Trajectories of problem behaviors from 4 to 23 years in former preterm infants

Abstract

Premature infants have significant risk for later behavior problems. This study examined growth trajectories of three problem behaviors across five developmental age points from preschool to early adulthood in a well-characterized sample of premature infants. The effects of neonatal risk, gender, and socioeconomic context were modeled on these trajectories. The longitudinal sample was comprised of preterm infants (N = 160) with full variation of neonatal morbidity and birth weight (640–1,950 g). Trajectories of externalizing, internalizing and attention problem behaviors from 4 to 23 years, measured by the Child Behavior Checklist, were tested using latent growth curve modeling. The results indicate individual variation in the number of externalizing and internalizing problems over time. Externalizing problems were not significantly different for males and females, but male scores were consistently higher. Neonatal risk was significantly associated with higher internalizing problems at age 4, but was not predictive at school age and beyond. Attention-problem scores increased from early preschool through adolescence for males, but females had little change over the same ages. SES was not predictive of any problem behavior trajectories and no significant two-way interactions were found. The results advance understanding of stability and change of three important problem behaviors through preschool, childhood, and adolescence to young adulthood in prematurely born infants in order to inform clinicians about timely assessment and the refinement of effective interventions.

Keywords

developmental trajectory growth curve analysis neonatal risk preterm infants problem behaviors

Each year, 15 million infants worldwide (World Health Organization, 2013) and 1 in 10 (9.6%) US infants are delivered prematurely (Hamilton, Martin, Osterman, Curtin, & Mathews, 2015). Preterm infants are at risk for long-term neurological disabilities and later developmental difficulties, and the risk is increased when prematurity is coupled with neonatal illness (Anderson, 2014; Saigal, 2014). Compared to full-term peers, preterm infants who have suffered neonatal illnesses (e.g. septicemia, necrotizing enterocolitis, intraventricular hemorrhage, sepsis) have lower cognitive and academic achievement, poorer social competence and more problem behavior during childhood and early adolescence (Aarnoudse-Moens, Weisglas-Kuperus, van Goudoever, & Oosterlaan, 2009; Quigley et al., 2012; Winchester et al., 2009). Specific problem behaviors of externalizing (e.g., aggression, oppositional), internalizing (e.g., social withdrawal, depression, anxiety), and attention including ADHD, have been reported in children, adolescents and adults born prematurely (Bhutta, Cleves, Casey, Cradock, & Anand, 2002; Dahl et al., 2006; Farooqi, Hagglof, Sedin, Gothefors, & Serenius, 2007; Loe, Lee, Luna, & Feldman, 2011; Msall & Park, 2008; Spittle et al., 2009). Aylward (2005) estimated that subtle, “high prevalence/low severity dysfunctions,” including attention and behavior problems occur in 50–75% of very low birth weight (VLBW; 1,000–1,499 g) infants.

While term-born boys and girls differ in externalizing and internalizing behaviors, this has not been widely studied in preterm samples (Castelao & Kroner-Herwig, 2014; Dahl et al., 2006; Dekker et al., 2007; Keiley, Bates, Dodge, & Pettit, 2000; Samara, Marlow, & Wolke, for the EPICure Study Group, 2008). Less is known about preterm-adult outcomes, but recent reviews of the Canada cohort show mixed results (Saigal, 2014; Van Lieshout, Boyle, Saigal, Morrison, & Schmidt, 2015). The early and continuing physical, motor, and behavioral difficulties put former premature infants at risk for later socially appropriate adult behaviors across settings of family, peers, and work. There is also broadening recognition that preterm outcomes are not determined by neonatal factors alone and that the environmental context must be considered as well as the interplay of both biology and context (Conrad, Richman, Lindgren, & Nopoulos, 2010; Msall, Sullivan, & Park, 2010).

The majority of preterm problem-behavior research has used cross-sectional or short-term longitudinal designs. So, from a life course perspective, the research is limited as to whether behavior problems found in former premature toddlers and preschoolers improve, lessen, or are sustained into adolescence and adulthood. By examining trajectories of problem behaviors from preschool through childhood and into early adulthood, we can advance better understanding of at-risk outcomes that have been identified in short-term studies. Specifically, we can see how neonatal history affects the developmental course of problem behaviors reported in former premature infants, whether this differs between boys and girls, and whether socioeconomic context plays a role. Therefore, the purpose of this study was to examine growth trajectories of externalizing, internalizing, and attention-problem behaviors across five developmental age points from preschool to early adulthood in a well-characterized sample of premature infants. Secondly, we examined the effect of neonatal risk, gender, and socioeconomic context on these trajectories.

Problem behaviors

Problem behavior represents fundamental psychosocial processes that underlie behavior and shape (both positively and negatively) the course of development. The majority of developmental researchers agree on a classification of behavior problems associated with various aspects of children’s emotional, behavioral, and social maladjustment. Externalizing problem behaviors, for example, are characterized by disruptive and aggressive behaviors, such as rule-breaking, irritability, belligerence, and aggression towards others (Castelao & Kroner-Herwig, 2014; Fanti & Henrich, 2010). The number of externalizing problems that children exhibit has been shown to differ between genders in both preterm and full term studies, where boys exhibit more aggressive, rule-breaking behaviors than girls (Castelao & Kroner-Herwig, 2014; Dahl et al., 2006; Samara et al., 2008). In a study of children born full-term aged 9 to 14 years, three different groups of externalizing behavior trajectories (“low,” “moderate,” and “high-decreasing”) were modeled, with females comprising the majority of subjects in the “low” class and a larger percentage of males comprised the “moderate” and “high-decreasing” groups (Castelao & Kroner-Herwig, 2014). For the vast majority of healthy, full-term children, externalizing behaviors are expected to increase during early childhood and adolescence (Bongers, Koot, van der Ende, & Verhulst, 2004; Fanti & Henrich, 2010; Keiley et al., 2000).

Internalizing behaviors are characterized by emotional dysregulation, social withdrawal, depression, anxiety, and feelings of worthlessness or inferiority. High-internalizing problems are an important predictor of later psychosocial maladjustment affecting scholastic attainment, and career employment opportunities in adulthood (Hauser-Cram & Woodman, 2016; Masi, Muratori, Manfredi, Pisano, & Milone, 2015). Gender differences where females are more likely to exhibit internalizing behaviors, and males more likely to exhibit externalizing behaviors, has been observed in both term-born and preterm samples (Boyle et al., 2011; Dekker et al., 2007; Hack et al., 2005; Keiley et al., 2000). When compared, children born prematurely had higher internalizing behaviors than their term-born peers (Conrad et al., 2010; Farooqi et al., 2007; Spittle et al., 2009), but others reported no differences (Aarnoudse-Moens et al., 2009; Gurka, LoCasale-Crouch, & Blackman, 2010).

