Household exposure to violence and human rights violations in western Bangladesh (I): prevalence, risk factors and consequences

Torture and other serious human rights violations were committed during the Bangladesh Liberation War with Pakistan in 1971, when the authorities specifically targeted political dissidents in Bangladesh [2]. In independent Bangladesh, both the Awami League and the Bangladesh Nationalist Party (BNP) have used armed groups and militias against their political opponents. They have also systematically used torture and ill-treatment as a means to suppress opposition parties and criminals [3‐5]. Torture and violent attacks against women have been widespread as well [6‐9]. A previous study of refugees from Bangladesh treated at the Centre for Torture and Trauma Survivors in Stockholm reported that all of the female refugees and one-third of the male refugees had been raped [10].

The BNP won the election in October 2001 and launched Operation Clean Heart in October 2002 as a drive against growing crime: a rise in murder, rape, and acid throwing. In March 2004, the anti-crime and anti-terrorism elite force 'Rapid Action Battalions' was founded. Up to December 2007, Rapid Action Battalions had seized a total of 23,632 people including two senior terrorists and recovered 3,592 illegal weapons based on official data from their website [11]. Their lack of accountability has been questioned, however, by the Human Rights Watch and Amnesty International [3, 4]. Daily violent attacks on the press by political militants, criminal gangs and security services were also reported by Reporters Without Borders [12]. A local non-governmental organisation (NGO), Odhikar, documented human rights abuses by gathering information through a nationwide network of trained human rights workers and from media reports: 65,057 injuries and 2,339 killings due to political clashes were identified during the period January 2002-June 2009 (Figure 1). The number of reported incidents peaked 2 months before the general election in October 2006 at 3,096 injured and 77 killed. The Awami League alleged electoral bias in favour of its rival, the BNP, and boycotted the elections. The president of the caretaker government announced a state of emergency and curfews in January 2007, due to weeks of riots and protests over erroneous voter lists, and postponed the general election until December 2008. Pre- and post-election violence was observed at the end of 2008 and the beginning of 2009.

Khulna division, which borders India, is the province most affected by violence in Bangladesh. Operation Spider Web was launched in 2004 by the BNP-led government in the south-western districts of Bangladesh after Operation Clean Heart failed to restore public order. Its primary target was the left-wing extremists along the Indian border. Substantial numbers of extrajudicial killings in Khulna division have been reported since 2004 [3, 13, 14]. The Bangladesh Rehabilitation Centre for Trauma Victims (BRCT) also reported 1,102 torture cases and 414 torture-related deaths in Khulna division in 2004 [15].

Meherpur district, with its area of 716.08 square kilometres, is the smallest district in Bangladesh. It is located in Khulna division, and borders West Bengal in India to the north. In 1971, the leaders of the Awami League declared the independence of Bangladesh in the town of Mujibnagar. Throughout the liberation war with Pakistan, the Mujibnagar upazila served as the base for the pro-independence militias. Consequently, the area was severely affected by the war. Meherpur district is currently under the threat of left-wing and Islamic extremism. In the south-western districts of Kushtia, Meherpur, and Chuadanga, where at least six outlawed parties with 4,000 armed members are active, many people were killed and frequent cases of abduction, robbery, rape, murder and other crimes were recorded. In the last 10 years, the BRCT identified and treated 126 cases of torture and organised violence in the Meherpur district. The Task Force against Torture has been working in Meherpur district, and a local NGO, Manab Unnayan Kendra, provides legal and social support to female victims of rape and domestic violence.

This is a multisite epidemiological study whose initial focus is to assess OPV and human rights violations at the population level and to study their causes and consequences at the individual and family level. This study was commissioned by the Rehabilitation and Research Centre for Torture Victims (RCT) for three countries and was implemented in Bangladesh and Kosovo in 2008. A separate review will describe this work in a wider context. The population-based survey was aimed at: 1) describing the magnitude and patterns of OPV and human rights violations; 2) estimating the prevalence of injury, OPV-related injury and pain in the violence-affected area; 3) identifying the risk determinants of victimisation, 4) quantifying the association between OPV exposure and general health conditions, as well as family financial burden. Our hypotheses are as follows: 1) there are spatial variations in levels of OPV and number of reported cases of human rights violations; 2) mortality and prevalence of injury and pain correlate with the level of OPV and human rights violations in the community; 3) a family could be at increased risk of becoming victims of OPV and human rights violations if any of the family members is active in a social or political movement or has trouble in interpersonal relationships.

