Bulletin of the World Health Organization

Exposure to physical and sexual violence and adverse health behaviours in African children: results from the Global School-based Student Health Survey

David W Brown a, Leanne Riley a, Alexander Butchart a, David R Meddings a, Laura Kann b & Alison Phinney Harvey a

a. World Health Organization, 20 avenue Appia, 1211 Geneva 27, Switzerland.
b. Centers for Disease Control and Prevention, Atlanta, GA, United States of America.

Correspondence to David W. Brown (e-mail: dwbrown.6@gmail.com).

(Submitted: 10 September 2007 – Revised version received: 10 September 2008 – Accepted: 19 September 2008 – Published online: 05 May 2009.)

Bulletin of the World Health Organization 2009;87:447-455. doi: 10.2471/BLT.07.047423


Exposure to violence and traumatic stressors among children is common1 and has both short- and long-term effects on multiple health behaviours (e.g. smoking, substance abuse, physical inactivity) and health outcomes (e.g. higher prevalences of heart, lung, and liver disease, diabetes, and depression).2,3 Moreover, such exposure appears to influence health behaviours and outcomes through a cumulative process.2

While our understanding of the burden of violence and its relationship with adverse health behaviours has increased globally, studies on such questions in children in Africa are lacking and until recently few data were available. An understanding of such relationships is important for policy and programme planning efforts. With this in mind, we estimated the prevalence of exposure to physical violence (PV), sexual violence (SV) or both among children from five African countries and examined the association between exposure to violence and several adverse health behaviours during childhood using data from the Global School-based Student Health Survey (GSHS).


The GSHS is a self-administered, school-based survey developed by WHO in collaboration with the United Nations Children’s Fund, the United Nations Educational, Scientific and Cultural Organization, and the Joint United Nations Programme on HIV/AIDS, and with technical and financial assistance from the United States Centers for Disease Control and Prevention in Atlanta, GA. The survey is conducted primarily among students 13–15 years of age and can be administered during one regular class period. In each country, the questionnaire comprises multiple core modules, core-expanded questions and country-specific questions, and a standardized scientific sample selection process and common school-based methods are followed. Further details of the GSHS can be obtained at http://www.who.int/chp/gshs and http://www.cdc.gov/gshs

In this analysis, we pooled data from five African countries – Namibia, Swaziland, Uganda, Zambia, Zimbabwe – where the survey was administered during 2003 or 2004 (Table 1). These countries were selected because each included questions on exposure to PV and SV as well as questions on mental health, tobacco use, the use of alcohol and other drugs, and sexual behaviours, though not all countries asked all questions under each domain. For example, Swaziland and Zambia had no questions on cigarette smoking, Swaziland had none on suicidal ideation, and Namibia and Uganda had none on the history of sexually transmitted infections (STI).

Exposure to PV and SV

Survey questions about exposure to PV and SV and variable definitions are available in Appendix A, available at: http://www.who.int/bulletin/volumes/87/6/07-047423/en/index.html). To examine associations with the combined occurrence of exposure to PV and SV, we created a score variable that was assigned the values 0 in the absence of exposure to both PV and SV, 1 in the presence of exposure to one form of violence or the other, and 2 in the presence of exposure to both PV and SV.

Adverse health behaviours or events

We compared the relative frequency and likelihood of several adverse health behaviours or events in children who had and had not been exposed to PV and/or SV. Survey questions and variable definitions for these behaviours are available in Appendix B, available at: http://www.who.int/bulletin/volumes/87/6/07-047423/en/index.html.

Statistical analysis

When judging the appropriateness of combining GSHS data across countries, we took into consideration sampling design, sampling error and nonsampling error. Prior research from the United States has shown that pooling state-based survey data to obtain national estimates is feasible depending on sampling and nonsampling error.4 In the GSHS, sampling designs were similar across countries (Table 1). The sampling of students within each country was conducted at two levels: the school and the class. Schools and classes included in the sampling frame for each country are provided in Table 1. We examined country-level sampling errors (large sampling errors imply imprecise survey estimates) by using the coefficient of variation of survey weights and the design effect averaged across all survey items for each country. Averaged design effects ranged from 1.5 for Zimbabwe (Harare) to 2.2 for Namibia; coefficients of variation ranged from 0.3307 for Uganda to 0.7677 for Namibia. We examined nonsampling errors using survey response rates. Response rates ranged from 69% (Uganda) to 96% (Swaziland) (Table 1). Of the 24 845 observations available for analysis after pooling country data sets, 92% (n = 22 656) had complete information on age, sex and the variables for exposure to PV and SV.

