Behavior Problems in Low-Income Young Children Screened in Pediatric Primary Care

Key Points

Question

What are the trajectories of emotional and behavioral adjustment among children 2 to 6 years of age visiting pediatric primary care and what variables are associated with differential courses?

Findings

Of 4 trajectory groups identified in this cohort study of 15 218 children, 3 reflected clinically elevated levels of emotional and behavioral problems over time. Relative to the nonelevated group, children in the elevated groups were more likely to be male, White, receive public insurance, and have a social need and a caregiver with depression.

Meaning

Differential patterns of trajectories over time underscore the importance of early identification of emotional and behavior problems in pediatric primary care; mitigation of social needs and depression among caregivers may lead to improved trajectories.

Abstract

Importance

Screening of behavior problems in young children in pediatric primary care is essential to timely intervention and optimizing trajectories for social-emotional development. Identifying differential behavior problem trajectories provides guidance for tailoring prevention and treatment.

Objective

To identify trajectories of behavior problems in children 2 to 6 years of age screened in pediatric primary care.

Design, Setting, and Participants

This retrospective cohort study identified trajectories of behavior problems and demographic and clinical correlates. Data were collected as part of routine care in 3 pediatric primary care offices and 3 school-based health centers in Ohio serving a primarily low-income population. In total, 15 218 children aged 2 to 6 years with well-child visits between July 13, 2016, and January 31, 2022, were included.

Exposure

Caregivers completed the Strengths and Difficulties Questionnaire (SDQ) at annual well-child visits.

Main Outcomes and Measures

Trajectory groups were identified using latent growth mixture modeling of SDQ total difficulties scores, and relative risk ratio (RRR) of various demographic (eg, race) and clinical (eg, depression in caregiver) variables were assessed by multinomial logistic regression analysis.

Results

Of 15 281 children (51.3% males), 10 410 (68.1%) were African American or Black, 299 (2.0%) were Asian, 13 (0.1%) were American Indian or Alaska Native, 876 (5.7%) were multiracial, 26 (0.2%) were Native Hawaiian and Other Pacific Islander, 2829 (18.5%) were White, and 39 (0.02%) were categorized as other. In addition, 944 (6.2%) identified as Hispanic and 14 246 (93.2%) as non-Hispanic. Four behavior problem trajectory groups reflecting severity were identified: low-stable (LS; 10 096 [66.1%]), moderate-decreasing (MD; 16.6%), low-increasing (LI; 13.1%), and high-increasing (HI; 4.3%). Relative to the LS group, patients in each elevated group were more likely to be male (HI RRR, 1.87 [95% CI, 1.55-2.26]; MD RRR, 1.55 [95% CI, 1.41-1.71]; and LI RRR, 1.94 [95% CI, 1.70-2.21]), White (HI RRR, 2.27 [95% CI, 1.83-2.81]; MD RRR, 1.28 [95% CI, 1.13-1.45]; and LI RRR, 1.54, [95% CI, 1.32-1.81]), publicly insured (HI RRR, 0.49 [95% CI, 0.28-0.84]; MD RRR, 0.56 [95% CI, 0.43-0.73]; and LI RRR, 0.50 [95% CI, 0.35-0.73]), have a social need (HI RRR, 3.07 [95% CI, 2.53-3.73]; MD RRR, 2.02 [95% CI, 1.82-2.25]; and LI RRR, 2.12 [95% CI, 1.84-2.44]), and have a caregiver with depression (HI RRR, 1.66 [95% CI, 1.38-2.00]; MD RRR, 1.44 [95% CI, 1.31-1.58]; and LI RRR, 1.39 [95% CI, 1.23-1.58]). Relative to the LI group, patients in the MD group were less likely to be male (RRR, 0.80; 95% CI, 0.68-0.93).

