key: cord-0764802-qomhaso3
authors: Bott, Sarah; Ruiz-Celis, Ana P; Mendoza, Jennifer Adams; Guedes, Alessandra
title: Co-occurring violent discipline of children and intimate partner violence against women in Latin America and the Caribbean: a systematic search and secondary analysis of national datasets
date: 2021-12-09
journal: BMJ Glob Health
DOI: 10.1136/bmjgh-2021-007063
sha: 3fbb0c93b409d71be40fb3a5e40f02a4d043f318
doc_id: 764802
cord_uid: qomhaso3

INTRODUCTION: Intersections between violent discipline (physical punishment and/or verbal aggression) of children and intimate partner violence (IPV) against women have received growing international attention. This study aimed to determine how many Latin American and Caribbean (LAC) countries had national data on co-occurring IPV and violent discipline in the same household, how estimates compared and whether violent discipline was significantly associated with IPV. METHODS: A systematic search (following Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines) was used to identify which LAC countries had eligible, national co-occurrence data. The most recent eligible dataset in each country was obtained and reanalysed for comparability. Standardised national estimates were produced for prevalence of violent discipline, physical and/or sexual IPV and co-occurrence among ever partnered women of reproductive age living with a child aged 1–14. Bivariate analyses and logistic regressions produced levels and odds ratios (ORs) of physical punishment and verbal aggression in households affected by IPV (past year and before past year) compared with never, adjusted for sociodemographic characteristics. RESULTS: Nine countries had eligible datasets. Co-occurring physical punishment with past year IPV ranged from 1.7% (Nicaragua) to 17.5% (Bolivia); and with IPV ever from 6.0% (Nicaragua) to 21.2% (Haiti). In almost all countries, children in IPV affected households experienced significantly higher levels and ORs of physical punishment and verbal aggression, whether IPV occurred during or before the past year. Significant adjusted ORs of physical punishment ranged from 1.52 (95% CI 1.11 to 2.10) in Jamaica to 3.63 (95% CI 3.26 to 4.05) in Mexico for past year IPV; and from 1.50 (95% CI 1.23 to 1.83) in Nicaragua to 2.52 (95% CI 2.30 to 2.77) in Mexico for IPV before past year. CONCLUSIONS: IPV is a significant risk factor for violent discipline, but few national surveys in LAC measure both. Co-occurrence merits greater attention from policymakers and researchers.

, Ana P Ruiz-Celis, Jennifer Adams Mendoza, Alessandra Guedes 

Research and programmes addressing violence against children and violence against women historically proceeded along parallel but separate pathways, leaving important

What is already known?

► Intimate partner violence (IPV) against women and violence against children often co-occur in the same household. ► Co-occurrence estimates vary widely, are difficult to compare across sites due to methodological diversity and come primarily from high-income countries. ► Some studies suggest that IPV is associated with a higher risk of violent discipline of children (physical punishment and/or verbal aggression), with mixed findings about whether associations persist beyond the past year.

What are the new findings?

► Most national data on IPV and violent discipline in Latin America and the Caribbean come from separate streams of research; 9 of 35 countries had national surveys measuring both, providing a rare opportunity for a standardised multicountry analysis. ► IPV and violent discipline co-occurred in a substantial proportion of households, ranging from 6.0% to 21.2% for co-occurring physical punishment and IPV ever. ► In almost all countries, IPV was significantly associated with a higher risk of physical punishment and/ or verbal aggression, whether IPV occurred during or before the past year, even after controlling for sociodemographic characteristics.

What do the new findings imply? ► Co-occurrence merits greater attention from policymakers and service providers, with implications for children's and women's health, well-being and rights. ► Until countries expand the number of high-quality, comparable surveys measuring both forms of violence, evidence needed to inform prevention strategies will remain incomplete.

