| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Contact author: Gina Conti-Ramsden, Human Communication and Deafness, School of Psychological Sciences, The University of Manchester, Humanities Devas Street Building, Oxford Road, Manchester, M13 9PL, United Kingdom. Email: gina.conti-ramsden{at}manchester.ac.uk
 |
Abstract |
|---|
 Top Abstract DiscussionConcluding Remarks and Clinical...References |
|---|
METHOD: Using these 2 methods, the prevalence of language and literacy disorders was investigated in the immediate family (n = 271) of 93 children with a history of SLI (mean age 13;11 years).
RESULTS: The overall proportion of relatives with reported language or literacy difficulty was similar for both methods (34.5% for reported difficulties compared with 35% on direct assessment). The present study further explored maternal, paternal, and sibling prevalence rates and strength of agreement between parental interview and direct assessment. When a low cutoff score was used, good agreement (of true negatives and true positives) for reading and spelling difficulties and expressive language between the 2 types of case identification method was found.
CONCLUSIONS: Parents can be effective identifiers when the impairment is severe (below 2 SDs from the population mean). Poor agreement was observed between report and assessment of receptive language difficulties.
KEY WORDS: specific language impairment (SLI), prevalence rates, direct assessment, parental interview
Substantial evidence suggests that specific language impairment (SLI) aggregates in biological families (e.g., Choudhury & Benasich, 2003; Flax et al., 2003; Lahey & Edwards, 1995; Spitz, Tallal, Flax, & Benasich, 1997; Tallal et al., 2001; Tomblin, 1989). However, a study by Whitehurst et al. (1991) indicates no strong familial component of developmental expressive language delay. Although the aggregation effect is robust (with the exception of the study above), there is sizable variation in prevalence estimates. Stromswold (1998), in a review of 18 family aggregation studies of language impairment, found that the rates of family members who were impaired ranged from 20% (Neils & Aram, 1986) to 42% (Tallal, Ross, & Curtiss, 1989). It is of both theoretical and clinical importance to accurately estimate patterns of familial manifestation of language impairment. This potentially impacts on accuracy of early detection of affected individuals, provision of specific remediation, and more general policy formation.
Methodological Differences in Estimating Prevalence
Different methods of ascertaining prevalence have been used by studies in this area. The majority of investigations have used parental report of history of speech, language, or learning problems. There are various report methodologies used to ascertain prevalence. These can be divided into two main types: questionnaire and interview. Questionnaire methodologies typically require parents to independently complete checklists for the presence or absence of problems amongst relatives. For example, Neils and Aram (1986) gave a questionnaire format to the parents of affected children with a checklist of speech pronunciation, stuttering, reading, and language items that a family member either had or continued to present. Lahey and Edwards (1995) used a questionnaire format that was either filled out independently by a family member or read out by a researcher over the phone. In the analysis of their data, no distinction was made between speech, language, or learning problems, as the authors were not clear that parental report could give valid distinctions. Van der Lely and Stollwerck (1996) used case history notes taken on probands during school entry and a follow-up parental questionnaire with yes/no response for speech-language problems, provision of speech-language therapy, and reading/writing problems. In the studies of Plante, Shenkman, and Clarke (1996) and Tomblin and Buckwalter (1994), family case history reports/questionnaire data were used to identify individuals with a history of speech-language therapy, poor reading, and school grade failure (being kept behind a year in school). Tallal et al. (2001) used a family history questionnaire with yes/no questions followed by open-ended questions. This was completed in the presence of a member of the research team, giving the opportunity to ask for clarification or help in reading or interpreting a question. Whitehurst et al. (1991) used a posted questionnaire for parents to indicate whether, for immediate and extended family members, there was or was not a history of late talking, speech problems, or school problems.
A smaller number of studies have used interview techniques where parents are interviewed in person concerning problems within their families. For example, Beitchman, Hood, and Inglis (1992) used a semistructured, in-depth diagnostic interview with parents to ascertain the presence of impairment among relatives of the proband. Specific problems were categorized into either speech-language difficulties (stuttering, articulation problems, and speech-language delay) or reading difficulties (reading delay, learning disability, and dyslexia). Rice, Haney, and Wexler (1998) used a combination of personal interviews and telephone interviews to ask respondents about the status of each family member. Similarly to the methodology of the present study, they provided examples of surface symptoms for each category of impairment. Informants considered the symptoms when determining whether individuals should be classified as speech-language impaired.
In contrast to the parental report methodologies described above, direct testing has also been used to estimate family aggregation rates in the study of language impairments. Several studies have used current test data from direct psychometric assessment to arrive at prevalence rates (Felsenfeld, McGue, & Broen, 1995; Flax et al., 2003; Gopnik & Crago, 1991; Plante et al., 1996; Tallal et al., 2001; Tomblin, 1996; Tomblin & Buckwalter, 1994). Areas examined have usually included measures of expressive and receptive language. In some studies, examination of reading difficulties has also been reported (e.g. Flax et al., 2003).
