A systematic review of the global and regional estimates of the prevalence of sexual violence against children

Summary

Background:

Sexual violence against children (SVAC) is a critical global public health issue. Nonetheless, the availability of robust and comparable global and regional prevalence estimates remains limited. To address this gap, we aimed to provide the current and refined estimates of SVAC to enhance understanding of its magnitude worldwide.

Methods:

We conducted comprehensive literature searches in the six official UN languages across 10 English-language databases, 23 non-English databases, and more than 20 grey literature sources for records published between January 1, 2010, and August 1, 2024. Eligible studies included children under the age of 18 or adults retrospectively reporting SVAC. Studies were double screened and extracted by two authors independently, with risk of bias assessed. A Bayesian hierarchical model was used to estimate lifetime and past-year prevalence, accounting for methodological and definitional variations across studies. This study was registered with PROSPERO (CRD42024495116).

Findings:

Of 64,393 records found and screened, 1,412 studies across 147 countries were included in this systematic review and meta-analysis, encompassing responses from 4,070,693 females and 2,910,973 males. Globally, 20.1% (95% UI 19.8–20.4) of women and 16.8% (95% UI 16.5–17.1) of men aged 18 years and older reported experiencing at least one form of SVAC (including both contact and non-contact forms) prior to age 18. Within this broader category, 12.4% (95% UI 12.2–12.6) of women and 10.0% (95% UI 9.8–10.2) of men experienced contact childhood sexual violence. In 2024 alone, an estimated 72.6 million girls and 60.9 million boys experienced at least one form of sexual violence, including 43.3 million girls and 34.8 million boys subjected to rape or sexual assault. In most regions, the estimated prevalence was somewhat higher for girls compared with boys, although the global difference is smaller than reported in previous meta-analyses.

Interpretation:

Our findings highlight the alarming global burden of SVAC and the urgent need for evidence-based, multi-sectoral prevention strategies and intervention programs to safeguard children worldwide. More well-designed research using standardized approach to data collection is also needed, especially in regions under-represented by the current data.

1. Introduction

Sexual violence against children (SVAC) is a critical global public health and development issue, with devastating consequences for millions of children worldwide.¹ According to the UNICEF International Classification of Violence Against Children (ICVAC), SVAC includes deliberate, unwanted sexual acts against a child that results in or has a high likelihood of resulting in injury, pain or psychological suffering.² SVAC includes both completed and attempted acts and those perpetrated for exploitative purposes, as well as experiences perpetrated by adults and unwanted and coercive acts by peers and other children.³ The UNICEF ICVAC includes both contact and non-contact forms of SVAC. The consequences of SVAC extend far beyond the immediate harm, encompassing a heightened risk of severe mental, physical, and behavioural health disorders throughout the lifecourse.^4–9 SVAC is strongly associated with increased risk of sexually transmitted infections (including HIV), chronic health conditions, mental health disorders, suicide, self-harm, substance use disorders, poor employment outcomes, reduced earnings in adulthood, and subsequent violence experiences.^4–8 Experiencing SVAC is also associated with heightened risk of perpetrating violence, engaging in criminal behavior, and enduring a diminished quality of life.⁷⁻⁹

These profound adverse outcomes underline the urgency of global efforts to prevent SVAC and mitigate its effects. Achieving Target 16.2 of the United Nations 2030 Sustainable Development Goals (SDGs), which aims to eliminate all forms of violence against children, is essential for safeguarding children and enabling them to reach their full social and economic potential.^8,10 Under this target, Indicator 16.2.3 focuses specifically on tracking the proportion of young women and men aged 18 to 29 years who experienced sexual violence before the age of 18.¹⁰

Efforts to quantify the global prevalence of SVAC, however, are constrained by significant data limitations.^11,12 While notable progress has been made through multi-country nationally representative surveys such as the Demographic and Health Surveys (DHS), Multiple Indicator Cluster Surveys (MICS), and Violence Against Children and Youth Surveys (VACS), many countries still lack such comprehensive data collection. In some contexts, national or subnational surveys using non-standardized instruments provide critical insights but often vary in their definitions, methods, and measures of SVAC.^13–15 Discrepancies in factors such as the age groups assessed, the timeframes considered (e.g., lifetime before 18, lifetime before 15, past year, or past three months), frequency thresholds (e.g., at least once, at least twice, or at least five times), and whether the violence is perpetrated by intimate partners, caregivers, or others make comparability across studies challenging.^13–15

Although the number of studies that measure SVAC prevalence has grown significantly in the past two decades, few studies have systematically reviewed and synthesized this body of literature to produce global prevalence estimates. For example, Stoltenborgh and colleagues conducted a meta-analysis of 217 publications in 2011 and estimated a global SVAC prevalence of 11.8%.¹⁶ However, this analysis relied on literature published between 1982 and 2008, limiting its relevance to current trends and the expanded availability of population data from more recent studies.^11,17 A more recent meta-analysis, which searched databases through April 2021, estimated a pooled prevalence of sexual violence against female children at 24%.¹⁸ However, this study included only 28 cross-sectional studies, was limited to females, and did not account for methodological or definitional variations.¹⁸ Another recent systematic review and meta-analysis, incorporating 165 studies from 80 countries, estimated global prevalence for specific types of child sexual abuse (CSA) among individuals aged ≤19 years.¹⁹ While this analysis represents progress in estimating global and regional CSA prevalence, it did not adjust for methodological or definitional differences. The reported global prevalence is essentially a simple weighted average of all estimates, without harmonization of key factors. For example, prevalence estimates based on whether violence was perpetrated by family members or peers were combined with violence perpetrated by anyone, without adjustment for comparability. Furthermore, the study did not incorporate nested random effects for superregions, subregions, and countries, nor did they generate country-level prevalence estimates and then derive regional and global estimates by weighting those country-level estimates by population.¹⁹ Consequently, countries with a high volume of studies, such as the US, may have disproportionately influenced global and regional prevalence estimates, while countries with fewer studies, even those with large populations like India, were underrepresented.

Meanwhile, UNICEF’s recent estimates, based primarily on DHS and MICS data and supplemented by other national and regional surveys, provided valuable insights.¹² However, the methodology differed for males and females, and estimates were limited to lifetime prevalence, without capturing past year prevalence, suggesting areas where methodological consistency could be further strengthened. Separately, the most recent study drew on 460 sources from publicly available data repositories to estimate the SVAC prevalence.²⁰ Although it adjusted for differences in case definitions and survey modes, it focused exclusively on contact sexual violence and did not provide past-year prevalence estimates for children.²⁰

The objective of this study is to produce baseline reliable, reproducible, and internationally comparable global and regional prevalence estimates of lifetime and past-year SVAC, disaggregated by sex. Unlike previous efforts,^12,18,19 this study draws on an extensive multilingual search across the six UN languages using both general and language-specific databases, resulting in a more comprehensive synthesis of the evidence. In addition, we account for definitional and methodological variation across studies and apply a Bayesian multilevel framework capable of accommodating sparse and unbalanced data. By providing robust global and regional estimates of contact and overall (both contact and non-contact, excluding online) SVAC, this research offers critical baselines for understanding the magnitude of SVAC, monitoring global progress, and informing evidence-based policies and interventions to protect children from sexual violence and its devastating consequences.

