Bulletin of the World Health Organization

Severe physical punishment: risk of mental health problems for poor urban children in Brazil

Isabel A Bordin a, Cristiane S Duarte b, Clovis A Peres a, Rosimeire Nascimento a, Bartira M Curto a & Cristiane S Paula c

a. Universidade Federal de São Paulo, Escola Paulista de Medicina, Rua Botucatu 572, São Paulo, SP, 04023-061, Brazil.
b. Columbia University, New York State Psychiatric Institute, New York, NY, United States of America.
c. Universidade Presbiteriana Mackenzie, São Paulo, SP, Brazil.

Correspondence to: Isabel A Bordin (e-mail: fbordin@dialdata.com.br).

(Submitted: 11 April 2008 – Revised version received: 28 May 2008 – Accepted: 13 June 2008.)

Bulletin of the World Health Organization 2009;87:336-344. doi: 10.2471/BLT.07.043125


Few epidemiological surveys have focused on child mental health in developing countries. A literature review (1980–1999) on population studies in Latin America1 identified 10 studies in which prevalence rates of mental health problems of 15% to 21% were found. Only nine studies in Brazil that used more rigorous methods can be found in the literature (1980–2006); all of them used probabilistic sampling and standardized instruments specifically developed to identify mental problems or disorders in children and/or adolescents.2 Based on screening instruments, prevalence rates varied according to informant, from 8.3% for teachers to 35.2% for parents; based on diagnostic interviews, such rates varied from 7.0% to 12.7%.

Several types of factors may increase a child’s vulnerability to mental health problems: biological factors (central nervous system abnormalities due to injury, infections or malnutrition), genetic factors (family history of depression), psychosocial factors (marital discord, maternal psychopathology or paternal criminality), stressful life events (parental death or separation from parents) and exposure to physical or sexual abuse.3 The cumulative effect of multiple factors is more important in determining a child’s emotional or behavioural problems than the presence of an isolated stressor, regardless of its magnitude.4 In Brazil, the negative impact of different stressors has been shown to affect child mental health; stressors include witnessing marital violence,5 living with a mentally disturbed mother5,6 and suffering physical aggression at home.57 However, Brazil lacks a comprehensive examination of the relationship between factors that can be prevented (e.g. corporal or physical punishment) and child mental health problems.

Corporal punishment is associated with mental health problems during childhood and adolescence (antisocial behaviour, depression) and in adult life (aggressive or criminal acts, antisocial behaviour, alcoholism, depression).8 This phenomenon occurs worldwide, but most studies on child maltreatment have been conducted in developed countries.9 According to the Pan American Health Organization’s Team on Violence and Injury Prevention,10 corporal punishment is more prominent in Latin America, whereas sexual abuse and exploitation are more common in the Caribbean. Nevertheless, methodologically sound data on child abuse in Latin America and the Caribbean are lacking, so that the magnitude of the problem remains poorly defined.

The United Nations Secretary General’s Study on Violence against Children (2006)11 recommended the development of a national research agenda on violence against children across settings where violence occurs. This study is part of WorldSAFE (World Studies of Abuse in the Family Environment), a project involving researchers from six countries.12 In our pilot study7 we found high rates of child mental health problems (22.4% in children aged 4–17 years) and severe physical punishment (10.1% in children aged 0–17 years), which confirms the need to explore the potential association between these factors. This study represents the first use of a comprehensive approach for examining – in a probabilistic household sample of children from a developing country (Brazil) living in an impoverished and violent area – whether and in what way severe physical punishment may be related to different types of mental health problems in children.


Study design and sampling

We conducted a cross-sectional study in south-eastern Brazil, in a typical urban poor neighbourhood of the city of Embu. The city has 238 891 inhabitants and is located in the boundaries of São Paulo city in the great São Paulo area, where more than 10% (19 million) of the Brazilian population lives. Embu is totally urban, and 38.1% of its inhabitants are under 20 years of age. It is characterized by neighbourhoods of small households and slums and was considered one of the most violent cities in the country at the time the data were collected (April 2002 to February 2003).