Conclusions

The substantial portion of young children with increased behavior problems observed in this cohort study underscores the need for screening in pediatric primary care. Caregivers with depression and family social needs warrant prioritization in early prevention and treatment to alter elevated trajectories.

This cohort study identifies trajectories of behavior problems and variables associated with differential trajectories as assessed during pediatric primary care well-child visits among children 2 to 6 years of age.

Introduction

Approximately 20% of children 17 years or younger experience a psychiatric disorder each year.^1,2 This worsened during the COVID-19 pandemic, with increases in anxiety and depression.³ Most psychiatric disorders begin in childhood and continue into adulthood.⁴ Early recognition of emotional and behavioral problems is essential to intervene and mitigate worsening trajectories.⁵ The American Academy of Pediatrics recommends regular screening of youth in primary care using measures of behavioral health to facilitate identification and intervention.⁶

Pediatric primary care is an ideal location for screening given the common utilization of pediatric services, an emphasis on prevention, and the high level of trust that caregivers have in pediatricians.⁷ Several measures of behavioral adjustment have been recommended.⁸ The Strengths and Difficulties Questionnaire (SDQ) is a widely used behavioral and emotional health screening tool.⁹ Psychometrics of the SDQ have been examined, including with low-income and marginalized populations.^10,11 Given the desire of pediatricians to detect early manifestations of behavior problems, the total difficulties score of the SDQ is most relevant, as it captures the full range of symptoms that may warrant intervention.¹²

Pediatric primary care also provides an opportunity for tracking behavioral problems longitudinally. The universality of primary care permits population estimates of prevalence rates and trajectories of behavior problems. Several studies examining trajectories of emotional and behavioral problems have included children 6 years or older, had small sample sizes, and focused on specific types of behavior problems rather than broad indices of behavioral adjustment.^13,14 That research shows that a substantial proportion of children have behavior problems early in life, and many problems have a deteriorating course over time.^13,14 To our knowledge, no study has examined trajectories of behavior problems in young children assessed in pediatric primary care.

Given the need to better understand the course of emotional and behavioral problems in young children visiting pediatric primary care, this study used the SDQ to track caregiver-reported behavioral health among children 2 to 6 years of age screened during well-child visits. We sought to delineate trajectories over time and to identify risk factors with a particular focus on social needs and caregiver depression.

Methods

Setting and Participants

This cohort study was conducted in 3 pediatric primary care practices and 3 school-based health centers affiliated with an academic medical center in Ohio. The practices were in urban and suburban settings serving primarily low-income patients. Caregivers completed the SDQ on a tablet during annual well-child visits for their child ages 2 to 6 years as part of routine care. Participants included all children aged 2 to 6 years of all caregivers who completed the SDQ between July 13, 2016 (the date at which regular SDQ screening began), and January 31, 2022. Some children in the sample had their first SDQ screening at age 2 years, whereas other children had their first screening between 3 and 6 years of age. The study was approved by the Institutional Review Board of Cincinnati Children’s Hospital Medical Center, which waived the requirement for obtaining informed consent. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline was used in preparing this article.

Measures

The SDQ consists of 25 items that yield a total difficulties score and 5 subscale scores (emotional symptoms, conduct problems, hyperactivity, peer problems, and prosocial behavior). Each item is endorsed using a 3-point scale reflecting the presence of the behavior (not true, somewhat true, certainly true). We used total difficulties scores (range, 0-40, with higher values indicating greater severity of problems) as they reflect the broad expression of emotional and behavioral problems that are of clinical concern in the pediatric setting.⁹ Two versions of the SDQ were used: the version for children aged 2 to 4 years was used for children aged 2 to 3 years, and the version for children aged 4 to 10 years was used for children aged 4 to 6 years. The 2 versions differed on 3 items that overlapped in content but used slightly different language to reflect developmental differences. Cutoff values were used to classify children’s scores into the following categories: close to average, slightly raised, high, and very high.