2. Results

A total of 53,035 records in English and 11,358 records in other UN languages were identified through database searches and grey literature exploration. After removing duplicates, we screened 39,358 records at the title and abstract level for eligibility. Subsequently, we reviewed 2,464 records at the full-text stage. We excluded records that lacked sufficient research data (e.g., non-disaggregated results, lack of representative sampling), duplicated survey data, or focused on special populations. Ultimately, 1,159 records, comprising 1,412 studies from 147 countries and representing 93% of the global population, met the inclusion criteria and were selected for analysis (Figure 1). Collectively, these studies encompassed responses from 4,070,693 females and 2,910,973 males. Sample characteristics are detailed in Table 2.

Fig.1|. PRISMA flow diagram of systematic review.

Although 1,159 records were included, they contained 1,412 extractable studies. Multi-country surveys contributed multiple studies because each country-level estimate was treated separately. Some records also included multiple distinct datasets from the same country (e.g., different years or independent samples), each counted as a separate study in the meta-analysis. Six additional non-UN language reports were retained because they were nationally comparable survey reports (e.g., DHS), included to maximize country coverage.

Table 2:

Characteristics of included studies on lifetime and past-year sexual violence against children (SVAC)

	Lifetime			Past-year
Sample characteristics	Female	Male	Both	Female	Male	Both
Number of participants interviewed	3,486,254	2,397,826	7,941,152	858,136	678,186	2,128,833
Number of studies	796	532	705	289	170	186
Number of specific observations	1,632	1,119	1,513	583	417	565
Nationally representative studies	540/796 (67.8%)	329/532 (61.8%)	370/705 (52.5%)	209/289 (72.3%)	113/170 (66.5%)	107/186 (57.5%)
Sub-nationally representative studies	93/796 (11.7%)	73/532 (13.7%)	148/705 (21.0%)	35/289 (12.1%)	25/170 (14.7%)	27/186 (14.5%)
City/District/County level representative studies	163/796 (20.5%)	130/532 (24.4%)	187/705 (26.5%)	45/289 (15.6%)	32/170 (18.8%)	52/186 (28.0%)
Number of countries represented	142	94	84	98	60	55
Countries with one study	30/142 (21.1%)	23/94 (24.5%)	22/84 (26.2%)	35/98 (35.7%)	27/60 (45.0%)	24/55 (43.6%)
Countries with two studies	30/142 (21.1%)	26/94 (27.7%)	19/84 (22.6%)	30/98 (30.6%)	17/60 (28.3%)	16/55 (29.1%)
Countries with three studies	29/142 (20.4%)	13/94 (13.8%)	11/84 (13.1%)	11/98 (11.2%)	4/60 (6.7%)	10/55 (18.2%)
Countries with four or more studies	53/142 (37.3%)	32/94 (34.0%)	32/84 (38.1%)	22/98 (22.4%)	12/60 (20.0%)	5/55 (9.1%)
Number of SDG sub-regions represented	20/20 (100.0%)	19/20 (95.0%)	18/20 (90.0%)	19/20 (95.0%)	16/20 (80.0%)	17/20 (85.0%)
Population coverage	93.4%	87.1%	84.8%	83.6%	66.2%	61.0%
Median date of data collection	2013	2013	2013	2013	2013	2012
Studies conducted prior 2005	82/796 (10.3%)	35/532 (6.6%)	48/705 (6.8%)	23/289 (8.0%)	7/170 (4.1%)	14/186 (7.5%)
Studies conducted 2005–2009	151/796 (19.0%)	100/532 (18.8%)	127/705 (18.0%)	33/289 (11.4%)	25/170 (14.7%)	23/186 (12.4%)
Studies conducted 2010–2014	248/796 (31.2%)	211/532 (39.7%)	233/705 (33.0%)	112/289 (38.8%)	72/170 (42.4%)	80/186 (43.0%)
Studies conducted 2015–2019	214/796 (26.9%)	141/532 (26.5%)	214/705 (30.4%)	92/289 (31.8%)	54/170 (31.8%)	57/186 (30.6%)
Studies conducted 2020–2023	101/796 (12.7%)	45/532 (8.5%)	83/705 (11.8%)	29/289 (10.0%)	12/170 (7.1%)	12/186 (6.5%)

Lifetime prevalence

Globally, 20.1% (95% UI 19.8–20.4) of women and 16.8% (95% UI 16.5–17.1) of men aged 18 years and older reported experiencing at least one form of SVAC (including both contact and non-contact forms) prior to age 18 (Table 3). Regional variations, based on SDG classifications, indicate that the lifetime prevalence of overall SVAC among women is highest in Oceania ((33.9%; UI 33.1–34.7)), followed by Sub-Saharan Africa (28.4%; UI 28.0–28.9), Europe and Northern America (25.6%, UI 25.3–25.9), Latin America and the Caribbean (23.0%; UI 22.6–23.3), and Northern Africa and Western Asia (19.3%; UI 17.9–20.9). The lowest prevalence rates were observed in Eastern and South-Eastern Asia (16.9%, UI 16.7–17.2) and Central and Southern Asia (15.8%; UI 14.9–16.6).

Table 3:

Regional prevalence estimates of lifetime sexual violence against children (SVAC) by United Nations Sustainable Development Goal (SDG) super region and sub region