A generic protocol was developed based on a WHO document "Guideline for conducting community surveys on injury and violence" [16]. It was modified to focus specifically on OPV-related injury and mortality. Optimal sample sizes were calculated using the standard statistical formula developed by the United Nations Children's Fund (UNICEF) [17]. The formula was as follows: n = [4 (r) (1-r) (f) (1.1)]/[(e²) (p) (nh)] with an anticipated prevalence of 15-30% (r), a design effect of 2 (f), a factor necessary to raise the sample size to allow for an estimated non-response rate of 10%: 1.1, a margin of error of 10% (e) of r, the proportion of the total population that the smallest subgroup comprises (p): 1, and an average household size of 4.9 (nh) in Bangladesh. The estimated number of households that needed to be surveyed in order to estimate the prevalence of lifetime experience on violence-related injury in the whole population was around 500-1,000 households.

When the household survey was finished, the list of families to be invited to the mobile clinic was prepared based on the survey results. The interviewers were told to return to the villages for which they were responsible and distribute the vouchers to the victims' families who reported incidents of: 1) torture and other cruel, inhuman or degrading treatment or punishment; 2) sexual harassment, molestation, rape or insertion of a blunt object into a genital organ and/or rectum; 3) arrest or detention; 4) extrajudicial execution of family member, perpetrated by members of law enforcement agency. The victims were invited to the mobile clinic for free examination and treatment. Two-day mobile clinics were organised at four sites in Meherpur district (Figure 2). The victims were screened and interviewed, and then underwent physical examination and consultation. The results are presented in Part II.

Only four households refused or did not complete the interview. During the household survey, every tenth to twentieth of the respondents was interviewed twice by different interviewers or randomly selected for spot-check by deputy team leaders. The data set was checked three times for discrepancies.

Data were entered and validated using Microsoft Access 2000 and Epi Info™ 6.04 (CDC Atlanta, USA, 2001). The data were analysed using Stata 9.2 (StataCorp LP, Texas, USA, 2003) at the individual, household, and village levels. The major explanatory variables were age group, sex, religion, area of residence, education and occupation of head of household, political involvement and social participation of family members, as well as the household income level: 0-1 United States dollars (USD) per day, 1-2 USD per day, 2-4 USD per day, and higher than 4 USD per day. A generalised linear model was used to assess the association between outcomes and explanatory variables, adjusted for the cluster effect (village) and the confounding effects of age groups. P values < 0.05 were considered significant.

The study complied with the Declaration of Helsinki and Danish law. The Bangladesh NGO Bureau gave approval for the study. In the household survey, the head (or the spouse of the head) of a household provided the informed consent. Confidentiality and privacy protection was respected. Basic treatment was offered to simple cases of torture or ill-treatment identified during the subsequent OPV screening at the mobile clinic. Severe cases were referred to the BRCT. The sponsor had no role in the study design, data collection and analysis, interpretation of the data, or writing of the report.

The total study population was 4,870: 50 households (4.5%) with 250 persons were located in the capital of Meherpur Sadar upazila (sub-district), 50 households (4.5%) with 217 persons were located in the suburban area in Gangni upazila, and 1,001 households (91.0%) with 4,403 persons were located in rural areas. The average size of a household was 4.4 people (range: 1-15).

The demographic profile of the study population is provided in Tables 1 and 2. The mean age of the study population was 28.0 (range: 0-102), and 8.2% were under 5 years old. Over 96% of heads of households were Muslim and 2.3% were Hindu. More than 50% of heads of households were engaged in agricultural, fishing, animal husbandry or hunting activities; nearly one-fifth were engaged in business, which is much higher than the percentage from the 2001 national census in Meherpur district. The possible explanation is that, in the border area, many people were engaged in trade and smuggling between Bangladesh and India. The total household income ranged from 0 to 36,000 Bangladesh taka per month (0 to 525 USD). Overall, 22.6% of households lived below the poverty line (1 USD per day).

Table 2 shows the prevalence of injury within 12 months, lifetime experience of violence-related injury, and pain complaints within 2 weeks. The average number of people in each family who reported injury within 12 months was 1.6 (95% CI: 1.51-1.68); for lifetime experience of violence-related injury, the number was 2.2 (95% CI: 2.08-2.32); for pain complaints within 2 weeks, the number was 2.5 (95% CI: 2.4-2.56). Almost half of the population had been injured because of violence at some point in their lifetime and 18.8% of violence-related injury was due to OPV.