BMJ Global Health gaps. 1 2 Recently however, researchers and policymakers have paid growing attention to intersections between violent discipline of children and intimate partner violence (IPV) against women-the two most common forms of violence in the household. 3 United Nations (UN) Member States have recognised both forms of violence as global public health and human rights problems and agreed to work towards eliminating them as part of 2030 Sustainable Development Goals (SDGs). 4 As of February 2020, at least 10 Latin American and Caribbean (LAC) countries had prohibited corporal punishment of children in all settings including the home, and six others had expressed a commitment to full prohibition. 5 Large proportions of children aged 1-14 in the LAC region experience violent discipline (defined by SDG indicator 16.2.1 as physical punishment and/or verbal aggression by caregivers in the past month), with estimates ranging from 45% in Panama to more than 80% in Haiti, Jamaica and Suriname. 6 IPV against women and girls is also widespread in the region, with national estimates of physical and/or sexual IPV ever against ever partnered women and girls ranging from 7.6% in Uruguay to more than 58.5% in Bolivia. 7 Both violent discipline and child exposure to IPV against women have negative consequences for children. Evidence indicates that physical punishment is both ineffective and detrimental to children's health, well-being and human rights. 8 Verbal aggression by caregivers also has negative consequences for children's health, wellbeing and development. 9 Studies link child exposure to IPV to emotional impairment and mental health disorders, 10 malnutrition and stunting, 11 and aggression towards peers and siblings. 12 Research and programmes have paid increasing attention to co-occurrence, 2 3 defined for purposes of this study as caregiver violence against children and IPV against women that has occurred in the same household (ever, not necessarily concurrently). Evidence suggests that co-occurrence may compound negative effects of direct violence against children or exposure to IPV alone. 13 For example, research from Uganda found that children who witnessed IPV in the home and experienced violence had about two times the odds of mental health difficulties as children who experienced violence but did not witness IPV. 14 The international evidence on co-occurrence has limitations. A 2020 review identified 132 studies on co-occurring IPV and child maltreatment, with co-occurrence rates ranging from 1% to 89%. 15 Studies used highly diverse research designs, operational definitions and respondent characteristics, however. Some limited the definition of co-occurrence to households affected by both forms of violence in the past year; others did not include clear temporal bounds. These differences make it difficult to compare co-occurrence estimates across sites. The vast majority of co-occurrence studies in that 2020 review came from Europe and the USA; 15 however, some studies from LAC have explored levels and/or correlates of co-occurring IPV and violent discipline [16] [17] [18] and/or IPV as an independent risk factor for violent discipline of children. [19] [20] [21] Globally, co-occurrence research has been hampered by a lack of nationally representative surveys measuring both IPV and child discipline. UNICEF-supported Multiple Indicator Cluster Surveys (MICS) are considered the gold standard for measuring child discipline, 22 but they do not usually measure IPV against women. Multicountry analyses of MICS data exploring correlates of violent discipline in low-and middle-income countries have found significant associations with caregiver attitudes supporting violence, younger age of children, lower caregiver education, parental depression and socioeconomic status; 23-25 however, because MICS do not measure IPV, they have not been able to explore associations with IPV. Similarly, Demographic and Health Surveys (DHS) and World Health Organization (WHO) surveys, the most common sources of national IPV prevalence estimates, 26 often ask about violent discipline in women's childhood but not in current households, with some exceptions. 19 20 This article presents a secondary analysis of nine national datasets from LAC identified through a systematic search. The study aimed to determine which countries had national data on IPV and child discipline in the current household, how co-occurrence estimates compared across countries, and whether children in IPVaffected households had a higher risk of violent discipline than other children. Specific research questions were: (1) How many LAC countries have nationally representative, population-based surveys that gathered data on co-occurring IPV against women of reproductive age and violent discipline of children in the current household? (2) How comparable are these data in terms of sample characteristics, violence measures and risk of bias? (3) How do co-occurrence estimates vary across countries after standardising variables for comparability? (4) Are the odds of physical punishment and verbal aggression by caregivers significantly higher in households affected by IPV (past year and before past year) compared with households in which women did not report IPV, after adjusting for sociodemographic characteristics? Henceforth, this article refers to women and girls of reproductive age as 'women' for simplicity, while acknowledging that the Convention on the Rights of the Child considers girls aged 15-17 to be children, not adults, regardless of marital status. 27 

A systematic search was carried out in two stages (December 2018 and December 2019) following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (figure 1). At all stages, eligibility was assessed independently by at least BMJ Global Health two authors (ARC, JAM, SB), with differences resolved by consensus.

