Physical Abuse,Emotional Abuse,and Neglect and Childhood Behavior Problems: A Meta-Analysis of Studies in Mainland China

Abstract

The relationship between three types of child maltreatment, including physical abuse, emotional abuse and neglect, and childhood behavior problems in Mainland China, has not been systematically examined. This meta-analysis reviewed findings from 42 studies conducted in 98,749 children in Mainland China and analyzed the pooled effect sizes of the associations between child maltreatment and childhood behavior problems, heterogeneity in study findings, and publication bias. In addition, this study explored cross-study similarities/differences by comparing the pooled estimates with findings from five existing meta-analyses. Equivalent small-to-moderate effect sizes emerged in the relationships between the three types of maltreatment and child externalizing and internalizing behaviors, except that emotional abuse related more to internalizing than externalizing behaviors. Considerable heterogeneity exists among the 42 studies. Weak evidence suggests that child gender and reporter of emotional abuse may moderate the strengths of the relationships between child maltreatment and behavior problems. No indication of publication bias emerged. Cross-study comparisons show that the pooled effect sizes in this meta-analysis are about equal to those reported in the five meta-analyses conducted in child and adult populations across the world. Findings urge relevant agencies in Mainland China to build an effective child protection system to prevent child maltreatment.

Keywords

child maltreatment behavior China meta-analysis cross-study comparison

Child maltreatment, especially physical abuse, emotional abuse, and neglect, is highly prevalent in Mainland China. Researchers estimated that in China, one in three Chinese experienced physical abuse, one in five experienced emotional abuse, and two in five experienced neglect in their childhood (Fang et al., 2015; Ji & Finkelhor, 2015). In particular, the prevalence of physical abuse is higher in Mainland China (40.6%) than the estimated prevalence in other Chinese societies (19.5%; Ji & Finkelhor, 2015) and the pooled estimate globally (17.7%; Stoltenborgh, Bakermans-Kranenburg, IJzendoorn, & Alink, 2013).

Accumulating evidence shows that child maltreatment by parents, guardians, or family members relates to aggressive and violent behaviors, depression, anxiety, and suicide in Chinese societies (for reviews, see Dunne, Chen, & Wan Yuen, 2008; Fry, McCoy, & Swales, 2012; Gershoff, 2010; Kwok, Chai, & He, 2013). However, existing systematic reviews or meta-analyses of the relationship between child maltreatment and behavioral outcomes in the Chinese context are limited in several ways. First, they focused on studies conducted in child and adult populations together and failed to differentiate between them. Childhood behavior problems are a major risk factor for adulthood crime, violence, and mental disorders (Leschied, Chiodo, Nowicki, & Rodger, 2008; Reef, Diamantopoulou, van Meurs, Verhulst, & van der Ende, 2011). Investigation of child maltreatment–related childhood behavior problems is informative for prevention of behavior problems at an earlier life stage. Second, whether different types of child maltreatment relate to different behavior problems equally or differently is still unclear. Third, many possible factors that could contribute to the heterogeneity in relationships between child maltreatment and behavior problems have not been analyzed in the existing meta-analyses. For example, emerging evidence shows that child and parent gender matter in the relationship between child maltreatment and behavior problems (Cui, Deatrick, & Liu, 2018; Xing & Wang, 2013; Xing, Wang, Zhang, He, & Zhang, 2011); however, whether child or parental gender moderates the average effect size of the association of child maltreatment and behavior problems is unclear. Lastly, the cross-study similarities and differences in the pooled estimates of the associations between child maltreatment and behavior problems among existing meta-analyses have not been analyzed.

The primary objective of the present meta-analysis was to estimate the effect size of the relationships between three types of child maltreatment, including physical abuse, emotional abuse, and neglect, and childhood behavior problems. Childhood behavior problems refer to child externalizing and internalizing behaviors and their subtypes including aggression, delinquency/antisocial behavior, hyperactivity/attention deficit, anxiety, depression, somatic complaints, suicidal attempts, suicidal ideation, and self-injury behavior (J. Liu, 2004; J. Liu, Chen, & Lewis, 2011). In addition, the meta-analysis also (1) examined whether different types of child maltreatment relate to different behavior problems equally or differently, (2) tested the heterogeneity in study findings and the contributions of sample characteristics and methodological factors to the study heterogeneity, and (3) explored cross-study similarities/differences in the effect sizes of the relationship between child maltreatment and behavior problems by comparing our findings with those in recent existing meta-analyses.

Method

Literature Search

A computerized literature search was conducted using the two largest Chinese databases (the Chinese National Knowledge Infrastructure and Wanfang Data) and three English databases (PubMed, PsychoInfo, and Scopus) in May 2016. Chinese and English search terms related to child maltreatment (i.e., child maltreatment, child abuse, victimization, trauma, neglect, emotional abuse, psychological abuse, physical abuse, or physical punishment), behavior problems (i.e., externalizing behaviors, internalizing behaviors, aggression, violence, crime, delinquency, antisocial behavior, hyperactive, attention deficit, anxiety, depression, somatic, suicide attempt, suicidal ideation, or self-injury), and Mainland China were used. The search terms were restricted to English or Chinese language but not restricted to any time period. Reference lists of eligible articles and five review articles (i.e., Dunne, Chen, & Wan Yuen, 2008; Fang et al., 2015; Fry, McCoy, & Swales, 2012; Ip et al., 2016; United Nations Children’s Fund, 2012) were also examined to identify additional relevant publications.

A separate computerized literature search was conducted in Scopus to identify recent existing meta-analyses on the relationship of child maltreatment and behavior problems for the cross-study comparison. Search terms were the same as described previously (without Mainland China) and meta-analysis in January 2017 and time period restricted to 2012 and beyond.

Eligibility Criteria and Study Selection

To meet inclusion criteria for this meta-analysis, studies had to (1) be quantitative original research, (2) include children under 18 years old (i.e., maximum age or 1 standard deviation above mean age was below 18 years old, or participants were from senior high school¹ or lower if age range or mean was not reported) from Mainland China, and (3) have the independent variable of child maltreatment and the dependent variable of behavior problems that meet the definitions described in Table 1. Studies were excluded if they (1) presented child maltreatment as dependent variable, (2) used data that were duplicated in another article, or (3) did not provide sufficient information in articles directly or through contact with the authors to calculate effect sizes. Studies solely on child sexual abuse or maltreatment contaminated by sexual abuse were also excluded because this meta-analysis focused on maltreatment practiced by parents or family members, whereas very few Chinese studies on child sexual abuse specified whether parents or family members were the abusers.

Table 1.

Definitions of Child Maltreatment and Behavior Problems in the Meta-Analysis.

Terms	Definitions
Child maltreatment	All forms of physical and/or nonphysical abuse, neglect, or negligent treatment by parents, guardians, or other family members in the household resulting in actual or potential harm to the child’s health, survival, development, or dignity. The meta-analysis focuses on its three forms: physical abuse, emotional abuse, and neglect.
Physical abuse	The physical force against a child by parents, guardians, or other family members in the household that results in or has a high likelihood of resulting in harm to the child’s health, survival, development, or dignity. This includes hitting, beating, kicking, shaking, biting, strangling, scalding, burning, poisoning, suffocating, and other violent acts.
Emotional abuse	Nonphysical forms of rejection or hostile treatment by parents, guardians, or other family members that may have a high probability of damaging the child’s physical or mental health or the child’s physical, mental, spiritual, moral, or social development, such as the restriction of movement, belittling, blaming, threatening, frightening, discriminating against, humiliating, or ridiculing.
Neglect	Failure to provide for the development and well-being of the child in health, education, emotional development, nutrition, shelter, or safe living conditions by parents, guardians, or other family members.
Externalizing behaviors	Acts directed to the social environment and characterized by disinhibitory control. These include aggression, delinquency/antisocial behavior, and hyperactivity/attention deficit.
Aggression	Physical or verbal behaviors that harm or threaten to harm others including children, adults, and animals.
Delinquency/antisocial behavior	Acts (not including violent acts) that break rules or laws such as lying, cheating, stealing, and committing antisocial acts with bad companions.
Hyperactivity/attention deficit	An excess of motor activity, restlessness, or attention deficits in which the children are unable to sustain and modulate their attention in a controlled setting such as the classroom.
Internalizing behaviors	Negative behaviors directed inward to children themselves. These include anxiety, depression, somatic complaints, and suicide (attempts, ideation, or nonsuicidal self-injury behavior).
Anxiety	Symptoms of worry, nervousness, and apprehension without cause.
Depression	Symptoms characterized by excessive sadness and loss of interest in usually enjoyable activities.
Somatic complaints	Physical symptoms with no identifiable, specific physiological cause.
Suicide	The act (suicidal attempt) and thoughts (suicidal ideation) of intentionally causing one’s own death or the act of deliberately harming the surface of one’s own body without the intention of causing death (self-injury).

Note. The definitions of child maltreatment, physical abuse, emotional abuse, and neglect were adapted from World Health Organization (2006, pp. 9–10). The definitions of externalizing and internalizing behaviors and their subtypes were adapted from J. Liu (2004) and J. Liu, Chen, and Lewis (2011).

