Prevalence of childhood conduct problems in Brazil
Sixteen studies eligible for the review reported the prevalence of childhood conduct problems. Four of the studies used a diagnostic assessment tool (DAWBA: Development and Wellbeing Assessment) to estimate rates of conduct disorder and oppositional defiant disorder. These studies are summarised in Table
1. Rates of conduct disorder ranged from 0.6 to 2.2 %, and had a weighted average of 1.4 % (95 % CI 0.5–3.6). Rates of oppositional defiant disorder ranged from 2.0 to 3.2 % and had a weighted average of 2.4 % (95 % CI 1.7–3.5). Rates of any CD or ODD ranged from 2.6 to 7.0 % and had a weighted average of 4.1 % (95 % CI 2.1–7.9). Because there were few studies with results using diagnostic instruments, we did not analyse moderators that might explain variation in their results.
Table 1
Population-based studies reporting prevalence of childhood conduct disorder (CD) and oppositional defiant disorder (ODD) in Brazil
| Urban: Pelotas, RS | All children born in maternity hospitals in 1993 | 4,448 screening (85 %) 280 diagnostic interview (95 %) | 11–12 | 50 | Parent and adolescent combined in clinical assessment | Diagnostic interview (DAWBA) after screening questionnaire (SDQ with impact) | CD 2.2 % |
ODD 2.1 % |
CD/ODD 4.4 % |
| Urban: Pelotas, RS | All children born in maternity hospitals in 2004 | 3,585 (90.2 %) | 7 | 51 | Parent | Diagnostic interview (DAWBA) | CD 0.6 % |
ODD 2.0 % |
CD/ODD 2.6 % |
Fleitlich-Bilyk and Goodman [ 37] | Urban/Rural: Taubaté, SP | Random sample of students in stratified sample of public/private schools | 1,251 (83 %) | 7–14 | 53 | Parent, adolescent and teacher combined in clinical assessment | Diagnostic interview (DAWBA) | CD 2.2 % |
ODD 3.2 % |
CD/ODD 7.0 % |
| Mainly rural: Ilha de Maré, BA | Random sample of students within random sample of schools | 430 screening (100 %) 100 diagnostic interview (100 %) | 7–14 | 50 | Parent, adolescent and teacher combined in clinical assessment | Diagnostic interview (DAWBA) after screening questionnaire (SDQ with impact) | CD/ODD 3.4 % |
Fourteen studies (summarised in Table
2) reported the prevalence of children with conduct problems using screening questionnaires. (Two of the studies using screening questionnaires also used diagnostic instruments to estimate CD and ODD, as shown in Table
1.) Eight of the screening studies used the Strengths and Difficulties Questionnaire (SDQ); four studies used questionnaires from the Achenbach System of Empirically Based Assessment (the Child Behavior Checklist, Teacher’s Report Form or Youth Self-Report); and one study used the Mini-International Neuropsychiatric Interview (MINI).
1 The prevalence of conduct problems estimated using screening questionnaires ranged from 6.5 to 48.8 %. The weighted average was 20.8 % (95 % CI 15.9–26.9), but there was significant heterogeneity in the results (
Q = 577.5, df = 13,
p < 0.001). Given this heterogeneity, we examined moderating variables that might explain the variation.