In the first stage, described elsewhere, 7 a systematic search used the following a priori inclusion criteria: (a) nationally representative, population-based survey, (b) from the Americas, (c) with a report published (at least online) between January 2000 and July 2018 in any language, (d) measured IPV against women of reproductive age, not just adolescents or young adults, (e) adequate information to clarify data quality (instruments were requested directly from research teams if unavailable from published documents), (f) using measures that met basic quality criteria, for example, no filter questions about violence 'by anyone' before asking about partner violence and no emotionally laden terms such as 'violence', 'abuse' or 'assault'-known to reduce disclosure. 28 The search included databases of journals (Pubmed, Lilacs, SciELO, EBSCO, Web of Science and Google Scholar), international organisations and research programmes; national institutes of statistics websites; article bibliographies; and personal contacts with researchers and government officials throughout the region. Search terms included names of each Member State of the Pan American Health Organization (PAHO) combined with terms such as: gender-based violence, intimate partner violence, domestic violence, spouse abuse, violence against women and family violence (and Spanish equivalents, online supplemental table A). After removing duplicates, 1046 records were screened. Full text records (e.g. articles, reports, manuals and survey instruments) were retrieved and reviewed for 72 surveys from 24 countries.

The second stage (December 2019) updated the search using all stage one criteria, modified as follows: (a) publication deadline was extended to December 2019, (b) surveys had to measure both IPV prevalence and child discipline in the current household, (c) from LAC, not Canada or the USA and (d) the most recent eligible 37 Henceforth, each survey is referred to by country name.

As noted earlier, the search limited eligible datasets to population-based surveys using violence measures that met basic international quality standards. All nine eligible surveys included in the secondary analysis were assessed in duplicate (ARC, SB) for other potential risks of bias/ data quality limitations as required by PRISMA guidelines. The following risk of bias checklist was adapted from existing tools, 38 informed by good practice guidelines 6 28 : (1) measured IPV with a module, not a dedicated violence survey, (2) collected data >10 years ago, (3) used a child discipline or IPV subsample that could not be fully standardised, (4) did not use the MICS child discipline module, considered the gold standard, 22 (5) gathered child discipline data among household questionnaire respondents who were not always the same as women's questionnaire respondents asked about IPV, (6) measured IPV during the past year but not ever, (7) measured IPV by any partner in life but not by the current/ most recent partner specifically and (8) deviated from WHO ethical and safety guidelines. 39 Results are reported alongside prevalence estimates.

All surveys used ethical and safety measures for IPV research, including informed consent, privacy, confidentiality and interviewer training on violence against women, although reports provided different levels of detail on these measures. Some surveys did not adhere to all WHO ethical and safety recommendations for researching IPV, 39 however. For example, Bolivia strongly recommended but did not require privacy. Bolivia and Colombia asked all women in the household about IPV, not just one woman (posing a risk to confidentiality). And, informed consent instructions in Mexico did not emphasise women's right to refuse any question. Published reports did not explain how interviewers responded to disclosures of violence against children (e.g. whether they provided referrals to local services).

All surveys were developed with the participation of civil society organisations and women's advocates working on violence against women. In most countries, all stages of design, implementation and dissemination were guided by specialised committees comprised of a wide array of government, academic, civil society and international organisations and advocates with expertise in violence research, prevention and services.

All surveys used multistage, probability samples. Primary sampling units (PSUs) were selected from master sampling frames, usually a census. Households were randomly selected within each PSU. All surveys except Bolivia and Colombia randomly selected one woman in the household for violence questions. Women were interviewed face to face, usually in or around the woman's residence. All surveys had response rates >85% and used weights for producing nationally representative estimates for women, children and households.

The types of respondents eligible for child discipline questions varied by survey (online supplemental table B). In some surveys, women were asked about child discipline even if their children were infants, already adults and/or not living in the household. Jamaica and Mexico asked all women but included 'no children' as a response option. Haiti gathered child discipline data among male and female household questionnaire respondents, only some of whom were women asked about IPV or even a primary caregiver of the selected child. Most surveys limited female respondents to women of reproductive age, but Mexico had no upper age limit.

The secondary analysis eliminated most of these differences by restricting datasets to subsamples of everpartnered women of reproductive age (15-49) living with a daughter or son aged 1-14 with data on both IPV against women and child discipline in the current household (online supplemental table B). Full standardisation was not always possible, however. Women in the Haiti subsample were usually but not always the same respondent who provided child discipline data. Bolivia limited IPV questions to women with a partner in the past year. Jamaica and Paraguay did not gather enough data to definitively exclude all women whose children aged 1-14 all lived elsewhere. And, Paraguay capped women's age at 44 years.