The Brazilian Institute of Geography and Statistics randomly selected 24 clusters in the chosen neighbourhood, based on census units. In these clusters, we identified 996 eligible households (those having a woman aged 15–49 years with at least one child < 18 years of age). One mother–child pair was randomly selected per household. At the time of the interview, nine pairs had to be excluded (mother and child not living together due to separation or death, mother with health problems that impaired communication, or mother unable to be interviewed because of death threats by local drug dealers). From the initial sample (n = 987), 813 mothers completed the WorldSAFE Core Questionnaire (attrition rate: 17.6%). Sample size was calculated based on the lowest expected outcome prevalence (5%; 95% confidence interval, CI: 3%–7%) with an adopted relative precision of 40%.13 The present study was restricted to children aged 6–17 years of age, with complete information on the three outcomes and 12 independent variables of interest (n = 480). Five subjects were excluded due to missing data.

Variables and instruments

Variables were measured with standardized instruments delivered to mothers by trained interviewers. These included the Child Behaviour Checklist, the Self Reporting Questionnaire and a family economic classification questionnaire developed by the Brazilian Association of Research Companies. Child mental health problems (outcomes) were defined as internalizing problems only, externalizing problems only, and both internalizing and externalizing problems (comorbidity). “Internalizing” corresponds to inwardly-directed emotional and subjective symptoms that are hard to observe and cause suffering to the individual affected (e.g. ideas of suicide); “externalizing” corresponds to outwardly-directed behavioural symptoms that disturb other people (e.g. robbery, vandalism). Anxiety and depression at a clinical level are thus considered internalizing mental health problems, whereas aggressive behaviour and breaking of rules at a clinical level are considered externalizing mental health problems. We examined potential correlates from four domains: the child (gender, age, severe physical punishment, ever witnessing physical marital violence); the mother (education, unemployment, anxiety or depression, severe physical marital violence); the mother’s husband or partner (not residing in the household in the last 12 months, unemployment, getting drunk at least once in the last 12 months); and the family (socioeconomic status). Child severe physical punishment was defined as the child being hit with an object (e.g. a stick, broom, cane or belt); being kicked, choked, smothered, burnt, scalded, branded, beaten (i.e. hit repeatedly with an object or fist) or threatened with a weapon (such as a knife or gun) by the mother or her husband or partner in the last 12 months. Severe physical marital violence in the last 12 months was defined as the mother being kicked, hit, beaten, or threatened or assaulted with a weapon by a residing husband or partner.

The WorldSAFE Core Questionnaire on Domestic Violence investigates intrafamilial violence and associated factors (the original English questionnaire was developed by the WorldSAFE steering committee). The child-rearing behaviour section included parental behaviours usually noted in developing countries and items from the Parent–Child Conflict Tactics Scales14 (included with the authors’ permission). The steering committee decided that mothers would not be questioned about violence in the last 12 months from non-residing husbands or partners. The Core Questionnaire was translated to Portuguese, back translated to confirm the quality of the translation, field tested and applied in a pilot study before being used in the full study. (The Brazilian version of the WorldSAFE Core Questionnaire was developed by Bordin & Paula in 1999).

The Child Behaviour Checklist (CBCL/6–18) is a standardized parent report screening questionnaire that is used to identify emotional and behavioural problems in children and adolescents; it has adequate psychometric properties.15 The Brazilian version of CBCL/6–18 was developed by Bordin et al. in 2002. The last version of CBCL/6–18 was CBCL/4–18; validity studies demonstrated the high sensitivity of the Brazilian version of CBCL/4–18 compared to the “gold standard” psychiatric diagnosis based on the International statistical classification of diseases and related health problems, 10th revision and the Diagnostic and statistical manual of mental disorders, 4th edition criteria.16,17 CBCL/4–18 and CBCL/6–18 are very similar (only 6 out of 118 items were changed from the 1991 English version to the 2001 English version), so we assumed the validity of the Brazilian version of CBCL/6–18. In our study, children and adolescents with T scores > 63 (above the 90th percentile according to American normative data) on internalizing or externalizing scales were classified as clinical cases.15 Borderline cases (T scores 60–63) were considered non-clinical.

The Self Reporting Questionnaire (SRQ-20) is a screening instrument developed by WHO; it has 20 items that can be used in community and primary care settings, especially in developing countries, to identify symptoms that may be indicative of mental disorders.18 The current version detects probable cases of anxiety and depression. The Brazilian version of SRQ-20 has good validity and high reliability.19,20 Mothers with a total score > 7 were considered cases.18,20

The Brazilian Association of Research Companies developed a family economic classification questionnaire to determine socioeconomic classes according to family purchase power.21 The instrument is based on the number of home appliances, the existence of private bathrooms inside or outside the dwelling, the educational level of the head of the household, and the number of home employees working at least 5 days a week. Total scores were used to determine families’ socioeconomic status (low: 0–16; middle–low: 17–34).