To meet the inclusion criteria for the cross-study comparison, meta-analyses must (1) focus on the relationship between at least one of the three types of child maltreatment (i.e. physical abuse, emotional abuse, or neglect) and behavior problems, (2) have comparable definitions of child maltreatment and behavior problems with those in the present meta-analysis, and (3) have available information of pooled estimates of effect size and their confidence intervals (CIs).

Data Extraction and Coding

Two raters screened the search results independently based on the eligibility criteria. Disagreements were resolved by consensus to reduce bias and errors (Buscemi, Hartling, Vandermeer, Tjosvold, & Klassen, 2006). After obtaining all eligible studies, we used the first author’s last name and publication year to identify each study. Based on the literature, we extracted the following data from each study using the predetermined codebook.

Publication type (peer-reviewed journal article, conference paper, or thesis/dissertation).

Sample characteristics including sample size (large if n > 1,000, otherwise small), percentage of girls, age (mean and standard deviations or age range or grade if mean age was unreported), study region (northeast, central, southwest, northwest, north, south, east, multiple regions, or unspecified regions; X. Zhou et al., 2015), residential areas² (rural, urban, or mixed if participants were from both rural and urban areas), and response rate (RR; the number of usable responses returned divided by the total number eligible in the sample chosen) or usable return rate (URR; the number of usable responses returned divided by the total number of questionnaires distributed) if RR was unreported.

Methodological factors including study design (cross-sectional or longitudinal), sampling methods (probability sampling or not), subgroups (e.g., girls or boys, father report, or mother report), number of study sites (single site or multiple sites), validated assessment of child maltreatment (yes or no), assessment tool of child maltreatment (Conflict Tactics Scale for Parent and Child [CTSPC], Childhood Trauma Questionnaire [CTQ], self-developed instrument, or others), reporter of child maltreatment (child, parent, or others), abuser of child maltreatment (father, mother, parents if the parental gender was not specified, or mixed with parents, guardians, and family members), and measurement of behavior problems (questionnaire, diagnosed, or others). These factors may contribute to the heterogeneity in study findings and hence were used as moderators separately (Ji & Finkelhor, 2015).

Information of child maltreatment (physical abuse, emotional abuse, neglect, or mixed) and behavior problems (Level-1 coding: externalizing behaviors, internalizing behaviors, or mixed if externalizing and internalizing behaviors were not differentiated; Level-2 coding: aggression, delinquency/antisocial behavior, hyperactivity/attention deficit, anxiety, depression, somatic complaints, suicidal attempts, suicidal ideation, and self-injury behavior) and information to calculate effect sizes were also collected. Missing information was requested from authors and was coded as missing if it could not be obtained. Additionally, after obtaining eligible existing meta-analyses for the cross-study comparison, information regarding types of child maltreatment, types of behavior problems, pooled effect sizes, and CIs of the relationship between child maltreatment and behavior problems were also extracted.

Quality Assessment

The same two raters appraised the quality of each study using the modified Quality Assessment Tool for Observational Cohort and Cross-sectional Studies (National Institute of Health, 2014) based on each study’s objective(s), sampling description, measurement, and analysis. Each item is worth one point and a higher summed score indicates better study quality. Study quality was regarded as high if the point sum was higher than half of the total number of items that applied to the study. Otherwise, it was rated as low. Disagreements in the assessment between the raters were resolved by consensus.

Statistical Analysis

Calculation of effect sizes

To calculate the effect size for each study, Cohen’s d was first calculated from bivariate correlation coefficients, unadjusted odds ratios (ORs), means and standard deviations, or frequencies and/or proportions. Then, Hedges’s g was calculated from Cohen’s d multiplying J, where

J = 1 - (\frac{3}{4 (N - 2) - 1}),

to correct for potential bias introduced by studies with small sample sizes (N) to estimate the effect size of each study. Similar to Cohen’s d, Hedges’s g = .2 is considered a small effect size, .5 represents a medium effect size, and .8 a large effect size (Lipsey & Wilson, 2001).

Effect sizes within a study were averaged in the following cases when necessary: (1) studies that reported two or multiple Level-2 behavior problems by the same participants that can be categorized into the same Level-1 behavior problems, (2) studies that assessed multiple conditions of a single type of child maltreatment (e.g., physical abuse reported by fathers and mothers separately, or information about mild physical abuse and severe physical abuse separately), and (3) longitudinal studies that reported cross-sectional relationships between child maltreatment and behavior problems at multiple time points. The variance of the averaged effect size was calculated using:

var (\frac{1}{m} \sum_{i = 1}^{m} Y_{i}) = {(\frac{1}{m})}^{2} var (\sum_{i = 1}^{m} Y_{i}) = {(\frac{1}{m})}^{2} (\sum_{i = 1}^{m} V_{i} + \sum_{i \neq j} (r_{i j} \sqrt{V_{i}} \sqrt{V_{j}})),

as described in Borenstein, Hedges, Higgins, and Rothstein’s (2009) study, where Y_i denotes the effect sizes from different behavior problems, conditions of child maltreatment, or time points within a study; Y_j denotes the outcome variable i , i = 1,…m, Vi and Vj denotes the variance of Y_i and Y_j Y_j when i ≠ j (j = 1, ...m), respectively; and r_ij is the correlation between Y_i and Y_j. When r_ij was not reported in the study, correlation equal to 1 was used.

Considering that only two studies reported the longitudinal relationships between child maltreatment and behavior problems and the remaining studies were cross-sectional, the effect sizes of all the cross-sectional relationships were submitted to analyze the pooled estimates across studies, moderator effects, and publication bias. Sensitivity analysis was conducted by adding the effect sizes from the longitudinal relationships.

Separate meta-analyses using random effects model (DerSimonian & Kacker, 2007) were run for each type of child maltreatment and each type of behavior problems at the two levels. The differences in the associations of each type of child maltreatment and each behavior problems were first screened if there were any pairs of 95% CIs were not overlapped. Then, for nonoverlapped 95% CIs, we further tested if the heterogeneity was statistically significant using the Q statistics (Higgins, Thompson, Deeks, & Altman, 2003).

Heterogeneity

Firstly, heterogeneity across studies was identified by the I² statistic that was calculated for each pooled estimate to determine the proportion of the observed variance that reflects variance in true effect sizes. I² values <25%, 50%, and ≥75% represent low, moderate, and high levels of heterogeneity, respectively (Higgins et al., 2003).

Next, heterogeneity among studies due to the categorical moderators was estimated using the Q statistic. Meta-regression was used to estimate the possible effect of mean age, RR, or URR on the relationship between child maltreatment and behavior problems. Because the number of studies reporting information for boys and girls separately was small, the effect of child gender was analyzed in two ways: (1) meta-regression using the proportion of girls as the independent variable and (2) Q statistics comparing the effect sizes in boy and girl subgroups, respectively, using studies that reported such information.

Publication bias

Publication bias was estimated using three methods such as the classic fail-safe N, funnel plot, and the trim-and-fill method. The classic fail-safe N calculates the number of potential unpublished studies with insignificant findings that would be needed to reduce the pooled effect size in the meta-analysis below the level of significance. A file-drawer effect, selective publication bias, exists when the fail-safe N is less than 5 times the number of published studies plus 10 (Rosenthal, 1979). A funnel plot is a plot of the effect sizes in studies included in meta-analysis against their standard errors (Sterne & Egger, 2001). It is expected to be symmetrical. An asymmetrical funnel plot might arise if larger studies with nonsignificant results or with an effect size in the nonhypothesized direction are preferentially published, while smaller studies with such findings are less likely to be published (Sterne & Egger, 2001). Based on the funnel plot, the trim and fill method imputes the missing effect sizes to fill in asymmetrical areas of the funnel plot and then recalculate the overall effect size to test publication bias (Duval & Tweedie, 2000).

Sensitivity analysis

Sensitivity analyses were conducted to test the uncertainty in the study findings. For studies without information about the correlations among multiple behavior problems or child maltreatment conditions, the variance of the average effect size was recalculated using correlations of 0 and 0.5, respectively, and then all analyses were rerun. The correlations of 0 and 0.5 were chosen conventionally to represent no correlation and moderate correlation. Next, all analyses were rerun after adding in the longitudinal relationship between maltreatment and behavior problems. Finally, the relationships between child maltreatment and behavior problems were tested in the high and low study quality subgroups, respectively.

Converting effect sizes in the eligible existing meta-analyses for cross-study comparison

The pooled effect sizes and CIs were converted to Cohen’s d if other forms (e.g., OR, risk ratio) of effect sizes were reported in the eligible meta-analyses to ensure comparability.

All the formulae employed in these analyses but not displayed can be found in the work by Lipsey and Wilson (2001). Comprehensive Meta-Analysis Version 3 (Biostat, Englewood, NJ) was used to complete the meta-analysis.

Results

Study Selection

The literature search yielded a total of 4,466 unique citations; 4,313 were excluded after title and abstract screening. Of the remaining 153 articles, 105 were further excluded upon full-text screening based on the eligibility criteria. One additional eligible article was identified from the reference list of a retrieved article. This process resulted in 46 articles from 42 studies that met the eligibility criteria (Figure 1). Articles from the same study were identified based on the author names, study location, and sample characteristics and were coded as one study if they focused on different maltreatment or behavior problems.

