publishing Group; 1993. 23. Rescorla L. The Language Development Survey: a screening tool for delayed language in toddlers. J Speech Hear Disord. 1989;54: 587 599. 24. Miller JF, Sedey AL, Miolo G. Validity of parent report measures of vocabulary development for children with Down syndrome. J Speech Hear Res 1995;38:1037 1044. 25. Feldman HM, Dale PS, Campbell TF, et al. Concurrent and predictive validity of parent reports of child language at ages 2 and 3 years. Child Dev 2005;76:856 868. 26. Grimm H, Doil H. Elternfragebögen für die Früherkennung von Risikokindern. Göttingen: Hogrefe 2000. 27. Ellis Weismer S, Murray-Branch J, Miller JF. A prospective longitudinal study of language development in late talkers. J Speech Hear Res 1994;37:852 867. 28. Hellbrügge T, Lajosi F, Menara D, et al. Münchener Funktionelle Entwicklungsdiagnostik. Lübeck: Hansisches Verlagskontor; 1985. 29. Horwitz SM, Irwin JR, Briggs-Gowan M, et al. Language delay in a community cohort of young children. J Am Acad Child Adolesc Psychiatry. 2003;43:932 940. 30. Grimm H. Sprachentwicklungstest für sweijährige slindes-setk-2 und für dreibis fünfjährige Kinder-SETK 3-5. Göttingen: Hogrefe; 2000. 31. Dale PS. The validity of a parent report measure of vocabulary and syntax at 24 months. J Speech Hear Res. 1991;34:565 571. 32. Ireton H, Glascoe FP. Assessing children s development using parents reports. The Child Development Inventory. Clin Pediatr. 1995;34:248 255. 33. Westerlund M, Sundelin C. Can severe language disability be identified in three-year-olds? Evaluation of routine screening procedure. Acta Paediatr. 2000;89:94 100. 34. Dollaghan CA, Campbell TF, Paradise JL, et al. Maternal education and measures of early speech and language. J Speech Lang Hear Res. 1999;42:1432 1443. 35. Hoff E. The specificity of environmental influence: socioeconomic status affects early vocabulary development via maternal speech. Child Dev. 2003;74:1368 1378. 36. Landry SH, Smith KE, Swank PR. Environmental effects on language development in normal and high-risk child populations. Semin Pediatr Neurol. 2002;9:192 200. 37. Paul R. Predicting outcomes of early expressive language delay: ethical implications. In: Bishop DVM, Leonard LB, eds. Speech and Language Impairments in Children: Causes, Characteristics, Intervention and Outcome. Hove, East Sussex: Psychology Press Ltd.; 2001:195 209. 38. Rescorla L, Bascome A, Lampard J, et al. Conversational patterns in late talkers at age 3. Appl Psycholinguistics. 2001;22:235 251. 39. Thal D, Tobias S, Morrison D. Language and gesture in late talkers: a 1-year follow-up. J Speech Hear Res. 1991;34:604 612. Vol. 29, No. 1, February 2008 2008 Lippincott Williams & Wilkins 41