Demographic and clinical information were collected as part of routine documentation accessed from the electronic health record. These included patient sex (male or female), race (African American or Black, American Indian or Alaska Native, Asian, multiracial, Native Hawaiian and Other Pacific Islander, White, or other [defined as a racial category different from the aforementioned options that was unavailable in the documentation system]), ethnicity (Hispanic or non-Hispanic), gestational age, insurance type (Medicaid or self-pay vs private), caregiver spoken language (English or other), and caregiver depression scores on the Patient Health Questionnaire 2.¹⁵ The questionnaire items were scored as yes or no through October 1, 2017, and afterward on a 4-point scale reflecting the frequency of each symptom (anchors of “not at all” and “nearly every day”). A positive screen was recorded if caregivers endorsed at least 1 yes in the first scoring system and had a score of 3 or more in the second system. We also included caregiver responses to routine health-related social needs documenting the presence of at least 1 of 5 needs (food insecurity, problems accessing benefits, financial strain, housing instability, and household trauma). A positive screen for caregivers with depression or the presence of a social need at any point from age 2 to 6 years was documented as an occurrence in data analyses.

Statistical Analysis

Latent growth mixture modeling (LGMM) was used to identify subgroups of patients with different trajectories from 2 to 6 years of age. Other longitudinal methods, such as conventional growth models, assume that individuals come from a single population and that a single trajectory can adequately summarize the entire population. However, a single trajectory would be an oversimplification of the complex growth pattern that characterizes social and behavioral development in this population. The use of LGMM can accommodate missing data at any time point using full information maximum likelihood estimations, allowing us to define trajectories for the full sample.

Model estimation proceeded in 3 steps. First, an unconditional latent growth model was estimated to determine the overall form of change over time (linear or quadratic). Second, LGMMs were estimated for 2 to 6 trajectory groups. The following 6 criteria were used to identify the number of trajectory groups: (1) the smallest Bayesian Information Criterion (BIC); (2) entropy, with values close to 1.0 denoting excellent fit; (3) class total count no less than 1%; (4) high posterior probabilities for each group (close to 1.0); (5) significant likelihood ratio tests (the Vuong-Lo-Mendell-Rubin likelihood ratio test, the Lo-Mendell-Rubin adjusted likelihood ratio test, and the bootstrap likelihood ratio test, with a significant test favoring the model with a larger number of groups); and (6) substantive interpretation of the groups and clinical usefulness.

After establishing the number of groups from the LGMM, we created a variable that assigned each patient to their most likely trajectory group membership. Then, multinomial logistic regression was used to identify the sociodemographic and clinical characteristics that distinguished among the latent trajectory groups. The group showing the best outcomes over time was selected as the reference category and was compared with the other groups, and then the reference class was changed until all groups were compared. To control for possible inflated type I error rates in the analysis, we used the Benjamini-Hochberg false discovery rate procedure to calculate a corrected overall critical P value.¹⁶ Statistical significance was defined as a 2-sided P < .05 or a 95% CI excluding 1. All analyses were performed using Stata, version 17 (StataCorp) and Mplus, version 8.8 (Muthén & Muthén).

Results

Study Population

During the study period, 15 281 patients between the ages of 2 and 6 years attended an annual well-child visit at least once, and their caregivers completed an SDQ at that visit. Table 1 shows the included patients’ characteristics. Caregivers were predominantly mothers, although greater specificity was not available in the electronic health record. The children were predominantly male (51.3% vs 48.7%); 10 410 (68.1%) were African American or Black, 299 (2.0%) were Asian, 13 (0.1%) were American Indian or Alaska Native, 876 (5.7%) were multiracial, 26 (0.2%) were Native Hawaiian and Other Pacific Islander, 2829 (18.5%) were White, and 39 (0.02%) were categorized as other, and 14 456 (94.6%) had public insurance or were self-pay. Mean (SD) gestational age was 38.15 (2.35) weeks (13.2% were premature). During the study period, at least 1 social risk factor was reported by 21.3% of caregivers, and 41.7% of caregivers screened positive for depression. In total, 7831 patients had an SDQ score for 1 visit between 2 and 6 years of age, 4879 had 2 visits with SDQ scores, 2117 had 3 visits with SDQ scores, 384 had 4 visits with SDQ scores, and 70 had SDQ scores at all 5 visits between 2 and 6 years of age. Of 7831 patients who had only a single time point, 2698 had data at age 2 years, 918 had data at age 3 years, 944 had data at age 4 years, 1298 had data at age 5 years, and 1973 had data at age 6 years. Overall, there were 5751 observations at age 2 years, 4988 observations at age 3 years, 5627 observations at age 4 years, 5060 observations at age 5 years, and 4400 observations at age 6 years.