Attention problems are frequently reported in preterm follow-up studies with boys displaying excessive hyperactivity (Bhutta et al., 2002; Dahl et al., 2006; Samara et al., 2008). At age 11, the rate for attention problems for children born at less than 26 weeks’ gestation was 30% (OR: 3.5; p ≤ .007) (Farooqi et al., 2007). There are some reports that attention problems in premature samples are associated with cognitive and academic performance particularly for ELBW (extremely low birth weight; < 1000 g) groups (Farooqi et al., 2007; Loe et al., 2011) but this evidence is inconsistent (Conrad et al., 2010; Samara et al., 2008).

There is reason for concern when externalizing, internalizing, and attention problem behaviors are observed at a young age because a persistent life-course trajectory of behavioral problems is possible, leading to a greater prevalence of aggressive behavior and conduct problems in adulthood (Fanti & Henrich, 2010; Sterba, Prinstein, & Cox, 2007). Former premature infants are not a homogeneous group; they have a broad spectrum of growth, health, and developmental outcomes, and there is significant variability in the emergence, persistence, and consequences of behavior problems. In a meta-analysis of school age outcomes of prematurity, internalizing and externalizing behavior problems were found in 81% of the studies, while the pooled relative risk was more than twice for ADHD (RR 2.6, 95% CI 1.85–3.78) (Bhutta et al., 2002). The majority of research documenting the behavioral problems of premature infants has focused on the birth weight of ELBW and VLBW infants, excluding premature infants with higher birth weights. In the only study of behavior problems in larger, healthy, late preterm infants (34–36 weeks gestation) no differences were found in externalizing and internalizing behaviors from ages 4–15 when compared to a term-born group (Gurka et al., 2010). More internalizing behaviors have been found in small for gestational age (SGA) premature infants (Boyle et al., 2011). Also, there is limited examination of neonatal illness as a factor in behavior problems. However, Hack and colleagues (Hack et al., 2005) did not find that perinatal factors predicted internalizing behaviors in young women at age 20 years who were born VLBW.

Socioeconomic status (SES) seems to be related to psychopathology, where individuals with lower SES demonstrate greater problem behavior (van Oort, van der Ende, Wadsworth, Verhulst, & Achenbach, 2011; Wadsworth & Achenbach, 2005). In one US preterm study, higher social risk was associated with higher externalizing behaviors at age 2 years (Spittle et al., 2009). In school-age children aged 6, 11, and 17 years who were low birth weight (LBW; < 2,500 g), externalizing and internalizing behaviors were slightly elevated (adjusted odds ratios = 1.53 and 1.28, respectively) and attention problems were noted in the urban community only (adjusted odds ratio = 2.78; Bohnert & Breslau, 2008). In 7–11-year-old children born ELBW and VLBW, birth weight explained more variance in negative behaviors than SES, which was not significant (Conrad et al., 2010). Loe and colleagues (2011) followed a sample aged 9–16 years who were born preterm and found that SES was not significantly related to behavior problems.

The present study

The present study aimed to longitudinally examine the trajectories of externalizing, internalizing, attention problem behaviors from age 4–23 years and the impacts of neonatal acuity, gender, and SES in a sample of infants born preterm. Growth curve modeling (GCM), often used to assess change in individuals’ development across the entire lifespan (Nagin, 1999), was used to examine the growth trajectories for each problem behavior. The GCM approach allows the flexible specification of both fixed (i.e., average scores) and random effects (i.e., individual departures from the average scores) within a mixed-model framework to estimate the degree to which interdependencies among dimensions of behavior changed over several phases of development on the same children (Nagin & Tremblay, 2005; Singer & Willet, 2003).

First, we examine the functional form of time that best describes the developmental trajectories of problem behaviors for the sample of preterm infants. We expected that curvilinear growth in externalizing and internalizing scales would increase over time, and that attention-problem scores would be above the standardized mean and increase through adolescence. Next, the effects of neonatal acuity, gender, and SES on the intercept and slopes were modeled in order to determine whether the main effects and cross-level interactions between variables would have substantial effect at age 4 and the change in problem behaviors over time. We expected to find significant effects of neonatal acuity and SES on the three growth trajectories and a significant effect of gender for externalizing behavior.

Methods

Participants

A total of 160 premature infants were recruited during 1985–1989 while hospitalized after birth in a New England specialty hospital. The recruitment criteria were neonatal diagnoses, birth weight, gestational age, absence of maternal mental illness, and English as a primary language. All infants were less than 37 weeks’ gestational age, birth weight < 1,850 g and comprised four a priori groups with a full range of neonatal complications. The preterm groups were: (1) no medical or neurological illness; (2) medical illness including bronchopulmonary dysplasia (defined as oxygen requirement at 28 days of life), respiratory distress syndrome, necrotizing enterocolitis (Bell et al., 1978), and/or sepsis; (3) neurological illness such as Grade III & IV intraventricular hemorrhage (Papile, Burstein, Burstein, & Koffler, 1978), meningitis, and/or shunted hydrocephalus; or (4) were small for gestational age (SGA; birth weights below the 10th percentile for their gestational age with/without neonatal illness, see Lubchenco, Hansman, & Boyd, 1966). SES was equally distributed across preterm groups at recruitment. Fewer than 10% of the parents approached declined participation in the study.

Procedures

Infants and their parents were followed in a series of research studies at home and hospital laboratory at developmentally specific time points: birth, 18 months, 30 months (toddler age), and ages 4 (preschool age), 8 (school age), 12 (early adolescence), 17 (late adolescence), and 23 years (emerging adulthood). The neonatal hospital course and demographic information were collected at hospital recruitment. As part of a larger research protocol, demographics and age-appropriate behavior questionnaires were completed at ages 4, 8, 12, 17, and 23 years. Written informed consent was obtained from parents at recruitment and study participants at ages 17 and 23 years. University and hospital Institutional Review Boards approved each age point study.

Measures

Problem behaviors scales

The Child Behavior Checklist (CBCL; Achenbach, 1991; Achenbach & Rescorla, 2003) is designed to assess adaptive and function and behavior problems. The Adult Self-Report (ASR) (Achenbach & Rescorla, 2003) extends the measurement of the CBCL to ages 18–59. Both measures use a 3-point scale ranging from 0 (not true), 1 (somewhat true), and 2 (very true) for each behavior. Scores are reported as t scores (mean = 50, standard deviation = 10) with higher scores indicative of higher levels of behavior problems. For ages 4 to 17 years, reliability coefficients of .95 on behavior problems and .99 on social competence were reported. Test-retest reliability demonstrated a strong positive relationship with a mean of .90. Content, criterion-related, and construct validity were supported (Achenbach, 1991). For age 23 years, reliability was excellent: 1-week test-retest correlations were .80–.90 and 2 years was .69. Internal consistency ranged from .78 to .85. Discriminant validity was demonstrated in referred and non-referred samples (Achenbach & Rescorla, 2003).

Hobel neonatal risk

The medical record was the source for birth and neonatal data. The Hobel was designed to assess perinatal factors in the mother and neonate. There are 51 prenatal items, 40 intrapartum items, and 46 neonatal items, each weighted according to assumed risk, then summed. The Neonatal Risk Score was used in this study. Construct and predictive validity were established in a series of studies (Hobel, Hyvarinen, Okada, & Oh, 1973; Hobel, Youkeles, & Forsythe, 1979).