Combining Report and Assessment Methodologies Concurrently
A few studies have compared prevalence rates obtained from the same sample using both methods described above. Plante et al. (1996) compared impairment rates between the parents of children with SLI and control parents using case history methods and language test methods. It was found that classification by case history methods resulted in fewer affected adults than classification through standardized testing. A case control study determining prevalence through both parental report and direct testing was conducted by Tallal et al. (2001). Rates of impairment as determined in current testing were compared directly with impairment rates estimated from family history questionnaires collected from the same families. Group data showed impairment rates estimated from the family history questionnaires to be similar to the rates based on actual testing. However, case-by-case analysis showed poor intrasubject agreement on classification as language impaired on the basis of current testing as compared with history information.
Aims of the Present Study
The aim of the present study was to compare direct testing versus parental interview methodologies in examining prevalence rates of impairments in a large sample of the immediate family of a group of adolescents with a history of SLI. Rates of maternal, paternal, and sibling impairment were investigated in relation to the type of estimation method (interview vs. testing). Relative rates were also examined in relation to types of impairment such as receptive and expressive language, reading, spelling, and provision of speech therapy or special education.
Method
Participants
The participants in this study were the relatives of children originally part of a wider study; the Conti-Ramsden Manchester Language Study (Conti-Ramsden & Botting 1999a, 1999b; Conti-Ramsden, Crutchley, & Botting, 1997). This cohort was recruited from 118 language units attached to English mainstream schools. All language units for Year 2 children were contacted, and any centers enrolling children with global delay or hearing impairments were excluded. The remaining language units provided a list of Year 2 children attending for at least 50% of the week. Across England, approximately 500 children fitted this criterion. All language units were asked to participate, and two declined this invitation. Subsequently, approximately half of the eligible children in each unit were randomly sampled. This resulted in an initial study cohort of 242 children. The age range was 7;5 [years;months] to 8;9 and consisted of 186 boys and 56 girls (girls formed 23.1% of the cohort). These children were reassessed as part of the original study at 8 years and 11 years of age.
From the original cohort of 242 children described above, 4 families (2%) in which the proband was adopted were not contacted, no response could be obtained from 59 (24%) families, and 55 (23%) families refused consent. Of the 124 (51%) families who agreed to take part in the present phase of the study, 118 (95%) were assessed, and 6 (5%) were not assessed due to alterations in family circumstances.
From this pool of 118 consenting, assessed families, 93 were selected for participation in the present aggregation study based on longitudinal data using the following proband (i.e., child in the original study) criteria:
In the present study, the probands had a mean age of 13;11. Probands were classed as currently impaired if, at the time of the study, they met the following criteria for SLI: Performance IQ (Wechsler Intelligence Scale for Children—Third Edition—Revised, UK; WISC–III; Wechsler, 1992) of 80 or more and concurrent Expressive or Receptive Language standard score (Clinical Evaluation of Language Fundamentals—Revised UK; CELF–R ELS/RLS; Semel, Wiig, & Secord, 1987) less than 85. Note that these concurrent criteria are identical to the criteria used for the selection of study participants using the longitudinal data, except that to be concurrently impaired these criteria have to apply at age 14 years (the time of the study).
These children are referred to as "current SLI status" in Table 1. As can be seen from this table, just over half of the probands (51.7%) were classified as currently meeting criteria for SLI at the time of the study. The remainder (48.3%) all met the established SLI criteria at some point in the last 7 years. These remaining children were divided into two groups: those with both impaired language and nonverbal skills at the time of the study (n = 35, 39.3% of the total) and those with current normal language and nonverbal ability (n = 8, 9% of the total). As would be expected, there were no cases of probands with current impaired nonverbal abilities but normal language scores. Therefore, at the time of the study, a total of 81 probands (i.e., the current SLI status group plus the current impaired language and nonverbal skills group, 91% of tested probands) had current language difficulties indicated by scores at least 1 SD below the mean on standardized tests of expressive and receptive language.
|
The immediate family of the probands participated in the present study. Minimum age for participation in the study was 6 years. Participation was defined as being directly assessed, being reported on, or both. In terms of the total number of family members participating in the present study analyses, there were 271 immediate family relatives (constituting 73 fathers, 74 mothers, 37 male/28 female siblings over the age of 16 years, and 27 male/32 female siblings between the ages of 6 and 16 years). Of the 93 families participating in the study, 22 had the proband only, 30 had the proband and one sibling, 25 had the proband and two siblings, 14 had the proband and three siblings, 1 had the proband and four siblings, and 1 had the proband and eight siblings. The majority of siblings were full siblings of the proband. There were 5 half-siblings between the ages of 6 and 16 years and 15 half-siblings older than 16 years.