In 2007, the crude mortality rate (CMR) in the study population was 17.9 per 1,000 and the under 5 mortality rate (U5MR) was 75 per 1,000 (Table 2). The survey participants reported that at some point in their lifetime, 457 people had been injured and 15 had been killed because of OPV. Of the 87 people who died within 12 months of the survey, five (5.7%) were reported to be killed because of violent attacks; this is 1.02 per 1,000 people. We estimate, therefore, that in 2007 there were approximately 679 deaths in Meherpur district due to violent attacks.

The reported pain complaints increased with age, reaching a plateau over the age of 35; there was little difference between men and women in the study population (Figure 3). More than 80% of the population over 35 years old reported experiencing pain within 2 weeks preceding the survey. With cluster effect (village) taken into account, the association between pain complaints and the lifetime experience of violence-related injury was positive for males in all age groups except for the 45-54 age group.

Overall, 83.4% of households reported being exposed to at least two categories of OPV or human rights violations (Table 3): two-thirds of the households had experienced gunshots, shelling, or bombing in the neighbourhood; 75% of the households had witnessed a violent action against their friends or neighbours. Nearly 30% (n = 317) of families reported that at least one family member had been arrested or detained without warrant. Over 30% (n = 343) of families reported that at least one family member had been subjected to torture and other cruel, inhuman or degrading treatment or punishment and among them 182 (53.1%) were arrested or detained without warrant, 266 (77.4%) had been exposed to at least four categories of OPV and human rights violations, and 337 (98.3%) had been exposed to at least two categories. Thus, torture co-exists with other forms of OPV and human rights violations.

Variations in reporting torture and other cruel, inhuman or degrading treatment or punishment between different villages are presented in Figures 2 and 4. In Mahabbatpur, Gangni-Fataipur, and Tangi, no family reported a case of torture and other cruel, inhuman or degrading treatment or punishment, while over 30 families reported cases in Kasba, Kamarpur and Shipur, which are near the border with India. Variations in annual injury rate, prevalence of lifetime experience of violence-related injury, and pain complaints were also observed among the villages (Figure 4). Higher numbers of lifetime experience of violence-related injury were also reported from Meherpur Sadar upazila.

Meherpur Sadar upazila had a higher number of reported arrests and detentions (odds ratio [OR] = 1.58, 95% CI: 1.00-2.49, P = 0.05) and combat or cross-fire incidents (OR = 26.09, 95% CI: 27.5-247.28, P < 0.005). Meherpur Sadar and Mugibnagar upazila had significantly higher odds ratio of reported extrajudicial executions than Gangni upazila: Meherpur Sadar OR = 6.38, (95% CI: 1.01-40.40, P < 0.05); Mugibnagar OR = 5.63, (95% CI: 1.41-22.45, P < 0.05 (Table 4). There was fighting in Teragharua in Meherpur Sadar upazila, and 43 cases of extrajudicial execution were reported on this occasion.

A multivariate model was built up and tested in a stepwise manner, taking parameters of univariate associations into account. Household serves as the unit of analysis and weighting of family size was adjusted. The risk for a family member to be arrested or detained was higher if the income level of the household was higher (Table 4). Adjusted to the household income level and random cluster effect of village, the likelihood for a family member to be arrested or detained was significantly higher if a member of this family: 1) was involved in party politics; 2) had ever participated in a demonstration, a strike or a human rights rally; 3) had a personal or financial conflict with other families (Table 4). A statistically significant increased risk was observed for families involved in the Awami League and for families whose members had ever participated in a demonstration, strike or human rights rally. Family members were more likely to be arrested or detained if the family had a conflict with other families.

The families whose members were involved in the Awami League or the BNP were also more likely to report experiences of torture and other cruel, inhuman or degrading treatment or punishment (Table 4). There was no correlation, however, between having conflicts with other families and the reporting of torture or other abuse experience.

Families whose members were involved in politics or had a conflict with other families in the community were also more likely to report that a family member experienced combat or cross-fire incidents or was extrajudicially executed by members of a law enforcement agency (Table 4). The odds ratio of reporting combat or cross-fire incidents was 19.53 for Awami League supporters and 21.62 for BNP supporters. The odds ratio of reporting extrajudicial execution of a family member was 9.22 for Awami League supporters, 9.15 for BNP supporters, and 3.97 for Jamaat-e-Islami Party supporters. An association was found between reporting combat or cross-fire incidents or extrajudicial execution, and participation in a political or social movement (a demonstration, a strike or a human rights rally). The reporting was also significantly higher among families that had conflicts with other families. There was a strong protective effect if relatives or friends of the family worked with a law enforcement agency: OR = 0.49, (95% CI: 0.29-0.81, P < 0.01).

