Customize your JAMA Network experience by selecting one or more topics from the list below.
Copyright 2010 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2010
To examine the relationship between characteristics of the cognitive environment at age 10 to 18 months and vocabulary at age 18 to 30 months.
Analysis of baseline and follow-up data on 157 families participating in a comparison of 2 anticipatory guidance programs.
Children's Hospital outpatient department serving low-income families.
Parents of children aged 10 to 18 months at baseline who participated in a follow-up telephone interview at age 18 to 30 months.
Main Outcome Measures
Three subscales of the StimQ (reading, parental involvement in developmental activities, and parental verbal responsivity [PVR]) and the short form of the MacArthur Communicative Development Inventories.
Vocabulary score percentiles dropped significantly between baseline and follow-up, with scores for bilingual families showing a greater decrease than those for English speaking–only families. StimQ subscale scores increased with maternal education and increased between baseline and follow-up. Multiple regression analysis showed that baseline variables accounted for 25% of the variance in follow-up vocabulary score percentile, with significant contributions from baseline expressive vocabulary (P < .001), PVR (P = .01), and home language (P = .03) scores. Seventy-seven percent of children with PVR scores less than 4 had follow-up vocabulary scores at or less than the 25th percentile, with an associated likelihood ratio of 4.33. However, 35% of children with a PVR score of 4 also had vocabulary scores less than the 25th percentile at follow-up, with an associated likelihood ratio of 0.67.
The StimQ is a clinically useful method for assessing early environmental factors that influence vocabulary development. The PVR subscale score was the best StimQ predictor of later vocabulary delay and may be useful in identifying children needing referral for evaluation.
Vocabulary at age 3 years is a major determinant of readiness for and success in school.1 According to Hart and Risley,1 the number of words a child hears, especially in conversation, and the amount of encouragement are the 2 most important environmental influences on vocabulary development prior to age 3 years. Unfortunately, the amount of talking and encouragement are typically more limited in families of lower compared with higher socioeconomic status, as are other factors supporting development of early literacy. The gap in vocabulary associated with socioeconomic status can be detected as early as 9 months of age2 but becomes more obvious by 2 years of age.3,4 The gap continues to widen, and by age 3 years, many children are so far behind that school failure becomes almost assured.1 A goal of prevention is to identify and intervene before the decline becomes more evident. Developmental screening tests may not identify children at risk for delay secondary to environmental influence prior to about 18 months of age, but an environmental assessment might.
A large number of studies have demonstrated that assessment of parent attitudes5,6 and practices in infancy7,8 are highly predictive of later vocabulary development and school performance. The most widely used instrument for assessing relevant features of the home environment for predicting school failure is the HOME interview of Bradley and Caldwell.9 This instrument can be used to identify homes that are unlikely to support development, but it requires extensive training and an hour of interaction in the child's home. The Coons et al10 Home Screening Questionnaire was an early attempt to reduce the time and training involved in obtaining the same information. In at least 1 study,11 the Home Screening Questionnaire was a better predictor of developmental quotient after age 18 months than the Bayley Scales of Infant Development I administered at 12 months of age. More recently, Dreyer et al12 have developed a brief questionnaire, the StimQ, which they term a measure of the cognitive environment13 and which is also based on the HOME interview. In addition, its reliability and concurrent validity data are substantial, showing significant relationships with factors in the HOME interview and with the Bayley Scales of Infant Development II administered at 21 months of age.14 There are, however, limited data describing the relationship between the StimQ and vocabulary development.
This article examines vocabulary development during the second year of life in children living in low-income families and how it relates to the cognitive environment as measured by the StimQ. The data represent a secondary analysis of a randomized controlled study designed to assess the effect of education about language stimulation or injury prevention on child vocabulary, parenting practices, and injury prevention knowledge and behavior.15-17 There were no significant differences between the intervention groups on StimQ subscale or vocabulary scores at baseline, follow-up, or across time. This article presents the longitudinal StimQ and vocabulary data from this study. The questions addressed in this analysis were (1) Were there differences between vocabulary score percentiles at baseline and follow-up, and if so, what parent characteristics influenced the differences? (2) Can information about parent characteristics and the cognitive environment at baseline predict vocabulary score status at follow-up? (3) Can this information be used to identify children at risk for being in the lowest vocabulary score percentile at follow-up?