Figure 1.

A PRISMA (Preferred Reporting Items for Systematic Review) diagram for the steps of article search and selection for the meta-analysis. CNKI = China National Knowledge Infrastructure.

Study Characteristics

Study characteristics of the 42 studies were summarized in Table 2. These studies were conducted among 98,749 children with an average mean age of 12.13 years and an average of 48.7% girls in all the seven regions of Mainland China. Among the 42 studies, 34 were peer-reviewed publications, 7 were theses/dissertations, and 1 was conference paper. Nineteen studies had sample sizes over 1,000. Sixteen studies used probability sampling methods. Eleven studies reported RR (>90% in eight studies), 17 studies reported URR (>90% in 16 studies), while 14 studies did not specify RR or URR. Fourteen studies recruited samples from both rural and urban areas, 2 from rural areas only, 2 from urban areas only, and the remaining 24 did not specify this information.

Table 2.

Summary of Study Characteristics and Effect Sizes.

1. Study No.	2. Study Name	3. Region	4. Sample Size	5. Age (Years)	6. Gender (% Girls)	7. Probability Sampling	8. RR/URR	9. Child Maltreatment: Measurement	10. Behavior Problem: Measurement	11. Study Quality
1	Chang, Schwartz, Dodge, and McBride-Chang (2003)	SC	325	4.6 ± 1.0	44.6	UNSP	UNSP	Mixed (PA and EA) by father and mother, respectively: PARQ (FR, MR)	Aggression: SDI (TR)	Low
2	Chen (2015)	NE (Heilongjiang)	1,113	Middle school Grade1–4	50.6	No	95.62	EA and Neglect: CPANS	Depression: CES_D	High
3	Cheng et al. (2010)	EC (Jiangxi)	223 LBC	12.3 ± 2.0	49.0	Yes	99.4^U	EA and Neglect: CPANS	Hyperactivity, INTER, and EXTER: SDQ	High
4	Du, Shi, Zou, Sun, and Yang (2014)	unspecified	393	11.0 ± 1.2	52.4	Yes	91.4^U	Neglect: CPANS	EXTER behavior: RCBQ (PR)	High
5	Eisenberg, Chang, Ma, and Huang (2009)	NC (Beijing and Hebei)	697	8.3 ± 0.6	44.5	UNSP	UNSP	PA: PSDQ (PR)	EXTER and INTER: SDI (TR)	Low
6	Gao, Wong, and Yu (2016)	SC (Guangdong)	1,163	13.8 ± 1.0	43.3	Yes	UNSP	Mixed (PA and EA): CTSPC	Delinquency and Depression: SDI	High
7	Ge et al. (2013), Yu et al. (2013), and Zhang et al. (2013)	EC (Anhui)	1,417	14.3 ± 1.3	46.6	No	98.5^U	PA and EA: SDI	Aggression: BPAQ	Low
									Delinquency: SDI
									Depression: SRDS
									Self-Injury: SDI
									Suicidal attempt: SDI
									Suicidal Ideation: SDI
8	Hesketh et al. (2011)	EC (Zhejiang)	2,203	7–12	43.7	Yes	80	PA: SDI	EXTER and INTER: RCBQ	High
9	Kwok, Chai, and He (2013)	EC (Shanghai)	560	14.0 ± 1.2	47.2	No	UNSP	PA and EA: CTSPC	Suicidal ideation: C-SIS	High
10	Lansford et al. (2005)	NC (Beijing)	50	10.6 ± 1.9	46.0	No	UNSP	PA: Interview (child and MR)	Aggression and anxiety: YSR and CBCL (MR)	Low
11	Lansford et al. (2014)	EC (Shandong and Shanghai)	239	8.3 ± 0.6	51.0	No	UNSP	PA: Interview (MR)	Aggression and Anx/Dep: YSR and CBCL (MR)	Low
12	P. Li (2013)	NC	647	Middle school	52.4	UNSP	93.0^U	EA and neglect: CPANS	Aggression: MCP	Low
13	Y. Li, Cao, Cui, and Li (2012)	EC (Shandong)	816	14.6 ± 2.1	52.2	Yes	UNSP	Neglect: CTQ	EXTER, hyperactivity, and INTER: SDQ	High
14	L. Li et al. (2016)	NC, EC, and SE	ODD: 259 Control: 269	ODD; 9.6 ± 1.6 Control: 9.1 ± 1.5	ODD: 27.4% Control: 46.8%	No	ODD: 84.9 Control: UNSP	PA and EA: CTQ (PR)	Aggression: CBS (TR)	ODD; high control: low
15	X. Liu, Sun, and Yang (2008)	EC (Shandong)	1,920	13.6 ± 1.7	45.2	Yes	UNSP	PA, EA, and neglect: SDI	Suicidal attempt	High
16	B. Liu, Wei, Xing, and Wang (2012)	EC (Shandong)	1,070	13.8	52.1	Yes	91.9	PA and EA by father and mother: CTSPC	EXTER: YSR	Low
17	J. Liu, Wan, Yuan, Li, and Tao (2016)	EC (Anhui)	5,116	15.8 ± 1.9	45.7	Yes	94.7^U	Neglect: CTQ	Suicidal attempt and suicidal ideation: SDI	Low
18	A. Liu and Nian (2012)	NE (Heilongjiang)	503	9.7 ± 0.9	49.9	No	96.3^U	Neglect: CPANS	Aggression: MCP	High
19	Luo, Deng, Tang, and Wei (2013)	CC (Hunan)	940	8–12	49.9	No	83.9^U	EA and neglect: CPANS	Somatic: CSI	High
20	Ma, Chen, and Huang (2011)	SC (Guangdong)	3,037	13.4 ± 0.5	49.9	Yes	96.1^U	PA: CTSPC	Depression: DSRSC	High
21	Olson et al. (2011)	NC (Beijing)	59	4.4 ± 0.3	45%	UNSP	96.70^U	Mixed (PA and EA): SOMA-PP (PR)	EXTER: CBCL/1.5-5	Low
22	Pan (2010)	CC (Hunan)	7,248	7–17	46.9	Yes	94.8^U	EA: CPANS	Anxiety: SCARED	High
22	Pan (2010)	CC (Hunan)	7,248	7–17	46.9	Yes	94.8^U	EA: CPANS	Depression: DSRSC	High
23	Su et al. (2015)	SW (Guizhou)	4,617	15.2 ± 1.8	54.1	Yes	96.0	Neglect: CTQ	Self-harm: SDI	High
24	Tan (2014)	CC (Hunan)	280	10.1 ± 0.9	51.1	No	93.3^U	Neglect: CTQ	Aggression: CBCL (PR)	Low
									Delinquency: CBCL (PR)
									Anx/Dep/: CBCL (PR)
									Anxiety: SCARED
									Depression: CDI
									Somatic: CBCL (PR)
25	Tang et al. (2011) and Wong et al. (2009)	SC (Guangdong)	6,593	14.2 ± 1.0	49.9	Yes	99.5^U	PA: CTSPC	BPmixed: SDI	Low
									Suicidal ideation: SDI
									Somatic: SDI
26	Tao et al. (2006)	EC (Anhui)	5,141	15.0 ± 1.6	48.3	No	94.3^U	PA and EA: SDI	Anxiety: SCL-90	Low
									Depression: SCL-90
									Somatic: SCL-90
27	Wan (2015)	CC (Henan), SW (Guizhou), and EC (Anhui)	14,820	15.4 ± 1.8	50.2	UNSP	95.6^U	Neglect: CTQ	Self-Injury: SDI	Low
28	Wan et al. (2015)	SC (Guangdong), CC (Henan), NC (Liaoning), and SW (Chongqing)	14,211	15.1 ± 1.9	52.8	No	96.9	PA and EA: SDI	Self-injury: SDI	Low
29	Y. Wang, Qian, and Wang (2008)	CC (Henan)	355	16.2 ± 1.1	56.3	Yes	88.7^U	EA and neglect: CPANS	Depression: SDS	High
30	M. Wang, Wang, and Liu (2016)	EC (Shandong)	1,971	7–17	49.4	No	95.80	PA and EA by father and mother: CTSPC (FR)	Anxiety: SCAS (FR, MR)	Low
31	Wu, Deng, and Pan (2011)	CC (Hunan)	1,555	13.1 ± 1.1	45.3	No	96.9	EA: CPANS	Conduct: SDQ	High
									Hyperactivity: SDQ
									INTER: SDQ
32	Xiao, Tao, Xu, Su, and Huang (2008)	EC (Anhui)	10,894	15.4 ± 2.0	47.5	No	UNSP	PA and EA: SDI	Self-Injury: SDI	Low
33	Xing, Wang, Zhang, He, and Zhang (2011) and Xing and Wang (2013)	EC	486 at T1 and 454 at T2	10.7 ± 1.0	49.8	UNSP	UNSP	PA: CTSPC	Aggression: YSR	High
									Delinquency: YSR
									Anx/Dep/: YSR
34	Xu (2015)	NW (Shanxi)	597	5.1 ± 0.8	48.2	No	96.3	PA and EA: CTSPC (FR, MR)	BPmixed: RCBQ (PR)	High
35	Yang (2012)	CC (Hubei)	324	Elementary school Grades 3–6	36.1	No	92.6	EA and neglect: CPANS	Anxiety: SCS	High
36	Z. Zhang and Guo (2003)	SW (Sichuan)	1,393	15.0 ± 2.0	51.5	Yes	98.0^U	Mixed (PA and EA): ASLEC	Suicidal ideation: SDI	Low
37	L. Zhang, Qian, and Wang. (2008)	CC (Henan)	258	15.2 ± 1.7	57.7	Yes	UNSP	Mixed (EA and neglect): CPANS	Aggression: AAQ	Low
38	Zhao et al. (2014)	E (Anhui)	2,917 LBC and non-LBC	7–17	47.4	Yes	UNSP	PA and neglect: CTSPC	Anxiety: SASC	High
39	Q. Zhou, Eisenberg, Wang, and Reiser (2004)	N (Beijing)	425	7.7 ± 0.6	55.5	No	72.20	PA and EA: PSDQ (PR)	EXTER: CBCL (PR)	High
									Aggression: RCP (peer)
									EXTER: PCSC (TR)
40	Zhu, Xu, Su, and Lin (2010)	S (Guangxi)	659	Middle school Grades 1–3	52.8	No	UNSP	EA and neglect: CPANS	Depression: CES_D	High
41	Zhu, Xu, and Lin (2013)	S (Guangxi)	866	11.4 ± 1.0	49.4	No	97.4^U	Neglect: CTQ	Anxiety: SAS	High
42	Zou et al. (2012)	CC (Henan)	74	14.2 ± 0.7	0	No	UNSP	PA: CECAQ	Aggression: AQ	Low
42	Zou et al. (2012)	CC (Henan)	74	14.2 ± 0.7	0	No	UNSP	EA and neglect by father and mother: CECAQ	Aggression: AQ	Low