Information was gathered on parental level of education and occupation to provide further description of the sample used in the study. It was found that 28.5% of parents had no educational qualifications, 29.2% had achieved GCSE/O levels (comparable to an American high school graduation diploma), 25.5% achieved A level qualifications (comparable to the American university freshman year), 11.7% were university graduates, and 5.1% had achieved postgraduate qualifications.
Occupation was coded as per the Standard Occupational Classification manual (Office of Population Censuses and Surveys, 1991) and referred to the person's full-time or part-time job/profession at the highest level usually held (i.e., not in a temporary capacity) during their working life. It was found that 5.1% of parents were classed as professional, 19% were managerial/technical, 16.8% were non-manual skilled, 21.2% were manual skilled, 7.3% were partly skilled, and 4.4% were unskilled. It was found that 1.6% were chronically unemployed, and 25% were homemakers who had not worked for the majority of their adult life.
Measures for Immediate Family
Direct Assessment Battery
Language measures. Adult relatives (parents and siblings 16 years of age or over) were administered the Verbal short form of the Wechsler Adult Intelligence Scale–Revised (WAIS-R VIQ; Wechsler, 1986). Child relatives (siblings between 6 and 16 years of age) were administered the full WISC–III VIQ (Wechsler, 1992). Child relatives were also assessed using the CELF–R (Semel et al., 1987). This yielded the RLS and ELS scores.
It is important to comment on the rationale for the instrumentation used to identify language impairments. In the case of the children's measures, the use of the CELF–R provides a clear advantage to the study. This instrument has been shown to have high sensitivity (88%) and specificity (88%) at a cutoff score of below 1 SD when using the RLS and ELS (Semel et al., 1987). Therefore, these data provide a benchmark against which the parental report data can be evaluated.
Since standardized psychometric IQ test scores may be available for children who lack formal language test scores, we have included VIQ as a measure in the assessment of the SLI phenotype. In the case of the children's data (i.e., siblings between 6 and 16 years), the VIQ measure is moderately correlated with CELF–R total score (r = .80), ELS (r = .74), and RLS (r = .80); which is about as well as many language scores correlate with each other. An examination of these data suggests that VIQ in children quite accurately reflects the prevalence of receptive language impairments in siblings but underestimates the prevalence of expressive language impairments in comparison with the more specific test of language, the CELF–R. Further examination of the data from the children who had a validated CELF–R identifier (i.e., they performed below 1 SD on the CELF–R) suggests that overall VIQ underestimates the level of the language of the child by about 10 points (mean VIQ for affected children = 83.51, SD = 19.79; mean CELF–R total language score for affected children = 73.78, SD = 15.30).
Among adults, formal language test scores are rare, and currently no standardized tests are available for the United Kingdom. Can VIQ be used in adults as a measure of language skill and therefore an indication of possible impairment? We would like to argue that with the appropriate caution, the use of VIQ with adults can provide preliminary indications of impairment. Thus, the option to use VIQ in this study was based on both practical considerations and the supportive available correlational data on children up to 16 years presented above. Having said this, it is important to emphasize that VIQ is a global omnibus test that was not designed to identify residual language deficits. Thus, caution is required in the use and interpretation of findings from VIQ.
Literacy measures. Adults were administered the short form of the Wechsler Objective Reading Dimensions (WORD; Rust, Golombok, & Trickey, 1993). This constituted Basic Reading and Spelling subtests. Child relatives were also administered the WORD Basic Reading and Spelling subtests and additionally the WORD Reading Comprehension subtest.
General cognitive measures. Adult relatives were administered the Performance short form of the WAIS–R. This gave an overall PIQ score (WAIS–R PIQ; Wechsler, 1986).
Children in the study completed all the WISC–III Performance subtests. This gave an overall standardized PIQ score (WISC–III PIQ; Wechsler, 1992). All standardized tests above had a standard score with a mean of 100 (SD = 15).
Parental Interview
The FHI (Bolton et al., 1994, subsequently amended for use in language studies) was completed for all immediate family members. The informant was usually the mother (74%) but sometimes the father (7%) or a combination of both (19%). The FHI uses an investigator-based rather than a respondent-based approach to elicit information and was administered by trained interviewers. Each coding is structured in terms of a definition that specifies the focus and scope of the item, with criteria to set the severity threshold used for coding. In each case, there are one or more mandatory probes to provide a comparable orienting introduction to the item for the informant. However, the interviewer's task is to obtain a description of behavior that is sufficiently precise for a decision to be made on whether the specified criteria for the item are met. A selection of follow-up probes is suggested in the schedule, but it is the responsibility of interviewers to select questions appropriately to obtain the requisite descriptions. For example, the informant is asked the following mandatory probe regarding intelligibility: "In everyday situations, how much of what X says can you understand? How able is she or he at communicating a message?" For the same item, the suggested follow-up probes include the following.