Hollingshead four-factor index of socioeconomic status

Social status was based on four domains: marital status, retired/employed status, educational attainment, and occupational prestige (Hollingshead, 1975). The parent education code is rated on a 7-point scale and the occupational code is rated on a 9-point scale. The total SES score was calculated at each age point. Since there were no significant differences in SES by group over time, SES at birth was used in the present study.

Gender

Infant gender was determined at birth and dummy coded (0 = male, 1 = female) for analysis.

Analysis Plan

Prior to analyzing the GCMs, Little’s missing data analysis (p > .05) was conducted in SPSS 23 to indicate whether the data were missing at random. Mplus 7.11 was used to assess the growth trajectories of externalizing, internalizing, and attention problem behaviors from ages 4 to 23 years using all available case data. The default method for handling missing data in Mplus is full information maximum likelihood (ML), however, in addition to using the default method, we employed multiple imputation across 100 datasets to ensure we obtained robust estimates of all the parameters and standard errors in our models.

The unconditional growth models were constructed in three stages. First, the null model, also called the baseline unconditional means model, estimated the average behavior problem scores at age 4 (i.e. intercept) for the externalizing, internalizing and attention problems scales. The unconditional means model included the random intercept parameter, without any other growth factors or covariates included in the null model. The second stage of models included the growth factors (i.e. linear and quadratic slope parameters) to estimate the average rate of change and average acceleration or deceleration of change over time. The correlations among repeated observations on the same individual were accounted for via random intercepts and slopes that were included in each model, but autocorrelations between adjacent time points were excluded from the models. The last stage included neonatal risk, gender, and SES covariates to investigate their effects on the growth parameters in the GCMs.

Model selection

The best-fitting unconditional models were evaluated using several model fit statistics (Hox & Stoel, 2005). We used several conventional cut-off criteria, the Comparative Fit Index (CFI), the root mean square error of approximation (RMSEA), and standardized root mean squared residual (SRMR), to assess the adequacy of the model used to represent the data. Models that met the cut-off criteria for CFI ≥ .90 (Hu & Bentler, 1999) and values ≤ .10 for the RMSEA and SRMR were retained (Kline, 2010). Additionally, information-based indices such as the Akaike Information Criterion (AIC) and Bayes Information Criterion (BIC) were used to evaluate the goodness-of-fit between two or more nested models, where the lowest AIC and/or BIC indicated the better-fitting model.

Results

Descriptive findings

The preterm infants in the sample (N = 160) had experienced a wide range of neonatal illnesses, including bronchopulmonary dysplasia, necrotizing enterocolitis, sepsis, intraventricular hemorrhage grades III/IV, and long hospital stays as shown in Table 1. Although sample size varied across time, sample retention was high from birth to ages 4 years (N = 131, 82%), 8 years (N = 139, 87%), 12 years (N = 131, 82%), 17 years (N = 128, 80%), and 23 years (N = 128, 80%). The mean age in months for each time point was 47.34 (SD = 1.89) at 4 years, 96.75 (SD = 3.90) at 8 years, 148.73 (SD = 6.18) at 12 years, 206.69 (SD = 4.42) at 17 years, and 277.42 (SD = 11.66) at 23 years. No differences were found between those who participated and those who dropped at age 23 on neonatal illness, SES, parent education, marital status, or race. Approximately 88% of the sample reported being White, 8% African-American, 2% more than one race, < 1% Native American, and < 1% Asian.

Table 1.

Sample characteristics for preterm groups at birth.

	Total	Healthy preterm	Medical preterm	Neurological preterm	Small for gestational age preterm	Male	Female	Test statistic	Effect size
Characteristics	N / M (SD)	n / M (SD)	n / M (SD)	n / M (SD)	n / M (SD)	n / M (SD)	n / M (SD)	F, t or X ² (df)	η² or Cramer’s V	p
Gender
Male	80	16	31	24	9			9.55 (3)^a*	0.24	0.02
Female	80	17	29	12	22
Race
White	114	19	43	26	26			13.075 (12)^a	0.18	0.36
African American	10	2	5	2	1
Native American	1	1	0	0	0
Asian	1	0	1	0	0
Mixed Race	3	2	0	1	0
SES
High	33	6	10	6	11			21.71 (12)^a*	0.22	0.04
Moderate High	41	5	21	7	8
Average	37	13	12	8	4
Moderate Low	26	4	6	10	6
Low	20	5	10	4	1
Neonatal Characteristics
Birthweight (gr) (range 640–1850)	1270.20 (327.95)	1518 (216.29)	1268.03 (312.37)	1148.75 (277.61)	1151.29 (371.85)	1318.13 (307.96)	1222.28 (342.02)	11.07 (156)^b***	0.17	0.00
								1.86 (158)⁁^c		0.06
Total Days Hospitalized (range 9–137)	53.59 (32)	31.12 (9.34)	53.70 (25.83)	69.25 (22.61)	59.13 (50.33)	52.84 (27.82)	54.35 (35.80)	10.11 (156)^b***	0.16	0.00
								−0.30 (158)⁁^c		0.77
Duration of Oxygen (hours) (range 0–3288)	439.01 (826.22)	12.61 (17.72)	478.93 (721.33)	741.03 (770.68)	464.90 (1260.20)	518.19 (805.83)	359.82 (843.72)	4.93 (156)b***	0.09	0.00
								1.21 (158)⁁^c		0.28
Gestational Age (weeks) (range 24–36)	30.04 (2.69)	31.45 (1.68)	29.33 (2.32)	28.25 (2.20)	31.97 (2.86)	29.99 (2.83)	30.09 (2.55)	20.68 (156)^b***	0.29	0.00
								−0.24 (158)⁁^c		0.82
Hobel Neonatal score (range 8–160)	86.13 (32.22)	55.74 (20.93)	90 (25.88)	113.97 (20.74)	78.66 (34.42)	88.91 (31.70)	83.35 (32.71)	30.47 (156)^b***	0.37	0.00
								1.09 (158)⁁^c		0.28
Sepsis (y/n)	19	0	6	9	4	13	6	10.61 (3)^a**	0.26	0.01
								2.926 (1)⁁^a		0.09
Necrotizing Enterocolitis (y/n)	17	0	9	5	3	7	10	5.57 (3)^a	0.19	0.14
								0.592 (1)⁁^a		0.44
Bronchopulmonary Dysplasia (y/n)	32	0	14	14	4	20	12	17.67 (3)^a***	0.33	0.00
								2.5 (1)⁁^a		0.11
Intraventricular Hemorrhage (y/n)
IVH Grade I/II	21	2	2	15	2	8	13	4.98 (3)^a*	0.41	0.03
IVH Grade III/IV	16	0	0	12	4	12	4

Note. ⁁ grouping variable by Gender: 1 = Male, 2 = Female. ^aChi Square statistic; ^b F statistic; ^c t statistic. + score. *p < .05; **p < .01; ***p < .001.