Child discipline measures Surveys used diverse child discipline measures (table 1) . Haiti used the MICS module, asking closed-ended questions about whether one randomly selected child in the household (specified by name) was disciplined with specific acts in the past month. Mexico asked closedended questions about whether she or her partner hit the children when they became 'angry' or 'desperate' ("se enoja o desespera"). The other seven asked openended questions about how children were disciplined, 

coding spontaneous answers using predetermined categories. Bolivia, Colombia, Guatemala and Peru asked who punishes ("castiga") children (e.g. mother, father/ partner, anyone else) before asking how. No survey other than Haiti asked about a timeframe. Bolivia, Guatemala, Mexico and Peru asked about respondents' own daughters and sons. Colombia asked about biological, step and adopted children. Jamaica, Nicaragua and Paraguay asked about children in the household generally. Some surveys used wording such as 'punish', likely to elicit disclosure of negative discipline; others used neutral language such as 'teach' designed to elicit disclosure of positive and negative discipline. 40 For example, the MICS preamble used in Haiti (in English) reads: "Adults use certain ways to teach their children the right behaviour". For the secondary analysis, standardised violent discipline variables were constructed using operational definitions that aligned with SDG indicators 22 as much as possible within limits of available data. Although specific acts measured by each survey varied (table 1), all surveys measured physical punishment (1=1+ act; 0=none) defined as any act of physical discipline except 'pouring/throwing water' ("echándoles agua"), which did not clearly meet the UN definition of corporal punishment. 41 Only Bolivia and Haiti measured verbal aggression (1=1+ act; 0=none), defined as shouting or insults, separate from more ambiguous acts such as 'verbal reprimands' or 'scolding' ("reprimenda verbal'/'regaño"). Otherwise eligible women who reported that children were 'not punished' were retained in denominators and coded as 'no physical punishment' (or 'no verbal aggression'), in keeping with the approach used by the MICS, but in contrast to the 2018 Peru DHS report. 42 Measures of IPV against women Measures of physical IPV were highly comparable across surveys. All nine surveys used a modified conflict tactics scale, asking about behaviourally specific acts (e.g. slapped, punched, etc) by an intimate partner, considered good practice for measuring IPV. 28 Five surveys (Bolivia, Colombia, Haiti, Mexico and Peru) measured violence by the current or most recent partner. Four (Guatemala, Jamaica, Nicaragua and Paraguay) measured violence by any partner in life. Sexual IPV measures were fairly comparable, with some exceptions. All surveys asked about forced sex. In addition, some DHS surveys also asked about forced sex 'acts', and some RHS surveys also asked about sex that occurred due to fear of what a partner might do if she refused. Mexico measured a wider range of acts, including being forced to watch pornography and to have unprotected sex. Emotional/psychological IPV was measured in highly diverse ways across surveys, and therefore, was not included in this analysis.

For the secondary analysis, IPV was defined as any act of physical and/or sexual violence by an intimate partner. Dichotomous variables were constructed for the prevalence of IPV ever (1=ever; 0=never) and past year (1=past year; 0=not in the past year). In addition, a trichotomous variable for IPV by timeframe was constructed with three mutually exclusive categories: 0=never (reference category); 1=past year; 2=before (but not during) the past year.

Two co-occurrence variables were created: one for women who reported both physical child punishment in the current household and IPV against themselves ever (1=yes; 0=no), and one for women who reported both physical child punishment in the current household and IPV against themselves in the past year (1=yes; 0=no). Women who completed the violence module (or equivalent in Mexico) but were missing responses to select child discipline or IPV items (online supplemental table C) were retained in denominators and classified as 'no' for that act, in keeping with the DHS and MICS.

Partnership was defined as currently versus previously married or cohabited with a partner, except in Jamaica, which also included 'visiting partners', a common form of long-term, non-cohabiting partnership in that setting. Women's age was grouped into 5-year categories. Residence was defined as urban or rural, except in Mexico where three original categories were collapsed into urban (≥1 00 000 inhabitants) and rural/semi-urban (<1 00 000 inhabitants). Education was defined as the highest level reached (not necessarily completed), including: lower primary (≤3 years) or less, upper primary (>3 years), lower secondary (≤3 years except in Bolivia, where it included seventh and eighth years of 'primary school'), upper secondary and postsecondary. In Jamaica, few (nine) women reported less than upper primary, so they were combined into a category of primary and below for regression analyses.