	Contact SVAC		Overall SVAC
	Female	Male	Female	Male
Central and Southern Asia	9.9% (9.2–10.5%)	9.8% (9.2–10.4%)	15.8% (14.9–16.6%)	15.7% (14.8–16.5%)
Central Asia	3.4% (2.0–5.4%)	3.3% (2.0–5.3%)	5.7% (3.3–8.7%)	5.6% (3.3–8.7%)
Southern Asia	10.0% (9.4–10.6%)	9.9% (9.3–10.5%)	16.0% (15.1–16.8%)	15.8% (14.9–16.7%)
Eastern and South-Eastern Asia	8.9% (8.8–9.1%)	8.6% (8.5–8.8%)	16.9% (16.7–17.2%)	16.4% (16.2–16.7%)
Eastern Asia	8.8% (8.7–9.0%)	8.6% (8.5–8.7%)	16.8% (16.7–17.0%)	16.3% (16.2–16.5%)
South-Eastern Asia	9.1% (8.7–9.5%)	8.7% (8.3–9.2%)	17.1% (16.4–17.9%)	16.6% (15.9–17.4%)
Europe and Northern America	16.5% (16.3–16.8%)	9.7% (9.5–9.9%)	25.6% (25.3–25.9%)	15.8% (15.5–16.0%)
Eastern Europe	12.3% (11.6–12.8%)	8.6% (8.1–9.1%)	19.2% (18.4–20.0%)	13.9% (13.1–14.6%)
Northern America	20.8% (20.6–20.9%)	10.2% (10.1–10.3%)	31.6% (31.4–31.9%)	16.8% (16.6–16.9%)
Northern Europe	15.2% (15.1–15.4%)	10.2% (10.1–10.4%)	24.0% (23.7–24.2%)	16.7% (16.6–16.9%)
Southern Europe	15.6% (15.0–16.2%)	14.8% (14.2–15.5%)	24.4% (23.6–25.3%)	23.3% (22.5–24.3%)
Western Europe	16.4% (16.0–16.8%)	5.9% (5.7–6.1%)	25.6% (25.0–26.2%)	10.0% (9.6–10.3%)
Latin America and the Caribbean	15.2% (15.0–15.5%)	11.4% (11.2–11.6%)	23.0% (22.6–23.3%)	17.6% (17.3–18.0%)
Caribbean	21.9% (21.0–22.9%)	17.1% (16.3–18.0%)	31.6% (30.3–33.0%)	25.4% (24.3–26.6%)
Central America	16.1% (15.6–16.7%)	8.4% (7.9–8.8%)	24.2% (23.4–24.9%)	13.2% (12.6–13.9%)
South America	14.3% (14.1–14.6%)	12.4% (12.2–12.6%)	21.8% (21.4–22.2%)	19.1% (18.8–19.5%)
Northern Africa and Western Asia	14.1% (12.9–15.3%)	10.3% (9.4–11.4%)	19.3% (17.9–20.9%)	14.5% (13.3–15.9%)
Northern Africa	20.7% (18.0–23.6%)	10.9% (8.6–13.3%)	27.9% (24.6–31.5%)	15.4% (12.3–18.7%)
Western Asia	10.2% (9.4–11.2%)	10.0% (9.3–11.0%)	14.3% (13.2–15.6%)	14.1% (13.1–15.3%)
Oceania	24.2% (23.5–24.9%)	11.1% (10.7–11.6%)	33.9% (33.1–34.7%)	16.7% (16.1–17.4%)
Australia and New Zealand	24.5% (23.8–25.2%)	11.3% (10.8–11.7%)	34.2% (33.5–35.0%)	17.0% (16.4–17.6%)
Oceania (excluding Australia and New Zealand)	23.1% (21.7–24.6%)	10.6% (9.8–11.5%)	32.5% (30.8–34.4%)	16.0% (14.9–17.3%)
Sub-Saharan Africa	18.6% (18.3–18.9%)	14.0% (13.7–14.3%)	28.4% (28.0–28.9%)	22.1% (21.6–22.6%)
Eastern Africa	21.2% (20.9–21.5%)	16.0% (15.7–16.3%)	32.1% (31.6–32.6%)	25.1% (24.6–25.6%)
Middle Africa	21.3% (20.7–21.9%)	16.2% (15.7–16.8%)	31.7% (30.9–32.6%)	25.0% (24.3–25.9%)
Southern Africa	10.7% (10.5–11.0%)	7.8% (7.6–8.0%)	17.4% (17.0–17.9%)	13.0% (12.6–13.4%)
Western Africa	16.3% (15.9–16.6%)	12.1% (11.8–12.4%)	25.3% (24.7–25.9%)	19.5% (18.9–20.0%)
World	12.4% (12.2–12.6%)	10.0% (9.8–10.2%)	20.1% (19.8–20.4%)	16.8% (16.5–17.1%)

Data presented as % (95% uncertainty interval).

Among men, the regional variation in the prevalence of overall SVAC presents a different pattern. The highest prevalence was found in Sub-Saharan Africa (22.1%; UI 21.6–22.6), followed by Latin America and the Caribbean (17.6%; UI 17.3–18.0), Oceania (16.7%; UI 16.1–17.4), Eastern and South-Eastern Asia (16.4%; UI 16.2–16.7), Europe and Northern America (15.8%; UI 15.5–16.0), and Central and Southern Asia (15.7%; UI 14.8–16.5). The lowest prevalence was recorded in Northern Africa and Western Asia (14.5%; UI 13.3–15.9).

Regarding the global lifetime prevalence of childhood contact sexual violence, 12.4% (95% UI 12.2–12.6) of women and 10.0% (UI 9.8–10.2) of men aged 18 years and older have experienced contact sexual violence at least once before the age of 18 (Table 3). Regional variations in the lifetime prevalence of contact SVAC among women largely mirror the pattern observed for overall SVAC, except that Eastern and South-Eastern Asia (8.9%; UI 8.8–9.1) replaced Central and Southern Asia (9.9%; UI 9.2–10.5) as the region with the lowest contact SVAC prevalence. Oceania remained the region with the highest prevalence of contact SVAC (24.2%; UI 23.5–24.9).

For men, the regional variation in contact SVAC shows a somewhat different pattern from that observed for overall SVAC. Sub-Saharan Africa had the highest prevalence (14.0%; UI 13.7–14.3), followed by Latin America and the Caribbean (11.4%; UI 11.2–11.6), Oceania (11.1%; UI 10.7–11.6), Northern Africa and Western Asia (10.3%; UI 9.4–11.4), Central and Southern Asia (9.8%; UI 9.2–10.4), Europe and Northern America (9.7%; UI 9.5–9.9), and Eastern and South-Eastern Asia (8.6%; UI 8.5–8.8).

Figure 2 presents a map illustrating the lifetime prevalence of contact SVAC and overall SVAC at the subregion level for both women and men. Among the subregions, for women, Australia and New Zealand reported the highest prevalence of both contact SVAC (24.5%; UI 23.8–25.2) and overall SVAC (34.2%; UI 33.5–35.0), whereas Central Asia recorded the lowest prevalence for both contact SVAC (3.4%; UI 2.0–5.4) and overall SVAC (5.7%; UI 3.3–8.7). For men, the Caribbean exhibited the highest prevalence of both contact SVAC (17.1%; UI 16.3–18.0) and overall SVAC (25.4%; UI 24.3–26.6), while Central Asia again recorded the lowest prevalence for both contact SVAC (3.3%; UI 2.0–5.3) and overall SVAC (5.6%; UI 3.3–8.7), as shown in Figure 2 and Table 3.