We compared the relative frequency of each of the adverse health behaviours or events noted above in children who reported exposure to PV or SV and in those who did not report such exposure. A χ² test was used to compare differences in categorical variables across groups. We examined associations with exposure to PV and SV individually and to both PV and SV by using a score variable approach. After adjusting for age and sex, we used unconditional logistic regression to obtain adjusted odds ratios (aORs) and 95% confidence intervals (CIs) for associations between PV and/or SV and each of the adverse health behaviours. Ordinal trend tests used logistic regression with the dependent variable of interest and an ordinal independent variable. To make the GSHS data representative of each country included in the analysis, sample weights were used. To account for the complex sampling design and to obtain accurate variance estimates, we used Stata version 9 (Stata Corp., College Station, TX, USA) to complete all analyses. All statistical inferences were based on a 0.05 significance level.


Of the children sampled, 42% (standard error, SE: 0.95) had been exposed to PV during the 12 months preceding the survey; and 23% (SE: 0.80) had been exposed to SV during their lifetime (Table 2). Exposure to both PV and SV was reported in 12% (SE: 0.66); to PV or SV but not both, in 41% (SE: 0.74); and to no violence, in 47% (SE: 1.00), with differences by age group noted (P ≤ 0.05). After age adjustment, the odds of having been exposed to PV were greater among boys than girls (aOR: 1.21; 95% CI: 1.06–1.39; P ≤ 0.01). Similarly, the odds of having been exposed to SV were greater among girls than boys (aOR: 1.29; 95% CI: 1.12–1.48; P ≤ 0.001) after age adjustment, although differences between genders were not large (Table 2). Exposure to violence differed across countries (P ≤ 0.001); exposure to PV ranged from 27% to 50% and exposure to SV, from 9% to 33% (Fig. 1).

Fig. 1. Exposure to PVa during the 12 months preceding the survey and exposure to lifetime SVb in a sample of children from five African countries that participated in the GSHS, 2003–2004
Fig. 1. Exposure to PV<sup>a</sup> during the 12 months preceding the survey and exposure to lifetime SV<sup>b</sup> in a sample of children from five African countries that participated in the GSHS, 2003–2004
GSHS, Global School-based Student Health Survey; PV, physical violence; SV, sexual violence.
a Exposure to PV was defined by response to the question, “During the past 12 months, how many times were you involved in a physical fight?”.
b Exposure to SV was defined by response to the question, “Have you ever been physically forced to have sexual intercourse?”

Mental health and suicidal ideation

Approximately 16% (SE: 0.55) of the children reported feeling loneliness always or most of the time during the 12 months preceding the survey; 33% (SE: 0.67) reported never feeling lonely; 14% (SE: 0.68) reported rarely feeling lonely, and 37% (SE: 0.71) reported feeling lonely sometimes. The distribution did not vary by age or sex. Compared to unexposed children, those exposed to PV or SV had greater odds of reporting persistent loneliness (Table 3). Compared to children who did not experience either PV or SV, those exposed to both forms of violence had more than twice the odds of feeling lonely most or all of the time (Table 3).

Persistent sleep problems as a result of worry were reported by 17% (SE: 0.61) of children, and their frequency was similar across age groups and in boys and girls. A third (35%; SE: 0.81) of the children reported no such sleep problems and 14% (SE: 0.68) reported rare occurrences. The odds of persistent sleep problems due to worry were three times greater among children with ≥ 6 exposures to PV and two times greater among children exposed to SV than among unexposed children (Table 3).

Nearly one in four (24%; SE: 0.84) children reported having considered suicide and 29% (SE: 0.93) reported having planned suicide during the 12 months preceding the survey. Associations between exposure to PV and both measures of suicidal ideation were strong and dose-related (P for trend < 0.001), and the odds of having had suicidal ideation were twice as high among children exposed to SV as among unexposed children (Table 4).