What makes her or him difficult to understand—pronunciation, grammar, rate of speech? Does she or he tend to wander off the point? Are there any situations that cause particular problems? Do you frequently have to ask her or him to repeat things? Do you need to refer to context or use gesture, e.g., pointing and miming?
The coding is then made by the interviewer according to the item definitions and criteria in the schedule.
The severity threshold for coding the FHI items is set relatively high to ensure that it is likely that most informants will be aware of the disorder in question if it has been present. In each case, a coding of 2 (definite) means that the precise criteria as specified in the schedule have been met. A coding of 1 (probable) is used when the description of behavior obtained suggests that if fuller information had been available, the criteria would have been met (but they are not met fully on the information given). The 1 coding is provided in recognition of the fact that it will not always be possible to obtain sufficient detail to be sure, one way or the other, whether the criteria are met. Items are coded for behavior at different times in an individual's development, namely, at 4 to 5 years, currently or ever. Several language-related items were selected from the FHI for examination in the present study (see Table 2 for a list).
|
Procedure
Each consenting family was telephoned to complete a pedigree and arrange convenient appointments for assessment. Where possible, children were tested at school on the battery of language, literacy, and general cognitive measures for either a morning or afternoon session. In the majority of cases, testing took place in a room with just the examiner and child present. However, where access to empty rooms was not possible, testing took place in a quiet area such as the library. If the parents had requested the child be assessed at home, testing took place in a quiet room with only the child and examiner present.
Parents were seen at home for a minimum of two visits. Each session usually lasted for a maximum of 3 hr. Psychometric tests and interviews were carried out on all consenting immediate family members. In all cases, testing was started and completed within 1 month for each separate family.
Analysis
All point and period prevalences were calculated on an available case basis. Matched significance tests were used to compare rates of impairment across instruments. Standard statistical methods assume that observations are independent one from another, but with a common informant for each item and each member of a family, such independence could not be assumed here. Two approaches were used to tackle this problem. The first recognized the lack of independence by the use of estimation and testing methods for matched samples, such as conditional logistic regression, a method that can be thought of as a generalization of McNemar's test (Hosmer & Lemeshow, 1989). The second calculated estimates of the precision of statistics and constructed significance tests by methods that assumed only that observations between families were independent. We used two methods for this; a robust covariance matrix estimator (Huber, 1967; White, 1982) and bootstrap resampling (Efron, 1993). The latter samples sets of families (without replacement) from the study data and examines how the estimated statistic of interest varies from sample to sample. In some instances, expected cell frequencies of some analyses were low, making the use of standard large sample test statistics inappropriate. Under such circumstances we report exact p values from calculations based on combinations that have better small sample behavior. Currently, such tests cannot easily be adjusted for intrafamilial clustering. Fortunately, such adjustments for clustering are commonly unimportant for rare quantities that typically give rise to low cell frequencies. All analyses were undertaken in Stata 8.0 (Statacorp, 2004). Between-group differences were assessed using dummy variables in a regression framework for continuous outcomes and logistic regression for binary variables. Difference-score regression, conditional logistic regression, and McNemar's test were used for testing differences in matched quantities with only the latter not adjusted where necessary for intrafamilial clustering. Agreement is reported by means of kappa statistics, sensitivity, specificity, positive predictive value, and negative predictive value statistics, their precision being estimated by the use of the bootstrap method.
Results
Descriptive statistics for age, measures of verbal and nonverbal IQ, literacy, and expressive and receptive language, where applicable, for all immediate family members are presented in Table 3. Group mean standard scores were in the normal range (
85) on all tests apart from the ELS of the male siblings under 16 years old (M = 82.9). The equivalent score for female siblings less than 16 years was in the borderline normal range (M = 85.7).
|
|
In terms of assessment method used and prevalence by relative type, no differences in prevalence rates were reported (via parental interview) between fathers and mothers (p = .686) or between young brothers and young sisters (p = .523). However, a significant difference was observed between older brothers and older sisters (p = .037) with a larger number of older brothers than older sisters reported to exhibit a current language or literacy difficulty. In terms of direct assessment and relative type, there were no significant differences in prevalence between any relative types.
When the six relative types described above are collapsed into three categories (namely, parents, older siblings, and young siblings), significant differences are observed between prevalence rates reported for parents and young siblings (p = .008) with young siblings having greater probability of report of impairment. No significant differences were observed between parents and older siblings (p = .454) or older and young siblings (p = .138). Direct assessment measures did not reveal any significant differences among the three collapsed relative type categories.