Fig.2|. Map of prevalence estimates of lifetime sexual violence against children (SVAC) by SDG subregion.

a, Female lifetime prevalence of contact SVAC. b, Male lifetime prevalence of contact SVAC. c, Female lifetime prevalence of overall SVAC. d, Male lifetime prevalence of overall SVAC. Darker shading indicates higher prevalence, whereas lighter shading indicates lower prevalence. The base maps were adapted from Natural Earth (public domain, https://www.naturalearthdata.com/ ).

Across Asia (Central, Southern, Eastern, South-Eastern, and Western Asia), as well as in Southern Europe and South America, lifetime SVAC prevalence is similar between men and women, with less pronounced sex differences than those observed in other regions.

Past-year prevalence

Globally, 6.2% (UI 6.0–6.3) of female children and 4.9% (UI 4.8–5.0) of male children experienced any form of SVAC (including both contact and non-contact forms) at least once in the past 12 months (Table 4). The prevalence of overall SVAC in the past year increases with age. The past-year prevalence is highest among children aged 14–17 years (female: 13.3%, UI 13.1–13.6; male: 10.7%, UI 10.5–10.9), followed by children aged 10–13 years (female: 9.5%, UI 9.3–9.7; male: 7.5%; UI 7.4–7.7). Among children aged 0–9 years, the past-year prevalence of contact SVAC is 1.9% (UI 1.9–1.9) for females and 1.4% (UI 1.4–1.5) for males. For contact SVAC, 3.7% (UI 3.6–3.7) of female children and 2.8% (UI 2.8–2.9) of male children under the age of 18 experienced contact SVAC at least once in the past 12 months (Table 4). Similar to overall SVAC, the prevalence of contact SVAC increases with age. The highest prevalence is observed among children aged 14–17 years (female: 8.1%, UI 7.9–8.2; male: 6.3%, UI 6.2–6.5). In 2024, the global prevalence resulted in an estimated 72.6 million girls and 60.9 million boys who experienced at least one form of sexual violence, including 43.3 million girls and 34.8 million boys who were subjected to rape or sexual assault.

Table 4:

Global prevalence estimates of past-year sexual violence against children (SVAC)

	Contact SVAC		Overall SVAC
	Female	Male	Female	Male
0–9 years	1.1% (1.1–1.1%)	0.8% (0.8–0.8%)	1.9% (1.9–1.9%)	1.4% (1.4–1.5%)
10–13 years	5.6% (5.5–5.7%)	4.4% (4.3–4.5%)	9.5% (9.3–9.7%)	7.5% (7.4–7.7%)
14–17 years	8.1% (7.9–8.2%)	6.3% (6.2–6.5%)	13.3% (13.1–13.6%)	10.7% (10.5–10.9%)
10–17 years	6.8% (6.7–7.0%)	5.3% (5.2–5.4%)	11.4% (11.2%−11.6%)	9.1% (8.9–9.3%)
0–17 years	3.7% (3.6–3.7%)	2.8% (2.8–2.9%)	6.2% (6.0–6.3%)	4.9% (4.8–5.0%)

Data presented as % (95% uncertainty interval).

Regional variations in the past-year prevalence of contact and overall SVAC are detailed in Appendix 5 and Appendix 6 (pp 10–11). Figure 3 provides a map illustrating the past-year prevalence of contact SVAC and overall SVAC at the subregion level for both females and males aged 0–17 years.

Fig.3|. Map of prevalence estimates of past-year sexual violence against children (SVAC) by SDG subregion.

a, Female past year prevalence of contact SVAC. b, Male past year prevalence of contact SVAC. c, Female past year prevalence of overall SVAC. d, Male past year prevalence of overall SVAC. Darker shading indicates higher prevalence, whereas lighter shading indicates lower prevalence. The base maps were adapted from Natural Earth (public domain, https://www.naturalearthdata.com/).

In addition to presenting regional prevalence estimates based on SDG regional classifications, we also estimated prevalence by World Health Organization (WHO), World Bank (WB) income group, United Nations Children’s Fund (UNICEF), and Global Burden of Disease (GBD) regions. These results are provided in Appendices 7–10 (pp 12–17). The references for the studies included in this systematic review and meta-analysis are shown in Appendix 11 (pp 18–76).

3. Discussion

Our study confirms that SVAC is highly prevalent globally. One in five (20.1%) adult women experienced sexual violence as children. Among them, one in eight (12.4%) women experienced contact sexual violence in childhood. Among adult men, around one in six (16.8%) experienced sexual violence as children, with one in ten (10.0%) experiencing childhood contact sexual violence. In 2024 alone, an estimated 72.6 million girls and 60.9 million boys - a total of 133.5 million children - experienced any type of childhood sexual violence excluding online sexual violence. Among them, 43.3 million girls and 34.8 million boys experienced childhood contact sexual violence in the past year.

Comparisons with recent global syntheses indicate that our findings are broadly aligned with existing evidence while extending it in important ways (Table 5). For females, our lifetime prevalence estimates for both overall and contact SVAC closely match the 2024 UNICEF assessment. For males, our lifetime prevalence for overall SVAC is modestly higher than UNICEF’s. This divergence likely reflects methodological differences: UNICEF inferred male prevalence by applying a male–female ratio derived from a subset of countries with nationally representative data, whereas we synthesized all available male data directly.¹² Despite drawing on a substantially larger and more linguistically diverse body of evidence—1,412 studies across six UN languages—our lifetime estimates remain closely comparable. In contrast to our study, however, UNICEF reported only lifetime prevalence and did not provide past-year estimates for contact or overall SVAC.¹²

Table 5:

Comparison of prevalence estimates and methodological characteristics across five recent studies of sexual violence against children (SVAC)