Substance use

The frequency of current cigarette use (8%; SE: 0.52) was similar across age groups but slightly greater among boys (10%) than girls (6%) (P < 0.001). The associations between exposure to PV and current cigarette use were strong and dose-related (Table 5). Similarly strong associations were observed for current alcohol use (frequency, 24%; SE: 0.90) and lifetime drug use (frequency, 19%; SE: 1.03) (Table 5). Finally, current frequent alcohol use also was associated with exposure to PV (≥ 6 episodes versus 0, aOR: 14.00; 95% CI: 9.39–20.89) or SV (aOR: 3.65; 95% CI: 2.63–5.06) (data not shown).

Risky sexual behaviour

One in five (SE: 0.95) boys and one in 10 (SE: 0.90) girls reported multiple sex partners. More than one in five (SE: 1.19) children reported having had an STI. After adjustment for age and sex, children exposed to PV or SV had significantly greater odds of reporting risky sexual behaviours or a history of STI (Table 6).


Through secondary analyses we explored whether the associations described above between exposure to PV or SV and the adverse health behaviours differed for girls and boys (i.e. whether the effect was modified by gender). Effect modification by gender (evaluated at P ≤ 0.05) was observed for the relationship between violence and some adverse health behaviours, but not all behaviours. For instance, the odds of using drugs were 9.53 (95% CI: 7.14–12.73) times greater among girls exposed to 6 or more physical fights than among girls exposed to none, while in boys exposed to 6 or more physical fights the odds were only 5.65 (95% CI: 4.04–7.89) times greater than in unexposed boys (P for interaction = 0.033). Stronger associations were also observed among girls between exposure to PV and multiple sex partners (girls, ≥ 6 versus 0 fights: OR: 8.36; 95% CI: 4.89–14.27; boys, ≥ 6 versus 0 fights: OR: 3.64; 95% CI: 2.52–5.26) (P for interaction = 0.039) and between exposure to SV and multiple sex partners (girls: OR: 4.45; 95% CI: 3.28–6.03; boys: OR: 2.37; 95% CI: 1.89–2.98) (P for interaction = 0.001). Only the association between forced sexual intercourse and planned suicide was slightly stronger for boys (OR: 2.43; 95% CI: 2.04–2.90) than for girls (OR: 1.63; 95% CI: 1.37–1.95) (P for interaction = 0.002).


In the current study, exposure to PV or SV was found to be common among boys and girls in five African countries, where 1 in 10 children were exposed to both PV and SV. We found strong associations between exposure to PV, SV or both and multiple adverse health behaviours during childhood. In the case of exposure to PV, associations with health behaviours were dose-related, so that increases in exposure were associated with increased odds of showing the behaviours. In addition, the presence of multiple forms of violence was associated with increased odds of adverse health behaviours.

Our findings from five African countries contribute to the understanding of the relationship between violence and adverse health behaviours among children in several ways. First, they lend further support to research findings from developed countries that show exposure to violence is related to adverse health behaviours. Many of the health behaviours examined in our study are known to have implications for adult health behaviours and health outcomes later in life (e.g. smoking and cardiovascular and respiratory disease). Health behaviours such as cigarette smoking, alcohol use, or drug use may serve, either consciously or unconsciously, as coping mechanisms in the presence of stress resulting from exposure to violence. To the extent that such behaviours feel like effective and immediate solutions (through coping processes), they may become chronic and affect health in adulthood. Therefore, continued research is needed on the long-term implications of childhood exposure to violence. Researchers and programme managers trying to understand and prevent adverse health behaviours such as those discussed here may benefit from considering exposure to violence during childhood as a point of entry to intervene.

Second, we observed an association between exposure to PV among peers and adverse health behaviours. Studies of childhood maltreatment and its relationship to adverse health behaviours most often focus on maltreatment at the hands of parents or caretakers. Our findings expand this literature by showing that in some African countries peer violence is associated with similar adverse health behaviours, both among boys and girls.

Our findings also contribute to the understanding of the burden of exposure to PV and SV among boys and girls in Africa. First, few data exist on the burden of childhood violence outside of South Africa. Second, the data from the five African countries we studied suggest that physical fighting is more common among girls and SV more common among boys than data from other more developed countries would lead one to predict. For example, we observed exposure to PV (based on involvement in any physical fight during the 12 months preceding the survey) among 36–47% of the girls, depending on age (Table 2). A cross-sectional survey of grade 8 students conducted during 1997 and 2004 in Cape Town showed involvement in physical fighting during the 12 months preceding the survey in 16% (95% CI: 14–18%) of girls.5 Similarly, physical fighting during the 12 months preceding the survey varied from 13% among girls participating in the 2001/2002 Health Behaviour in School-Aged Children survey in Finland to 32% among those participating in Hungary.6