A total of 276 children between the ages of 10 and 18 months were initially enrolled during a well-child visit to The Children's Hospital primary health care clinic located in Denver, Colorado, from October 4, 2002, to August 15, 2003. Inclusion criteria consisted of the parent's ability to read the consent forms in English or Spanish and presence of a telephone. Families were excluded from enrollment if (1) the child was premature (<37 weeks' gestation) or had major medical or genetic abnormalities involving neurosensory disabilities or limitations in the child's mobility, breathing, or swallowing; (2) the child had a prolonged hospitalization (>7 days); or (3) the child was identified as or suspected of being developmentally delayed.
One hundred fifty-seven of these families had both baseline and follow-up assessments. Among these 157 families, 24 did not have useable vocabulary assessment data at follow-up because 13 were non–English speaking, 8 were too old for the assessment, and 3 had technical errors in data collection. An additional child was missing vocabulary scores at baseline. Therefore, 157 families were included in the analyses of the StimQ and family characteristics and 132 families were included in the analyses involving both baseline and follow-up vocabulary scores. Table 1 shows the demographic characteristics of the 157 families. There were no significant differences between families with follow-up and the original sample with respect to demographic variables or any of the specific measures described later.
Maternal education (MOED) and home language(s) were selected to characterize the child's sociocultural environment. An ordinal variable MOED was created as follows: (1) some high school or less, (2) high school diploma or General Equivalency Diploma, (3) some college, or (4) college degree or some postgraduate education. For the language variable, children were classified based on the parent's report of the language(s) to which the child was exposed at home. The resulting groups were English, Spanish, English plus Spanish language/bilingual, and other (English or Spanish plus another non-English language).
The StimQ12 was used to assess the “cognitive environment.”13 This is a validated 30-item questionnaire, available in both English and Spanish. We administered an abbreviated StimQ consisting of the 3 subscales most closely targeting behaviors fostered by the language stimulation program. These were reading (READ), parent involvement in developmental activities (PIDA), and parent verbal responsivity (PVR). The READ subscale consists of 11 items such as frequency of bedtime reading and number of books at home, with raw scores ranging from 0 to 18. The PIDA subscale consists of 10 items such as teaching colors, counting, and manipulating objects, with raw scores ranging from 0 to 10. The PVR subscale consists of 4 questions about the parent playing with the child in the bath, playing finger/rhyming games, playing peek-a-boo or hide-and-seek, and talking about the day when the child is eating, with raw scores ranging from 0 to 4.
The MacArthur Communicative Development Inventories short form18 is a validated instrument used to measure vocabulary acquisition based on parent report. Two short forms are available in English, 1 for infants (ages 8-18 months, using 89 words) and 1 for toddlers (ages 16-30 months, using 100 words) with 2 equivalent forms, A and B. The parent is asked to indicate whether the child speaks each word (expressive vocabulary) and, in the infant form, whether the child understands the word (receptive vocabulary). Receptive vocabulary is not reported herein because it is not included in the toddler form and showed no unique relationship to follow-up variables. The total number of words spoken was converted to a percentile score representing the child's standing in expressive vocabulary relative to norms for his or her age in months. The MacArthur short forms can be administered during a telephone survey as well as face to face and have been used in other similar types of research.19,20 A Spanish version, only available for the toddler form, was administered to children from monolingual Spanish families, but results are not included herein because only provisional norms were available at the time. Parents were asked which language they preferred to use for the vocabulary assessment.