Note. In Study 14, the ODD and control groups were treated as two independent samples. The CTQ measured five types of CM including physical abuse, emotional abuse, sexual abuse, physical neglect, and emotional neglect in childhood. Only data related to emotional neglect were included in the analysis because the abusers of the other types might include people other than parents, guardians, or family members in the household. UNSP = unspecified or unclear; ODD = oppositional defiant disorder; LBC = left-behind children, defined as children who were left at home for a long period with their extended family members by their parents who migrate to urban areas to work. Left-behind children are a unique phenomenon in Mainland China. Child maltreatment and behavior problems were self-report unless otherwise specified. FR = father report; MR = mother report; PR = parent report; father report MR = mother report; TR = teacher report; Peer = peer report. In Column 3: CC = Central China; EC = East China; NC = North China; NE = Northeast China; NW = Northwest China; SC = South China; and SW = Southwest China. Geographical regions can be visualized in X. Zhou et al. (2015). In Column 8: RR = response rates were calculated by dividing the number of usable responses returned by the total number eligible in the sample chosen. URR = usable return rates, indicated by superscript U, were calculated by dividing number of usable responses returned by the total number of questionnaires distributed. In Columns 9 and 10: Delinquency is short for delinquency/antisocial behavior, hyperactivity is short for hyperactivity/attention deficit, somatic is short for somatic complaints, and Anx/Dep refers to combined anxiety and depression. Child maltreatment measurement tools: ASLEC = Adolescent Self-Rating Life Events Checklist; CECAQ = Childhood Experience of Care and Abuse Questionnaire; CPANS = Child Psychological Abuse and Neglect Scale; CTQ = Childhood Trauma Questionnaire; CTSPC = Parent–Child Conflict Tactics Scale; PARQ = Parental Acceptance Rejection Questionnaire; PSDQ = Parenting Styles and Dimensions; SOMA-PP = Socialization of Moral Affect Questionnaire-Preschool Parent. Child behavior problems measurement tools: AAQ = Adolescent Aggression Questionnaire; AQ = Aggression Questionnaire; BPAQ = Buss-Perry Aggression Questionnaire; CBCL = Child Behavior Checklist; CBS = Child Behavior Scale; CDI = Children’s Depression Inventory; CES_D = Center for Epidemiological Survey Depression Scale; CSI = Children’s Somatization Inventory; DSRSC = Depression Self-rating Scale for Children; MCP = Master Class Play; RCBQ = Rutter Child Behavior Questionnaire; RCP = Revised Class Play; SAS = Self Rated Anxiety Scale; SASC = The Social Anxiety Scales for Children; SCARED = The Screen for Child Anxiety Related Emotional Disorders; SCAS = Spence Children’s Anxiety Scale; SCS = Child Self-Concept Scale; SDQ = Strength and Difficulties Questionnaire; SDS = Self-Rating Depression Scale; SIS = Suicidal Ideation Subscale; STDS = State Depression Scale in the State-Trait Depression Scale; YSR = Youth Self Report; SDI = self-developed instrument.

Most of the studies were cross-sectional; two were longitudinal. Almost all studies were conducted in multiple sites except three studies that did not specify study sites. Nine studies used the CTSPC, 6 used the CTQ , 5 used self-developed instruments, and the remaining 22 studies used other tools to assess child maltreatment. Nine studies collected child maltreatment data reported by parents, while the remaining studies used child-report data. Four studies reported the relationships between child maltreatment and behavior problems for boys and girls, respectively. Four studies differentiated father abuser and mother abuser. Eight studies collected maltreatment information from parents, while the remaining used child-report data. All studies measured behavior problems using questionnaires.

The included studies analyzed the relationships between physical abuse and externalizing behaviors (n = 10), physical abuse and internalizing behaviors (n = 13), emotional abuse and externalizing behaviors (n = 8), emotional abuse and internalizing behaviors (n = 16), neglect and externalizing behaviors (n = 7), and neglect and internalizing behaviors (n = 15). Six studies focused on mixed maltreatment or mixed behavior problems without differentiating their types. Twenty-one and 20 studies were classified as high quality and low quality, respectively. The study quality of L. Li et al. (2016) was evaluated for the two subsamples separately (one was classified as high and the other was low).

In addition, two studies reported information specific to left-behind children (children who were left at home for a long period with their extended family members by their parents who migrate to urban areas to work); therefore, a code to indicate left-behind children or not was added to capture this unique phenomenon in Mainland China.

Physical Abuse

Among all the studies that examined the relationships between physical abuse and externalizing behaviors, all but one study reported increased externalizing behaviors in relation to physical abuse, Hedges’s g = .503 (95% CI [0.358, 0.649]). All studies but one examining the relationships between physical abuse and internalizing behaviors reported increased internalizing behaviors related to physical abuse, Hedges’s g = .493 (95% CI [0.469, 0.518]). The two studies that reported relationships between physical abuse and mixed behavior problems showed a small effect size, Hedges’s g = .376 (95% CI [0.330, 0.422]). The separate analyses for the relationships between physical abuse and each Level-2 behavior problems showed similar small-to-medium effect sizes. All the 95% CIs were overlapped, indicating there were not significant differences in the relationships between physical abuse and Level-1 or Level-2 behavior problems (see Figure 2 and Table 3).

Figure 2.

Forest plot of the association of child physical abuse (PA) and externalizing (EXTER), internalizing (INTER), and mixed behavior problems (MixedBP, which refers to both externalizing and internalizing behaviors). The square shape and bar line represent the estimate of effect size and 95% confidence interval for each study, respectively. The diamond shapes represent the pooled estimates of the effect sizes for different behavioral outcomes. Favors B indicates that child maltreatment increases behavior problems, while Favors A indicates child maltreatment decreases behavior problems.

Table 3.

The Pooled Estimates of Effect Size and Heterogeneity (I²) From the Meta-Analysis.

Behavior Problems	Physical Abuse			Emotional Abuse			Neglect			Mixed Child Maltreatment
Behavior Problems	n	Hedges’s g (95% CI)	I ²	n	Hedges’s g (95% CI)	I ²	n	Hedges’s g (95% CI)	I ²	n	Hedges’s g (95% CI)	I ²
Externalizing	10	.503 [0.358, 0.649]	77.6	8	.466 [0.422, 0.510]	94.5	7	.266 [0.087, 0.444]	83.46	2	.713 [0.436, 1.989]	87.2
Aggression	10	.480 [0.319, 0.640]	80.9	8	.232 [0.062, 0.402]	84.7	6	.199 [−0.084, 0.482]	89.2	2	.639 [0.395, 0.883]	95.7
Delinquency	2	.559 [0.291, 0.827]	82.5	2	.568 [0.013, 1.123]	94.7	2	.616 [0.187, 1.046]	85.7	1	.603 [−0.500, 1.707]	NA
Hyperactivity	0		NA	2	.519 [0.003, 1.036]	92.4	2	.386 [0.229, 0.542]	23.5	0		NA
Internalizing	13	.493 [0.469, 0.518]	92.6	16	.583 [0.562, 0.604]	94.5	15	.460 [0.340, 0.580]	93.3	4	.626 [0.260, 0.902]	68.9
Anxiety	5	.496 [0.353, 0.639]	78.0	3	.746 [0.321, 1.172]	96.0	4	.388 [0.329, 0.448]	0	0		NA
Depression	2	.531 [0.112, 0.949]	91.4	7	.631 [0.432, 0.829]	95.8	4	.823 [0.580, 1.066]	85.5	1	.742 [−0.362, 1.846]	NA
Anx/Dep	3	.336 [−0.05, 0.723]	93.9	0		NA	1	.524 [0.281, 0.767]	NA	0		NA
Somatic	2	.466 [0.052, 0.881]	87.1	2	.663 [0.441, 0.884]	57.0	2	.446 [0.292, 0.601]	32.5	0		NA
Suicidal attempt	2	.836 [0.235, 1.438]	88.3	2	.561 [0.336, 0.786]	0	2	.281 [0.090, 0.472]	0	0		NA
Suicidal ideation	3	.402 [0.325, 0.479]	13.5	2	.564 [0.335, 0.794]	69.8	1	.483 [0.170, 0.796]	NA	1	.488 [−0.6371.613]	NA
Self-injury	3	.599 [0.543, 0.655]	87.1	3	.586 (0.550, 0.623]	8.4	2	.232 [0.191, 0.274]	17.0	0		NA
MixedBP	2	.376 [0.330, 0.422]	99.1	1	.874 [0.750, 0.998]	NA	0		NA	0		NA