Table 1. Demographics and Clinical Features of 15 281 Included Participants.

Variable	Participants, No. (%)
Child’s sex
Male	7843 (51.3)
Female	7438 (48.7)
Child’s race
African American or Black	10 410 (68.1)
American Indian or Alaska Native	13 (0.1)
Asian	299 (2.0)
Native Hawaiian and Other Pacific Islander	26 (0.2)
White	2829 (18.5)
Multiracial	876 (5.7)
Othera	3 (0.002)
Unknown or missing	825 (5.4)
Child’s ethnicity
Hispanic	944 (6.2)
Non-Hispanic	14 246 (93.2)
Missing	91 (0.6)
Child’s gestational age, mean (SD), wk	38.15 (2.35)
Parent’s spoken language
English	14 564 (95.3)
Other	716 (4.6)
MIssing	1 (0.1)
Insurance type
Public or self-pay	14 456 (94.6)
Private	778 (5.1)
Missing	47 (0.3)
PHQ-2 elevatedb
Yes	6373 (41.7)
No	8908 (58.3)
Social riskb
Yes	3259 (21.3)
No	12 022 (78.7)

Abbreviation: PHQ-2, Patient Health Questionnaire 2.

^a Other race is defined as a racial category different from the listed options that was unavailable in the documentation system.

^b Derived from caregiver report at any administration from 2 to 6 years.

Model Selection and Characterization of Trajectories

The LGMM with 4 groups was selected over the models with 2, 3, 5, or 6 groups to achieve a balance between model fit, parsimony, and clinical relevance. Specifically, model fit indicated that the 5-group model was preferable to the 6-group model. However, the fit between the 4- and 5-group models was similar, and the 4-group model was more clinically interpretable and parsimonious. The 4-group model produced relatively distinct SDQ trajectories, including 4.3% of the cases in the smallest group. The posterior probability of membership for each allocated group ranged between 0.62 and 0.87. Overall model fit statistics for each of the models are presented in Table 2. The final model parameters are based on 60 random start values and 100 iterations.

Table 2. Model Fit and Classification Information for the Latent Growth Mixture Models of 2 to 6 Trajectory Groups.

Model	BIC	P value			Entropy	Trajectory group probabilities	Lowest total group percentage
		VLMRa	LMR-LRTa	BLRTa
2 Groups	157 125	<.001	<.001	<.001	0.77	0.85-0.95	14
3 Groups	156 626	<.001	<.001	<.001	0.71	0.75-0.90	6
4 Groups	156 282	<.001	<.001	<.001	0.66	0.62-0.87	4
5 Groups	156 146	.003	.003	<.001	0.67	0.60-0.85	1
6 Groupsb	156 037	.14	.14	<.001	0.69	0.64-0.86	<1

Abbreviations: BIC, Bayesian information criterion; VLMR, Vuong-Lo-Mendell-Rubin likelihood ratio test; LMR-LRT, Lo-Mendell-Rubin adjusted likelihood ratio test; BLRT, bootstrap likelihood ratio test.

^a Tests of the k-1 group vs k-group solution.