One-way ANOVAs demonstrated significant mean differences for birth weight, total days hospitalized, oxygen duration, gestational age, and Hobel Neonatal Risk between preterm groups (all p’s < .05; see Table 1). The independent samples t tests on the same neonatal variables indicated no significant differences by gender and non-significance between gender and sepsis, necrotizing enterocolitis, and bronchopulmonary dysplasia (Chi Square p’s > .05). However, the Chi Square demonstrated a statistically significant association between gender and IVH Grade I & II and IVH Grade III & IV, X ²(3) = 4.98, p < .05. There were more males (n = 13) than females (n = 4) who had IVH grade III or IV. Socioeconomic status did not differ across preterm groups, F (3,159) = 1.95, p = .12.

The unadjusted means for the externalizing, internalizing and attention t scores at each wave (ages 4, 8, 12, 17, 23 years) are shown in Table 2. We used multiple imputation and ML estimation strategies for handling missing data. There were small differences in parameter estimates between MI and ML methods that did not influence results, therefore, the ML parameter estimates were reported.

Table 2.

Problem behaviors subscale t scores* over time by gender.

		Total		Males		Females
Syndrome	Time	M	SD	M	SD	M	SD
Externalizing problems	Age 4	48.56	8.56	49.38	8.97	47.68	8.08
	Age 8	48.02	10.53	48.55	11.16	47.53	9.95
	Age 12	48.08	10.7	49.88	11.71	46.56	9.6
	Age 17	47.62	9.76	49.28	9.74	46.24	9.63
	Age 23	48.45	9.92	50.6	9.81	46.42	9.66
Internalizing problems	Age 4	45.39	7.9	46.44	8.07	44.25	7.62
	Age 8	50.27	9.63	51.79	9.71	48.85	9.4
	Age 12	50.2	10.78	51.22	10.25	49.34	11.2
	Age 17	50.73	10.96	50.69	8.75	50.77	12.66
	Age 23	50.04	11.58	47.82	11.18	52.12	11.65
Attention problems	Age 4	53.42	7.66	52.68	8.61	54.33	6.25
	Age 8	56.17	8.2	56.47	8.14	55.85	8.31
	Age 12	56.16	8.34	56.93	8.71	55.36	7.92
	Age 17	54.83	5.86	56.13	6.69	53.25	4.22
	Age 23	54.58	6.17	53.95	5.92	55.21	6.4

Note. Total N: 4 years = 131, age 8 years = 139, 12 years = 131, 17 years = 128, 23 years = 128. *T score (M = 50, SD = 10; range 31–100) higher scores indicate higher levels of behavior problems. Borderline range is 60–63 and clinical range is > 63.

GCM model results

We estimated the fit of the null model with the random intercept parameter and no other growth parameters included in the model. Table 3 contains the goodness of fit statistics for the GCMs for the three problem behavior scales. As expected, the intercept only model (i.e., null model) had poor fit to the data for all three scales (all CFI ≤ .75, RMSEA ≥ .13, SRMR ≥ .22). The linear slope was added to the GCMs and slightly improved model fit, but those models did not meet all the cut-off criteria for adequate model fit (all CFIs ≤ .79, RMSEA ≥ .13, SRMR ≥ .11). The quadratic slope parameter was included in the GCMs and provided the best fit to the data for all three problem behaviors (All CFIs ≥ .99, RMSEA ≤ .03, SRMR ≤ .07). Based on the change in the raw means (Table 2), and the excellent model fit, we elected to keep the quadratic change model for all three problem behavior scores (see Table 3). It is important to note that none of the models we evaluated included autocorrelations across any of the adjacent time points in the GCMs, which means that our models assume each time point was independent of the rest. Next, we tested the effects of gender, Hobel Neonatal Risk score, and SES on the intercept, and the linear and quadratic slope parameters in each GCM. The parameter estimates for the final models are reported in Table 4.

Table 3.

Unconditional GCM model fit statistics and model comparison results.

Syndrome	Model	CFI	RMSEA	SRMR	AIC	BIC
externalizing problems	1. Null model^a	0.02	0.24	0.27	4881.10	4899.55
	2. Linear change model	0.79	0.13	0.22	4782.36	4813.10
	3. Quadratic change model	1.00	0.00	0.03	4756.97	4800.02
internalizing problems	1. Null model^a	0.00	0.24	0.26	4938.89	4957.34
	2. Linear change model	0.72	0.13	0.11	4841.36	4872.41
	3. Quadratic change model	0.99	0.03	0.07	4817.34	4860.39
attention problems	1. Null model^a	0.00	0.05	0.10	4348.23	4366.37
	2. Linear change model	0.00	0.06	0.10	4352.01	4376.20
	3. Quadratic change model*	1.00	0.00	0.05	4346.56	4379.83

Note. ^a = Unconditional means model where only the fixed intercept is fit to the data and the variance is set to zero. All other models included a random intercept and/or slope estimates. * indicates intercept covariance was negative (not significant), thus the covariance was fixed to zero. Bold indicates the best unconditional growth model chosen for each problem behavior scale that met model fit criteria.

Table 4.

Parameter estimates for the final models (N = 160).

	Between-subjects				Within-subject
		95% C.I.				95% C.I.
Syndrome	Parameter	Estimates	Upper	Lower	Parameter (co)variances	Estimates (SE)	Upper	Lower
Externalizing problems	Intercept	50.96***	56.41	45.51	Intercept	51.773***	83.23	20.32
	Linear slope	1.73	8.30	−4.84	Linear slope	56.44***	90.19	22.69
	Quadratic slope	−0.413	1.19	−2.02	Quadratic slope	2.77***	4.61	0.93
	Hobel	0.02	0.06	−0.02	Intercept × linear slope	−18.44	10.12	−47.00
	SES	−0.087	0.01	−0.19	Intercept × quadratic slope	1.76	7.60	−4.08
	Gender	−1.23	1.61	−4.07	Linear slope × quadratic slope	−12.10**	−4.51	−19.69
	Hobel × linear slope	−0.006	0.05	−0.06
	Hobel × quadratic slope	0.002	0.02	−0.02
	SES × linear slope	−0.87	−0.77	−0.97
	SES × quadratic slope	0.008	0.03	−0.01
	Gender × linear slope	−0.25	3.16	−3.66
	Gender × quadratic slope	−0.12	0.72	−0.96
Internalizing problems	Intercept	44.20***	49.22	39.18	Intercept	29.14	59.01	−0.73
	Linear slope	7.13*	13.56	0.70	Linear slope	37.79*	69.74	5.84
	Quadratic slope	−1.44	0.17	−3.05	Quadratic slope	2.26*	4.16	0.36
	Hobel	0.04*	0.08	0.00	Intercept × linear slope	−8.78	18.23	−35.79
	SES	−0.04	0.08	−0.16	Intercept × quadratic slope	1.81	7.36	−3.74
	Gender	−1.75	0.84	−4.34	Linear slope × quadratic slope	−8.88*	−1.49	−16.27
	Hobel × linear slope	−0.004	0.05	−0.06
	Hobel × quadratic slope	0.001	0.01	−0.01
	SES × linear slope	−0.04	0.08	−0.16
	SES × quadratic slope	0.009	0.04	−0.02
	Gender × linear slope	−1.1	2.23	−4.43
	Gender × quadratic slope	0.62	1.46	−0.22
Attention problems	Intercept	52.35***	56.35	48.35	Intercept^a	0	0.00	0.00
	Linear slope	2.16	7.39	−3.07	Linear slope	8.18	18.59	−2.23
	Quadratic slope	−0.29	1.00	−1.58	Quadratic slope	0.7	1.62	−0.22
	Hobel	0.007	0.05	−0.03	Linear slope × quadratic slope	−2.35	0.65	−5.35
	SES	0.016	0.08	−0.05
	Gender	1.3	3.30	−0.70
	Hobel × Linear slope	−0.15	−0.11	−0.19
	Hobel × quadratic slope	0.002	0.01	−0.01
	SES × linear slope	0.07	0.15	−0.01
	SES × quadratic slope	−0.02	0.00	−0.04
	Gender × linear slope	−3.98**	−1.35	−6.61
	Gender × quadratic slope	0.96**	1.61	0.31