		Current Study	Qu et al. (2022)18	Piolanti et al. (2025)19	UNICEF (2024)12	Cagney et al. (2025)20
Global Lifetime Prevalence	Contact SVAC (Female)	12.4%	Not reported	9.5%	About 12.5% (1 in 8)	18.9%
	Contact SVAC (Male)	10.0%	Not reported	5.5%	About 9.1% (1 in 11)	14.8%
	Overall SVAC (Female)	20.1%	24.0%	Not reported	About 20.0% (1 in 5)	Not reported
	Overall SVAC (Male)	16.8%	Not reported	Not reported	About 14.3% (1 in 7)	Not reported
Global Past-Year Prevalence (0–17 years)	Contact SVAC (Female)	3.7%	Not reported	7.6%	Not reported	Not reported
	Contact SVAC (Male)	2.8%	Not reported	4.2%	Not reported	Not reported
	Overall SVAC (Female)	6.2%	Not reported	Not reported	Not reported	Not reported
	Overall SVAC (Male)	4.9%	Not reported	Not reported	Not reported	Not reported
Study Design	Number of Studies Included	1,412	28	165	Not reported	460
	Systematic Review Conducted	Yes	Yes	Yes	No	No
	Databases Searched	PubMed/Medline, Global Health, PsycINFO, Embase, Web of Science, Sociological Abstracts, CINAHL, ERIC, Criminal Justice Abstracts, Google Scholar, + 23 non-English databases + >20 grey literature sources	PubMed, Embase, Web of Science	PubMed, Embase, CINAHL, Web of Science, PsycINFO, ERIC, APA PsycArticles	Not applicable; estimates derived from Tier 1 (DHS/MICS) and Tier 2 (other nationally representative surveys) data sources.	Global Health Data Exchange, WHO Global Database on the Prevalence of Violence against Women, UN Women Global Database on Violence against Women
	Search Terms Language	Search terms in 6 UN languages	English search terms only	English search terms only	Not applicable	Not applicable
	Search Period	Publications in any of the six UN languages from Jan 1, 2010 - Aug 1, 2024	From inception of each database through April 2021	From inception of each database through March 2022	Data collected between 2010 – 2021	Data collected between 1980 – 2023
	Statistical Model Applied	Bayesian hierarchical model	Random-effects model	Random-effects model	Not reported (UNICEF notes “several adjustments” but does not describe the model)	Spatiotemporal Gaussian process regression
	Adjustment for Definitional Variation	Yes	No	No	Yes	Yes
	Adjustment for Age Effects	Yes	No	No	Yes	Yes
	Adjustment for Data Collection Year	Yes	No	No	No	No
	Use of Population Weighting for Regional/Global Estimates	Yes	No	No	Yes	Yes

Methodological heterogeneity across other recent studies also contributes to variation in reported prevalence. Cagney et al. (2025) relied on three predefined databases rather than a systematic search and estimated only lifetime contact SVAC (Table 5). Their global estimates (18.9% for females; 14.8% for males) exceed both UNICEF’s and ours, likely owing in part to upward adjustments applied to estimates obtained via face-to-face interviews.²⁰ Conversely, the systematic review by Piolanti et al. (2025), based primarily on English-language sources and including 165 studies, reported lower lifetime contact SVAC prevalence (Table 5). This pattern is expected given their lack of adjustment for definitional differences, categories of relationship with person who perpetrated violence, respondent age, and data-collection period. Their analytic approach pooled studies reporting violence perpetrated by family members only or peers only with those reporting any relationship, which mathematically biases pooled estimates downward.¹⁹ Moreover, their past-year estimates predominantly reflect adolescents aged 14–17, since the mean respondent age for most included studies falls within this range, rather than all children aged 0–17.¹⁹ Accordingly, their past-year prevalence estimates for contact SVAC among both females and males are lower than our estimates for adolescents aged 14–17, likely reflecting their lack of adjustment for definitional differences. However, their past-year estimates exceed our prevalence estimates for the full 0–17 age range. This difference arises because contact SVAC prevalence is substantially lower among younger children, particularly those aged 0–9, who constitute more than half of the child population. When this age distribution is incorporated, our population-weighted past-year prevalence of contact SVAC becomes lower than theirs despite higher estimates within the 14–17 age group (Tables 4 and 5). The prevalence estimates of overall SVAC were not provided in their study.

The Qu et al. (2022) meta-analysis provides the narrowest scope, based on only 28 studies from three databases.¹⁸ Their sole estimate, lifetime overall SVAC prevalence among females (24%), is modestly higher than our female lifetime estimate (20.1%). No estimates for contact SVAC, for males, or for past-year prevalence were provided. Taken together, these comparisons underscore the distinct contribution of the present study. By drawing on an extensive multilingual search, harmonizing definitional and methodological variation, and applying a Bayesian multilevel framework capable of accommodating sparse and unbalanced data, we provide the most comprehensive and internally consistent set of global SVAC prevalence estimates to date, spanning both sexes, both contact and overall SVAC, and both lifetime and past-year indicators (Table 5).

Girls are generally considered to be at significantly higher risk of sexual violence, a pattern observed in earlier studies and meta-analyses primarily based in Europe and North America.^16,21 However, our findings suggest less pronounced sex differences in some regions, particularly across Asia, as well as in Southern Europe and South America. The reasons for these regional variations have not been widely explored.

Inhibited disclosure, in the face of more intense stigma in some places and for some individuals who experience SVAC, has often been hypothesized. But this explanation for sex patterns needs to include why inhibition varies by sex. Some research suggests that boys are less likely to disclose abuse and may experience greater inhibition than girls.²² Over time, as stigma declines, disclosure rates tend to increase, with little sex difference observed across generations.²²

Another possible explanation is that, in some regions, boys and girls may interpret or encounter sexual violence items differently. Non-contact forms of SVAC such as being told unwanted sexual jokes, being shown pornographic materials, or being made to expose or view private parts may occur more frequently among boys in certain cultural or social contexts.^23–26 As a result, when surveys include these behaviourally specific items, boys may be more likely than girls to endorse such experiences, contributing to higher overall SVAC prevalence among males in some regions.

Cultural and sociological factors may also play a role. Patterns of supervision and household sleeping arrangements vary across cultures and sexes, potentially influencing exposure to risk.²⁷ Additionally, access to pornography, societal expectations for males to gain sexual experience, masculinity norms, and attitudes toward male-to-male sexual activity may contribute to differential risks and vulnerabilities for boys across various contexts.²⁸ Further research is needed to understand the complex interplay of cultural, societal, and structural factors shaping these sex dynamics, as well as how they vary across different contexts including within regions where the sex differences appear minimal, such as in Asia.

Our study includes estimates of both contact sexual violence and overall SVAC (including both contact and non-contact forms of sexual violence), providing a more complete picture of children’s experiences of sexual harm, in line with the UNICEF International Classification of Violence Against Children (ICVAC). While contact forms such as rape and sexual assault represent the most severe physical violations, non-contact forms—including sexual harassment, exposure, and coercive or degrading sexualized behaviours—also cause significant psychological and developmental harm. By presenting estimates for both, this study captures the continuum of experiences encompassed within internationally recognized standards, facilitating cross-study comparability and ensuring that prevention and response strategies address the full range of behaviours that constitute sexual violence against children. Nonetheless, fewer studies have included non-contact forms of SVAC, and these experiences may be more challenging to measure using behaviourally specific approaches. Presenting both estimates allows for a more comprehensive assessment of the burden of sexual violence against children, capturing both the severe physical forms and the broader spectrum of non-contact experiences that contribute to psychological harm and social vulnerability.

Importantly, the adoption of the ICVAC framework provides conceptual clarity and harmonization relative to many earlier syntheses that relied on broader or less well-defined child sexual abuse (CSA) definitions. By using standardized and developmentally grounded categories, our study aims to ensure that the pooled prevalence estimates reflect a consistent conceptualization of SVAC. At the same time, this standardization may limit direct comparability with prior meta-analyses^16,21 that applied less harmonized definitions. The implications of this definitional alignment for interpreting differences in prevalence remain uncertain. Future meta-analytic work could explore how estimates differ when studies are classified according to ICVAC-aligned definitions versus legacy CSA definitions, helping clarify whether this framework substantively changes epidemiological understanding or primarily reorganizes existing constructs.