Agreement Between Parental Interview and Test Scores at Different Levels of Severity
Concurrence between parental interview and direct assessment was further explored through examining agreement on the presence of impairment between the two aforementioned estimation methods. For direct assessment, impairment was defined as scores falling below each of three cutoff levels: 1 SD, ` SDs, and 2 SDs from the normal population mean. For parental interview data, impairment was defined as report of either probable or definite difficulties.
Kappa agreement statistics were used to examine which of these three cutoff levels on psychometric assessments best matched reported impairment. The kappa statistic measure of agreement is scaled to be 0 when the amount of agreement is what would be expected to be observed by chance and 1 when there is perfect agreement.
Analyses were also completed for sensitivity and specificity. Sensitivity is the proportion of impaired relatives (using the psychometric test scores as the gold standard) that are correctly identified by the interview item. Specificity is the proportion of unimpaired relatives who are correctly identified by the interview item. A further analysis of positive and negative predictive values was carried out to ascertain what the probability was of the interview giving the correct diagnosis, whether it was positive or negative. The positive value was the proportion of relatives with reported impairment who were correctly diagnosed. The negative predictive value was the proportion of relatives with no reported impairment who were correctly diagnosed. Both of these values depend on the prevalence of impairment found in this particular sample that is felt to be representative of clinical samples used in previous studies.
Agreement was investigated in three areas: expressive language, receptive language, and literacy. As before, FHI interview items concerning expressive language (intelligibility, narrative skills, complexity of utterances, instructions, and word retrieval) were collapsed to form a composite measure of FHI expressive language. Interview items involving current language and literacy were selected against the relevant current psychometric assessments. Thus, the FHI expressive language composite was compared with the CELF–R and also with the WISC–III and WAIS–R VIQ scores, which both require expressive language. The FHI comprehension interview item was compared with the CELF–R receptive language score and also with the WISC–III and WAIS–R VIQ scores, which both require receptive language. Finally, as the reading impairment interview item covered reading in general terms (difficulties in learning to read and also difficulties, e.g., reading the sports page of a newspaper), measures of agreement for both WORD basic reading and reading comprehension were examined. Parental report of spelling difficulties was compared with the WORD spelling subtest.
Expressive Language
The agreement between performance of relatives on language tests and interview report of their expressive language difficulties is presented in Table 5. Recall that the CELF–R and WISC–III assessments were administered to children in the study (between 6 and 16 years of age) and the WAIS–R was administered to adults.
|
= .63) can be described as substantial (Landis & Koch, 1977). It needs to be noted, nonetheless, that the descriptors of substantial, moderate, and poor developed by Landis and Koch do not reflect clinical utility but describe the size of the statistical effect observed. With this in mind, the use of this terminology is meant to be indicative only. Moderate agreement was observed with the WISC–III VIQ score at both 1 SD and 1.25 SDs below the population mean ( = .50). The lowest agreement was observed between parental report of expressive difficulties and WAIS–R VIQ score. The kappas obtained at 1 SD and 1.25 SDs were poor (s = .20 and .14, respectively). Thus, higher strength of agreement was observed for parental report of expressive language difficulties in children rather than adults. These findings were mirrored in the sensitivity and specificity analysis with the best results being obtained with a cutoff for impairment at 2 SDs below the population mean and the CELF–R ELS (78% sensitivity, 80% specificity). Interestingly, when prevalence is taken into consideration through the analysis of positive and negative predictive value, the cutoff of 1.25 SDs appears to provide the most favorable results for predicting positive presence and negative absence of impairment (e.g. CELF–R ELS, 73% and 80%, respectively).
Receptive Language
The agreement between performance of relatives on language tests involving receptive language and interview report of their comprehension difficulties is presented in Table 6.
|
= .22). Both the WAIS–R and WISC–III demonstrate poor agreement with parental report of receptive language difficulties. The same pattern of results is observed in the specificity and sensitivity analysis, with specificity rates no higher than 25% and in the positive/negative predictive value analysis, with positive predictive values not exceeding 60%.
Literacy
The agreement between performance of relatives on tests measuring literacy and interview report of their reading and spelling difficulties is presented in Table 7. Agreement between parental report and tests used for direct assessment was fair to moderate when considering reading and spelling difficulties in relatives. Reported reading difficulty was most closely associated with WORD Reading Test scores using a cut off of 1 SD ( = .54). Reported reading difficulty also had a moderate strength of association with assessment of reading comprehension ( = .48). Reported spelling difficulty was most closely associated with WORD Spelling test scores at a cutoff of 1.25 SDs ( = .50). Recall that the WORD Basic Reading and Spelling subtests were administered to all relatives and the WORD Reading Comprehension subtest was completed by siblings between 6 and 16 years of age only.
|
In summary, agreement between parental report and direct assessment was good when considering reading and spelling and less so when considering language. Specifically in terms of language, report of expressive language difficulties in children was moderate to good, while agreement between report and assessment of receptive language difficulties was poor and the lowest of the domains examined.
