Symposium
Chair: Margaret Burchinal
The Phase I public-release analytic data set from the NICHD Study of Early Child Care contains a vast array of longitudinal information about demographic, family, and child care characteristics and children's behavior and development during the first three years of life for a diverse sample of 1364 children and their families. The release of these data scheduled for January, 2000 will provide infant researchers with a new extant data set that will be useful for addressing a wide variety of developmental issues.The purpose of this symposium is to describe the organization of the data sets and the measures of the family, child care experiences, and child collected between the child's birth and their third birthday included in those data sets. The measures will be described in general and discussed within the context of the published papers from the Early Child Care Research Network. Example programs will be distributed to assist individuals with using the data base. Audience questions about study measures, the available data, and research opportunities afforded by the data set will be encouraged and addressed by investigators from the Network.
Details of individual items:
paper
A rich array of longitudinally measured family variables is in the Phase I analytic dataset of the NICHD Study of Early Child Care. The purpose of this paper is to describe the family measures collected in the first three years of the study. Findings from several papers of the Early Child Care Network will be briefly highlighted in order to convey evidence for the construct and convergent validity of the study's measures. Family measures were collected when children were 1-, 6-, 15-, 24-, and 36-months of age from maternal interview, questionnaires, naturalistic observation, and videotaped mother-child interactions in home and laboratory visits. Demographic measures of family income and their sources, parental work hours, and marital status of the child's mother were measured repeatedly. All household members and their relation to the target child and mother were also tracked over time allowing the derivation of 11 categories of household type and the measurement of stability and change in the child's household over time. For example, 796 of the families at child age 24 and 36 months were traditional nuclear families with married mothers and fathers. 184 experienced various changes in household members between the 24- and 36-month home visits; for example, 50 mothers separated or divorced from their husbands or partners in this time interval. Maternal characteristics included child-rearing attitudes measured at 1 month, personality factors at 6 months, and an index of mothers' intelligence at 36 months. Maternal depressive symptoms and separation anxiety were measured at 1, 6, 15, 24, and 36 months. In one study utilizing the repeated measurement of maternal depressive symptoms, we have related various child outcomes at 36 months to chronic, episodic and no maternal depression over the child's first three years and examined the mediation of these associations through qualities of mother-child interaction and their moderation by family income. Family factors of social support, stressful life events, maternal role satisfaction, marital quality, job stress, and financial stress have also been measured repeatedly. Although these measures generally have shown moderately strong stability over time, with correlations between measurements of approximately .50, patterns of continuity and discontinuity over time can be examined as well as the family and child correlates of different patterns over time. The home environment was measured using the infant-toddler HOME at 6 and 15 months, and the preschool HOME at 36 months. Home visitors were extensively trained and certified to collect HOME data on the basis of common training sessions and their rating reliability in both live observations and their agreement with gold standard ratings of videotaped observations. Mother-child interaction was measured in semi-structured 15-minute mother-child interactions collected four times over the three years. The interactions were reliably rated at a central non-data collection location. Qualitative ratings of maternal and child behavior and dyadic behavior are included in the dataset. The maternal behavior ratings denoting sensitivity and cognitive stimulation have been among the strongest predictors of child social and cognitive outcomes at 24 and 36 months in the study.
paper
The data set from Phase I of the NICHD Study of Early Child Care, from birth through 36 months of age, contains a wealth of longitudinal measures of the child-care context. These measures will be described with respect to their development, their psychometric properties, and their structure within the analysis data set. Various findings from Phase I will be outlined to illustrate the ways in which the child-care data have been used Child-care measures were collected when the children were 6, 15, 24, and 36 months old. A major innovation of the study was the development of an observational instrument, the ORCE (Observational Record of the Caregiving Environment), to assess the quality of caregiving for individual children. The ORCE is valuable because it can be used across settings (e.g., home- or center-based care) and across types of caregivers (e.g., relative, nanny, teacher). The ORCE yielded frequency counts of specific caregiver acts (positive caregiving frequencies) as well as ratings of the quality of caregiver behaviors (positive caregiving ratings) These variables were associated with smaller child-adult ratios and group sizes and with caregivers' child-centered beliefs (NICHD Early Child Care Research Network [ECCRN], 1996; in press-a). The ORCE variables also were related to the observed quality of the physical environment, as measured by a modified HOME inventory for home-care settings (Caldwell & Bradley, 1984), and the Assessment Profile for Early Childhood Programs (Abbott-Shim & Sibley, 1987) for child-care homes and centers (NICHD ECCRN, 1996). The ORCE variables have been used extensively in the major substantive analyses of the effects of child care on children's development in Phase I2E For example, after controlling for relevant family and child characteristics, the quality of child care, as measured by the ORCE, was related to greater social competence and cooperation and less problem behavior (NICHD ECCRN, 1998) and to cognitive and language development (NICHD ECCRN, in press-b). The mothers of the study children also provided child-care information every 3 months via phone interviews regarding amount of care, type of care, age of entry into care, and changes in child-care arrangements. These variables also have been used in the major outcome analyses for Phase I. For example, we found that an elevated rate of insecure infant-mother attachment was associated with insensitive mothering combined with more-than-minimal hours in child care, low-quality care or relatively unstable care (NICHD ECCRN, 1997). Caregivers also have provided data about their demographic characteristics as well as their beliefs about childrearing. The belief variables, in particular, have proven to be significant markers of child-care quality (NICHD ECCRN, 1996; in press-a). Finally, the 'regulable' aspects of care-such as caregiver training and education, observed group size, and observed child-adult ratio-were measured at each major assessment point. Analyses of these variables have indicated that child outcomes are better when children attend center classrooms that meet professional recommendations for child-adult ratios and caregiver training and education (NICHD ECCRN, 1999).