Note. Hedges’s g was calculated from Cohen’s d multiplying J, where $J = 1 - (\frac{3}{4 (N - 2) - 1})$ to correct for potential bias introduced by studies with small sample sizes (N). Delinquency = delinquency/antisocial behavior; hyperactivity = hyperactivity/attention deficit; Anx/Dep = anxiety/depression; somatic = somatic complaints; MixedBP = behavior problems mixed by both externalizing and internalizing behaviors; NA = not applicable; CI = confidence interval.

Emotional Abuse

Among all the studies focused on emotional abuse and externalizing behaviors, all but one study reported increased externalizing behaviors in relation to emotional abuse. The pooled estimate Hedges’s g = .348 (95% CI [0.150, 0.547]). All studies on emotional abuse and internalizing behaviors reported positive relationships and the pooled estimate was 0.592 (95% CI [0.489, 0.695]). One study on the relationship between emotional abuse and mixed behavior problems reported a large effect size, Hedges’s g = .874 (95% CI [0.750, 0.998]). The effect size of the relationship between emotional abuse and internalizing behaviors was larger than the relationship between emotional abuse and externalizing behaviors (Q = 5.961, p = .015). Specifically, the associations of emotional abuse with Level-2 behavior problems, including depression, self-injury behavior, and somatic complaints, were higher than that of emotional abuse with aggression ( p values < .05; see Figure 3 and Table 3).

Figure 3.

Forest plot of the associations of child emotional abuse (EA) and externalizing (EXTER), internalizing (INTER), and mixed behavior problems (MixedBP, which refers to both externalizing and internalizing behaviors). The square shape and bar line represent the estimate of effect size and 95% confidence interval for each study, respectively. The diamond shapes represent the pooled estimates of the effect sizes for different behavioral outcomes. Favors B indicates that child maltreatment increases behavior problems, while Favors A indicates child maltreatment decreases behavior problems.

Neglect

The majority of studies on neglect and externalizing or internalizing behaviors reported that neglected children had more behavior problems except for the five studies. The pooled estimates for the relationship between neglect and externalizing behaviors and internalizing behaviors were 0.266 (95% CI [0.087, 0.444]) and 0.460 (95% CI [0.340, 0.580]), respectively. These were not significantly different. Neglect was also associated with all Level-2 behavior problems about equally because all 95% CI overlapped (see Figure 4 and Table 3).

Figure 4.

Forest plot of the associations of child neglect and externalizing (EXTER) and internalizing (INTER). The square shape and bar line represent the estimate of effect size and 95% confidence interval for each study, respectively. The diamond shapes represent the pooled estimates of the effect sizes for different behavioral outcomes. Favors B indicates that child maltreatment increases behavior problems, while Favors A indicates child maltreatment decreases behavior problems.

Mixed Child Maltreatment

All studies that examined the relationship between child maltreatment without differentiating its subtypes reported increased externalizing (Hedges’s g = .713, 95% CI [0.436, 1.989]) and internalizing (Hedges’s g = .626, 95% CI [0.260, 0.902]) behaviors about equally (Figure 5 and Table 3).

Figure 5.

Forest plot of the associations of mixed child maltreatment (MIXED refers to combined multiple types of child maltreatment) and externalizing (EXTER) and internalizing (INTER). The square shape and bar line represent the estimate of effect size and 95% confidence interval for each study, respectively. The diamond shapes represent the pooled estimates of the effect sizes for different behavioral outcomes. Favors B indicates that child maltreatment increases behavior problems, while Favors A indicates child maltreatment decreases behavior problems.

Heterogeneity

Despite some exceptions, there was substantial heterogeneity among most of the studies evidenced by the majority of I² values >75% ( p values < .001, Table 3). Child gender and reporter of child maltreatment were identified as potential moderators that may contribute to the heterogeneity. However, separate analyses of other moderators did not show significant results.

Child gender

Meta-regression analysis did not find a significant linear relation between the proportion of girls and the estimated effect sizes. A further analysis using the four studies that reported information stratified by child gender revealed that the relationship between physical abuse and externalizing behaviors was stronger among girls (Hedges’s g = .935, 95% CI [0.712, 1.160]) than boys (Hedges’s g = .568, 95% CI [0.354, 0.781]), Q = 5.443, p = .020. The association of emotional abuse with internalizing behaviors was stronger among boys (Hedges’s g = .651, 95% CI [0.584, 0.718]) than girls (Hedges’s g = .507, 95% CI [0.447, 0.567]), Q = 9.835, p = .002.

Reporter of child maltreatment

The effect size of the relationship emotional abuse and internalizing behaviors in one study that used parent-report data (Hedges’s g = .191, 95% CI [−0.059, 0.441]) was smaller than the pooled estimate of this relationship in the 14 studies that used child-report data (Hedges’s g = .632, 95% CI [0.554, 0.709]), Q = 10.920, p = .001.

Sensitivity Analyses

The same analyses were rerun using variance of the average effect sizes among multiple outcomes/conditions calculated from correlations of 0 and 0.5, respectively. These analyses produced equivalent results. The results remained unchanged after adding the effect sizes of the longitudinal relationships. The effect sizes of the relationships between child maltreatment and behavior problems did not significantly differ between high and low study quality subgroups.

Publication Bias

The present meta-analysis did not find evidence of publication bias. The funnel plots were symmetrical (Figure 6). The fail-safe N numbers were very large to reduce the mean effect size below the level of significance for the relationship between physical abuse and externalizing (N = 845) or internalizing behaviors (N = 3,500), between emotional abuse and externalizing (N = 582) or internalizing behaviors (N = 8,190), and between neglect and externalizing (N = 171) or internalizing behaviors (N = 2,366). The trim-and-fill method did not identify any missing studies for these relationships.

Figure 6.

Funnel plots of the included studies. (A) Studies of physical abuse and externalizing behaviors, (B) studies of physical abuse and internalizing behaviors, (C) studies of emotional abuse and externalizing behaviors, (D) studies of emotional abuse and internalizing behaviors, (E) studies of neglect and externalizing behaviors, (F) studies of neglect and internalizing behaviors, and (G) all studies included in the meta-analysis. Hedges’s g was calculated from Cohen’s d multiplying J, where $J = 1 - (\frac{3}{4 (N - 2) - 1})$ to correct for potential bias introduced by studies with small sample sizes (N).

Cross-Study Comparison

Five meta-analyses (Ip et al., 2016; Lindert et al., 2014; Mandelli, Petrelli, & Serretti, 2015; Norman et al., 2012; Wilson, Norris, Shi, & Rack, 2010) were yielded from the secondary literature search and compared with the present findings. As shown in Table 4, the point estimates (Hedges’s g) in the present meta-analysis were slightly higher than those (Cohen’s d) in the five meta-analyses conducted in child or both child and adult populations across the world. However, all 95% CIs overlapped.

Table 4.

Cross-Study Comparisons.