Wealth quintiles (poorest, poorer, middle, richer, richest) were precoded by original research teams, except in Mexico and Nicaragua, which were produced by authors (ARC, JAM) using SPSS (V.26) following DHS methodology. 43 Household assets and other characteristics were assigned weights generated through principal components analysis. Scores were standardised in relation to a normal distribution with a mean of zero and SD of one. Standardised scores were summed to create the household wealth index. Households were ranked and divided into quintiles, applying household weights, adjusted for household size.

Statistical analyses were done with Stata V.16 (StataCorp LP). Prevalence of each form of violence (physical punishment, verbal aggression, any violent discipline, past year IPV, IPV ever, co-occurring physical punishment and past year IPV and co-occurring physical punishment and IPV ever) was estimated with percentages and 95% confidence intervals (CIs). To explore whether children in households affected by IPV experienced higher levels of physical punishment and verbal aggression than other children, bivariate analyses were carried out by BMJ Global Health IPV timeframe (never, past year, before past year), with significance testing using Pearson's χ 2 test corrected for survey design effects and converted into an F-statistic. Logistic regression produced odds ratios (ORs) of physical punishment and verbal aggression in households affected by IPV past year and before past year compared with never, adjusted for partnership, age, residence, education and wealth. Sensitivity testing was carried out for physical versus physical and/or sexual IPV.

Survey design effects were taken into account for all CIs, regressions and significance testing, using domestic violence module weights (if available) or women's individual weights, which were normalised to equalise weighted and unweighted numbers of women in each study subsample. Each survey was analysed separately, not pooled because: (a) the diversity of child discipline measures posed a risk of bias, (b) surveys were conducted over a 10-year period from a limited number of countries, so the reference population would have been unclear and (c) violence was not rare, so pooling could not be justified by a need to increase statistical power.

As of December 2019, 25 LAC countries had nationally representative data on the prevalence of IPV against women that met basic quality and reporting criteria, but only nine (Bolivia, Colombia, Guatemala, Haiti, Jamaica, Mexico Nicaragua, Paraguay and Peru) had a survey that met those criteria and measured both IPV and child discipline in the current household. National data on co-occurring IPV and verbal aggression against children were even more limited, measured by only two of nine eligible surveys. All MICS surveys in the region measured child discipline, but none measured IPV against women. Among eligible surveys, Mexico was dedicated specifically to violence against women; all others were DHS or RHS surveys using violence modules within larger health surveys. Eligible surveys were carried out between 2008 and 2018. Four (Bolivia, Guatemala, Jamaica and Paraguay) were >10 years old.

Reanalysed subsample sizes ranged from 3291 women in Haiti to 43 095 in Mexico. Women's sociodemographic characteristics varied by country (table 2). In all countries, most women were currently partnered. Relatively few (1%-5%) were adolescent girls aged 15-19. Education levels varied widely; more than half of women in the Guatemalan subsample had less than 4 years of primary education, but only nine women in Jamaica fell into that category.

In all countries, except Nicaragua, levels of violent discipline were higher than levels of IPV ( Associations between IPV and violent discipline Bivariate analyses found significant associations between IPV and physical punishment in all nine countries and between IPV and verbal aggression in Bolivia and Haiti, the only two countries that measured this indicator (table 4) . In all countries, ORs of physical punishment were higher in households affected by IPV (past year and before past year) than in households where women reported no IPV, before and after adjusting for partnership, age, residence, education and household wealth. All elevated ORs were statistically significant, except in Haiti, where only the unadjusted OR for past year IPV was significant, and in Nicaragua, where ORs (adjusted and unadjusted) were significantly elevated for children in households affected by IPV before past year but not by IPV in the past year. Significant adjusted ORs of physical child punishment ranged from 1.52 (95% CI 1.11 to 2.10, p=0.010) in Jamaica to 3 In Bolivia and Haiti, ORs of verbal aggression were significantly elevated for both past year IPV and IPV before past year, before and after adjusting for sociodemographic characteristics (p<0.001 for both surveys and timeframes, except IPV before past year in Haiti, which was p=0.001).

Sensitivity testing did not find notable statistical differences between physical versus physical and/or sexual IPV for any survey or type of discipline.