Test Score Differences Equivalent to Parental Reports of Probable Versus Definite Impairment
To explore the difference between reports of probable and definite impairment on the questionnaire language and literacy items, a standard score on the psychometric tests was calculated that gave equivalent prevalence to the two levels of reported difficulty. Figure 1 presents this information for all relatives.
|
|
Discussion |
|---|
|
TopAbstractDiscussionConcluding Remarks and Clinical...References |
|---|
A closer examination of family aggregation rates by sex and relative type revealed no significant differences in relative's rate of affectedness as ascertained by the direct assessment method and few differences as revealed by the parental interview method. Significant differences in prevalence rates were found between older brothers and sisters (higher prevalence for older brothers) and between parents and younger siblings (higher prevalence in younger siblings) using the parental interview method. Previous literature has not revealed consistent differences in prevalence rates across sex and relative type. Thus, Flax et al. (2003) found significant sex ratio differences, with more males than females being affected overall regardless of relative type. In contrast, Tallal et al. (2001) did not find a consistent effect of sex. She found prevalence rates to be equivalent for mothers and fathers but brothers were significantly more affected than sisters. In the present study, we found the same pattern for older brothers versus older sisters with the parental interview method only. This male bias in male:female ratio by parental report of affectedness did not remain when direct assessment was used. Thus, the results of the present study seem to indicate that the underidentification of girls may be a problem that extends beyond school personnel biases (Shaywitz, Shaywitz, Fletcher, & Escobar, 1990; Wadsworth, DeFries, Stevenson, Gilger, & Pennington, 1992) and includes a reduced recognition of the disorder in girls by their parents. Furthermore, Tomblin (1989) found that siblings had a higher prevalence rate of impairment than parents. In the present study, we also found this to be the case (but specifically for younger siblings between 6 and 16 years of age) using the parental report method only.
At present, the differences in methodology used complicate the picture of the possible effects of sex and type of relative on the familiality of language impairment. What is clear is that these two factors (sex and type of relative) need to be explored further in future studies in this area.
Concurrence Between Parental Interview and Direct Testing
Interestingly, the present study found similar rates between methodologies for overall data when all immediate family members were considered. However, this group data masked patterns of differences based on type of relative and language/literacy domains. Plante et al. (1996) found that classification by case history methods resulted in fewer affected adults than classification through standardized testing. Thus, previous research has not always found consistency between different methods even at the level of overall affectedness in relatives. When consistency has been found in overall estimation of affectedness using parental report versus direct assessment (Tallal et al., 2001), caveats have been put forward. Specifically, Tallal et al. found that case-by-case analysis showed poor intrasubject agreement on classification of language impairment on the basis of current testing as compared with family history information. The present study concurred with these results and found that group data showed impairment rates to be similar regardless of method of collection, but patterns of agreement were different for particular relative types and for particular domains of language and literacy considered.
Consequently, the present investigation adds to the above body of research in that it specifies particular areas where reporting procedures can be particularly problematic. Thus, it was found that parental interview underestimates the prevalence of language and literacy difficulties in fathers. This is important information particularly given the possible role that gender appears to play in the familiality of language impairment (Choudhury & Benasich, 2003; Tomblin & Buckwalter, 1994). Having said this, it needs to be acknowledged that in the present study, most of the informants were mothers (74%). Future research comparing self-report of fathers versus mothers reporting on partners may be of interest.
In addition, it was found that parental interview is difficult when we are aiming to ascertain receptive language difficulties. The results for this area were consistently poor. Parental interview was found to have poor agreement with direct assessment, poor sensitivity and specificity for diagnosing the presence or absence of impairment, and poor predictive value for both positive identification and negative absence of impairment. Interestingly, manipulating the severity of impairment (below 1 SD, 1.25 SDs, and 2 SDs below the population mean) did not appear to improve any of the measures above. In the same vein, parental report of adults' expressive language was poor compared with direct assessment.
In contrast, a more positive picture for parental interview was revealed for the assessment of expressive language in children between 6 and 16 years and literacy (reading and spelling) for both children and adults, but even these results are not without a caveat. Parental report of children's expressive language abilities revealed good-to-moderate agreement with direct assessment and adequate sensitivity and specificity (78% and 90%) only when a low cutoff score was used, that is, below 2 SDs from the population mean. As far as literacy measures of reading and spelling were concerned, the pattern of results was similar, with the levels of agreement being in the moderate range and adequate sensitivity and specificity obtained, but once again only at the 2 SD level of severity. The predictive values for literacy were modest and ranged at best between 45% and 77% for positive predictive value. Overall, these results suggest that parents are effective identifiers only when the impairment is severe (below 2 SDs) relative to where scores of language-impaired children typically fall (below 1 SD). Such findings put into question the possible clinical usefulness of parental interview in identifying the presence/absence of difficulties in their children.