CM	BP	The Present Study	Ip et al. (2016)	Lindert et al. (2014)	Mandelli et al. (2015)	Wilson et al. (2010)	Norman et al. (2012)
PA	EXTER	.503 (0.358, 0.649)	.242 (0.005, 0.478)			.290 (−0.467, 1.047)
	Delinquency	.559 (0.291, 0.827)					.360 (0.284, 0.436)
	Anxiety	.496 (0.353, 0.639)	.374 (0.084, 0.664)	.293 (0.156, 0.429)
	Depression	.531 (0.112, 0.949)	.286 (0.169, 0.403)	.22 (0.141, 0.299)	.377 (0.286, 0.467)		.238 (0.083, 0.393)
	Suicidal attempts	.836 (0.235, 1.438)					.675 (0.428, 0.922)
EA	Depression	.631 (0.432, 0.829)			.564 (0.351, 0.777)		.617 (0.49, 0.744)
	Delinquency	.568 (0.013, 1.123)					.189 (0.058, 0.32)
	Suicidal attempts	.561 (0.336, 0.786)					.670 (0.492, 0.848)
Neglect	EXTER	.266 (0.087, 0.444)				.300 (−0.289, 0.889)
	Depression	.823 (0.580, 1.066)			.558 (0.257, 0.859)		.412 (0.263, 0.561)
	Delinquency	.616 (0.187, 1.046)					.170 (0.103, 0.237)
	Suicidal attempts	.281 (0.090, 0.472)					.368 (0.066, 0.67)

Note. Ip et al. (2016) used 24 studies of Chinese children and adults in Chinese societies including Mainland China, Taiwan, Hong Kong, Macau, and Singapore. Lindert et al. (2014) used 19 articles about child and adult studies in 14 countries around the world, including one Chinese study conducted in Beijing and Shanghai. Mandelli et al. (2015) used 26 adult studies conducted in the United States, Canada, European countries, and South American countries. Norman et al. (2012) used 124 studies conducted in child and adult populations across the world, with the majority in the United States. Wilson et al. (2010) used child studies conducted in the United States, Canada, and Spain. The present study reported Hedges’s g as the effect size, whereas the effect sizes of the other five meta-analyses were Cohen’s d. For studies reporting odds ratios (ORs), ORs were converted to Cohen’s d using relevant formulae described by Lipsey and Wilson (2001). Hedges’s g was calculated from Cohen’s d multiplying J, where $J = 1 - (\frac{3}{4 (N - 2) - 1})$ to correct for potential bias introduced by studies with small sample sizes (N). Hedges’s g and Cohen’s d are almost identical for large studies, whereas Hedges’s g is generally smaller than Cohen’s d in studies with relatively small sample sizes. Blank cells indicate corresponding relationship was not examined in corresponding meta-analysis. Delinquency is short for delinquency/antisocial behavior. CM = child maltreatment; PA = physical abuse; EA = emotional abuse; BP = behavior problems; EXTER = externalizing behaviors.

Discussion

Findings from the 42 studies conducted among 98,749 Chinese children supported the hypothesis that physical abuse, emotional abuse, and neglect by parents, guardians, and other family members in the household are undoubtedly associated with a broad spectrum of behavior problems in childhood. The pooled estimates suggest that different types of child maltreatment relate to different types of behavior problems approximately equally, with small-to-moderate effect sizes, except that emotional abuse related more to internalizing than externalizing behaviors. Moderator analysis showed that child gender and reporter of child maltreatment may contribute to the heterogeneity among study findings, but the evidence is still weak. No indication of publication bias emerged. In addition, the cross-study comparison showed that the present findings are equivalent to findings from the five identified meta-analyses conducted in child and adult populations across the world.

Pooled Estimates of Effect size

The present meta-analysis found that all three types of child maltreatment related to a broad spectrum of childhood behavior problems, including externalizing and internalizing behaviors in Mainland China. This is consistent with evidence that child maltreatment may be a nonspecific risk factor for health problems due to its universal adverse effects on multiple biological, neurological, and epigenetic pathways underlying behavioral, mental, and physical health problems (Kelly-Irving et al., 2013; McCrory, De Brito, & Viding, 2010; Vachon, Krueger, Rogosch, & Cicchetti, 2015).

The estimated small-to-moderate effect sizes were considered as of practical significance. Methodologically, public health, social, and psychological studies are more likely to produce smaller effect sizes due to attenuation in validity of measures in comparison to studies with good experimental control like clinical random trials (McCartney & Rosenthal, 2000). Empirically, they are of practical importance because of the high economic burden and the long-term adverse consequences related to child maltreatment and childhood behavior problems. Based on the modest associations of physical and emotional abuse with mental health problems in China, Fang et al. (2015) estimated that the economic cost of physical and emotional abuse was 0.47% and 0.84% of the gross domestic product (i.e., about US$27 and US$50 billion) in 2010, respectively. Also, childhood externalizing and internalizing behaviors are major risk factors for many social, societal, and health outcomes such as adulthood crime, violence, and mental disorders (for meta-analyses, see Leschied et al., 2008; Reef et al., 2011).

Heterogeneity

Considerable heterogeneity exists among study findings. Although most of the predetermined sample characteristics and methodological factors did not significantly contribute to the heterogeneity, weak evidence emerged that child gender may contribute to the heterogeneity. This notion was supported in the literature that child maltreatment and behavior problems are not equally prevalent in boys and girls (Cui, Xue, Connolly, & Liu, 2016; J. Liu, 2004; J. Liu et al., 2011) and that child gender moderates the relationship between child maltreatment and behavior problems (Braza et al., 2015). In addition, parent-report emotional abuse showed different strength of relationship with internalizing behavior from child-report emotional abuse. This could be because Chinese parents and children have discrepant perceptions of emotional abuse (Chan, 2012), but the exact explanation needs more studies to clarify.

Cross-Study Comparison

The positive relationship between child maltreatment and more externalizing and internalizing behaviors in Chinese children is consistent with findings from five recent meta-analyses conducted in a variety of populations from different countries and regions. The point estimates of the pooled effect size in the present study were consistently and slightly larger than the estimates in the other five meta-analyses, especially regarding physical and emotional abuse. This outcome may be because the relationships between child maltreatment and behavior problems are relatively stronger in child populations than in adult populations. All studies included in the present meta-analysis measured behavior problems using subjective questionnaires rather than diagnostic standards, whereas the other meta-analyses included studies using diagnostic behavior problems as outcomes. Therefore, measurement error introduced by subjective measures may contribute to the difference. Also, it is possible that the effect sizes calculated from unadjusted estimates in the present meta-analysis may produce larger effect sizes than other meta-analyses with effect sizes calculated from estimates that were adjusted for covariates (Norman et al., 2012). Nevertheless, the 95% CI overlapped across all meta-analyses, indicating that the observed differences were not significant. The cross-study comparisons provided initial evidence of the equal strength of the relationship between child maltreatment and behavior problems across populations and countries. Researchers can make greater effort to directly compare similarities and differences in the relationship between child maltreatment and behavior problems across social, cultural, and societal contexts using comparable methodology.

Limitations

Findings should be interpreted cautiously in light of study limitations. First, considerable variations exist in how researchers defined and measured physical abuse, emotional abuse, and neglect across the 42 studies. As discussed in the five existing meta-analyses, the present meta-analysis found that some studies defined physical abuse using multiple specific parental violent behaviors such as hitting, beating, and kicking (e.g., Kwok et al., 2013; Tang et al., 2011), whereas others simply referred to “physical punishment” or “being hit” (Hesketh et al., 2011; X. Liu, Sun, & Yang, 2008). Some studies (e.g., Ge et al., 2013; Tao et al., 2006; Yu et al., 2013; W. Zhang et al., 2013) may have excluded physical abuse in its mild form that are conventionally accepted as ordinary discipline but still related to behavior problems (Gershoff, 2002). Emotional abuse was measured by public humiliation (Olson et al., 2011), verbal abuse (X. Liu et al., 2008), or multiple behaviors including both and threatening to hit (e.g., B. Liu, Wei, Xing, & Wang, 2012; Pan, 2010; Tao et al., 2006). The majority of studies on neglect applied the Child Psychological Abuse and Neglect Scale developed by Chinese researchers based on the CTQ, the CTSPC, and the Chinese sociocultural context and assessed physical, psychological, and educational neglect (A. Liu & Nian, 2012; Yang, 2012), whereas others applied the CTQ and only measured psychological and physical neglect.

Second, child participants in the 42 studies were diverse in gender, age, and geographic region. However, they were relatively homogenous considering that all were students attending regular schools in the regions where Han Chinese people populate. Attention is still needed for disadvantaged children who do not attend regular schools (e.g., disabled or critically ill children who reside at home or in special education schools) and children in the 55 minority ethnic groups in Mainland China.

Third, the review focused on studies using child maltreatment as the independent variable and behavior problems as the dependent variable. However, all studies but two were cross-sectional. Hence, no causal relationship can be inferred, and the possibility that behavior problems increase the risk of child maltreatment should not be ruled out.

Fourth, the moderator analyses for child gender, reporter of child maltreatment, and abusers of maltreatment were performed only in a limited number of studies. Therefore, the estimate may be not precise enough to draw any conclusion. Future studies should stratify analysis by child and parent gender and collect child maltreatment information from children and from parents to generate more evidence.

Lastly, the present meta-analysis calculated effect sizes using the unadjusted rather than adjusted relationship between child maltreatment and behavior problems because estimate adjusting covariates may be affected by possible mediating mechanisms which will downwardly bias estimates of the overall effect of child maltreatment. It is possible that the relationship can change after adjusting for confounding variables, especially co-occurring maltreatment, family dysfunction, and parental behavioral and mental health (Norman et al., 2012).

Conclusion

This meta-analysis contributes to the existing literature by examining the associations between physical abuse, emotional abuse, and neglect and childhood behavior problems in the Mainland China context. Findings showed child maltreatment was undoubtedly related to a broad spectrum of childhood behavior problems among Chinese children. The associations of different types of child maltreatment and different types/subtypes of childhood behavior problems were about equivalent. In addition, associations between child maltreatment and behavior problems in Mainland Chinese children were equal to the estimates in some existing meta-analyses of studies of child and adult populations across the world.