^b Nonpositive definite latent variable covariance matrix in this model due to a negative slope variance, leading to an improper solution.

Using the total difficulties score of the SDQ, 4 trajectories emerged, with 33.9% of children being members of groups in which elevated scores were found over time. Final group counts (and percentages) based on the estimated model indicated that the largest trajectory group, termed low-stable (LS), included 10 096 (66.1%) patients who had low SDQ scores initially (mean intercept, 6.38) that remained low and slightly decreasing over time (mean slope, −0.44). The second largest group, termed moderate-decreasing (MD), included 2529 patients (16.6%) who had higher SDQ scores initially (mean intercept, 14.95]) but scores decreased over time (mean [SD] slope, −1.12). The third trajectory, termed low-increasing LI), included 2002 (13.1%) patients who initially had lower SDQ scores (mean intercept, 8.12), and SDQ scores increased over time (mean slope, 2.37). The final group, termed high-increasing (HI), included 654 (4.3%) patients who had high SDQ scores initially (mean intercept, 20.86) that remained high and slightly increasing over time (mean slope, 0.76). The intercept variance was 4.94 (P < .001) and the slope variance was 0.21 (P = .04), both of which were held constant across the 4 trajectory groups. The Figure shows the observed mean SDQ total difficulties scores over time by group. Table 3 gives the observed means and SDs of scores for each trajectory group by age. Mean total difficulties scores for each of the groups with elevated levels were above the normative cutoff values at some point across their trajectories.

Figure. Observed Mean Strengths and Difficulties Questionnaire (SDQ) Total Difficulties Scores in 4 Trajectory Groups Over Time.

Table 3. Descriptive Statistics for SDQ Total Difficulties Scores for the Full Sample and 4 Trajectory Groups at Ages 2 to 6 Years.

Age, y	Total	Trajectory Group
		Low-stable	Moderate-decreasing	Low-increasing	High-increasing
2 a
No.	5751	4339	1024	237	151
Mean (SD)	8.62 (5.31)	6.56 (3.45)	15.55 (2.98)	7.72 (3.17)	22.12 (4.70)
3 a
No.	4988	3650	844	317	177
Mean (SD)	8.38 (5.83)	5.86 (3.51)	15.26 (3.75)	10.82 (4.54)	23.05 (4.18)
4 b
No.	5627	4071	896	471	189
Mean (SD)	7.91 (5.65)	5.34 (3.32)	13.48 (3.52)	13.51 (3.52)	22.78 (3.74)
5 b
No.	5060	3540	731	594	195
Mean (SD)	8.14 (6.15)	5.09 (3.17)	11.92 (3.37)	16.43 (4.23)	24.21 (3.67)
6 b
No.	4400	2917	653	658	172
Mean (SD)	8.51 (6.57)	4.78 (3.15)	11.41 (2.46)	18.01 (2.88)	24.35 (3.98)

Abbreviation: SDQ, Strengths and Difficulties Questionnaire.

^a Classification cutoff scores: close to average, 0 to 12; slightly raised, 13 to 15; high, 16 to 18; very high, 19 to 40.

^b Classification cutoff scores: close to average, 0 to 13; slightly raised, 14 to 16; high, 17 to 19; very high, 20 to 40.

Sociodemographic and Clinical Characteristics of Patients In the 4 Groups

Table 4 gives the results of multinomial logistic regressions identifying the sociodemographic and clinical characteristics that differentiated the trajectory groups. The corrected overall critical P value for these analyses was P = .03. Compared with patients in the LS trajectory group, patients assigned to the HI trajectory group were more likely to have younger gestational age (relative risk ratio [RRR], 0.93; 95% CI, 0.90-0.96), identify as White (RRR, 2.27; 95% CI, 1.83-2.81), and be male (RRR, 1.87; 95% CI, 1.55-2.26). In addition, HI group members more often had a caregiver endorse 1 social risk (RRR, 3.07; 95% CI, 2.53-3.73), were publicly insured or self-pay (RRR, 0.49; 95% CI, 0.28-0.84), and had a caregiver screen positive for depression (RRR, 1.66; 95% CI, 1.38-2.00).