paper
The Phase I analytic data set of the NICHD Study of Early Child Care contains an immense and diverse amount of information about children's behavior and development during the first three years of life. The data are longitudinal and permit the study of both normative change and individual differences. The purpose of this paper is to describe the measures of the child at each age of assessment and within each major developmental domain. This will include how the data were collected as well as the basic psychometric properties of the measures, including their longitudinal stability. Their use in predictive analyses will be illustrated from several of the papers published by the NICHD Early Child Care Research Network [ECCRN].Children's behavior and development were assessed in four major domains: 1) socioemotional; 2) cognitive; 3) language; 4) physical. Within each domain multiple constructs were assessed: temperament, attachment security, self-regulation, peer competence, and behavior problems in the socioemotional domain; play complexity, sustained attention, Bayley mental development index, and school readiness in the cognitive domain; vocabulary, verbal comprehension, sentence complexity and expressive language in the language domain; growth, general health, GI and ear infections, injuries, and sleep problems in the health domain. Assessments were conducted when children were 6, 15, 24, and 36 months old. Data were collected in the laboratory, in the children's homes, and in their child care settings, as well as from parents and child-care providers. Data collection included videotaped and real-time observations that yielded counts or ratings of behavior, laboratory-based standardized tests, questionnaires, interviews, and medical records searches. A particular strength of this study is that nearly every child outcome was measured in multiple contexts, at multiple ages, and using multiple data collection strategies. Data collectors were extensively trained using centrally produced video- and audio-tapes and face-to-face training by the developers of the procedures. Videotaped data were coded centrally by intensively trained and reliable coding teams, with different sites responsible for coding different outcomes. The corpus of data thus meets extremely high standards for standardization and uniformity across data collection sites and for inter-rater reliability.We have shown several of these developmental outcomes to be predicted by family, child, and child care characteristics. Attachment classification was predicted at 15 months by maternal characteristics, and by child-care experience combined with insensitive mothering (NICHD ECCRN, 1997); by 36 months, mothering predicted security but child care did not (NICHD ECCRN, submitted). Children's compliance with adults, measured at multiple ages and in multiple contexts, and their self-control and behavior problems, were predicted by family characteristics and child care quality at 24 and 36 months (NICHD ECCRN, 1998). Children's peer competence, assessed in child care, in dyadic play with a friend, and by maternal and caregiver report, was predicted by family and child characteristics, as well as by child care experience at 24 months and amount of peer experience in child care at 36 months (NICHD ECCRN, submitted). Cognitive and language development were assessed observationally, by maternal report, and by standardized tests, and were predicted by both family characteristics and child care quality at 24 and 36 months (NICHD ECCRN, in press).
paper
The public-release data set from the Phase I of the NICHD Study of Early Child Care documents the demographic, family, and child care characteristics and child outcomes of 1364 children recruited at birth and followed through their third birthday. Participants were recruited from hospitals located in or near Little Rock, AR; Irvine, CA; Lawrence, KS; Boston, MA; Philadelphia, PA; Pittsburgh, PA; Charlottesville, VA; Morganton, NC; Seattle, WA; and Madison, WI. The families and child care setting were observed when the children were 6, 15, 24, and 36 months of age, and cognitive, social, and peer outcomes were collected at 15, 24, and 36 months. The public release data set consists of both 'analysis' and 'item-level' data. The analysis data sets contain the summary measures that displayed good psychometric properties and validity. They are organized by age of the child (1, 6, 15, 24, and 36 months) and type of data (demographic, family, child care, and child outcomes). Included also in the public release CD is the 'data dictionary' and Child Care Data Reports (CCDRs). The CCDRs describe the psychometric properties of these summary variables and provide information about how they were constructed. The data dictionary provides an index that links the variables, the data sets, and the CCDRs. Finally, there is a separate item-level data set for each instrument that was administered by age of assessment.The analysis data sets are likely to be most use to most users. These data sets are named according to type of data and age of assessment. The primary data sets include the demographic data sets (DEMO1, DEMO6, DEMO15, DEMO24, DEMO36), family data sets (FAM1, FAM6, FAM15, FAM24, FAM36), summary child care data sets (CC_CG01, CC_CG06, CC_CG15, CC_CG24, CC_CG36), and child outcome data sets (COUT1, COUT15, COUT24, COUT36). More specific information about all child care providers, not just the primary provider, can be found in the caregiver data sets (CARGIV6, CARGIV15, CARGIV24, CARGIV36). Finally, we have added a data set that contains the composite variables created for many of the primary ECCRN papers (COMBINE3). An index of the primary types of variables and their variable names will be provided for each of these data sets.To assist the audience with using this data set, we will also provide a short SAS and SPSS program that reads the primary data sets, merges them together, and analyzes the primary 36 month outcomes to estimate the effect sizes associated with child care quality, parenting, and child care hours. These findings were presented at SRCD (ECCRN, 1999) and APA (ECCRN, 1999) during the past year. Results suggested that effect sizes associated with child care quality tended to be about half of those associated with parenting. Child care hours was not related to any of the outcomes, whereas child care quality was significantly related to the cognitive outcomes and parenting was related to both cognitive and social outcomes.