The clinical trial was approved by the university and Children's Hospital combined institutional review board, and all families participated in a 2-phase informed consent process. Families were initially recruited to participate in a study assessing the benefits of an enhanced well-child visit in promoting the health and development of their child. In this study, the 2 programs were the Bright Beginnings of Colorado Language Power program21,22 and the American Academy of Pediatrics educational program around injury prevention. Families first signed a phase 1 consent form agreeing to complete the initial survey and be randomly assigned to 1 of the nondescribed educational programs. The initial survey consisted of the StimQ subscales, a safety assessment, and the MacArthur vocabulary short form appropriate for the child's age. For phase 2, they were then assigned randomly to receive 1 of the programs, which the research associate then described. The phase 2 consent for participation in that program (including consent for follow-up) was obtained and the selected program delivered as a 10- to 15-minute addition to the well-child visit.
Families became eligible for follow-up at least 6 months after the intervention but no earlier than 18 months of age. A research assistant unaware of program assignment contacted each family by telephone and completed the StimQ, the MacArthur vocabulary short form for toddlers, and the safety assessment. The time between baseline and follow-up ranged from 5 months to 19 months with a mean (SD) interval of 9.2 (3.1) months. The mean (SD) age at baseline was 13.4 (2.5) months and the mean (SD) age at follow-up was 22.6 (4.0) months.
A repeated-measures (time 1 and time 2) multivariate analysis of variance using SAS PROC GLM23 was performed to examine the differences between baseline and follow-up for vocabulary and StimQ subscale scores. Multiple regression examined how well baseline information predicted follow-up vocabulary scores. A classification analysis was performed on the 2 × 2 table derived from grouping children into low and typical StimQ scores to determine likelihood ratios and predictive values for follow-up vocabulary score in the lowest quartile.
A total of 157 families had baseline and follow-up data (Table 1). The majority of families were low income, 74% being enrolled in a public coverage program or having no coverage. One-third of enrollees were single-parent households. Slightly more than half of the mothers did not have any education beyond high school. More than one-third of the enrollees were Hispanic and the majority of these families were bilingual.
Table 2 shows the means and standard deviations for expressive vocabulary score percentiles at baseline and follow-up grouped by home language(s) and MOED for 132 children. The data in Table 2 were subjected to a repeated-measures multivariate analysis of variance of expressive vocabulary score percentiles at 2 times (baseline and follow-up) for 2 language groups (English and bilingual) and the 4 MOED groups. This was done for all variables combined and for each independent variable separately. A significant main effect for time (F1,124(time) = 4.89; P < .03) reflected that the scores for the total population dropped a mean (SD) 7.0 (37.4) percentile points between baseline (mean [SD], 46.8 [31.2]) and follow-up (mean [SD], 39.8 [31.2]). When only the effect of home language was examined, children from bilingual homes registered a significantly larger mean decrease (16.6) in score than children from English speaking–only homes (0.9) (F1,130(time × language) = 4.81; P < .03). The difference between unadjusted means for bilingual and English speaking–only groups across time in Table 2 suggests that only the bilingual children's scores dropped significantly. However, the majority (55%) of the total group's scores showed some decrease.
In the individual analysis of MOED, the main effect was not significant but the interaction of MOED with time showed a trend toward statistical significance (F3,124(MOED × time) = 2.49; P < .06), reflecting an increase in the means only for the highest educational group (college degree/postgraduate). In the analysis that included both home language and MOED simultaneously, the interaction of time with both language (F1,124(time × language) = 3.22; P < .08) and MOED (F3,124(time × MOED) = 2.49; P < .06) showed a trend toward statistical significance.
Preliminary analysis showed no significant effect of home language on StimQ scores at baseline, follow-up, or across time. Consequently, only the results of a multivariate repeated-measures analysis of variance for the effects of MOED and time are reported for StimQ scores. These are presented in Table 3. Significant increases were observed for READ (mean [SD], baseline, 11.0 [4.8]; follow-up, 12.4 [3.1]; F1,153(time) = 15.8; P < .001) and PIDA (mean [SD], baseline, 7.8 [1.9]; follow-up, 8.9 [1.5]; F1,153(time) = 47.91; P < .001) scores. A significant effect for MOED was also observed for both the READ (F3,153(MOED) = 4.95; P < .003) and PIDA (F3,153(MOED) = 5.78; P < .001) scores. These differences were present at both baseline and follow-up but the interaction between MOED and time was not statistically significant for either READ or PIDA scores.