Practice, Policy, and Research

Child protection persists as a concern in Mainland China. China still lacks an implemented and effective child protection system, and child maltreatment still does not receive much attention from the public, child welfare agencies, and the law enforcement (Man, Barth, Li, & Wang, 2017). Our findings provide empirical evidence that it is urgent and important to build an effective child protection system and launch evidence-based programs for the prevention and intervention of child maltreatment and childhood behavior problems in Mainland China. It is also important for practitioners and health professionals to develop strategies and take actions to prevent child maltreatment from occuring, and to prevent or treat maltreatment related behaivor problems. With respect to resarch, the present findings suggest that further research in minor groups (e.g., ethinicity minorities, disadvantaged families, children in underdevelopped areas) is warranted. In addition, future studies can differentiate child and parent gender to examine the gender effect in the relationship between child maltreatment and behavior problems.

Footnotes

Acknowledgments

We thank Fanghong Dong for her great help in literature screening and quality assessment. We also thank Drs. Ruth Lebet and Miranda Varrasse for proofreading this article.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by the Research Award from Office of Nursing Research at University of Pennsylvania School of Nursing and the Starting Grant for Young Investigator of Shandong University School of Nursing.

ORCID iD

Naixue Cui

Notes

References

Borenstein

Hedges

L. V.

Higgins

J. P. T.

Rothstein

H. R

. (2009) Multiple outcomes or time-points within a study. In Introduction to meta-analysis. Chichester, England: John Wiley & Sons, Ltd. doi:10.1002/9780470743386.ch24

Braza

Carreras

Muñoz

J. M.

Braza

Azurmendi

Pascual-Sagastizábal

… Sánchez-Martín

J. R.

(2015). Negative maternal and paternal parenting styles as predictors of children’s behavioral problems: Moderating effects of the child’s sex. Journal of Child and Family Studies, 24, 847–856.

Buscemi

Hartling

Vandermeer

Tjosvold

Klassen

T. P.

(2006). Single data extraction generated more errors than double data extraction in systematic reviews. Journal of Clinical Epidemiology, 59, 697–703.

Chan

K. L.

(2012). Comparison of parent and child reports on child maltreatment in a representative household sample in Hong Kong. Journal of Family Violence, 27, 11–21.

*Chang

Schwartz

Dodge

K. A.

McBride-Chang

(2003). Harsh parenting in relation to child emotion regulation and aggression. Journal of Family Psychology, 17, 598–606.

*Chen

(2015). The relationship among psychological maltreatment and neglect, resilience, and depression of junior middle students (Master’s thesis). Harbin Normal University, Heilongjiang, China.

*Cheng

Cao

Feng

Jiang

(2010). A comparative study on psychological abuse and neglect and emotional and behavioral problems of left-behind children and non-left-behind children in rural areas. Chinese Journal of Clinical Psychology, 18, 250–251.

Cui

Deatrick

J. A.

Liu

(2018). Maternal and paternal physical abuse: Unique and joint associations with child behavioral problems. Child Abuse & Neglect, 76, 524–532.

Cui

Xue

Connolly

C. A.

Liu

(2016). Does the gender of parent or child matter in child maltreatment in China? Child Abuse & Neglect, 54, 1–9.

10.

DerSimonian

Kacker

(2007). Random-effects model for meta-analysis of clinical trials: An update. Contemporary Clinical Trials, 28, 105–114.

11.

*Du

Shi

Zou

Sun

Yang

(2014). The relationship among child neglect, resilience and childhood behavior problems. Sichuan Journal of Mental Health, 27, 10–13.

12.

Dunne

M. P.

Chen

J. Q.

Wan Yuen

. (2008). The evolving evidence base for child protection in Chinese societies. Asia-Pacific Journal of Public Health, 20, 267–276. doi:10.1177/1010539508325047

13.

Duval

Tweedie

(2000). Trim and fill: A simple funnel-plot–based method of testing and adjusting for publication bias in meta-analysis. Biometrics, 56, 455–463.

14.

*Eisenberg

Chang

Huang

(2009). Relations of parenting style to Chinese children’s effortful control, ego resilience, and maladjustment. Development and Psychopathology, 21, 455.

15.

Fang

Fry

D. A.

Finkelhor

Chen

Lannen

Dunne

M. P.

(2015). The burden of child maltreatment in China: A systematic review. Bulletin of the World Health Organization, 93, 176–185C.

16.

Fry

McCoy

Swales

. (2012). The consequences of maltreatment on children’s lives: A systematic review of data from the East Asia and Pacific region. Trauma, Violence, & Abuse, 13, 209–233. Retrieved from http://dx.doi.org/10.1177/1524838012455873

17.

*Gao

Wong

D. S.

(2016). Maltreatment and delinquency in China: Examining and extending the intervening process of general strain theory. International Journal of Offender Therapy and Comparative Criminology, 60, 38–61.

18.

*Ge

Liu

Zhang

Wang

… Su

(2013). Study on the relationship between child abuse, parent–child separation in childhood and the aggressive behavior in adolescence among 1417 junior high school students. Chinese Journal of Epidemiology, 34, 5–9.

19.

Gershoff

E. T.

(2002). Corporal punishment by parents and associated child behaviors and experiences: A meta-analytic and theoretical review. Psychological Bulletin, 128, 539–579.

20.

Gershoff

E. T

. (2010). More harm than good: A summary of scientific research on the intended and unintended effects of corporal punishment on children. Law & Contemporary Problems, 73, 31–56.

21.

*Hesketh

Zheng

Jun

Y. X.

Xing

Z. W.

Dong

Z. X.

(2011). Behaviour problems in Chinese primary school children. Social Psychiatry and Psychiatric Epidemiology, 46, 733–741.

22.

Higgins

Thompson

S. G.

Deeks

J. J.

Altman

D. G.

(2003). Measuring inconsistency in meta-analyses [journal article as teaching resource, deposited by John Flynn]. British Medical Journal, 327, 557–560.

23.

Wong

R. S.

S. L.

Chan

K. L.

F. K.

Chow

(2016). Mental health consequences of childhood physical abuse in Chinese populations: A meta-analysis. Trauma, Violence, & Abuse, 17, 571–584.

24.

Finkelhor

(2015). A meta-analysis of child physical abuse prevalence in China. Child Abuse & Neglect, 43, 61–72.

25.

Kelly-Irving

Mabile

Grosclaude

Lang

Delpierre

(2013). The embodiment of adverse childhood experiences and cancer development: Potential biological mechanisms and pathways across the life course. International Journal of Public Health, 58, 3–11.

26.

*Kwok

S. Y.

Chai

(2013). Child abuse and suicidal ideation among adolescents in China. Child Abuse & Neglect, 37, 986–996.

27.

*Lansford

J. E.

Chang

Dodge

K. A.

Malone

P. S.

Oburu

Palmérus

… Zelli

(2005). Physical discipline and children’s adjustment: Cultural normativeness as a moderator. Child Development, 76, 1234–1246.

28.

*Lansford

J. E.

Sharma

Malone

P. S.

Woodlief

Dodge

K. A.

Oburu

… Tapanya

(2014). Corporal punishment, maternal warmth, and child adjustment: A longitudinal study in eight countries. Journal of Clinical Child & Adolescent Psychology, 43, 670–685.

29.

Leschied

Chiodo

Nowicki

Rodger

(2008). Childhood predictors of adult criminality: A meta-analysis drawn from the prospective longitudinal literature. Canadian Journal of Criminology and Criminal Justice, 50, 435–467.

30.

*Li

Lin

Chi

Heath

M. A.

Fang

Wang

(2016). Maltreatment and emotional and behavioral problems in Chinese children with and without oppositional defiant disorder: The mediating role of the parent–child relationship. Journal of Interpersonal Violence, 31, 2915–2939.

31.

*Li

(2013). Relational aggression, self-supporting personality, and psychological abuse and neglect (Master’s thesis). Qufu Normal University, Shandong, China.

32.

*Li

Cao

Cui

(2012). Child abuse and neglect, executive function, and emotional and behavioral problems in rural adolescents. Chinese Journal of Clinical Psychology, 20, 813–815.

33.

Lindert

von Ehrenstein

O. S.

Grashow

Gal

Braehler

Weisskopf

M. G.

(2014). Sexual and physical abuse in childhood is associated with depression and anxiety over the life course: Systematic review and meta-analysis. International Journal of Public Health, 59, 359–372.

34.

Lipsey

M. W.

Wilson

D. B.

(2001). Practical meta-analysis (Vol. 49).Thousand Oaks, CA: Sage.

35.

*Liu

Nian

(2012). Child neglect and peer acceptance: The mediating role of negative behavior. Journal of Psychological Science, 35, 911–915.

36.

*Liu

Wei

Xing

Wang

(2012). Relations between paternal and maternal harsh disciplines and junior middle school students’ externalizing problem behaviors. Chinese Journal of Clinical Psychology, 20, 842–845.

37.

Liu

(2004). Childhood externalizing behavior: Theory and implications. Journal of Child and Adolescent Psychiatric Nursing, 17, 93–103.

38.

Liu

Chen

Lewis

(2011). Childhood internalizing behaviour: Analysis and implications. Journal of Psychiatric and Mental Health Nursing, 18, 884–894.

39.

*Liu

Wan

Yuan

Tao

(2016). Correlations between adverse childhood experiences and suicidal behavior among middle school students. Chinese Journal of School Health, 37, 567–572.

40.