The Research Ethics Committee of Universidade Federal de São Paulo approved this study. Trained interviewers obtained written informed consent and then administered the instruments to mothers. All questions about intrafamilial violence were asked by female interviewers, following the WorldSAFE standardized procedure. All interviews were conducted individually at the local health centre to guarantee the privacy and safety of both mothers and interviewers. Conducting interviews at home in a violent community could have exposed women and interviewers to aggression; also, family members or neighbours could have overheard the interviews and provided the information to potential perpetrators. People involved in drug trafficking could also have become suspicious of strangers in the home asking questions and could have harmed or threatened the women and interviewers.

In Brazil, reporting of suspected child abuse and neglect is mandatory, but survey interviewers are exempt from this rule. At the end of the interviews, mothers were provided with a list of local institutions that could help with physical illness, mental disorders or intrafamilial violence.


Our sample of children (n = 480) was divided into four mutually exclusive groups: (i) children with internalizing problems only (Group 1; n = 92); (ii) children with externalizing problems only (Group 2; n = 33); (iii) children with both types of problems (Group 3; n = 52) and (iv) children without mental health problems (Group 4; n = 303). Groups 1–3 were each combined with Group 4 to create three subsets of data. For each combination of two groups, a logistic regression model was adjusted to the respective data using SPSS, version 10.0 (SPSS Inc., Chicago, IL, United States of America). When examining collinearity among 12 independent variables in all subsets, only two husband/partner variables were collinear (Cramer’s V > 0.60): residing in the household (for the last 12 months) and currently working for pay. All independent variables except the husband or partner’s unemployment were forced into the initial multivariate models.

All possible interactions with age, gender and severe punishment were initially explored by univariate analysis. When an interaction was present, the association between the risk factor and the outcome variable differed depending on age (11–17 years; 6–10 years), gender (female; male) or severe punishment (yes; no). Only interactions that reached a statistical level of significance of P < 0.08 when forced to enter the full model were retained in the model. Backward elimination of interactions (first) and independent variables (second) was applied, and interactions or independent variables were dropped manually, one at a time, according to the highest P-value. Final best associative models identified independent correlates and significant interactions (P < 0.05). A likelihood ratio test was applied to compare the full and the reduced models (where a non-significant test indicated no difference existed between models).


The mean age of the children was 11.0 ± 3.4 years. Informants (22–49 years of age; mean age: 37.1 ± 6.5 years) were biological mothers (97.9%) or female guardians (2.1%). The residing husband or partner was the child’s biological father in 88.1% of families with a husband or partner who had resided in the household for the last 12 months (388). Socioeconomic indicators revealed low living standards: 62.7% of mothers or female guardians did not complete 8th grade, 42.9% of mothers or female guardians and 18.3% of husbands or partners currently residing in the household (378) had no paid job, and 77.7% of families had low socioeconomic status (Table 1).

The data on mental health problems showed that 36.9% of children were affected by internalizing problems, externalizing problems or both, while 30.8% of mothers or female guardians had probable anxiety or depression. Also, 21.3% of mothers or female guardians reported that their husband or partner had been drunk at least once in the last 12 months. Exposure to physical aggression in the home was frequent: 20.0% of children had suffered severe physical punishment by one or both parents in the last 12 months, and 18.8% of children had witnessed marital physical violence (i.e. had ever seen or heard their mother being physically harmed or threatened by a resident husband or partner) (Table 1).

Table 2 summarizes the demographic characteristics, mental health problems and violence exposure of children and parents in the three subsets of data. The frequencies of all potential risk factors were similar across subsets.

The best associative model for internalizing problems only (model I, Table 3) identified four independent correlates: gender (female), age group (11–17 years), mother currently working for pay (yes) and maternal anxiety or depression (present), with no interactions. Girls were at greater risk than boys for internalizing problems only. Adolescents (11–17 years) were twice as likely to present internalizing problems as children (6–10 years). Offspring of women currently working for pay were at greater risk for internalizing problems alone than offspring of non-working women. Maternal anxiety or depression was strongly associated with internalizing problems only.