Table 4. Multinomial Logistical Regression Results Showing Differentiation of Trajectory Groups by Sociodemographic and Clinical Variables.

Variable	Comparison group
	High-increasing		Moderate-decreasing		Low-increasing
	RRR (95% CI)	P valuea	RRR (95% CI)	P valuea	RRR (95% CI)	P valuea
Reference group, low-stable
Gestational age	0.93 (0.90-0.96)	<.001b	0.95 (0.93-0.96)	<.001b	0.98 (0.95-1.01)	0.12
Ethnicity (0, non-Hispanic) vs 1, Hispanic	0.66 (0.37-1.19)	.17	0.91 (0.68-1.21)	0.50	0.71 (0.46-1.08)	0.12
Race (0, Black)
vs 1, Other	1.11 (0.76-1.61)	.60	1.25 (1.05-1.48)b	.01b	1.22 (0.96-1.55)	.10
vs 2, White	2.27 (1.83-2.81)	<.001b	1.28 (1.13-1.45)	<.001b	1.54 (1.32-1.81)	<.001b
Sex (0, female) vs 1, male	1.87 (1.55-2.26)	<.001b	1.55 (1.41-1.71)	<.001b	1.94 (1.70-2.21)	<.001b
Caregiver spoken language (0, English) vs 1, other	0.84 (0.43-1.62)	.60	1.32 (0.99-1.75)	.05	0.78 (0.49-1.25)	0.31
Positive social screen (0, no) vs 1, yes	3.07 (2.53-3.73)	<.001b	2.02 (1.82-2.25)	<.001b	2.12 (1.84-2.44)	<.001b
Insurance (0, public or self-pay) vs 1, private	0.49 (0.28-0.84)	.01b	0.56 (0.43-0.73)	<.001b	0.50 (0.35-0.73)	<.001b
Positive caregiver depression screen (0, no) vs 1, yes	1.66 (1.38-2.00)	<.001b	1.44 (1.31-1.58)	<.001b	1.39 (1.23-1.58)	<.001b
Reference group, low-increasing
Gestational age	0.95 (0.91-0.98)	.01b	0.96 (0.94-0.99)b	.03b
Ethnicity (0, non-Hispanic) vs 1, Hispanic	0.93 (0.46-1.86)	.83	1.27 (0.78-2.05)	.33
Race (0, Black)
vs 1, Other	0.90 (0.58-1.38)	.65	1.02 (0.78-1.33)	.89
vs 2, White	1.47 (1.14-1.89)	.003b	0.83 (0.69-0.99)	.05
Sex (0, female) vs 1, male	0.96 (0.77-1.20)	.76	0.80 (0.68-0.93)	.004b
Caregiver spoken language (0, English) vs 1, other	1.07 (0.49-2.33)	.87	1.68 (1.01-2.79)	.04
Positive social screen (0, no) vs 1, yes	1.45 (1.15-1.82)	.001b	0.95 (0.81-1.12)	.57
Insurance (0, public or self-pay) vs 1, private	0.97 (0.51-1.85)	.94	1.12 (0.73-1.71)	.61
Positive caregiver depression screen (0, no) vs 1, yes	1.18 (0.96-1.47)	.12	1.03 (0.89-1.19)	.68
Reference group, moderate-decreasing
Gestational age	0.98 (0.95-1.02)	.27
Ethnicity (0, non-Hispanic) vs 1, Hispanic	0.73 (0.39-1.36)	.33
Race (0, Black)
vs 1, Other	0.88 (0.59-1.32)	.55
vs 2, White	1.77 (1.41-2.24)	<.001b
Sex (0, female) vs 1, male	1.21 (0.98-1.47)	.07
Caregiver spoken language (0, English) vs 1, other	0.64 (0.32-1.26)	.20
Positive social screen (0, no) vs 1, yes	1.52 (1.23-1.87)	<.001b
Insurance (0, public or self-pay) vs 1, private	0.87 (0.48-1.57)	.65
Positive caregiver depression screen (0, no) vs 1, yes	1.15 (0.94-1.40)	.16

Abbreviation: RRR, relative risk ratio.