Table 2
Population-based studies reporting prevalence of childhood conduct problems (CP) using screening questionnaires in Brazil
| Urban: Pelotas, RS | All children born in maternity hospitals in 1993 (half of sub-sample studied at 4 years) | 634 (87 %) | 4 | 50 | Parent | CBCL (borderline-abnormal score) | 31.8 |
| Urban: Pelotas, RS | All children born in maternity hospitals in 2004 | 3,750 (89 %) | 4 | 52 | Parent | CBCL (abnormal score) | 21.9 |
| Urban: Embu, SP | Random sample of children in low-income households within random sample of census units | 480 (81 %) | 6–17 | 49 | Parent | CBCL (abnormal score) | 17.7 |
| Urban: São Carlos, SP | Random sample of children in 5 out of 8 selected public schools | 321 (85 %) | 6–10 | 49 | Parent | SDQ (abnormal score) | 39.3 |
| Urban: Pelotas, RS | All adolescents in selected housing blocks in 79 census districts | 1,145 (90 %) | 11–15 | 48 | Adolescent | MINI (≥2 symptoms) | 29.2 |
Cucchiaro and Dalgalarrondo [ 33] | Urban: Campinas, SP | All students in random sample of public school classes | 765 (77 %) | 10–16 | 52 | Adolescent and teacher combined with OR rule | SDQ (abnormal score with impairment) | 6.5 |
| Urban: Ribeirão Preto, SP | Students in one public school (selection method not reported) | 107 (75 %) | 6–11 | 63 | Parent and teacher combined with OR rule | SDQ (abnormal score with impairment) | 9.8 |
| Urban: Ribeirão Preto, SP | Children from families enrolled in a public health programme | 100 (79 %) | 6–12 | 48 | Parent | SDQ (abnormal score) | 25 |
| Urban: São Gonçalo, RJ | Random sample of students within random sample of classes in random sample of public schools | 372 (74 %) | 7–11+ | 49 | Teacher | TRF (borderline-abnormal score) | 12.6 |
| Urban: Barretos, SP | Random sample of students in public and private schools | 327 (93 %) | 11–15 | 43 | Adolescent | SDQ (abnormal score with impairment) | 8.6 |
| Urban: São Luís, MA | Random sample of singletons born in public and private maternity hospitals in 1997/98 | 805 (68 %) | 7–9 | 52 | Parent | SDQ (abnormal score) | 48.8 |
| Urban: Salvador, BA | All students in one public school | 344 (49 %) | 11-18 | 38 | Adolescent | YSR + Impact section of SDQ (borderline-abnormal score with impairment) | 11.6 |
| Urban: Riberão Preto, SP | All singletons born in 10 maternity hospitals during 4 months in 1994 | 784 (68 %) | 9–11 | 51 | Parent | SDQ (abnormal score) | 35.3 |
| Urban/Rural: Taubaté, SP | Random sample of students within stratified random sample of public/private schools | 454 (83 %) | 7–11 | 52 | Parent | SDQ (borderline-abnormal score) | 23.6 |
Results of analyses of moderator variables that were measured as categories are shown in Table
3. Significantly higher rates of conduct problems were found in studies that recruited children from the community (in household surveys, maternity hospitals in birth cohort studies, or public health programmes) compared with studies recruiting from schools. Higher rates of conduct problems were reported by parents, followed by children, then teachers, and lowest rates were found in two studies using multiple informants—which also used an impairment criterion to identify children with probable conduct disorder. Considering the assessment instrument used, the highest rate of conduct problems was reported in the single study that used the MINI questionnaire, with lower rates being reported in studies using the Strengths and Difficulties Questionnaire and Achenbach scales. A much lower prevalence of conduct problems was found among studies using an impairment criterion to identify children with conduct problems compared with studies using symptom scores only. There were no significant differences in the results according to the region of Brazil in which the study was conducted.
Table 3
Moderators explaining variance in the prevalence of conduct problems in studies using screening instruments
Region of Brazil | | | 3.2 | 0.2 |
Northeast | 2 | 26.2 % (4.8–71.3) | | |
Southeast | 9 | 17.6 % (11.2–26.5) | | |
South | 3 | 27.3 % (21.3–34.3) | | |
Recruitment location | | | 5.9 | <0.05 |
Schools | 7 | 13.9 % (7.7–23.8) | | |
Community | 7 | 29.3 % (22.1–37.7) | | |
Informants | | | 37.5 | <0.001 |
Parent | 8 | 29.7 % (22.4–38.2) | | |
Child | 3 | 14.8 % (6.0–32.3) | | |
Teacher | 1 | 12.6 % (9.6–16.4) | | |
Multiple | 2 | 7.3 % (5.1–10.5) | | |
Instrument | | | 10.2 | <0.01 |
SDQ | 8 | 21.3 % (12.8–33.4) | | |
Achenbach scales | 5 | 18.5 % (13.5–24.7) | | |
MINI | 1 | 29.2 % (26.6–31.9) | | |
Impairment required | | | 35.5 | <0.001 |
Yes | 4 | 8.8 % (6.5–11.7) | | |
No | 10 | 27.6 % (21.7–34.4) | | |
We used meta-regression to examine whether variation in study results was associated with two study characteristics measured at the interval-level: the study response rate and age of the study children (defined as the mid-point of the age range in years). Studies with higher response rates had smaller effect sizes (a lower proportion of children with conduct problems; B = 0.012, p < 0.001). Child age was not significantly associated with study results.
Risk factors for childhood conduct problems in Brazil
Twelve studies examined associations between individual, family or social risk factors and children’s conduct problems in Brazil. The only variable which was analysed frequently enough to justify a meta-analysis was child sex. For all other variables, we summarise findings from each study separately, reporting odds ratios and confidence intervals wherever possible for significant (p < 0.05) associations, and listing any non-significant results.