Much of the published evidence shows that more girls than boys are victims of sexual abuse. For instance, in a sample of Cape Town high school students King et al.7 found that 2.0% of male and 13.3% of female respondents had been victims of attempted rape; 5.0% of males and 6.0% of females had been victims of completed rape, and the odds of having been the victim of attempted or completed rape were four times as high among females as among males. Our findings highlight the importance of not neglecting SV among boys and show that among African children exposure to SV and its potential consequences are as common among boys as among girls, contrary to what the existing literature may lead one to assume.

In Africa, understanding the burden of PV or SV during childhood and its association with health problems is a challenge due to insufficient data for establishing these associations. While children homicide rates in Africa are known to be among the highest in the world,8 the magnitude of the problem of fatal and non-fatal violence in children is unclear for most African countries. Further investment is needed in information systems for routine monitoring of trends in violent behaviour, injuries and deaths. Surveillance systems being developed with uniform standards for defining and measuring violence should also incorporate information from other sources, including health services (e.g. emergency departments) and law enforcement, education and other authorities.8 In addition, a continued and expanded commitment to routinely collect survey data on exposure to violence and health risk behaviours will be required to provide the data needed to further understand the complex interrelationships between violence and health behaviours and outcomes throughout the lifespan.

Our results should be interpreted in light of the following limitations. The GSHS is a school-based survey, so these data do not include children who do not attend school or who were absent from school the day the survey was administered, who may have higher prevalences of adverse risk behaviours.9 Because of the cross-sectional nature of the data, determinations and statements of cause and effect are impossible. We adjusted for age and gender in multivariable logistic regression models but were unable to account for possible differences in socioeconomic status and other covariates that might be associated with the school one attends, the prevalence of PV and/or SV or the prevalence of adverse health behaviours. Also, the time frames for health behaviour questions varied, with reference to either 30 days preceding the survey, 12 months preceding the survey, or to the entire lifetime. Also, our measure of current smoking does not distinguish those who experiment with cigarettes from those who smoke on a regular basis; however, symptoms of serious nicotine addiction often occur just days or weeks after youths first begin to “experiment” with smoking.10,11

These data are self-reported. Conceivably, some respondents may have misreported their exposure to PV or SV or their adverse health behaviours either out of embarrassment or to provide a socially desirable response. In addition, the measure of STIs is intended to capture only those infections diagnosed by a doctor or nurse; thus, we may be underestimating the true prevalence of STIs. Despite the potential problems of self-reported data, there is no reason to believe that children would systematically misreport in a manner that reflects the associations observed herein. Any misreporting is likely to have been nondifferential and thus would have biased our results towards the null hypothesis.

We must also keep in mind that the survey question on exposure to PV is focused on physical fighting among peers. Other types of PV, such as beatings by parents or teachers or violence that occurs between boyfriend and girlfriend, may not have been captured, so the prevalence of exposure to PV may be underestimated. Similarly, the survey question on exposure to SV did not distinguish between a perpetrator who was a peer or someone older than the student (e.g. parent, family member, caretaker, etc.).

In summary, these are among the first estimates of the burden of PV or SV among African children and their associations with adverse health behaviours. Because childhood exposure to violence is common and has strong associations with multiple health behaviours, increased attention to primary, secondary and tertiary prevention is needed. Increased awareness of the frequency of exposure to violence among children and potential health consequences in developing countries may lead to improvements in health promotion and disease prevention programmes. Further research and training are needed to help health professionals and programme managers recognize and understand the linkages between childhood exposure to violence and adverse health behaviours during childhood and across the life span. Tools and information, such as the World report on violence and health,8 the World report on violence against children1 and Preventing child maltreatment: a guide to taking action and generating evidence,12 are available to guide organizations in their efforts to prevent and respond to childhood exposure to violence. ■


The authors thank the country coordinators from Namibia (Kornelia K. Abraham), Swaziland (Mildred Xaba), Uganda (Jermiahs Twa-Twa), Zambia (George Sikazwe) and Zimbabwe (Edwin Sithole) for their assistance in collecting the Global School-based Student Health Survey data.

This research was completed while David W Brown worked in the Department of Injuries and Violence Prevention at the World Health Organization.

Competing interests: None declared.