^a Nominal α level adjusted to .032 with the Benjamini-Hochberg false discovery rate procedure.

^b Statistically significant after adjustment using Benjamini-Hochberg false discovery rate procedure.

Compared with patients in the LS group, patients in the MD trajectory group were more likely to be younger gestational age (RRR, 0.95; 95% CI, 0.93-0.96), identify as White (RRR, 1.28; 95% CI, 1.13-1.45), identify as a racial category other than Black or White (RRR, 1.25; 95% CI, 1.05-1.48), and be male (RRR, 1.55; 95% CI, 1.41-1.71). In addition, MD group members more often had a caregiver endorse 1 social risk (RRR, 2.02; 95% CI, 1.82-2.25) and screen positive for depression (RRR, 1.44; 95% CI, 1.31-1.58). They were less likely to have private insurance (RRR, 0.56; 95% CI, 0.43-0.73).

Compared with patients in the LS group, patients in the LI trajectory group were more likely to be White (RRR, 1.54; 95% CI, 1.32-1.81) and male (RRR, 1.94; 95% CI, 1.70-2.21). They more often had a caregiver endorse 1 social risk (RRR, 2.12; 95% CI, 1.84-2.44) and were more likely to have a caregiver screen positive for depression (RRR, 1.39; 95% CI, 1.23-1.58). In addition, they were less likely to have private insurance (RRR, 0.50; 95% CI, 0.35-0.73).

We also compared the 3 elevated groups with one another. Compared with the LI trajectory group, patients in the HI trajectory group were more likely to be younger gestational age (RRR, 0.95; 95% CI, 0.91-0.98), White (RRR, 1.47; 95% CI, 1.14-1.89), and have a caregiver endorse at least 1 social risk (RRR, 1.45; 95% CI, 1.15-1.82). Compared with patients in the LI group, patients in the MD trajectory group were more likely to be lower gestational age (RRR, 0.96; 95% CI, 0.94-0.99) and less likely to be male (RRR, 0.80; 95% CI, 0.68-0.93). Patients in the HI group were more likely to be White than Black race (RRR, 1.77; 95% CI, 1.41-2.24) and have a caregiver endorse 1 social risk (RRR, 1.52; 95% CI, 1.23-1.87) compared with patients in the MD group.

Discussion

This cohort study derived trajectories of behavior problems among low-income children aged 2 to 6 years screened in pediatric primary care. Demographic and clinical variables were associated with elevated trajectories and distinct courses of increasing, decreasing, and stable behavior problems over time. The findings of this cohort study underscore the prevalence of behavior problems in young children in pediatric primary care, and the importance of recognizing and acting on the varied courses of adjustment starting at 2 years of age.

White race was associated with each of the elevated trajectories relative to the LS group. Moreover, contrasts between elevated groups indicated that White race was often associated with more severe levels of behavior problems. Although the SDQ has been used with different races and ethnicities, there are few studies with low-income Black or African American and White families with young children assessed in pediatric primary care.¹⁷ It is possible that Black or African American and White caregivers diverge in their interpretation of similar behaviors, and that their perceptions of what is problematic behavior may be different.^18,19 It is also possible that Black or African American caregivers are reluctant to report concerns about their child’s behavior for fear of the response of health care professionals (eg, may result in reports to child protective services as there is disproportionately higher contact among Black or African American families compared with White families due to structural racism).²⁰ Our findings underscore the need for increased cultural sensitivity and understanding of Black or African American families and other historically marginalized populations in pediatric primary care both in terms of clinical practice and research.^21,22

Being male was associated with the 3 elevated trajectories compared with the LS trajectory group. Males were also more represented in the HI vs MD groups although being female was associated with membership in the MD vs LI groups. The overrepresentation of males among children with clinically elevated behavior problems is well documented.²³ Females were more likely to be in the only trajectory group in which there was a reduction in clinical severity from ages 2 to 6 years. There is a pressing need for additional research to elucidate the contextual, racial, and ethnic drivers of caregiver reports of child behavior.