The best associative model for externalizing problems only (model II, Table 3) identified four factors, all of them part of an interaction: age group – by residing husband/partner, and severe punishment – by maternal anxiety/depression. In the age group 6–10 years, children without a residing father or male guardian were as much at risk for externalizing problems only as those with a residing father or male guardian (odds ratio, OR: 0.42; 95% CI: 0.09–2.02). However, in the age group 11–17 years, adolescents without a residing father or male guardian were five times more likely to present externalizing problems only than adolescents with a residing father or male guardian (OR: 5.04; 95% CI: 1.28–19.79). When maternal anxiety or depression was present, children and adolescents who were exposed to severe punishment were equally at risk of externalizing problems as those who were not exposed (OR: 1.25; 95% CI: 0.28–5.58), but when maternal anxiety or depression was absent, children and adolescents exposed to severe punishment were almost eight times more likely to present externalizing problems only than children and adolescents not exposed to severe punishment (OR: 7.86; 95% CI: 3.02–20.42). Alternatively, when children and adolescents were exposed to severe punishment, those whose mother had anxiety or depression were equally at risk for externalizing problems only as those whose mother did not have anxiety or depression (OR: 0.75; 95% CI: 0.18–3.24). However, when children and adolescents were not exposed to severe punishment, having a mother with anxiety or depression put them at greater risk for externalizing problems only (OR: 4.72; 95% CI: 1.72–12.97) (not shown in Table 3).

The best associative model for comorbidity (model III, Table 3) identified three independent correlates: gender (female), severe punishment (yes) and maternal anxiety/depression (present), with no interactions. Severe punishment was associated with a more than twofold increase in the risk of presenting both internalizing and externalizing problems. Female gender and maternal anxiety or depression were both strongly associated with comorbidity.


This study had limitations. For example, because of the cross-sectional design, causal relationships between correlates and outcomes could not be determined. It has three possible sources of bias: (i) mothers may have refrained from reporting severe punishment because they were afraid that authorities would be notified; (ii) mothers may have refrained from reporting marital violence because they were too frightened to reveal abuse; (iii) the study does not account for violence perpetrated in the last 12 months by non-residing husbands, partners or boyfriends. To increase the likelihood of obtaining truthful answers, interviewers made their neutral role clear, while offering information about local services for distressed families. Other study procedures, such as having female interviewers working in privacy at the health centre, rather than conducting in-home interviews that could be overheard by family members, may have minimized reporting bias.

A previous study reported on risk factors for child mental health problems in Brazil with multivariate analysis6,22. Conducted in the city of Taubaté, south-eastern Brazil, the study used a stratified probabilistic sample of students from public (urban and rural) and private (all urban) schools (n = 1251; attrition rate: 17%).22 A screening instrument (Strengths and Difficulties Questionnaire) was also used to identify emotional (or internalizing) problems and behavioural (or externalizing) problems. Parent, teacher and youth reports were combined to identify clinical cases. In contrast to our study, a school-based sample of the general population was used, so it did not focus on high-risk children or children who were not at school. Furthermore, because the study’s main focus was not child physical punishment, its assessment was necessarily less comprehensive than ours. In the previous study, specific risk factors were identified for different types of mental health problems. Emotional problems were associated with female gender, parental anxiety or depression, and poor general health, while behavioural problems were associated with harsh punishment (e.g. child beaten with an object such as a belt or stick), parental anxiety or depression, family alcohol abuse and not living with both biological parents. The study did not examine potential interactions. When parental reports from the Strengths and Difficulties Questionnaire (n = 454, children aged 7–11 years) were considered,6 hitting with a belt, parental anxiety or depression, and low socioeconomic status were independently associated with behavioural problems. In low-income families, girls were twice as likely to present emotional problems as boys; however, in middle-income families, both genders were at similar risk. Parental anxiety or depression was also an independent predictor of emotional problems.

Our study revealed that female gender, older age and maternal working status were unique risk factors for internalizing problems only. Previous studies have demonstrated that internalizing behaviours are stable during childhood but increase during adolescence.23,24 Girls show higher mean levels of internalizing symptoms than boys, as well as sharper increases in internalizing symptoms from childhood to adolescence.25,26 Maternal employment may have a negative effect on an adolescent’s mental health if employment leads to deficient parenting practices.27 In the poor urban community where our study was conducted, women are usually overloaded with activities and obligations at home and outside and are frequently the main financial provider in the family. Such stressful life circumstances may reduce the ability of working mothers to be emotionally supportive and available to their children.