Differences in PVR scores over time or between education groups did not reach statistical significance, but the interaction of MOED with time showed a trend toward significance (F3,153(MOED × time) = 2.10; P < .10). This reflected the fact that statistically significant differences were observed among educational groups in univariate analyses of PVR scores at follow-up but not at baseline (Table 3).
Table 4 shows the results of a multiple regression analysis using variables that showed a significant zero-order correlation between status at baseline and vocabulary score at follow-up plus MOED group. This analysis showed that baseline expressive vocabulary (P < .001), PVR (P = .01), and home language (P = .03) scores contributed significantly to the prediction of follow-up vocabulary score percentile (R2 = 0.252).
Baseline PVR score emerged as the best single StimQ predictor of vocabulary score percentile at follow-up. This suggested that PVR score between 10 and 18 months of age might be useful in identifying children at risk for later being in low vocabulary score percentiles.
The PVR subscale consisted of only the 4 simple questions noted previously. At baseline, 5% (7 of 132) of parents answered yes to 0 to 2 questions; 17% (22 of 132), yes to 3 questions; and 78% (103 of 132), yes to all 4 questions.
Table 5 shows the vocabulary score percentiles at baseline and follow-up of children grouped by baseline PVR score into very low (PVR score = 0-2), low (PVR score = 3), and typical (PVR score = 4). A multivariate repeated-measures analysis of variance for effects of time × 3 PVR score groups on vocabulary score percentiles was performed on these data. The main effect of baseline PVR score on vocabulary score percentile was statistically significant for both vocabulary score at baseline (F1,130 = 3.76; P < .05) and follow-up (F1,130 = 1.41; P < .001) as well as time and the interaction of PVR score with time (F1,130(PVR × time) = 6.01; P < .02). At baseline, the mean vocabulary score percentiles for the low (40.4%) and typical (49.9%) groups clustered together, while at follow-up, the percentiles for the low group (24.1%) dropped to cluster with the very low group (14.3%).
The possible usefulness of these results for decision making was examined using PVR score at baseline to identify children with scores at or less than the 25th percentile at follow-up. This analysis is presented in Table 6 with 2 PVR score groups, low (score = 0-3), which combined the very low and low groups, and typical (score = 4). Results showed a predictive value of 77% (23/30) for a follow-up vocabulary score percentile in the lowest quartile when the baseline PVR score was less than 4. The associated likelihood ratio for a positive screen of 4.33 indicated that the odds of being in a low quartile at follow-up were increased by a factor of 4.33. The predictive value for a typical PVR score at baseline was 35%. The associated likelihood ratio for a negative screen was 0.67, indicating a 33% decrease in the odds of having a low-quartile vocabulary score at follow-up.
The major findings in this study are the following: (1) Average vocabulary score percentiles in this predominantly low-income group dropped between the beginning and end of the second year of life, with greater drops in children from bilingual homes than from English speaking–only homes. This trend was reversed only in the children of the most highly educated mothers. (2) Twenty-five percent of the variance in follow-up vocabulary score percentiles was accounted for by baseline variables, with statistically significant contributions from baseline expressive vocabulary (P < .001), PVR (P = .01), and home language (P = .03) scores. (3) A child with a low PVR score at baseline was 4.33 times more likely to have a vocabulary score in the lowest quartile at follow-up.
The observed drop in mean vocabulary score percentile is consistent with previous observations on children from low-income homes. Small differences, detectable as early as 9 months of age in large samples,2 increase until they are large enough to be significant in smaller samples, around 18 months to 2 years of age.3
Why the score percentiles of children from bilingual homes dropped so much more than those from English speaking–only homes is an interesting question. One answer is that assessing status in only 1 language does not adequately reflect their ability. Another possibility is that many of these children entered a “silent period,” such as seen in children adopted from foreign countries as infants.24 As some children are attempting to master their second language, they quit talking for a period. One clinician in the present study reported comments along this line from parents. Further study is needed to clarify the basis for this observation.