*Liu

Sun

Yang

(2008). Parent-reported suicidal behavior and correlates among adolescents in China. Journal of Affective Disorders, 105, 73–80.

41.

*Luo

Deng

Tang

Wei

(2013). Children’s somatization symptoms and its relation with alexithymia and psychological maltreatment. Chinese Journal of Clinical Psychology, 21, 236–239.

42.

*Ma

Chen

Huang

(2011). Study on the association between emotional abuse and other psychosocial factors with depressive symptom in junior high school students. Chinese Journal of Epidemiology, 32, 130–135.

43.

Man

Barth

R. P.

Wang

(2017). Exploring the new child protection system in Mainland China: How does it work? Children and Youth Services Review, 76, 196–202.

44.

Mandelli

Petrelli

Serretti

(2015). The role of specific early trauma in adult depression: A meta-analysis of published literature. Childhood trauma and adult depression. European Psychiatry, 30, 665–680.

45.

McCartney

Rosenthal

(2000). Effect size, practical importance, and social policy for children. Child Development, 71, 173–180.

46.

McCrory

De Brito

S. A.

Viding

(2010). Research review: The neurobiology and genetics of maltreatment and adversity. Journal of Child Psychology and Psychiatry, 51, 1079–1095.

47.

National Institute of Health. (2014). Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies. Retrieved from: https://www.nhlbi.nih.gov/health-pro/guidelines/in-develop/cardiovascular-risk-reduction/tools/cohort.

48.

Norman

R. E.

Byambaa

Butchart

Scott

Vos

(2012). The long-term health consequences of child physical abuse, emotional abuse, and neglect: A systematic review and meta-analysis. PLoS Medicine, 9, e1001349.

49.

*Olson

S. L.

Tardif

T. Z.

Miller

Felt

Grabell

A. S.

Kessler

… Hirabayashi

(2011). Inhibitory control and harsh discipline as predictors of externalizing problems in young children: A comparative study of US, Chinese, and Japanese preschoolers. Journal of Abnormal Child Psychology, 39, 1163.

50.

*Pan

. (2010). The revision of Child Psychological Maltreatment Scale and the epidemiological investigation of child psychological maltreatment (PhD dissertation). Central South University, Hunan, China.

51.

Reef

Diamantopoulou

van Meurs

Verhulst

F. C.

van der Ende

(2011). Developmental trajectories of child to adolescent externalizing behavior and adult DSM-IV disorder: Results of a 24-year longitudinal study. Social Psychiatry and Psychiatric Epidemiology, 46, 1233–1241.

52.

Rosenthal

(1979). The file drawer problem and tolerance for null results. Psychological Bulletin, 86, 638–641.

53.

Sterne

J. A.

Egger

(2001). Funnel plots for detecting bias in meta-analysis: Guidelines on choice of axis. Journal of Clinical Epidemiology, 54, 1046–1055.

54.

Stoltenborgh

Bakermans-Kranenburg

M. J.

IJzendoorn

M. H.

Alink

L. R

. (2013). Cultural–geographical differences in the occurrence of child physical abuse? A meta-analysis of global prevalence. International Journal of Psychology, 48, 81–94.

55.

*Su

Chen

Wan

Zhong

Tao

… Chen

(2015). The relationship between childhood abuse and non-suicidal self-injury in middle school students. Chinese Journal of School Health, 36, 1326–1329.

56.

*Tan

(2014). The correlation study of child abuse and children’s emotional and behavioral problems in an elementary school of Changsha (Master’s thesis). Central South University, Hunan, China.

57.

*Tang

C. S.

Wong

W. C.

Leung

P. M.

Chen

Lee

Ling

D. C.

(2011). Health compromising behaviors among Chinese adolescents: The role of physical abuse, school experience, and social support. Journal of Health Psychology, 16, 457–466.

58.

*Tao

Huang

Kim

Sun

Zhang

… Zhu

(2006). Correlation between psychopathological symptoms, coping style in adolescent and childhood repeated physical and emotional maltreatment. Chinese Journal of Pediatrics, 44, 688–693.

59.

United Nations Children’s Fund. (2012). Measuring and monitoring child protection systems: Proposed core indicators for the East Asia and Pacific Region. Strengthening Child Protection Series No. 1, UNICEF EAPRO, Bangkok, Thailand.

60.

Vachon

D. D.

Krueger

R. F.

Rogosch

F. A.

Cicchetti

(2015). Assessment of the harmful psychiatric and behavioral effects of different forms of child maltreatment. JAMA Psychiatry, 72, 1135–1142.

61.

*Wan

(2015). The role of pain processing in the relationship between childhood physical abuse and non-suicidal self-injury behaviors among middle school students (PhD dissertation). Anhui Medical University, Anhui, China.

62.

*Wan

Chen

Sun

Tao

(2015). The impact of childhood abuse on the risk of non-suicidal self-injury in mainland Chinese adolescents. PLoS One, 10, e0131239.

63.

*Wang

Wang

Liu

(2016). Paternal and maternal psychological and physical aggression and children’s anxiety in China. Child Abuse & Neglect, 51, 12–20.

64.

*Wang

Qian

Wang

(2008). Child psychological abuse & neglect, self-consistency & congruence and depression of high middle school students. Psychological Research, 1, 85–88.

65.

Wilson

S. R.

Norris

A. M.

Shi

Rack

J. J.

(2010). Comparing physically abused, neglected, and nonmaltreated children during interactions with their parents: A meta-analysis of observational studies. Communication Monographs, 77, 540–575.

66.

World Health Organization. (2006). Preventing child maltreatment: A guide to taking action and generating evidence. Geneva: World Health Organization.

67.

*Wong

W. C.

Leung

P. W.

Tang

C. S.

Chen

W. Q.

Lee

Ling

D. C.

(2009). To unfold a hidden epidemic: Prevalence of child maltreatment and its health implications among high school students in Guangzhou, China. Child Abuse & Neglect, 33, 441–450.

68.

*Wu

Deng

Pan

(2011). Childhood psychological abuse and emotional and behavioral problems: Mediating effects of social support. Chinese Journal of Clinical Psychology, 19, 494–495, 502.

69.

*Xiao

Tao

Huang

(2008). Self -injurious behaviors in adolescents with repeated childhood abuse. Chinese Journal of School Health, 24, 1028–1030.

70.

*Xing

Wang

(2013). Sex differences in the reciprocal relationships between mild and severe corporal punishment and children’s internalizing problem behavior in a Chinese sample. Journal of Applied Developmental Psychology, 34, 9–16.

71.

*Xing

Wang

Zhang

(2011). Gender differences in the reciprocal relationships between parental physical aggression and children’s externalizing problem behavior in China. Journal of Family Psychology, 25, 699–708.

72.

*Xu

(2015). The research about parents’ harsh discipline and children’s temperament and behavior problems (Master’s dissertation). Shanxi University, Shanxi, China.

73.

*Yang

(2012). The psychological abuse and neglect and loneliness: The mediating effect of self-awareness. Chinese Journal of Special Education, 12, 64–69.

74.

*Yu

Liu

Zhang

Wang

… Su

(2013). The relationship between childhood abuse and injury and violence among junior middle school students. Child Journal of School Doctor, 27, 161–165.

75.

*Zhang

Qian

Wang

(2008). Adolescent aggressiveness: A childhood psychological abuse and neglect model. Medicine and Society, 21, 52–54.

76.

*Zhang

Liu

Wang

… Su

(2013). Relationship between childhood maltreatment and depression symptoms among 1417 junior high school students. Chinese Journal of Public Health, 29, 1165–1168.

77.

*Zhang

Guo

(2003). Suicide ideas of middle school students. Chinese Journal of Mental Health, 17, 852–855.

78.

*Zhao

Chen

M. C.

X. L.

Jiang

Y. H.

Sun

Y. H.

(2014). Left-behind children in rural China experience higher levels of anxiety and poorer living conditions. Acta Paediatrica, 103, 665–670.

79.

*Zhou

Eisenberg

Wang

Reiser

(2004). Chinese children’s effortful control and dispositional anger/frustration: Relations to parenting styles and children’s social functioning. Developmental Psychology, 40, 352.

80.

Zhou

Zhu

Mao

Zhang

Cai

(2015). Geographic regions of China. Northeast China (NE), North China (NC), Northwest China (NW), East China (EC), Central China (CC), South China (SC), and Southwest China (SW). Retrieved from https://figshare.com/articles/_Geographic_regions_of_China_Northeast_China_NE_North_China_NC_Northwest_China_NW_East_China_EC_Central_China_CC_South_China_SC_and_Southwest_China_SW_/1560916

81.

*Zhu

Lin

(2013). The mediating effect of resilience on the abuse of primary school students and their anxiety. Chinese Journal of Special Education, 10, 80–84.

82.

*Zhu

Lin

(2010). Moderating effects of social support on the associations between depression and psychological abuse and neglect of Chinese junior high school students. Paper presented at the 2010 International Conference on Psychology Psychological Sciences and Computer Science (PPSCS), Hubei, China.

83.

*Zou

Wang

Gao

Yan

Yang

(2012). Relationship between maltreatment history or parents’ rearing style and aggressive behavior among reform school boys. Chinese Journal of School Health, 33, 1077–1079.