Note. ^a Intercept variance estimated at zero. *p < .05; **p < .01; ***p < .001.

Externalizing problem behaviors

Figure 1 shows the normative developmental trajectories of externalizing behavior problems for males and females. The final model AIC (4756.97) is smaller than the null model AIC (4763.63). The CFI (1.0) and RMSEA (0.0) indicate that the model fit was excellent. The shape of the growth curve was dependent on an intercept, two slope parameters, Hobel Neonatal risk, gender, and SES covariates. The initial value at age 4 is significantly different from zero (intercept = 50.96, p < .001). The normative developmental trajectory had no gender differences on the initial level. The linear (b = 1.73, ns) and quadratic slope (b = −0.413, ns) effect indicates that the normal developmental trajectory was not significantly different from zero. Neonatal risk (b = 0.02, ns) and SES (b = −0.087, ns) were not significant on the initial level or developmental trajectory, which indicates that gender, neonatal risk, or SES did not have a significant impact on the development of externalizing behavior. The simple variation in the growth parameters for different individuals across the total group are shown with the estimated (co)variances in Table 4. This provides information about the deviations in individual scores from the normative trajectory. There is significant variation in the initial level of externalizing problems (51.773, p < .001), and the linear (56.44, p < .001) and quadratic (2.77, p < .001) slopes are significant, which suggests that there is individual variation across the sample in the number of externalizing behaviors that develop over time (Table 4). The covariance between the linear and quadratic slope (−12.10, p < .01) is the only covariance that is significant, indicating that children who tend to change at a slower rate also show a stronger decline in externalizing problems as they got older.

Figure 1.

Developmental trajectories of externalizing problem behavior adjusted for gender, neonatal risk, and SES.

Internalizing problem behaviors

Figure 2 shows the normative developmental trajectories of internalizing behavior problems for males and females. The final model AIC (4817.34) is smaller than the null model AIC (4865.69). The CFI (.99) and RMSEA (0.03) of the model indicated that the model fit was excellent. The shape of the growth curve was dependent on an intercept, two slope parameters, Hobel neonatal risk, gender, and SES covariates. The initial value at age 4 is significantly different from zero (intercept = 44.20, p < .001). The normative developmental trajectory had no gender differences on the initial level (b = −1.75, ns). The linear (b = −1.10, ns) and quadratic slope (b = 0.62, ns) effect indicates that the normal developmental trajectory was not significantly different from zero. Neonatal risk (b = 0.04, p < .05) had a significant effect on initial level, but no effect was found for the linear or quadratic slopes, which indicated higher neonatal risk at birth was associated with higher internalizing problems at age 4. SES (b = −0.04, ns) was not significant on the initial level or developmental trajectory, which indicates that gender and SES did not have a significant impact on the development of internalizing behavior. The covariances of the linear (37.79, p < .05) and quadratic (2.26, p < .05) slopes were significant, which suggests that there is individual variation in the number of internalizing problems over time. The covariance between the linear and quadratic slope (−8.88, p < .05) is the only covariance that is significant, which indicates that children show a stronger decrease in the rate of change in internalizing problems as they grow older.

Figure 2.

Developmental trajectories of internalizing problem behavior adjusted for gender, neonatal risk, & SES.

Attention

Table 3 describes the model for the developmental trajectory for attention problems. The final model AIC (4346.56) is smaller than the null model AIC (4348.23). The CFI (1.0) and RMSEA (0.0) of the model indicate that the model fit was excellent. The shape of the growth curve was dependent on an intercept, two slope parameters, Hobel neonatal risk, gender, and SES covariates. Figure 3 shows the normative developmental trajectory of attention problems for males and females. The initial value for the intercept is significantly different from zero (intercept = 52.35, p < .001). The normative developmental trajectory had no significant gender (b = 1.30), neonatal risk (b = 0.007), or SES (b = 0.016) effect on the intercept (all p’s > .05). However, there was a significant gender effect on the linear (gender × linear slope = −3.98, p < .01) and quadratic slope (gender × quadratic slope = 0.96, p < .01), which indicates that the normative developmental trajectory is different for females and males. The linear and quadratic slope effect indicate that the normative developmental trajectory first shows an average increase to age 12 and thereafter shows a decrease with age in males, however the females’ trajectory shows an average decrease and thereafter shows a small increase with age (Figure 3). The simple variation in the growth parameters for different individuals across the premature sample are shown with the estimated (co)variances of the individual growth parameters in Table 4. As these estimates were small and not significant, the intercept variance was fixed to zero (Chen, Bollen, Paxton, Curran, & Kirby, 2001), indicating that there was no significant variation in the growth parameters among individuals in the initial level of attention problems at age 4. The linear (b = 8.18, ns) and quadratic slope (b = 0.70, ns) variances were not significant, nor were the covariances between the slopes (linear × quadratic slope = −2.35, ns), which indicate no significant differences in the rate of change for children who start at a higher level of attention problems than those who start at a lower level.

Figure 3.

Developmental trajectories of attention problem behavior adjusted for gender, neonatal risk, & SES.

Discussion

Early problem behaviors such as externalizing and internalizing behaviors are not uncommon in typically developing children, but when they persist portend later difficulties with peers, asocial function, academic performance, and risky behaviors (Aarnoudse-Moens et al., 2009; Bongers et al., 2004; Fanti & Henrich, 2010; Quigley et al., 2012). The examination of problem-behavior trajectories in term-born children has enabled greater understanding of the stability and change over time in order to advance effective interventions (Keiley et al., 2000). To that end, researchers have reported that the normative trajectories of externalizing and internalizing behavior problems tend to change in frequency and expression over time, and these trajectories are influenced by gender, SES, and parenting factors (Bongers et al., 2004; Castelao & Kroner-Herwig, 2014). In contrast, children born prematurely have higher risk for a range of neuropsychological outcomes including attention, externalizing, and internalizing behavior problems. However, we could not find any research on the longitudinal trajectories of behavior problems in premature infants from early childhood to adulthood.