Results support the usefulness of the StimQ as a measure of environmental factors that significantly influence the emergence of expressive vocabulary. Dreyer et al13 demonstrated that 2 of the factors used herein (READ and PIDA scores) were highly correlated with the component on the HOME interview that best predicts later intellectual development. Previous studies with the StimQ have emphasized the impact of these factors on development at 21 months of age.14 At the earlier age studied herein, PVR score emerged as the best single StimQ predictor of vocabulary at 2 years of age. This is consistent with other studies describing verbal responsivity as a major developmental-enhancing environmental influence in the first year of life.4,25,26 In development of the StimQ, PVR was the only subscale designed to measure “unstructured verbal responsiveness of the parent.”13 It was also the only factor that related to the parent responsiveness subscale of the HOME interview.13
Clinicians may apply these findings to identify some low-income children at risk for language delay before it otherwise becomes apparent. With prevalence of scores in the lowest vocabulary quartile at 44%, the likelihood ratio for a positive screen of 4.33 raised the positive predictive value associated with any no answer to 77%. This is higher than the minimum level (60%) suggested by Camp27 as acceptable for using developmental screening test results to justify referring a child for further evaluation.
Predictive value of a typical PVR score at baseline was 35% with a likelihood ratio for a negative screen of 0.67. According to the Camp27 criteria, the negative predictive value should be no more than 10%, with an associated likelihood ratio for a negative screen less than 0.20, before dismissing the likelihood of delay. Both of these values were too high to dismiss the likelihood of a low vocabulary score percentile at follow-up in children with a typical PVR score at baseline without further information. Because both home language and baseline vocabulary scores contributed independently to predicting follow-up vocabulary score percentile, efforts were made to see if adding these variables would improve the predictive value in children with typical PVR scores. The best combination achieved a predictive value of 12% with an associated likelihood ratio for a negative screen of 0.42 if a child from an English speaking–only family was in the top vocabulary quartile at baseline. While these values show some improvement, they were still insufficient according to the Camp criteria.
Selection of variables in the study and methods used to study them have several limitations. All data were obtained by parent report and at follow-up, by telephone interview. Bias may also have been introduced by using different forms of the vocabulary test (infant and toddler) and being able to provide only percentile scores rather than standard scores. Unexpected findings in some groups suggest caution in interpretation. For example, children from the least-educated English speaking–only families had the highest vocabulary scores at both baseline and follow-up. The trend toward significant interaction between time and MOED on vocabulary scores between baseline and follow-up may have resulted from underestimation of baseline vocabulary score in the highest educational group with a subsequent regression toward the mean.
The StimQ appears to be an efficient and economical way to evaluate the cognitive environment of young children and to identify those at greatest risk of poor vocabulary in the second year of life before evidence of significant delay is usually observable. Answering no to even 1 of the PVR questions increases the likelihood of later delay and supports the recommendation that referring these children for further study should be considered. A referral and more comprehensive evaluation may identify maternal depression or other treatable problems, remediable family issues, better child care options, or intervention services such as speech therapy. At a minimum, these families and children should receive careful follow-up and monitoring. Studies are needed to validate our findings. Further, pediatricians need to continue seeking effective interventions that reduce the prevalence of low vocabulary in toddlers of low-income families with typical PVR scores.
Correspondence: Stephen Berman, MD, The Children's Hospital, 13123 E 16th Ave, B032, Aurora, CO 80045 (email@example.com).
Accepted for Publication: April 6, 2010.
Author Contributions:Study concept and design: Camp, Cunningham, and Berman. Acquisition of data: Cunningham and Berman. Analysis and interpretation of data: Camp and Berman. Drafting of the manuscript: Camp and Berman. Critical revision of the manuscript for important intellectual content: Camp, Cunningham, and Berman. Statistical analysis: Camp. Obtained funding: Berman. Administrative, technical, and material support: Camp, Cunningham, and Berman. Study supervision: Berman.
Financial Disclosure: None reported.
Camp BW, Cunningham M, Berman S. Relationship Between the Cognitive Environment and Vocabulary Development During the Second Year of Life. Arch Pediatr Adolesc Med. 2010;164(10):950–956. doi:10.1001/archpediatrics.2010.169