The health effects associated with physical, sexual and psychological gender-based violence against men and women: a Burden of Proof study

Abstract

The health impacts of exposure to physical, sexual or psychological gender-based violence (GBV) against men and women are substantial yet not well delineated. We systematically reviewed and meta-analysed 40 studies to evaluate the associations between GBV (including but not limited to intimate partner violence) and eight health outcomes: sexually transmitted infections excluding HIV, maternal abortion and miscarriage, HIV/AIDS, major depressive disorder, anxiety disorders, drug use disorders, alcohol use disorders and self-harm. Using the Burden of Proof methods, we generated conservative metrics of association-including star ratings from one to five-reflecting both effect size and evidence strength. Sexual violence was associated with six outcomes, with moderate, three-star evidence of association for sexually transmitted infections, maternal abortion and miscarriage, and major depressive disorder-increasing the risk by at least 104%, 101% and 50%, respectively. Seven outcomes were associated with physical GBV with two- or one-star associations, reflecting weak associations and/or inconsistent evidence. Of the four health outcomes analysed in relation to psychological GBV, one, major depressive disorder, had a significant association with a one-star rating. These findings emphasize the serious health consequences of GBV for survivors and the necessity of additional data to further our understanding of this complex public health issue.

Fig. 1. Flow diagram of a systematic review on health effects associated with physical, sexual and psychological GBV against men and women.

The PRISMA flow diagram shows the data-seeking approach undertaken as a part of our review, which sought to identify all literature reporting on the health effects of GBV, violence against women, and violence against children and young people. Studies specifically measuring the health impacts of physical, sexual and psychological GBV against men and women were used for this analysis.

Fig. 2. Forest plots for physical GBV and health outcomes identified through a systematic review of the literature.

The shape of each point indicates the gender of the sample (women only, men only, and combined women and men). Light blue shading corresponds to the 95% UI incorporating between-study heterogeneity; dark blue shading corresponds to the 95% UI without between-study heterogeneity. The black vertical dashed lines reflect the null RR value (one), and the red vertical lines show the BPRF at the fifth quantile for these harmful risk–outcome associations. The black data points and horizontal lines each correspond to an effect size and 95% UI from the included study identified on the y axis. We included multiple observations from a single study when effects were reported by gender and health outcome (anxiety disorders and PTSD). Supplementary Table 1 contains more details on the observations included from each study.

Fig. 3. Forest plots for sexual violence and health outcomes identified through a systematic review of the literature.

The shape of each point indicates the gender of the sample (women only, men only, and combined women and men). Light blue shading corresponds to the 95% UI incorporating between-study heterogeneity; dark blue shading corresponds to the 95% UI without between-study heterogeneity. The black vertical dashed lines reflect the null RR value (one), and the red vertical lines show the BPRF at the fifth quantile for these harmful risk–outcome associations. The black data points and horizontal lines each correspond to an effect size and 95% UI from the included study identified on the y axis. We included multiple observations from a single study when effects were reported by gender, health outcome (induced abortion and spontaneous abortion) and perpetrator. Supplementary Table 1 contains more details on the observations included from each study.

Fig. 4. Forest plots for psychological GBV and health outcomes identified through a systematic review of the literature.

The shape of each point indicates the gender of the sample (women only, men only, and combined women and men). Light blue shading corresponds to the 95% UI incorporating between-study heterogeneity; dark blue shading corresponds to the 95% UI without between-study heterogeneity. The black vertical dashed lines reflect the null RR value (one), and the red vertical lines show the BPRF at the fifth quantile for these harmful risk–outcome associations. The black data points and horizontal lines each correspond to an effect size and 95% UI from the included study identified on the y axis. We included multiple observations from a single study when effects were reported by gender, exposure definition and age group. Supplementary Table 1 contains more details on the observations included from each study.

Fig. 5. Mean RR and strength of the evidence for the association between multiple forms of GBV and eight health outcomes.