Our findings underscore the pervasiveness of SVAC globally, with evidence from every region where it has been measured. This high prevalence highlights the urgent need to strengthen child protection systems and invest in robust primary prevention programs, as well as secondary prevention efforts to facilitate justice and healing for those who experience sexual violence. In recent years, global efforts to reduce the burden of SVAC have increasingly emphasized evidence-based, scalable solutions. One major initiative advancing this agenda is the Safe Futures Hub, a global knowledge and evidence platform designed to synthesize, curate, and disseminate rigorous research on “what works” to prevent and respond to childhood sexual violence. By making high-quality evidence more accessible to policymakers, practitioners, and program implementers, the Hub plays a critical role in bridging the gap between research and practice, particularly in low- and middle-income countries where evidence needs are greatest.

Recent evaluations featured within the Safe Futures Hub and related global evidence platforms have identified several effective prevention and response interventions. The 2024 Building Safe Futures Evidence Review for low- and middle-income countries consolidates emerging evidence on effective strategies, highlighting promising approaches such as adolescent development clubs, parenting programs, teacher training initiatives, and cash-plus programs.²⁹ Likewise, the 2025 INSPIRE Evidence Update reports strong new evidence supporting the effectiveness of cognitive behavioral therapy for children exposed to violence and cash-plus life skills training for youth.³⁰

While these findings provide a strong foundation for evidence-based prevention strategies, these reviews also highlight critical gaps. A recent Outcome Measurement Rapid Review flags the ‘mismatch between problem and measurement’, where prevention programs measure outcomes related knowledge or skills rather than experiences of sexual violence.³¹ The Safe Futures Hub Evidence Review notes the lack of evaluations specifically assessing the efficacy of multisectoral coordination in preventing SVAC.²⁹ Despite the absence of rigorous evaluations, some country-specific evidence suggests the promise of such approaches. For example, in Kenya, repeat nationally representative Violence Against Children and Youth Surveys (VACS) show the prevalence of lifetime sexual violence against girls declined significantly from 36.2% in 2010 to 25.2% in 2019.³² A qualitative study identified key factors that may have contributed to the reductions following the government’s and partners’ response to the inaugural VACS and government response plan, including a strengthened legal framework, increased service provider capacity, and improved information systems.³³ Latent class analyses of adverse childhood experiences between the two survey periods revealed shifts in risk factor patterns. For example, the significance of orphanhood co-occurring with sexual and physical violence in males increased in the second VACS compared to the first, even as overall prevalence declined, suggesting that vulnerability to violence may become concentrated among specific populations when the overall population prevalence declines.³⁴ These findings highlight the crucial role of prevalence studies not only in quantifying the magnitude of violence against children but also in identifying shifts in its drivers and risk factors over time. Such insights can guide the reprioritization of prevention, support, and justice efforts for individuals who experienced SVAC, ensuring that limited resources are allocated as strategically and effectively as possible.

Some important limitations of our findings must be acknowledged. Sexual violence is a highly sensitive and stigmatized experience, often leading to significant underreporting in surveys. The extent of underreporting likely varies across countries and regions, potentially distorting global and regional prevalence estimates, particularly in regions with lower reported rates, such as highly populated areas of Asia. If inhibited disclosure differs by sex, this could also affect observed sex differentials.

Additionally, sexual behaviours, attitudes, and cultural norms surrounding sexual violence vary widely across the world. Broad survey prompts may be interpreted differently across regions, potentially capturing distinct types of experiences—such as bullying content, exposure to sexual images, early childhood sexual exploration, or consensual homosexual activity—thereby producing misleading impressions of regional differences. As of now, the field lacks a comprehensive understanding of the factors driving the sometimes-large differences in prevalence estimates within and across countries. Further research is urgently needed to investigate these disparities, which may lead to reinterpretations of the findings presented here.

More broadly, these definitional and methodological inconsistencies highlight that global syntheses, including our own and prior studies, are unavoidably constrained by variability in measurement quality and conceptual alignment. A larger pool of studies does not necessarily indicate stronger evidence; rather, expanding the evidence base may increase exposure to inconsistent designs, weak measures, or problematic question wording. These foundational uncertainties underscore the need for caution when interpreting pooled global estimates of SVAC.

Another key limitation is that almost all past-year prevalence estimates for children under 10 rely on parental or proxy reports, which are known to substantially underreport sexual violence against young children.^16,35 Consequently, past-year prevalence estimates for the 0–9 age group generated in this study may underestimate the true prevalence.

Moreover, the Bayesian hierarchical modelling approach used in this study applies partial pooling, meaning that estimates for subregions with limited data are influenced by broader regional or global averages. While this helps stabilize estimates, it may misrepresent reality if the true prevalence in these subregions is substantially higher or lower than the pooled estimate.

Furthermore, the current systematic review has not reported data on race or ethnicity, as many of the included studies either did not report those characteristics due to local restrictions or reported those as descriptive demographics. In addition, the surveys did not disaggregate the prevalence estimates by those characteristics, so we excluded them to prevent potential misinterpretation. This study did not include online SVAC which may have underestimated prevalence estimates. Online sexual violence was excluded due to limited and non-comparable data. Existing studies are few, recent, and often focus exclusively on online forms, with wide variation across countries in internet access, digital use, and definitions, precluding reliable estimation. While this exclusion preserved the validity of the Bayesian estimates, it represents an important limitation, as digital contexts are increasingly relevant to children’s experiences of sexual harm.

It is also important to note that throughout this manuscript, we aimed to use person-first and non-stigmatizing language when describing individuals affected by SVAC. Person-first terminology helps ensure that individuals are not defined by harmful experiences or behaviors and supports a more prevention-oriented framing of SVAC.³⁶ We recognize the importance of continued attention to language in future research to ensure accuracy, respect, and alignment with prevention goals, including prevention of perpetration of sexual violence.

Finally, our study estimated overall population prevalence using general population samples representative at national or subnational levels. While these samples included individuals of all sexual orientations and gender identities, subgroup-specific data were rarely available, limiting separate analyses. Future research should collect and report data that enable disaggregated analyses by sexual orientation and gender identity to enhance inclusivity and understanding of disparities.