This is the first study from LAC to present a systematic search and secondary analysis of national data on co-occurring IPV against women and violent discipline of children in the current household. The search confirmed that nationally representative, population-based data on co-occurring IPV and violent discipline remain limited in the region. National prevalence data on these forms of violence are often generated by separate research streams. Only nine LAC countries had eligible datasets, and only five had data gathered in the past decade. Published

analyses of national co-occurrence data from LAC were even more limited. Most co-occurrence data gathered by surveys in this study had not been previously analysed, although secondary analyses of co-occurrence have been published using data from Bolivia (2008) and previous rounds of DHS and RHS surveys from Colombia (2010), Nicaragua (2006) (2007) and Peru (2000 & 2012) . 19 20 44 45 As a result, this secondary analysis provided a rare opportunity to explore standardised national estimates of co-occurring IPV and violent discipline across different countries. High levels of violent child discipline (physical and/or verbal), physical and/or sexual IPV against women, and co-occurrence were found in all nine LAC countries. The range of co-occurrence estimates (1.7% to 17.5% for physical punishment and past year IPV and 6.0% to 21.2% for physical punishment and IPV ever) was narrower than the 1%-89% range found by Sijtsema and colleagues, 15 but wider than 5%-10% reported by reviews of community-based surveys with representative samples. 46 47 In almost all countries with data, IPV against women was significantly associated with a higher risk of physical punishment and verbal aggression against children, even after controlling for sociodemographic characteristics, compared with households in which women reported no IPV. This finding echoes many studies from the USA 48 and a more limited number from LAC countries, 15 including Nicaragua 19 and Peru. 17 20 Notably, most ORs of physical punishment and verbal aggression were significantly elevated whether IPV occurred during or before the past year, which NA, not available because not measured. All numbers are unweighted. *Secondary analysis subsamples were limited to ever-partnered women of reproductive age, living with a daughter or son aged 1-14. †Paraguay capped women's age at 44. ‡In Mexico, residence was urban (≥1 00 000 inhabitants) and rural/semi-urban (<1 00 000 inhabitants). §One woman in Mexico was missing education level; no other sociodemographic data were missing from any survey. ¶In Bolivia, lower secondary included seventh and eighth years of what they consider 'primary' school. to exclude all women without children age 1-14 in the household. (4) Did not use MICS questions to measure child discipline. (5) Gathered child discipline data from household questionnaire respondents. (6) . Measured IPV during the past year but not ever. (7) . Measured IPV by any partner in life not the current/most recent partner specifically. (8) Deviated from WHO ethical and safety guidelines. IPV, Intimate partner violence ; MICS, Multiple Indicator Cluster Survey; NA, not available, because not measured. 

suggests it may be a mistake to limit attention to recent IPV when studying intersections between violent discipline and IPV. Previous research produced mixed findings about whether associations between IPV and violence against children in the household persist beyond 1 year, but most research has come from high-income countries. 48 The question of why children in IPV-affected households face a higher risk of violent discipline than other children is beyond the scope of this study. It is possible that IPV and violent discipline share risk factors that independently affect the likelihood of each form of violence. 49 Some researchers theorise that women living with partner abuse are more likely to discipline their children harshly due to stress or anxiety. 48 It is also possible that men who abuse partners are more likely to abuse children. Research indicates that children may experience violent discipline by fathers/partners and other household members, not just mothers/stepmothers. 42 50 Five surveys eligible for this study examined who disciplined children and how, providing a future opportunity for research into how pathways linking IPV to violent discipline vary by caregiver.

This study had many limitations. Surveys used diverse measures of violent discipline. Most lacked a preamble introducing the idea of positive and negative discipline, used open-ended questions, coded a small number of acts and did not ask about a specific child or timeframe. Therefore, co-occurrence estimates in this analysis included some households in which physical punishment occurred only in the past year but IPV occurred before past year or vice versa. MICS estimates of physical child punishment in the past month were available from three countries in this analysis, including 68.4% for Jamaica 2011, 43.7% for Mexico 2015 and 39.7% for Paraguay 2016. 51 All three MICS estimates were higher than estimates from this analysis, even though MICS estimates were limited to only one child in the household in the past month. MICS reports do not include CIs however, so additional analyses would be required to confirm that differences were significant.