On externalizing behaviours only, our data revealed that having the mother’s husband or partner residing in the home was a protective factor for adolescents but not for children. It was found in a recent systematic review of longitudinal studies28 that a residing father or male guardian is a protective factor against externalizing behaviours for children, possibly through involved fathering, supporting the mother in her role and disciplining the children. We also found that severe punishment increased the risk of presenting externalizing problems alone, but only for children or adolescents not exposed to maternal anxiety or depression, while maternal anxiety or depression increased that risk only for children and adolescents not exposed to severe punishment. This suggests that when children perceive their relationship with parents as less warm and more rejecting (due to harsh punishment or as a consequence of maternal anxiety or depression), they may become aggressive as a means of externalizing anger, frustration or sadness. Severe physical punishment and maternal anxiety or depression do not strengthen each other’s influence; each risk factor alone is sufficiently powerful to have a deleterious effect on parent–child relations.

In our study, correlates of comorbidity included factors independently associated with internalizing behaviours only, and factors involved in significant interactions that were influencing the occurrence of externalizing behaviours only. According to the model of risk for the development of comorbidity proposed by Wolff & Ollendick,29 female gender can be considered a unique risk factor for internalizing problems, while maternal anxiety/depression can be interpreted as a common risk factor underlying the co-occurrence of internalizing and externalizing problems. Maternal depression is generally associated with lower levels of nurturance and affection, as well as with hostility and conflict.30 Depressed mothers are more verbally aversive, monitor and supervise their children’s activities less effectively, engage in fewer affectionate interactions with their children, and respond to them with less warmth.31 Children prone to externalizing problems only may respond to maternal anxiety or depression with anger and impulsivity, while children prone to internalizing problems only may respond with withdrawal and behavioural inhibition. Some researchers32 consider children with comorbidity similar to externalizing children, since both are high on anger and impulsivity but low on attention regulation and inhibitory control. Others33 suggest that children with comorbidity are unique, as externalizing children show elevated aggression in response to aversive events (e.g. warnings from adults) and internalizing children show elevated withdrawal in response to aversive peer events (e.g. teasing from peers), whereas children with comorbidity display elevated rates of aggression and withdrawal in response to positive events (e.g. friendly peer talk or adult praise).

Finally, aggressive behaviour from children may stimulate parental use of more strict disciplinary methods, and inadequate parenting behaviours may influence the development of child mental health problems. Similarly, chronically depressed mothers may use more relaxed and inconsistent discipline with their children, which favours the development and persistence of externalizing disorders,34 while the behaviour of children with conduct disorders may decrease a mother’s disposition to adopt nurturing attitudes, generating disappointment and depressive feelings.35 Psychogiou et al.36 noted that a mother’s expressed emotions were driven more by the child than by maternal characteristics, and these effects were specific to conduct and emotional problems. Therefore, physical punishment as well as parental psychopathology may be the cause or consequence of child externalizing problems, deserving attention in future longitudinal studies.


In this study we focused on a high-risk, relatively homogeneous low-socioeconomic status sample from a Latin American country, which allowed us to identify key child and family factors related to specific types of child mental health problems in a population that has rarely been systematically studied. Our study results are probably generalizable to other disadvantaged communities located in the outskirts of highly populated cities in developing countries. As such, they have important implications for designing effective interventions to tackle the development of child mental health problems among these populations. Maternal anxiety and depression and severe punishment are modifiable risk factors; therefore, intervention efforts to support parents emotionally may help them adopt more adequate child-rearing practices. ■


We thank all study participants, professionals and employees from the health centre where data were collected, Embu Health Secretariat and Mental Health Coordination, and all research team members of the Brazilian Studies of Abuse in the Family Environment (BrazilSAFE) for the valuable collaboration.

Funding: The full study was financially supported by the São Paulo State Research Foundation (Fundação de Amparo à Pesquisa do Estado de São Paulo/FAPESP – Process no. 00/14555-4). The pilot stage was funded by the International Clinical Epidemiology Network, Inc., and the Fundo de Auxílio aos Docentes e Alunos (FADA), a research fund from the Federal University of São Paulo.

Competing interests: None declared.