SupplementalMaterials - American Psychological Association

advertisement
1
Supplemental Materials
“Child Care and Cortisol Across Early Childhood: Context Matters,”
by D. Berry et al., 2013, Developmental Psychology.
http://dx.doi.org/10.1037/a0033379
Measurement Details
Child-Care Quality
We fitted a longitudinal confirmatory factor analysis (CFA) model across the four
measurement points. The residual variances for common indicators were freely estimated over
time, as were the covariances between the latent factors. The model fit the data well (χ2 = 89.41,
df = 73, p =.09; CFI = .96, RMSEA = .02). All the factor loadings were statistically significant
and in the expected direction, with the standardized loadings ranging between .54 and .85. Each
latent variance was statistically significant, and rank-order stability over time was moderate; the
standardized covariances between the factors over time ranged between .38 and .54. We
aggregated these data by estimating the factor scores at each point, standardizing, and averaging
across time points.
Cumulative Risk
When the target child was 7, 15, 24, and 36 months of age, family income-to-needs ratio
was measured by taking reported total household income and dividing by the federal poverty
threshold for family size for the given year. Rank-order stability in family income was strong
over time (r = .73 to .79, p < .001). Household caregiver education was based on the highest
level of education attained by either the primary or secondary caregiver (when available)
reported at 7, 15, 24, and 36 months and was scaled as raw years of schooling. Rank-order
stability over this period was strong (r = .90 to .98, p < .001). Across this same period, primary
caregivers were considered to have a partner, if (a) they were married with the spouse living in
the home or (b) they designated a romantic partner or relative (e.g., grandparent of target child)
2
as living in the home and helping with the child. Rank-order stability in partnering between
adjacent epoch was also quite strong (r = .77 to .92, p < .001).
Average weekly hours of work were based on primary caregiver reports of her and the
secondary caregiver’s (if present) weekly work hours at 7, 15, 24, and 36 months, and showed
moderate rank-order stability between 7 and 36 months (r = .41 to .49, p < .001). Occupational
prestige was scored using the National Opinion Research Center (NORC) coding system (Nakao
& Treas, 1994). Higher prestige jobs are scaled as larger numbers. There was substantial rankorder stability in prestige levels across the four measurement periods (r = .76 to .80, p < .001).
Household density was calculated based on caregiver reports regarding the number of
individuals living in the home, divided by independent observer ratings of the number of rooms
in the household. Larger numbers reflect more densely populated households. Independent
observers also rated neighborhood noise and neighborhood safety during these home visits.
Safety was scored on a 1 (“obviously dangerous”) to 4 (“above average safety”) Likert-type
scale. Noise was reverse scored as 1 (“very loud”) to 4 (“very quiet”). Neighborhood safety and
noise showed strong positive correlations (r = .51 to .61, p < .001) and were subsequently
averaged. Rank-order stability in both household density and neighborhood noise/safety over
time ranged from moderate to strong (r = .43 to .82, .47 to .80, p < .001, respectively).
Finally, primary caregiver sensitivity was based on observer ratings of positive parenting
during a semi-structured play task at 7, 15, 24, and 36 months of age (see Cox, Paley, Burchinal,
& Payne, 1999). In this task, the primary caregiver and infant were presented with a set of
developmentally appropriate toys, and caregivers were asked to play with their infants as they
would normally. Primary caregiver behavior was recorded and subsequently coded for levels of
sensitivity, detachment, intrusiveness, positive regard, negative regard, and animation during
their interactions with the child. Each was rated on a 5-point Likert-type scale (1 = not at all
characteristic, 5 = highly characteristic). Informed by prior work with these data (e.g., Blair et
al., 2011a), we created a positive parenting composite at each time point by taking the average of
the primary caregiver’s ratings across the sensitivity, animation, positive regard, and detachment
(reversed). Two independent raters rated approximately 30% of the caregiver-child interactions;
the intraclass correlations for the positive parenting composite ranged between .87 and .91.
3
Rank-order stability in positive parenting was rather strong across the four measurement periods
(r = .55 to .65, p < .001).
The cumulative risk index was calculated by first averaging scores within a given risk
over time. As shown in Table S1, each of the cumulative risk factors showed moderate to rather
strong correlations with each other, such that higher levels of risk (i.e., lower levels positively
framed variables) were associated with higher risk in the other variables. We then conducted a
principal components analysis (PCA) across the eight items. There was clear evidence that the
data were best represented by a single component; only one component showed an eigenvalue
greater than 1 (i.e., 3.62) and rose substantially above the scree, accounting for 45% of the
variability across the items. The PCA loadings were in the expected direction, and ranged from .52 (household density) to .83 (family income). Internal-consistency reliability across the items
was reasonable ( = .82). We created a continuous cumulative risk index by reverse-scoring the
positively framed variables, standardizing each risk measure, and averaging across the risks.
Control Covariates
Infant traits. Child gender and race were dummy coded, such that male and African
American children were each coded as 1.
Temperament. Infant temperament was measured using the estimated factor scores from
a CFA-based composite measure comprising three measures (four scales) of behavioral reactivity
at 7 months of age: (a) observer ratings of infant behavior during a home visit, using the
Irritability scale of an adapted version of the Infant Behavior Record (IBR; Bayley, 1969; Stifter
& Corey, 2001); (b) observer ratings of infant reactivity during three moderately stressful tasks
(see Blair et al., 2008); and (c) primary caregiver reports of infant behavior across the Distress to
Novelty and Distress to Limitations scales of the Infant Behavior Questionnaire (Rothbart,
1981). More detailed descriptions of the model are available from the first author.
Mental development. Infant mental development at 7 months of age was measured using
the Mental Development Index (MDI) scores on the Bayley Scales of Mental Development
(Bayley, 1969). Higher scores represent more developmentally advanced abilities.
4
Caregiver/family traits. When the child was 2 months of age, primary-caregiver reading
ability was measured using the KFAST literacy screener (Kaufman & Kaufman, 1994), a
standardized measure of functional literacy. Higher scores indicate better reading skills.
Caregiver age. The primary caregiver’s age at the birth of the target child as well as her
age when she had her first child were included as covariates. Each was scaled in years.
Subsidy receipt. Dummy variables representing whether the primary caregiver received
Aid to Families with Dependent Children (AFDC) funding or Medicaid funding were also
included in the model. For each, receipt of funding is coded as 1.
Research site. The NC site was dummy coded as 1.
Measurement traits. Time-of-day of the saliva collection was coded on a 24-hour clock
at both 7 and 48 months of age. Time since sleep/nap, average amount of sleep per night, and
amount of sleep the night prior to the saliva collection were each controlled and scaled in hours.
5
Table S1
Zero-Order Correlations and Factor Loadings From Principal Components Analysis of
Variables Included in the Cumulative Risk Index
Variable
Family income
Education
Work hr
Occ. prestige
Constant
partner
Density
Safety
Pos. parent
Income Education
Work
hr
Occ.
prestige
Partner
—
.56**
.57**
.56**
.50**
—
.42**
.58**
.41**
—
.32**
.40**
—
.36**
—
.41
.47**
.48**
.39**
.39**
.51**
.24**
.32**
.37**
.28**
.39**
.40**
.16**
.35**
.48**
Density
p < .10.
PCA
loading
.833
.782
.645
.725
.661
—
.25**
.31**
Note. Variables are on original scales, prior to being reflected to represent risk.
**
Safety
—
.33**
.524
.624
.712
Download