A positive screen for caregiver depression was associated with each of the elevated trajectories. Although based on a 2-item screen, this finding is consistent with extensive evidence documenting the deleterious association of caregiver depression with behavioral functioning in young children.²⁴ Depression impedes acquisition and expression of nurturing parenting behaviors.²⁵ As a result, children may struggle to develop self-regulation and cognitive capacities necessary for optimal behavioral development.²⁶ Although depressive symptoms in caregivers were associated with elevated trajectories, these symptoms did not differentiate between more severe trajectory groups. Our findings support the value of depression screening among caregivers in pediatric primary care.²⁷ Linkage of caregivers with depression to evidence-based treatment can facilitate recovery and has the potential to prevent behavioral problems in children and subsequent development of mental health disorders.²⁸

Public insurance or self-pay status and a positive social screen, reflecting social and financial challenges, were also associated with elevated trajectories. Positive social screens, indicating health-related social needs, such as food insecurity and housing instability, differentiated each elevated trajectory relative to LS or less severe trajectories except for MD vs LI groups. The link between poverty and child behavior problems is well-established.²⁹ Financial and social stressors interfere with the ability to provide effective and nurturing parenting, disrupting developmental processes that are associated with healthy child behavioral adjustment. Supporting families with social needs, through public policy and providing linkages to services in pediatric primary care, may prevent negative behavioral trajectories in young children.^30,31

Lower gestational age was associated with increased SDQ score trajectories in which there was a mean elevated score at 2 years of age. Prematurity and low birth weight contribute to the development of behavior problems early in life, and this finding may be a reflection of the 13.2% of the children in the present study with premature birth.³² Interventions and treatment focused on promoting healthy pregnancy and full-term birth are important efforts to prevent emotional and behavioral problems in young children.³³

Strengths and Limitations

The study has several strengths, including a large sample size. The sample was derived from pediatric primary care offices representing a diverse group of low-income children, permitting generalization of findings to an important underresourced and historically marginalized population. The SDQ is a widely used measure that has been examined internationally, thus adding to a large empirical base for comparison. This study has limitations. Missing data occurred due to inconsistent attendance for well-child visits and point in time sampling in which some children were not assessed because systematic measurement had not yet been implemented. Risk factors for trajectories were limited to those found in electronic health records data. Studies using other measures of behavior may yield different findings. Although the SDQ has been validated with children aged 2 to 6 years, some studies have found that developmental invariance is limited for specific symptom types in younger ages, necessitating caution in interpreting findings.^9,34 A brief screen was used to characterize depression among caregivers, and an in-depth assessment may provide a more comprehensive picture of clinical severity. The SDQ relied on caregiver reports, and other sources of information on child behavior were not available. Generalization of findings to other populations and settings is limited.

Conclusions

The results of this cohort study indicated that a substantial proportion of young children from low-income families visiting pediatric primary care have elevated levels of emotional and behavioral health problems across 2 to 6 years of age. For children with behavior problems at 2 years of age, intervention may prevent increasing severity over time. Screening, referral, and treatment of depression in caregivers is important and may reverse negative trajectories of child behavior. Similarly, greater attention to family social needs is imperative. Prevention and addressing social needs during these early ages has the potential to reduce disparities, improve behavior, and enhance the lives of children.

Supplement.

Data Sharing Statement