Despite these limitations, this study compellingly demonstrates the persistently high prevalence of SVAC worldwide. Our findings highlight the urgent need for standardized data collection protocols and expanded research efforts to close critical knowledge gaps. Comprehensive prevalence studies that include both boys and girls and measure both lifetime and past-year experiences are essential for accurately assessing the scope of this issue. The example from Kenya demonstrates that data-driven awareness, political will, and financial investment can drive impactful policy and programmatic responses.^32–34 Successful SVAC prevention efforts have shown that addressing harmful social and cultural norms can reshape the contexts in which violence occurs.²⁹ To achieve lasting progress, a co-ordinated global response should integrate standardized data collection, rigorous research, and culturally sensitive, evidence-based prevention strategies. With strong commitment from governments, policymakers, researchers, civil society, and communities, we can drive transformative change and significantly reduce the global burden of childhood sexual violence.

4. Methods

Search strategy and selection criteria

We initially conducted searches on March 1, 2023, and updated them on August 20, 2024. We employed a comprehensive set of search terms designed to identify studies of SVAC prevalence. These terms corresponded to act-based measures of sexual violence as defined by the UNICEF International Classification of Violence against Children,² combined with additional terms to specifically limit the search to childhood populations and studies reporting prevalence estimates (see Appendix 1, Table B, for the full list of terms in six UN languages, pp 4–6).

We searched PubMed/MEDLINE, the Social Science Citation Index (Web of Science), Global Health, Embase, PsycINFO, CINAHL, ERIC, Sociological Abstracts (ProQuest), Criminal Justice Abstracts, and Google Scholar, along with key journals in the field of child protection. Grey literature sources included research from international non-governmental organizations (INGOs), UN agencies, community-based organizations (CBOs), and reports from national governments. We also searched relevant language-specific databases (see Appendix 1, Table A, for a full list of databases and grey literature repositories searched, pp 2–3). For globally comparable national survey reports (for example, DHS and MICS), we retained eligible reports regardless of publication language, including those beyond the six UN languages, to maximize completeness and country coverage of prevalence estimates.

Studies were included if they: (1) were published in one of the six official UN languages (Arabic, Chinese, English, French, Russian, Spanish) between January 1, 2010, and August 1, 2024; (2) reported the prevalence of SVAC; (3) relied on self-reported data from children (individuals under the age of 18 years), adults recalling childhood experiences (experiences that occurred prior to age 18), or reports from parents/guardians of experiences of their minor children; (4) included a measure of contact or non-contact SVAC, as consistent with the UNICEF ICVAC definitions (Table 1); (5) used general population samples representative at national or sub-national levels; and (6) had a sample size of at least 100.

Table 1:

Operational definitions of sexual violence against children (SVAC) and indicators^*

	Definition
Overall SVAC	Overall sexual violence against children (SVAC) refers to any deliberate, unwanted, and non-essential sexual act, whether completed or attempted, involving either contact or non-contact, that is perpetrated against a child. This definition excludes acts of sexual exploitation and those occurring in an online context.
	Contact SVAC includes rape and sexual assault.
	Rape refers to vaginal, anal, or oral penetration of a sexual nature of a child’s body with any bodily part or object, with or without the use of force, and without consent because the child is too young to consent or consent is not given. Examples include physically forced rape, pressured or coerced rape, drug- and/or alcohol-facilitated rape, non-consensual sexual penetration without physical force or threat, incest involving a child, rape in the context of armed conflict, and gang-perpetrated rape.
Contact SVAC	Sexual assault refers to touching a child’s private parts or making a child touch someone else’s private parts (excluding penetration), with or without the use of force, and without consent because the child is too young to consent or consent is not given. This primarily includes unwanted groping, fondling, or other forms of touching; sexual acts (other than penetration) obtained through threats of physical violence; sexual acts (other than penetration) obtained through threats to the well-being of family members; the use of force or coercion to obtain unwanted sexual acts or any sexual activity that the child finds degrading or humiliating; and pulling a child’s clothing up or down to reveal intimate areas.
Non-contact SVAC	Non-contact SVAC refers to any form of verbal or non-verbal non-physical conduct, whether isolated or persistent, that involves unwanted references to the body, sexual organs, or sexuality of the child, excluding acts occurring in an online context. This includes, but is not limited to, sexual harassment, sexual threats, exposure of a child to sexual abuse and pornography, sexual bullying and/or unwanted sexual jokes, taunts, or comments, exposing sexual organs, trapping a child and subjecting him/her to sexual advances, subjecting a child to sexual rumours, persistent leering looks, sexual stalking, sexual extortion, coercion and blackmailing a child for sexual purposes, and non-consensual image-taking of the sexual organs or sexual activities involving a child.
Lifetime prevalence of SVAC	The proportion of adults aged 18 years or older who reported experiencing contact or non-contact SVAC at any point before reaching 18 years of age
Past-year prevalence of SVAC	The proportion of children under 18 years who reported experiencing contact or non-contact SVAC by anyone at least once within the previous 12 months.

^*The definition of SVAC employed for this study aligns with the operational definitions specified in the UNICEF International Classification of Violence against Children.²

Studies were excluded if they: (1) focused on specific sub-populations unlikely to generalise to the general population (e.g., psychiatric patients, individuals who identify as LGBTQ+, or samples comprising only individuals who experienced SVAC); (2) did not provide disaggregated data for children or adults’ experiences as children (e.g., findings for those under 18 years could not be determined); (3) collected data through controlled studies or qualitative methods; and (4) only included data on online SVAC.

Eight reviewers screened English studies, and two reviewers screened studies in each of the other UN languages. For all papers, two reviewers independently screened studies using Covidence software.³⁷ After initial title and abstract screening and removal of duplicates, studies meeting eligibility criteria underwent full-text review. Discrepancies were resolved by a third reviewer.

At least two authors independently extracted data for each selected study using a standardized data extraction tool that we piloted across all language teams (see Appendix 2 for extraction form, pp 7). Any modifications were agreed upon and shared with all authors. For studies using the same dataset and reporting the same type of prevalence, we retained only the study with the largest sample size. We assessed methodological quality using Hoy et al.’s risk of bias tool.³⁸ This tool comprises nine questions evaluating internal validity (e.g., case definition, data source, instrument quality, data collection method, numerators, and denominators) and external validity (e.g., population representativeness, sampling method, non-response bias). Each item was rated as “low risk” (score 0) or “high risk” (score 1), with total scores classifying studies as low (0–3), moderate (4–6), or high risk (7–9).³⁸

Data analysis

The lifetime prevalence of SVAC is defined as the proportion of adults aged 18 years or older who reported experiencing contact or non-contact SVAC at any point before reaching 18 years of age. The past-year prevalence is defined as the proportion of children under 18 years who experienced contact or non-contact SVAC within the previous 12 months. In both measures, consistent with the UNICEF ICVAC definition, individuals are counted as experiencing SVAC if they were subjected to at least one act of sexual violence perpetrated by anyone, regardless of the relationship of the individual who perpetrated violence to the child. For this estimation, we examined two SVAC metrics: contact SVAC (rape and sexual assault) and the broader category of overall SVAC, which includes both contact and non-contact forms, excluding online SVAC (see definitions in Table 1). Global or regional estimates for online sexual violence, a form of non-contact violence, were not produced due to significant limitations in data availability. Available studies are few, mostly recent, and often focus exclusively on online sexual violence without data on other forms, limiting comparability and precluding imputation.¹⁵ Prevalence varies substantially with national differences in internet access, digital adoption, and social media use.¹² Variability in legal definitions, reporting practices, and cultural norms further complicates measurement.¹⁵ Given these constraints, we excluded online sexual violence to preserve the validity of Bayesian estimates for childhood sexual violence.