The number in each box corresponds to the mean RR estimated as part of this study for associations for psychological, physical and sexual GBV and specific health outcomes. The coloured boxes represent all associations supported by at least three published studies, allowing for our Burden of Proof assessment, while empty white cells reflect risk–outcome pairs for which we did not have enough data (three or more studies) to examine. The shades of the blue boxes represent the strength of evidence supporting each association based on our conservative interpretation of the data that aligns with the Burden of Proof approach. The strength of the evidence is translated into a star rating from one to five stars on the basis of thresholds outlined in Zheng et al., where one star denotes weak evidence, and each additional star indicates progressively stronger evidence. An absence of stars (zero stars) signifies insufficient evidence of a significant association between the exposure and the outcome.

Extended Data Fig. 1. Mean relative risk and strength of evidence found in sensitivity analyses on the associations between physical GBV and various health outcomes.

This heatmap compares the estimated mean relative risk and the star rating for the association between physical GBV and each of the seven health outcomes evaluated across different sensitivity analyses. The left most column, titled “Primary analysis,” reflects the primary results of the present manuscript. Each consecutive column reflects a different sensitivity analysis (described in more detail in the Methods) specific model restrictions or restrictions to the input data. The cells marked with a ^ reflect a restriction does not change the data or modeling parameters from the primary analysis, resulting in matching models. Each row corresponds to a different outcome of interest. The number overlayed on each cell corresponds to the mean relative risk estimate derived from each analysis for the relationship between the health outcome listed on the y-axis and exposure to physical violence. Colored cells with no overlayed text reflect zero-star risk-outcome pairs in which the analysis in question did not find sufficient evidence of an association between the outcome and physical violence. The cells are colored in accordance with the star rating based on a conservative interpretation of the data in the sensitivity analysis of interest on a 1 (weak evidence) to 5 (strong evidence) star scale. Empty white cells reflect sensitivity analyses for which we did not have enough relevant data (three or more studies) to examine.

Extended Data Fig. 2. Mean relative risk and strength of evidence found in sensitivity analyses on the associations between sexual violence and various health outcomes.

This heatmap compares the estimated mean relative risk and the star rating for the association between sexual violence and each of the six health outcomes evaluated across different sensitivity analyses. The left most column, titled “Primary analysis,” reflects the primary results of the present manuscript. Each consecutive column reflects a different sensitivity analysis (described in more detail in the Methods) specific model restrictions or restrictions to the input data. Each row corresponds to a different outcome of interest. The cells marked with a ^ reflect a restriction does not change the data or modeling parameters from the primary analysis, resulting in matching models. The number overlayed on each cell corresponds to the mean relative risk estimate derived from each analysis for the relationship between the health outcome listed on the y-axis and exposure to sexual violence. The cells are colored in accordance with the star rating based on a conservative interpretation of the data in the sensitivity analysis of interest on a 1 (weak evidence) to 5 (strong evidence) star scale. Empty white cells reflect sensitivity analyses for which we did not have enough relevant data (three or more studies) to examine.

Extended Data Fig. 3. Mean relative risk and strength of evidence found in sensitivity analyses on the associations between psychological GBV and various health outcomes.

This heatmap compares the estimated mean relative risk and the star rating for the association between psychological GBV and each of the two health outcomes evaluated across different sensitivity analyses. The left most column, titled “Primary analysis,” reflects the primary results of the present manuscript. Each consecutive column reflects a different sensitivity analysis (described in more detail in the Methods) specific model restrictions or restrictions to the input data. Each row corresponds to a different outcome of interest. The cells marked with a ^ reflect a restriction does not change the data or modeling parameters from the primary analysis, resulting in matching models. The number overlayed on each cell corresponds to the mean relative risk estimate derived from each analysis for the relationship between the health outcome listed on the y-axis and exposure to psychological violence. Colored cells with no overlayed text reflect zero-star risk-outcome pairs in which the analysis in question did not find sufficient evidence of an association between the outcome and psychological violence. The cells are colored in accordance with the star rating based on a conservative interpretation of the data in the sensitivity analysis of interest on a 1 (weak evidence) to 5 (strong evidence) star scale. Empty white cells reflect sensitivity analyses for which we did not have enough relevant data (three or more studies) to examine.