Annual injury rates, lifetime experiences of violence-related injury, and pain complaints within 2 weeks were statistically associated with an increase of collective exposure to OPV or human rights violations. A generalised linear model was adjusted for the cluster effects of village, location of dwelling, household income level, and whether the family member had ever participated in a political party, a demonstration, a strike or a human rights rally. Compared with families exposed to less than four categories of OPV or human rights violations, the family exposed to at least four categories of OPV or human rights violations yielded higher regression coefficients for the proportion of household members reporting injury within 12 months: 14.10, (95% CI: 4.23-23.97, P < 0.005). The regression coefficients were also higher for the proportion of household members reporting lifetime experience of violence-related injury: 22.36, (95% CI: 12.48-32.24, P < 0.001) and pain complaints within 2 weeks: 6.65, (95% CI: 0.51-12.80, P < 0.05).

Among 1,101 households, 522 families (47.5%) reported declining household income levels, 404 (36.7%) had borrowed money because of a major injury incident with a family member, and 181 of those 404 families were still in debt at the time of the survey. Loan amounts ranged from 100 to 150,000 Bangladeshi taka (1.5-2,205 USD). The average loan amount was 9,791 Bangladeshi taka (144 USD), which amounted to approximately two to three times the monthly household income of most families (70.4% of families had a monthly income of less than 4,000 Bangladeshi taka, that is, 2 USD per day). There was a heavy social and financial burden of having a high injury rate, which was related to collective exposure to OPV and human rights violations: 416 families reported that a family member had stopped working at least once in order to take care of an injured person (Table 5). Adjusted for income level of household, education and occupation of the head of household, village, and number of family members who had a lifetime experience of violence-related injury, the results showed that families exposed to at least two categories of OPV and human rights violations were likely to have declining household income levels due to the presence of an injured person: OR = 2.29, 95% CI: 1.22-4.31, P < 0.01. It was also more likely that the family had borrowed money (OR = 2.10, 95% CI: 1.24-4.25, P < 0.05) or that family members had stopped working in order to take care of the injured person (OR = 2.89, 95% CI: 1.37-6.11, P < 0.005).

A concern has been raised that carrying out such a survey can put the oppressed population at risk of state-organised violence. Therefore, it was absolutely crucial to work with a local NGO that had the capacity to provide both legal and health support to the victims. The BRCT and Task Force against Torture have provided legal support to victims, including making bail arrangements, providing legal advice and lobbying for legislative changes for many years. Both of them will be following up on the legal and health needs of victims once our study is completed. During the household survey, the majority of respondents were female. One of the strengths of our study is that the interviewers from the local NGO Manab Unnayan Kendra are very experienced in interviewing and providing consultation to victims of domestic violence and rape. Their experience helped to increase the survey response rate (99.9%). The household members were asked to remain to confirm the information provided. On the other hand, it made it embarrassing for the victims of sexual violence to reveal such an attack in front of their family members. Previous experience showed us that the victims of forced sexual contact or sexual torture were often in denial of tragedy or not ready to speak about it openly at the first or second approach by the social workers. They were likely to reach out for help afterwards. The brochures of the BRCT and Manab Unnayan Kendra were provided for their reference as resources that offer assistance or legal consultation for such types of victims. We believed that the numbers of victims of sexual violence in our survey were under-reported because of social stigma. The major limitations of this study are the cross-sectional and retrospective nature of the survey and there is concern about the accuracy and reliability of the survey participants' responses. Memory bias exists although a 10-year-recall is considered reliable [45, 46]. There is selection bias in that only one person from each household was interviewed. The head or the spouse of the head of the household may not necessarily know everything that happened to family members. To reduce the bias in responding, the other available household members were asked to remain to verify information. It is also possible that people in the village misunderstood the definition of each category of OPV and human rights violations. Finally, many of these limitations stem from the difficulty of conducting research in a society that is polarised politically. Some people are afraid to talk about the political violence and some party supporters accuse each other. The victim families were self-reported, and they may have expected to receive humanitarian aid. To avoid such expectations, interviewers did not mention the vouchers and subsequent mobile clinic activity and explained that there would be no compensation for participation in the survey. The reporting of injury and pain would have likely increased if survey participants were expecting free medical treatment from the BRCT, which established a pain treatment centre in the neighbouring district in Khulna division.