Note also that the above comparative analyses was based on the use of psychometric test scores as the gold standard against which parent interview was compared. This decision, although defensible in the case of the child data given the availability of the CELF–R, is more debatable in the case of the adult data, which use verbal IQ as the indicator of language impairment. As mentioned in the Method section, the WAIS–R is a global omnibus test that was not specifically designed to identify residual language deficits and is likely to underestimate language impairments. Thus, the results of the adult data based on this measure should be considered preliminary and be interpreted with caution.
Prevalence Rates Based on Parental Reports of Probable Versus Definite Impairment
A particular strength of the present study was that the interview approach used (the FHI) was highly structured in terms of the concept and criteria but semistructured in terms of the wording of probes. This is in contrast to respondent-based interviews that are highly structured with respect to the wording of questions but rely on unstructured concepts (leaving it open to informants to make decisions that go beyond the specifications of the question), together with "yes/no" formats rather than descriptions of behavior. It will be appreciated that the strength of the investigator-based approach, as used here, lies in the specification of the concepts and criteria and the interviewer's understanding of them, so that questioning can be shaped appropriately to obtain the specific behavioral features needed to decide on coding. The FHI has therefore several advantages over respondent-based-type questionnaires, particularly for use with a potentially language impaired sample.
It is of interest to comment on the decisions informants make when reporting on the FHI's probable and definite impairment categories. In the present study, it was found that regardless of whether decisions on language and literacy impairments reported in the FHI are probable or definite, they equate to depressed psychometric test scores (>1 SD below the mean). This has wider ranging implications for coding the responses of parents during family history interviews and suggests that parents are picking up relatively more marked abnormalities when a probable coding is used, and thus impairment thresholds may need to be set lower to positively code responses of probable impairment. This has been thought to be a particularly crucial consideration in family studies in which there is likely to be variation in the degree to which informants have had the opportunity to observe, or receive reports on, the behavior of different relatives. It is also important to reiterate that the probable coding in the FHI is not intended to assume that SLI is a continuous trait and thus individuals can show a lesser degree of affection. Consequently, the probable coding is not intended to reflect mild abnormalities that fall below the specified threshold and is not used for that purpose. Future research can address more directly the issue of whether SLI involves continuous versus discrete traits, as these different models have implications for the design and analysis of family and genetic studies.
|
Concluding Remarks and Clinical Implications |
|---|
|
TopAbstractDiscussionConcluding Remarks and Clinical...References |
|---|
These findings have implications for future family studies involving parental interview methodologies in that clinicians and researchers need to be aware of the potential strengths and weaknesses of the reporting method in the evaluation of affectedness and familiality. In addition, within the clinical context, the ascertainment of clients' family histories usually follows an informant-based interview or questionnaire model. Thus, similar considerations of the strengths and weaknesses of this method need to be applied in the clinical context.
|
Acknowledgments |
|---|
Received October 21, 2004
Revision received March 17, 2005
Accepted June 9, 2005
|
References |
|---|
|
TopAbstractDiscussionConcluding Remarks and Clinical...References |
|---|
Beitchman, J. H., Hood, J., & Inglis, A. (1992). Familial transmission of speech and language impairment. Canadian Journal of Psychiatry, 37, 151–156.[Medline]
Bolton, P., MacDonald, H., Pickles, A., Rios, P., Goode, S., Crowson, & M., et al (1994). A case-control family history study of autism. Journal of Child Psychology and Psychiatry, 35, 877–900.[Medline]
Choudhury, N., & Benasich, A. (2003). A family aggregation study: The influence of family history and other risk factors on language development. Journal of Speech, Language, and Hearing Research, 46, 261–272.
Conti-Ramsden, G., & Botting, N. (1999a). Characteristics of children attending language units in England: A national study of 7-year olds. International Journal of Language and Communication Disorders, 34, 359–366.[CrossRef]
Conti-Ramsden, G., & Botting, N. (1999b). Classification of children with specific language impairment: Longitudinal considerations. Journal of Speech, Language, and Hearing Research, 42, 1195–1204.
Conti-Ramsden, G., Crutchley, A., & Botting, N. (1997). The extent to which psychometric tests differentiate subgroups of children with specific language impairment. Journal of Speech, Language, and Hearing Research, 40, 765–777.
Efron, B. (1993). An introduction to the bootstrap New York: Chapman, & Hall.
Felensfeld, S., McGue, M., & Broen, P. (1995). Familial aggregation of phonological disorders: Results from a 28-year follow-up. Journal of Speech and Hearing Research, 38, 1091–1107.[Medline]
Flax, K., Realpe, T., Hirsch, L. S., Brzustowicz, L. M., Bartlett, C., & Tallal, P. (2003). Specific language impairment in families: Evidence for co-occurrence with reading impairments. Journal of Speech, Language, and Hearing Research, 46, 530–543.