Haiti used the MICS module, but had other risks of bias. Consistent with MICS series 4 and 5, Haiti gathered child discipline data from household questionnaire respondents who were not always the women asked about IPV or even a primary caregiver of selected children. 6 Moreover, Haiti limited child discipline and domestic violence modules to a subset (about two-thirds) of households, many of which did not have both an eligible child and an eligible woman. As a result, the final subsample from Haiti included only 3291 of 14 371 women who participated in the full survey, the smallest subsample size of all nine surveys. These limitations made it hard to draw conclusions about why associations between physical punishment and IPV were not significant in that country after controlling for sociodemographic factors. Were levels of physical punishment so high (>80%) that IPV did not make a difference? Was the subsample size too small to detect significance differences? Or, was this an artefact of gathering some child discipline and IPV data from different respondents in the household? Future surveys that include both the MICS child discipline module and the DHS domestic violence module could avoid these limitations by including both modules in the women's individual questionnaire.

Another limitation was that study subsamples did not include all households with children aged 15-17. Caregiver violence against older adolescents remains a research gap that merits attention. Nor did study subsamples include households in which no child lived with an ever-partnered mother of reproductive age, and violence patterns may be different in those households. However, household data from all six surveys that measured family structure (not shown) suggest that the vast majority of children lived with mothers, and generally, researchers have not found a universal relationship between family structure and violent discipline across low-and middle-income countries. 23 Another limitation was that the nine surveys were carried out over a 10-year period, and some are more than 10 years old. A 2019 analysis found preliminary evidence that IPV against women may have declined in some LAC countries. 7 Meanwhile, other research suggests that the COVID-19 pandemic may have increased levels of violence in the household, 52 although that evidence is still preliminary. Levels of physical child punishment may be changing as well, and co-occurrence estimates based on older datasets may be out of date.

Finally, any survey on violence may underestimate true prevalence due to barriers to disclosure such as respondents' fear of reprisal or social stigma. This study relied on survivor reports of IPV and caregiver reports of child discipline. In contrast, Violence against Children Surveys in Colombia, El Salvador, Haiti and Honduras have asked children aged 13-17 about violence they experienced by caregivers and IPV against parents they witnessed. 53 Surveys have also gathered retrospective data from women 54 and men 55 about childhood experiences of violent discipline and exposure to IPV. Retrospective data do not shed light on violence in current households, and children of IPV survivors may underreport levels of IPV even more than survivors. Nonetheless, these are important data sources that complement this study.

This study suggests that policymakers and professionals working on children's health and well-being should be aware that IPV in the household is often associated with an increased risk of violent discipline of children, with implications for children's rights, health, well-being and development. This study also suggests that research on correlates and consequences of violent discipline may be incomplete if researchers do not include exposure to IPV (ever and past year) as a possible risk factor. This study supports growing evidence from the programme literature suggesting that long-term efforts to prevent violence against children may be ineffective without greater attention to violence against women-and vice versa. 3 

The finding that national data on violence against children and violence against women continue to come from parallel but segregated data collection efforts suggests a need to expand the number, quality and comparability of national surveys that measure both forms of violence. Until researchers bridge these gaps, our understanding of how to prevent and respond to violence against children will remain fragmented. Ideally, more national surveys would include the MICS child discipline module within women's questionnaires that ask about IPV. Intersections between violence against children and violence against women deserve greater attention, both for advancing public health and protecting human rights.

Acknowledgements Thanks are due to thousands of women and girls who provided data for this study and to the original nine survey research teams. Lori Heise, Karen Devries, Carolina Coll and Gwyther Rees kindly reviewed early versions of this work.

Contributors All authors participated in the conception, data analysis plans, interpretation of results and revision of the manuscript. ARC and SB carried out the search. ARC, JAM and SB conducted the statistical analysis. SB led the writing. All authors accepted responsibility for the overall content and approved the final manuscript.

Funding UNICEF, Office of Research-Innocenti.

Competing interests None declared.

Patient consent for publication Not applicable.

Ethics approval Demographic and Health Surveys were approved by the Institutional Review Board of ICF, previously known as Macro International (approval ID number FWA00000845). RHS surveys were approved by the United States Centres for Disease Control Institutional Review Board and specialised national committees in each country. The Mexico 2016 ENDIREH survey was approved by a specialised national committee established by the Instituto Nacional de Estadística y Geografía to provide scientific and ethical review for all stages of survey design and implementation; that committee was comprised of a wide array of government, academic, civil society and international organisations and advocates with expertise in violence research, prevention and services.

Provenance and peer review Not commissioned; externally peer reviewed.

Data availability statement Data are available in a public, open access repository. All nine datasets are open access, available online or by request.

Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

Open access This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http:// creativecommons. org/ licenses/ by-nc/ 4. 0/.

ORCID iD Sarah Bott http:// orcid. org/ 0000-0003-1128-8612

Understanding the linkages between violence against women and violence against children

Bridging the gaps: a global review of intersections of violence against women and violence against children

Gender dimensions of violence against children and adolescents: a discussion paper

ga/ search/ view_ doc. asp? symbol= A/ RES/ 70/ 1& Lang=E 5 Global Initiative to End All Corporal Punishment of Children. Progress towards prohibiting all corporal punishment in Latin America and the Caribbean. London: Global Initiative to End All Corporal Punishment of Children

UNICEF. Violent discipline: data and methodology. New York: UNICEF, 2020

Intimate partner violence in the Americas: a systematic review and reanalysis of national prevalence estimates

COUNCIL ON CHILD ABUSE AND NEGLECT. Effective discipline to raise healthy children

Psychological maltreatment: an international challenge to children's safety and well being

Children's exposure to intimate partner violence: an overview

Association between intimate partner violence and poor child growth: results from 42 demographic and health surveys

A familiar face: violence in the lives of children and adolescents

Intersection of child abuse and children's exposure to domestic violence

Witnessing intimate partner violence and child maltreatment in Ugandan children: a crosssectional survey

Unique risk factors of the co-occurrence between child maltreatment and intimate partner violence perpetration

Intimate partner violence against women and child maltreatment in a Brazilian birth cohort study: co-occurrence and shared risk factors

The co-occurrence of domestic and child violence in urban Peru: evidence from three regions

Violencia física contra adolescentes y estructura familiar: magnitudes, expresiones y desigualdades.] Physical violence against adolescents and family structure: Scope, expressions and inequalities

Exposure to intimate partner violence reduces the protective effect that women's high education has on children's corporal punishment: a populationbased study

Maternal violence, victimization, and child physical punishment in Peru

Civil conflict, domestic violence, and poverty as predictors of corporal punishment in Colombia

Sustainable Development Goal indicator 16.2.1 metadata

Child disciplinary practices at home: evidence from a range of low-and middle income countries

Childrearing discipline and violence in developing countries

Child discipline in low-and middle-income countries: socioeconomic disparities at the household-and countrylevel

Convention on the Rights of the Child (Article 19). Geneva: United Nations, Office of the High Commissioner for Human Rights

New York: United Nations, Department of Economic and Social Affairs Statistics Division

US Centers for Disease Control and Prevention

Morbidité et Utilisation des Services (EMMUS-VI)

US Centers for Disease Control and Prevention

DATASET] Encuesta Nacional sobre la Dinámica de las Relaciones en los Hogares (ENDIREH)

The development of a critical appraisal tool for use in systematic reviews addressing questions of prevalence

Putting women first: ethical and safety recommendations for research on domestic violence against women. Geneva: World Health Organization

Preventing violence: evaluating outcomes of parenting programmes. Geneva: World Health Organization

the right of the child to protection from corporal punishment and other cruel or degrading forms of punishment. Geneva: UN Committee on the Rights of the Child (CRC)

Informe final. Lima: Instituto Nacional de Estadística e Informática (INEI)

Wealth index construction. Calverton, Maryland: ICF, International

Associations between intimate partner violence, childcare practices and infant health: findings from Demographic and Health Surveys in Bolivia, Colombia and Peru

Neighborhood crime undermines parenting: violence in the vicinity of households as a predictor of aggressive discipline

The co-occurrence of spouse and physical child abuse: a review and appraisal

Prevalence and correlates of the co-occurrence of family violence: a meta-analysis on family polyvictimization

Intimate partner violence victimization and parenting: a systematic review

Pathways between childhood trauma, intimate partner violence, and harsh parenting: findings from the UN Multi-country Study on Men and Violence in Asia and the Pacific

Corporal punishment of children in nine countries as a function of child gender and parent gender

Multiple Indicator Cluster Surveys

Violence against children in the time of COVID-19: what we have learned, what remains unknown and the opportunities that lie ahead

Together for Girls. Violence against children surveys. Washington DC: Together for Girls

Intergenerational effects of violence against girls and women: Selected findings from a comparative analysis of population-based surveys from 12 countries in Latin America and the Caribbean

Risk factors for men's lifetime perpetration of physical violence against intimate partners: results from the International Men and Gender Equality Survey (IMAGES) in eight countries