Results on sex differences were inconsistent across seven studies
2 that used screening questionnaires: three found significantly higher rates of conduct problems among boys, two found significantly higher rates among girls, and two studies found no significant sex difference. Meta-analysing these screening studies, the weighted average odds ratio (comparing boys to girls) was not significant (OR = 1.2; 95 % CI 0.9–1.7), and heterogeneity in the results was significant (
Q = 25.5,
p < 0.001). In the only study [
37] that compared rates of CD/ODD between boys and girls, there was a significantly higher rate among boys (OR = 3.0; 95 % CI 1.8–5.0).
Brion et al. [
38] examined perinatal risk factors for conduct problems at age 4 years among 523 children in the Pelotas Birth Cohort Study 1993 (see the study by Anselmi et al. [
39] in Table
2). Pelotas is a city in the state of Rio Grande do Sul (RS) in the southern region of Brazil. The analyses were focused on effects of maternal smoking in pregnancy on children’s mental health. Higher conduct problem scores were associated with maternal smoking in pregnancy (OR = 1.7; 95 % CI 1.2–2.4) and with maternal psychiatric problems (OR = 3.1; 95 % CI 2.1–4.5). However, conduct problems were not significantly associated with paternal smoking during the mother’s pregnancy, maternal and paternal education, family income, or social class. Maternal smoking in pregnancy remained significantly predictive of conduct problems at age 4 after controlling for all other variables.
Using data on 4,423 participants in the same Birth Cohort Study in Pelotas (RS), Anselmi et al. [
40] examined perinatal and age 11 risk factors for conduct problems at age 15 (on the Strengths and Difficulties Questionnaire), focusing on the effects of family income change between birth and the age of 11 years. Higher conduct problem scores were associated with non-white skin colour, maternal smoking during pregnancy, young maternal age, and low maternal schooling in the perinatal assessment; family income change from birth to age 11; and stressful life events, poor maternal mental health, and the mother living without a partner when the child was aged 11. Controlling for all other variables, family income change was significantly predictive of conduct problem scores at age 15. Compared with children with high family incomes at both birth and age 11 (high–high), children in the following income groups had higher conduct problem scores at age 15: high–low, intermediate–intermediate, intermediate–low, low–intermediate, and low–low.
Caputo and Bordin [
41] conducted a case–control study comparing rates of conduct problems (clinical level problems on the Youth Self-Report questionnaire) between 207 primiparous pregnant adolescents and 308 sexually active but never-pregnant adolescent girls (13–17 years old) in Marília (SP). Conduct problems were less likely among pregnant girls (13.0 %) than among non-pregnant girls (20.8 %), equivalent to a significant odds ratio of 0.6 (95 % CI 0.4–0.9).
Cid [
42] compared levels of conduct problems by various parenting and family characteristics in a study of 321 elementary school children (6–10 years old) in São Carlos (SP) (see Table
2). Higher conduct problem scores were associated with receiving inconsistent discipline, relaxed discipline, physical abuse, lower levels of positive parenting, poor parent–child communication, the child having comorbid mental health problems, parental mental health problems, not living with both parents, fighting within the family, and repeating a school grade. Conduct problems were not significantly associated with positive parental monitoring, negative parental monitoring, moral parenting, negligent parenting, having clear family rules and responsibilities, or school performance.
Cruzeiro et al. [
29] investigated risk factors for conduct problems among 1,145 adolescents (11–15 years old) in a cross-sectional household survey in Pelotas (RS) (see Table
2). The following risk factors predicted higher conduct problem scores: higher age (13–15 compared with age 11–12 years), low social class (OR = 1.7; 95 % CI 1.1–2.5), having repeated school years (OR = 1.5 for twice or more; 95 % CI 1.1–2.1), no religion (OR = 1.3; 95 % CI 1.0–1.6), no participation in protestant service or catholic mass (OR = 1.4; 95 % CI 1.1–1.8), alcohol use/drunkenness (OR = 2.6; 95 % CI 2.0–3.5 and OR = 3.0; 95 % CI 1.8–5.1, respectively), smoking cigarettes (OR = 2.3; 95 % CI 1.3–3.9), drug use (OR = 5.9; 95 % CI 3.3–10.6), depression (OR = 5.1; 95 % CI 1.1–25.2), and victim of bullying (OR = 2.1; 95 % CI 1.6–3.0). Years of schooling were not significantly associated with conduct problem scores. In a multivariate model, significant independent predictors were male sex, age, lower socioeconomic status, use of alcohol or drugs, and victim of bullying.