In cross-sectional and short-term follow-up studies, preterm infants have been found to experience more behavior problems throughout childhood and adolescence (Aarnoudse-Moens et al., 2009; Gurka et al., 2010; Samara et al., 2008). In a meta-analysis of 1,759 school-age children born preterm, 81% (13 of 18) of the studies reported increased externalizing and internalizing behaviors (Bhutta et al., 2002). Commonly, follow-up studies on preterm children have two predictors of neonatal morbidity, gestational age or birth weight (Gurka et al., 2010; Hack et al., 2009; Loe et al., 2011). Our sample included a full range of prematurity in birth weight (640–1,950 g), neonatal illness (healthy, medical, neurological illnesses, SGA), and all levels of SES to evaluate the developmental trajectories of externalizing, internalizing, and attention problems from preschool to adulthood. We found individual variation within each trajectory. Neonatal risk had a significant effect on internalizing behavior at age 4, and gender was significant in trajectories of attention problems. Contrary to expectations, SES was not significant in any trajectories. Thus, our results offer new insights with a diverse premature sample into the change and stability of three problem behaviors from childhood to early adulthood that have been identified in premature samples.

Externalizing problem behaviors

Although the mean t scores were close to the standardized mean, gender differences were not significant on initial status or on the slopes in the GCM. Visually, males had higher externalizing problems on average at age 4 and this trajectory remained consistently higher than for females through age 23 (Figure 1). Females had slight decreases in externalizing behaviors over the same period. For those who tend to change, there was a stronger decline with age. Contrary to our expectations, SES and neonatal risk had no impact on the growth curves. These findings are in line with research of typically developing children showing that trajectories of externalizing behavior problems tend to gradually increase for males, while decreases are expected for females (Bongers et al., 2004). In contrast, externalizing behaviors for children with disabilities reveal group trajectories where ¾ of the children showed stability in low and moderate externalizing behavior levels during childhood, and ¼ had high levels which increased through childhood then leveled at young adulthood (Hauser-Cram & Woodman, 2016).

Internalizing problem behaviors

We found similar trajectories for males and females for internalizing behavior problems with males reporting slightly higher problems until age 12, when internalizing behaviors decreased steadily for males and increased for females to age 23 (Figure 2). Neonatal risk had a significant effect at age 4, that is, neonatal illness due to prematurity had a significant impact on internalizing problems at preschool age. The significant negative covariance between the linear and quadratic slopes suggest that the rate of change slows in internalizing problems with age. Dekker’s (Dekker et al., 2007) Dutch study of trajectories of depression in typically developing children from ages 4 to 18, found one group trajectory for both boys and girls to be increasing toward a high level at adolescence. In another group trajectory, similar to the results in our study, girls differed from boys in that boys’ depressive scores were at normative levels at late childhood to mid-adolescence whereas girls’ scores did not decrease. This differs from some reports of early and persistent internalizing behavior problems in premature samples. Birth weight was a strong predictor of inattention and internalizing (anxiety/depression) behaviors in two US studies of children aged 7–16 who were born prematurely (Conrad et al., 2010; Loe et al., 2011) and a meta-analysis (Aarnoudse-Moens et al., 2009). Our results also showed this effect in that birth weight is one indicator in the Hobel Neonatal Risk score (Hobel et al., 1973).

Attention problem behaviors

The mean attention t scores were higher than the standardized mean by 5 points on average for both boys and girls born prematurely over the five age points (Table 2). The developmental trajectory for attention showed a curvilinear increase for males from age 4 to 23 years, and a small curvilinear decrease for females (Figure 3). Females had slightly higher levels of attention problems at preschool-age but were stable through childhood and adolescence to age 23. For males, attention problems peaked during early adolescence then declined in late adolescence and by adulthood, men reported slightly fewer attention problems than women. This differs from the findings of Conrad and colleagues (Conrad et al., 2010) comparing ELBW and VLBW infants to ages 7–16 where no gender differences were found, only birth weight was significant in models of attention behaviors. In a total population study of boys and girls born ELBW in UK and Ireland, pervasive attention problems were found at ages 5–7 that were unrelated to cognitive deficits (Samara et al., 2008). Likewise, in a meta-analysis of nine behavioral studies of very preterm (< 32 weeks’ gestation)/VLBW infants’ inattention problems were more pronounced than internalizing behaviors (Aarnoudse-Moens et al., 2009). In the single trajectory analysis of LBW infants compared to normal controls at ages 7, 11, and 17, more attention problems were found (adjusted odds ratio 2.78) in urban locations, but not for those in suburban locations (Bohnert & Breslau, 2008).

Socioeconomic status

An unexpected finding was that SES was not a significant effect in any of the three problem behavior trajectories. In typically developing children, lower SES has been associated with more problem behaviors (Keiley et al., 2000; van Oort et al., 2011; Wadsworth & Achenbach, 2005). Our findings add to the growing evidence that SES play a lesser role in later problem behaviors in premature infants. Similar to our results, SES was not significant in a pooled sample aged 7–16 who were VLBW and ELBW (Conrad et al., 2010). Similarly, in a small preterm sample of 63 infants with wide birth weight variation (range 482–2,495 g), SES was not significant in models predicting internalizing and externalizing behaviors at ages 9–16 (Loe et al., 2011). In a study of internalizing behaviors of former preterm, now 20-year-old, women, an inverse relationship was found, however their composite measure was comprised of maternal education, marital status, and race, which makes comparison difficult (Hack et al., 2005).

Strengths and limitations

This prospective, longitudinal study enabled an examination of problem-behavior trajectories in premature infants from preschool to early adulthood. Typically, preterm samples are characterized by birth weight or gestational age alone, our heterogeneous sample represented a broad spectrum of birth weight and neonatal morbidity, equivalent genders and full range of SES, thus advancing knowledge of the development of problem behaviors to a broader group. We do acknowledge the limited racial and ethnic diversity of the sample that reflect the geographic distribution of southern New England at the time of sample recruitment. Recent preterm birth demographics show higher incidence in non-White races. Studies show persistent health disparities in birth outcomes between White Americans and African Americans which may set in motion developmental trajectories whereby early life experiences of stress lead to cumulative allostatic load over the life course (Lu & Halfon, 2003). More research of premature infants of racial and ethnic diversity is needed.

Repeated measures of parent-reported child behaviors using the widely known CBCL scales is a strength of this study that allows comparison with other premature samples, as well as term-born and clinical samples. Similar to other studies cited here, we used parent-report except at age 23 when self-report was used. Some studies have combined parent and teacher report of the CBCL scales instead of relying on one informant (e.g., Bohnert & Breslau, 2008). However, Keiley (Keiley et al., 2000) kept the two respondents separate in their study and noted that the correlation between parent and teacher is moderate at best and varies with other factors. Ideally, we would have liked to use multiple informants, but decided that the consistency of parent-report over time was an advantage. Also, the consistency of parent-report is especially important because parents are often the initiators of mental health services for their children (Choudhury, Pimentel, & Kendall, 2003). It was also possible that the younger child would not be able to distinguish behavior and feelings as an older adolescent would (Langley, Bergman, & Piacentini, 2002).