To estimate SVAC prevalence, we defined an optimal set of observations. These included data where the case definition encompassed all acts of sexual violence (e.g., at least one experience of any type of SVAC), did not restrict specific categories of individuals who perpetrated violence, and represented both rural and urban populations to ensure comprehensive coverage. Observations must also have reported prevalence separately by sex and have used clear, behaviour-specific questions. For lifetime prevalence, incidents must have occurred before age 18 (e.g., not before 15 or 13), and for past-year prevalence, within the prior 12 months (e.g., not six or three months).

If studies did not meet these criteria, we applied statistical adjustments to align data with the optimal definition. The analysis incorporated all available observations, differentiating between those aligning with the optimal definition and those requiring adjustment, thereby addressing missing data and inconsistencies in sampling and case definitions.¹³ Details of these statistical adjustments, particularly for the fixed effects component of the Bayesian hierarchical model, are provided in detail later.

We developed a Bayesian hierarchical (multilevel) framework to estimate the lifetime or past year prevalence of SVAC across different regions. We used a Bayesian multilevel model because it offers greater flexibility and robustness than traditional meta-regression for estimating pooled prevalence.^13,39,40 This approach accounts for multiple sources of variability, such as between-country, regional, temporal, and definitional differences, through random and mixed effects, allowing a more realistic representation of hierarchical data.^13,40 The Bayesian framework also handles sparse or unbalanced data effectively by borrowing strength across units, resulting in more stable estimates, particularly in settings with small sample sizes or limited observations for certain subgroups.^13,39 In addition, it accommodates the bounded nature of prevalence data by using appropriate likelihoods (e.g., binomial or beta binomial) and link functions.^13,41 Finally, Bayesian inference provides full posterior distributions and credible intervals with direct probability interpretations, offering clearer and more intuitive uncertainty quantification than conventional methods.^13,39

Our model is specified within the generalized linear regression (GLM) framework, using a binomial distribution with a logit link function. The response variable represents the number of SVAC cases observed out of the total trials.⁴¹ This framework has four nested levels: (1) countries, (2) subregions, (3) super-regions, and (4) the world. The SDG super-regions and subregions were defined by the United Nations Statistics Division (UNSD) to support reporting and monitoring of progress toward the SDGs. These groupings, based on geographical, economic, and social similarities, include 7 super-regions further divided into 20 nested subregions (see Appendix 3 for regional classifications, pp 8).⁴² The framework also accounts for various demographic and geographic factors and includes both fixed and random effects to pool data from different sources and capture variability at different levels. This multilevel approach of including both fixed and random effects enables us to “borrow strength” across units, ensure the degree of pooling is determined empirically by the data and not arbitrarily by the user, and adjust for definitional and methodological variations across studies.^13,39,40

The meta-regression model includes fixed effects for each type of violence (contact versus non-contact) and time frame (past-year versus lifetime) across the seven SDG super-regions. To account for the wide sex difference in SVAC prevalence across regions, the model also includes sex as both a main effect and an interaction term with the 20 SDG subregions. This interaction allows the model to capture regional variations in SVAC prevalence between males and females, ensuring that sex-specific differences are not assumed to be uniform across all subregions.

We modelled the non-linear effects of age and the year in which data collection for the SVAC prevalence study ended using natural splines with three degrees of freedom, with two knots evenly distributed across the observed range for each variable. Additional fixed effects included relationship with person who perpetrated violence (e.g., any, parents, partner, peers), frequency of acts (e.g., at least once vs. at least twice), prevalence recall period (e.g., prior to age 18, prior to age 15, past 12 months, past 3 months), subtype of violence, sample site, and SVAC question format (behaviour-specific inquiry, mixed, non-behaviour-specific inquiry). Non-behaviour-specific inquiries utilize broad or umbrella terms, placing reliance on respondents’ subjective interpretation of concepts such as “sexual harassment” or “sexual abuse.” For example, questions may include “Have you ever been sexually harassed?” or “Have you ever been sexually abused?”. The specific adjusted variables and categories can be found in Appendix 4 (pp 9).

We applied nested random effects to the geographical regions in the hierarchy of super-region, subregion, and country, assuming that a country should be closer to its regional prevalence than to that of other regions and allowing borrowing strength from other observations in the same region to improve prevalence estimates.^13,40 We modelled sub-national observations as having more variability than those representative at the national level, meaning less weight was given to sub-national observations.^13,39

The proposed Bayesian model employed weakly informative priors, specifying normal distributions for the regression coefficients and a half-t distribution for the standard deviations of the random effects.⁴⁰ This ensured reasonable regularization of the model parameters, improving convergence and interpretability. To specify the hyperparameters (including the number of knots in the splines, the prior variance for the regression coefficients of the fixed effects, and priors for the standard deviation of the random effects) in the model, we used the Widely Applicable Information Criterion (WAIC) as the model selection criteria.^13,39

Posterior inferences were obtained using the Markov chain Monte Carlo (MCMC) simulations with 10,000 iterations with 5,000 burn-ins. To ensure computational efficiency, we employed parallel processing. We assessed model convergence using Gelman-Rubin diagnostic statistics with multiple MCMC chains. Model validation was carried out using posterior predictive checks (including Bayesian p-values) and both in-sample and out-of-sample comparisons. These assessments examined median errors, median absolute errors, and the coverage of posterior predictive intervals.^13,39 The final model produced country-level estimates of the SVAC prevalence accompanied by 95% uncertainty intervals (UIs). To generate aggregate estimates (i.e., by subregion, super-region, and globally), we used population denominators for the 2024 calendar year from the World Population Prospects (2024 revision).⁴³ We performed all analyses in R (version 4.5.0) and fitted the model primarily using the “brms” package. Analysis code is publicly available to ensure reproducibility.

The protocol for this systematic review and meta-analysis was developed following the PRISMA-P standards⁴⁴ and has been registered on PROSPERO (number CRD42024495116). We adhered to PRISMA⁴⁵ standards for reporting the results.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All the data supporting the findings of this work can be accessed via the Open Science Framework at https://doi.org/10.17605/OSF.IO/EG8TN.

Code availability

All analysis scripts are openly available at https://doi.org/10.17605/OSF.IO/EG8TN.