Gopnik, M., & Crago, M. (1991). Familial aggregation of a developmental language disorder. Cognition, 39, 1–50.[CrossRef][Medline]
Hosmer, D. W., & Lemeshow, S. (1989). Applied logistic regression New York: Wiley.
Huber, P. J. (1967). The behavior of maximum likelihood estimates under non-standard conditions. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability Berkeley: University of California Press 221–233.
Lahey, M., & Edwards, J. (1995). Specific language impairment: Preliminary investigation of factors associated with family history and with patterns of language performance. Journal of Speech and Hearing Research, 38, 643–657.[Medline]
Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33, 159–174.[CrossRef][Medline]
Neils, J., & Aram, D. (1986). Family history of children with developmental language disorders. Perceptual and Motor Skills, 63, 655–658.[Medline]
Office of Population Censuses and Surveys. (1991). Standard occupational classification London: Her Majesty's Stationery Office Vol. 3.
Plante, E., Shenkman, K., & Clark, M. (1996). Classification of adults for family studies of developmental language disorders. Journal of Speech and Hearing Research, 39, 661–667.[Medline]
Rice, M., Haney, K., & Wexler, K. (1998). Family histories of children with SLI who show extended optional infinitives. Journal of Speech, Language, and Hearing Research, 41, 419–432.
Rust, J., Golombok, S., & Trickey, G. (1993). Wechsler Objective Reading Dimensions San Antonio, TX: The Psychological Corporation.
Semel, E., Wiig, E. H., & Secord, W. (1987). Clinical Evaluation of Language Fundamentals—Revised UK San Antonio, TX: The Psychological Corporation.
Shaywitz, S. E., Shaywitz, B., Fletcher, J. M., & Escobar, M. D. (1990). Prevalence of reading disability in boys and girls: Results of the Connecticut Longitudinal Study. The Journal of the American Medical Association, 264, 998–1002.
Spitz, R. V., Tallal, P., Flax, J., & Benasich, A. A. (1997). Look who's talking: A prospective study of familial transmission of language impairments. Journal of Speech, Language, and Hearing Research, 40, 990–1001.
Statacorp. (2004). Stata Statistical Software (Release 8.0) College Station, TX: Stata Corporation.
Stromswold, K. (1998). Genetics of spoken language disorders. Human Biology, 70, 297–324.[Medline]
Tallal, P., Hirsch, L. S., Realpe-Bonilla, T., Miller, S., Brzustowicz, L. M., Bartlett, C., & Flax, J. J. (2001). Familial aggregation in specific language impairment. Journal of Speech, Language, and Hearing Research, 44, 1172–1182.
Tallal, P., Ross, R., & Curtiss, S. (1989). Familial aggregation in specific language impairment. Journal of Speech and Hearing Disorders, 54, 167–173.
Tomblin, J. B. (1989). Familial concentration of developmental language impairment. Journal of Speech and Hearing Disorders, 54, 287–295.
Tomblin, J. B. (1996). Genetic and environmental contributions to the risk for specific language impairment. In M. L. Rice (Ed.), & Genetics of specific language impairment (pp. 191–210) Baltimore: Brookes.
Tomblin, J. B., & Buckwalter, P. R. (1994). Studies of genetics of specific language impairment. In R. V. Watkins, & M. L. Rice (Eds.), Specific language impairments in children (pp. 17–34) Baltimore: Brookes.
Van der Lely, H., & Stollwerck, L. (1996). A grammatical specific language impairment in children: An autosomal dominant inheritance? Brain and Language, 52, 484–504.[CrossRef][Medline]
Wadsworth, S. J., DeFries, J. C., Stevenson, J., Gilger, J. W., & Pennington, B. F. (1992). Gender ratios among reading-disabled children and their siblings as a function of parental impairment. Journal of Child Psychology and Psychiatry, 33, 1229–1239.[CrossRef][Medline]
Wechsler, D. (1986). Wechsler Adult Intelligence Scale—Revised, UK San Antonio, TX: The Psychological Corporation.
Wechsler, D. (1992). Wechsler Intelligence Scale for Children—Third Edition—Revised, UK San Antonio, TX: The Psychological Corporation.
White, H. (1982). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48, 817–830.[CrossRef]
Whitehurst, G. J., Arnold, D. S., Smith, M., Fischel, J. E., Lonigan, C. J., & Valdez-Menchaca, M. C. (1991). Family history in developmental expressive language delay. Journal of Speech and Hearing Research, 34, 1150–1157.[Medline]
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| All ASHA Journals | AJA | AJSLP | JSLHR | LSHSS |