The results from this study should be interpreted in light of several other limitations. Although sample attrition was minimal (80% retention at 23 years), over 30 patterns of missing data could have biased the results. To counteract this issue, we used ML and MI estimation strategies in Mplus to preserve all available information to reduce the bias in the parameter estimates and standard errors. We reported the ML estimates, but the MI method helped ensure that our conclusions drawn from the default GCM analysis represented the most likely parameter estimates.

SES was used as a proxy for the ecological context of behavioral development, which could be considered inadequate given the importance of developmental context (Boyle et al., 2011; Conrad et al., 2010; Capaldi, Pears, Kerr, Owen, & Kim, 2012; Treyvaud et al., 2012). The complete evaluation of health outcomes requires consideration beyond the neonatal factors of birth weight and neonatal morbidity to consider contextual factors known to influence health and development (Msall et al., 2010).

Since we considered a broader spectrum of prematurity, the apparent medical risks associated with low birth weight, gestational age, and higher levels of perinatal acuity in some people, may have been masked by others who were at less risk or not at all at risk. Thus, our models may lack some precision when predicting which infants will experience more aggressive, rule-breaking behaviors or emotional problems. Taken as a whole, our findings revealed that neonatal risk, gender, or SES has little impact on the developmental trajectories of three problem behaviors for a wide spectrum of prematurity.

Conclusion

This study sought to increase our understanding of the development of three problem behaviors, externalizing, internalizing, and attention, which have been identified in premature samples. To our knowledge, this was the first study to model the development of problem behaviors from early childhood into early adulthood, and examine effects of neonatal risk, gender, and SES. Our findings revealed that former premature infants may experience fewer problem behaviors during childhood, although attention-problem scores were above the norm with male trajectories different from female trajectories. SES was not a factor in the trajectories suggesting that proximal contextual factors should be explored. Greater understanding of stability and change of three important problem behaviors in prematurely born children can begin to inform clinicians about timely assessment and refinement of effective interventions.

Footnotes

Acknowledgements

The authors wish to thank the participating parents and children for their long-term commitment to the research project.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the National Institutes of Health National Institute of Nursing Research R01 003695.

References

Aarnoudse-Moens

C. S. H.

Weisglas-Kuperus

van Goudoever

J. B.

Oosterlaan

(2009). Meta-analysis of neurobehavioral outcomes in very preterm and/or very low birth weight children. Pediatrics, 124, 717–728.

Achenbach

T. M.

(1991). Manual for the Child Behavior Checklist/4–18 and 1991 profile. Burlington: University of Vermont, Department of Psychiatry.

Achenbach

T. M.

Rescorla

L. A.

(2003). Manual for the ASEBA Adult forms & Profiles: For ages 18-59. Burlington: University of Vermont, Department of Psychiatry.

Anderson

P. J.

(2014). Neuropsychological outcomes of children born very preterm. Seminars in Fetal & Neonatal Medicine, 19, 90–96.

Aylward

(2005). Neurodevelopmental outcomes of infants born prematurely. Journal of Developmental and Behavioral Pediatrics, 26(6), 427–440.

Bell

M. J.

Ternberg

J. L.

Feigin

R. D.

Keating

J. P.

Marshall

Barton

Brotherton

(1978). Neonatal necrotizing enterocolitis: Therapeutic decisions based on clinical staging. Annals of Surgery, 187, 1–7.

Bhutta

A. T.

Cleves

M. A.

Casey

P. H.

Cradock

M. M.

Anand

K. J. S.

(2002). Cognitive and behavioral outcomes of school-aged children who were born preterm: A meta-analysis. Journal of the American Medical Association, 288(6), 728–737.

Bohnert

K. M.

Breslau

(2008). Stability of psychiatric outcomes of low birth weight. Archives of General Psychiatry, 65(9), 1080–1086.

Bongers

I. L.

Koot

H. M.

van der Ende

Verhulst

F. C.

(2004). Developmental trajectories of externalizing behaviors in childhood and adolescence. Child Development, 75(5), 1523–1537.

10.

Boyle

M. H.

Miskovic

Van Lieshout

Duncan

Schmidt

L. A.

Hoult

(2011). Psychopathology in young adults born at extremely low birth weight. Psychological Medicine, 41, 1763–1774.

11.

Capaldi

D. M.

Pears

K. C.

Kerr

D. C. R.

Owen

L. D.

Kim

H. K.

(2012). Growth in externalizing and internalizing problems in childhood: A prospective study of psychopathology across three generations. Child Development, 83(6), 1945–1959.

12.

Castelao

C. F.

Kroner-Herwig

(2014). Developmental trajectories and predictors of externalizing behavior: A comparison of girls and boys. Journal of Youth and Adolescence, 43, 775–789.

13.

Chen

Bollen

K. A.

Paxton

Curran

P. J.

Kirby

J. B.

(2001). Improper solutions in structural equation models. Sociological Methods and Research, 29(4), 468–508.

14.

Choudhury

M. S.

Pimentel

S. S.

Kendall

P. C.

(2003). Childhood anxiety disorders: Parent-child (dis)agreement using a structured interview for the DSM-IV. Journal of the American Academy of Child and Adolescent Psychiatry, 42, 957–964.

15.

Conrad

Richman

Lindgren

Nopoulos

(2010). Biological and environmental predictors of behavioral sequelae in children born preterm. Pediatrics, 125(1), e83–89.

16.

Dahl

L. B.

Kaaresen

P. I.

Jorunn Tunby

Handegård

B. H.

Kvernmo

Rønning

J. A.

(2006). Emotional, behavioral, social, and academic outcomes in adolescents born with very low birth weight. Pediatrics, 118(2), e449–e459.

17.

Dekker

M. C.

Ferdinand

R. F.

van Lang

N. D. J.

Bongers

I. L.

van der Ende

Verhulst

F. C.

(2007). Developmental trajectories of depressive symptoms from early childhood to late adolescence: Gender differences and adult outcome. Journal of Child Psychology and Psychiatry, 48(7), 657–666.

18.

Fanti

K. A.

Henrich

C. C.

(2010). Trajectories of pure and co-occurring internalizing and externalizing problems from age 2 to age 12: Findings form the National Child Health and Human Development Study of Early Child Care. Developmental Psychology, 46(5), 1159–1175.

19.

Farooqi

Hagglof

Sedin

Gothefors

Serenius

(2007). Mental health and social competencies of 10- to 12-year-old children born at 23 to 25 weeks of gestation in the 1990s: A Swedish national prospective follow-up study. Pediatrics, 120(1), 118–133.

20.

Gurka

M. J.

LoCasale-Crouch

Blackman

J. A.

(2010). Long-term cognition, achievement, socioemotional, and behavioral development of healthy late-preterm infants. Archives of Pediatric and Adolescent Medicine, 164(6), 525–532.

21.

Hack

Taylor

H. G.

Schluchter

Andreias

Drotar

Klein

(2009). Behavioral outcomes of extremely low birth weight children at age 8 years. Journal of Developmental & Behavioral Pediatrics, 30(2), 122–130.