Cucchiaro and Dalgalarrondo [
33] examined whether school students in a poor, outer region of Campinas (SP) had different rates of conduct problems compared with students in the central area of the city, in a cross-sectional study of 424 children (10–13 years old; see Table
2). The outer region of the city was characterised by lower levels of paternal educational, lower indices of wealth, and a higher proportion of black children. Rates of conduct problems were similar between children in the two areas (6.0 % in central areas and 7.9 % in outer-city areas; OR = 1.3; 95 % CI 0.8–2.4; not significant).
Curto and colleagues [
32] examined risk factors for conduct problems among 248 adolescents (11–17 years old) in a cross-sectional study in Embu (SP) (see Bordin et al. [
31] in Table
2). Risk factors associated with higher levels of conduct problems were severe physical punishment (OR = 2.8; 95 % CI 1.4–5.8), adolescent internalising problems (OR = 7.8; 95 % CI 3.3–15.7), and maternal anxiety/depression (OR = 2.9; 95 % CI 1.5–5.6). Variables not significantly associated with conduct problems in bivariate tests were adolescent age, maternal education, maternal paid work, marital violence, father absence, and socioeconomic status. In a multivariate model, significant risk factors for conduct problems were severe physical punishment, internalising problems, father absence, and three interactions: age*internalising, age*maternal anxiety/depression, and maternal work*socioeconomic status. The three interactions showed that (1) younger adolescents had higher risk for conduct problems than older adolescents only if they also had internalising problems; (2) older adolescents had higher risk for conduct problems than younger adolescents only if their mother was anxious/depressed; and (3) adolescents with non-working mothers had higher risk for conduct problems than other adolescents only in situations of low socioeconomic status.
Ferriolli et al. [
43] assessed risk factors for conduct problems in a cross-sectional study of 100 children (6–12 years old) in Riberão Preto (SP) (see Table
2). The only variables significantly associated with conduct problems were not having a well-defined daily routine and not having a place to do homework in the house. Non-significant variables were poor parental relations, maternal depression, maternal stress, leisure activities, financial instability, and socioeconomic level.
Rodriguez et al. [
44] examined associations between perinatal and socioeconomic factors measured in the first year of life and conduct problems at ages 7–9 years in the São Luís (MA) Prospective Birth Cohort Study (see Table
2). Prevalence ratios (PR) were used to report the relative probability of conduct problems comparing children with and without risk factors. Conduct problems were more common among children whose mothers had 5–8 years of schooling compared with over 8 years of schooling (PR = 1.4; 95 % CI 1.2–1.7), and among children in middle income families compared with children in families with high income (PR = 1.3; 95 % CI 1.1–1.6). Nonsignificant variables were preterm birth, birth weight, maternal/paternal age, and mother’s marital status. In a multivariate model, the only significant predictors were male sex and lower maternal schooling.
Sherman et al. [
35] conducted a cross-sectional study of 263 adolescents (11−18 years old) in Salvador (BA) (see Table
2). Risk factors for conduct problems were analysed separately for boys and girls. For boys, conduct problems were associated with low religiosity (OR = 1.3; 95 % CI 1.0–1.6), low family cohesion (OR = 2.0; 95 % CI 1.2–3.5), and family conflict (OR = 2.5; 95 % CI 1.4–4.4). For girls, conduct problems were associated with parents not being married (OR = 5.9; 95 % CI 1.7–20.2), low family cohesion (OR = 2.1; 95 % CI 1.5–3.0), and family conflict (OR = 2.5; 95 % CI 1.7–3.8). Variables not significant for either sex were race, maternal/paternal education, and parental unemployment. In multivariate models, the only significant correlates of conduct problems were low family cohesion and family conflict for girls.
Vitolo et al. [
36] assessed risk factors for conduct problems among 454 children (7–11 years old) in a cross-sectional study in Taubaté (SP) (see Table
2). Children being hit with a belt was associated with increased risk for conduct problems (OR = 2.2; 95 % CI 1.2–2.3), as was parental mental health problems (OR = 2.1; 95 % CI 1.4–3.3) and low social class (OR = 1.6; 95 % CI 1.1–2.3). All these risk factors remained significant in multivariate models. Goodman et al. [
45] also examined correlates of conduct problems among 1,112 children in the same study (including children with a wider age range 7–14 years old). Risk factors independently associated with conduct problem symptoms in a multivariate model were not living with both biological parents, alcohol abuse in the family, parental stress, and harsh physical punishment. Non-significant variables were child age, general health, and IQ.