22.

Hack

Youngstrom

E. A.

Cartar

Schluchter

Taylor

G. H.

Flannery

D. J.

… Borawski

(2005). Predictors of internalizing symptoms among very low birth weight young women. Developmental and Behavioral Pediatrics, 26(2), 93–104.

23.

Hamilton

B. E.

Martin

J. A.

Osterman

M. J. K.

Curtin

S. C.

Mathews

T. J.

(2015). Births: Final Data for 2014. National Vital Statistics Reports (Vol. 64). Hyattsville, MD: National Center for Health Statistics.

24.

Hauser-Cram

Woodman

A. C.

(2016). Trajectories of internalizing and externalizing behavior problems in children with developmental disabilities. Journal of Abnormal Child Psychology, 44, 811–821.

25.

Hobel

C. J.

Hyvarinen

Okada

D. M.

(1973). Prenatal and intrapartum high-risk neonate. American Journal of Obstetrics & Gynecology, 117, 1–9.

26.

Hobel

C. J.

Youkeles

Forsythe

(1979). Prenatal and intrapartum high-risk screening. II. Risk factors reassessed. American Journal of Obstetrics and Gynecology., 135(8), 1051–1056.

27.

Hollingshead

A. B.

(1975). Four Factor Index of Social Status. New Haven, CT: Yale University Press.

28.

Hox

Stoel

R. D.

(2005). Multilevel and SEM approaches to growth curve modeling. Encyclopedia of Statistics in Behavioral Science, 3, 1296–1305.

29.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1–55.

30.

Keiley

M. K.

Bates

J. E.

Dodge

K. A.

Pettit

G. S.

(2000). A cross-domain growth analysis: Externalizing and internalizing behaviors during 8 years of childhood. Journal of Abnormal Child Psychology, 28(2), 161–179.

31.

Kline

R. B.

(2010). Principles and practice of structural equation modeling (3rd ed.). New York: Guilford Press.

32.

Langley

A. K.

Bergman

Piacentini

J. C.

(2002). Assessment of childhood anxiety. International Review of Psychiatry, 14, 102–113.

33.

Loe

I. M.

Lee

E. S.

Luna

Feldman

H. M.

(2011). Behavior problems of 9–16 year old preterm children: Biological, sociodemographic, and intellectual contributions. Early Human Development, 87, 247–252.

34.

M. C.

Halfon

(2003). Racial and ethnic disparities in birth outcomes: A life-course perspective. Maternal and Child Health Journal, 7, 13–30.

35.

Lubchenco

L. O.

Hansman

Boyd

(1966). Intrauterine growth in length and head circumference as estimated from live births at gestational ages from 26 to 42 weeks. Pediatrics, 37(3), 403–408.

36.

Masi

Muratori

Manfredi

Pisano

Milone

(2015). Child behaviour checklist emotional dysregulation profiles in youth with disruptive behavior disorders: Clinical correlates and treatment implications. Psychiatry Research, 225(1–2), 191–196.

37.

Msall

M. E.

Park

J. J.

(2008). The spectrum of behavioral outcomes after extreme prematurity: Regulatory, attention, social and adaptive dimensions. Seminars in Perinatology, 32, 42–50.

38.

Msall

M. E.

Sullivan

M. C.

Park

J. J.

(2010). Pathways of risk and resilience after prematurity: Role of socioeconomic status. In Nostari

Murray

Hack

(Eds.), Neurodevelopmental outcomes of preterm birth from childhood to adult life (pp. 224–236). Cambridge, UK: Cambridge University Press.

39.

Nagin

D. S.

(1999). Analyzing developmental trajectories: A semiparametric, group-based approach. Psychological Methods, 4, 139–157.

40.

Nagin

D. S.

Tremblay

R. E.

(2005). What has been learned from group-based trajectory modeling? Examples from physical aggression and other problem behaviors. The Annals of the American Academy of Political and Social Science, 602, 82–117.

41.

Papile

L. A.

Burstein

Koffler

(1978). Incidence and evaluation of subependymal and intraventricular hemorrhage: A study of infants with birth weights less than 1,500 gm. Journal of Pediatrics, 92(4), 529–534.

42.

Quigley

M. A.

Poulsen

Boyle

Wolke

Field

Alfirevic

Kurinczuk

J. J.

(2012). Early term and late preterm birth are associated with poorer school performance at age 5 years: A cohort study. Archives of Disease in Childhood Fetal Neonatal Edition, 97, F167–F173.

43.

Saigal

(2014). Functional outcomes of very premature infants into adulthood. Seminars in Fetal & Neonatal Medicine, 19, 125–130.

44.

Samara

Marlow

Wolke

, for the EPICure Study Group. (2008). Pervasive behavior problems at 6 years of age in a total-population sample of children born ≤ 25 weeks gestation. Pediatrics, 122, 562–573.

45.

Singer

J. D.

Willet

J. B.

(2003). Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. New York: Oxford University Press.

46.

Spittle

A. J.

Treyvaud

Doyle

L. W.

Roberts

Lee

K. J.

Inder

T. E.

… Anderson

P. J.

(2009). Early emergence of behavior and social–emotional problems in very preterm infants. Journal of the American Academy of Child and Adolescent Psychiatry, 48(9), 909–918.

47.

Sterba

S. K.

Prinstein

M. J.

Cox

M. J.

(2007). Trajectories of internalizing problems across childhood: Heterogeneity, external validity, and gender differences. Development and Psychopathology, 19, 345–366.

48.

Treyvaud

Inder

T. E.

Lee

K. J.

Northam

E. A.

Doyle

L. W.

Anderson

P. J.

(2012). Can the home environment promote resilience for children born preterm in the context of social and medical risk? Journal of Experimental Child Psychology, 112, 326–337.

49.

Van Lieshout

R. J.

Boyle

M. H.

Saigal

Morrison

Schmidt

L. A.

(2015). Mental health of extremely low birth weight survivors in their 30s. Pediatrics, 135(3), 452–459.

50.

van Oort

F. V. A.

van der Ende

Wadsworth

M. E.

Verhulst

F. C.

Achenbach

T. M.

(2011). Cross-national comparison of the link between socioeconomic status and emotional and behavioral problems in youths. Social Psychiatry and Psychiatric Epidemiology, 46, 167–172.

51.

Wadsworth

M. E.

Achenbach

T. M.

(2005). Explaining the link between low socioeconomic status and psychopathology: Testing two mechanisms of the social causation hypothesis. Journal of Consulting and Clinical Psychology, 73, 1146–1153.

52.

Winchester

S. B.

Sullivan

M. C.

Marks

A. K.

Doyle

DePalma

McGrath

M. M.

(2009). Academic, social, and behavioral outcomes at age 12 of infants born preterm. Western Journal of Nursing Research, 31, 853–871.

53.

World Health Organization. (2013). Preterm birth (Fact sheet No. 363). Retrieved from http://www.who.int/mediacentre/factsheets/fs363/en/index.html