Quantifying HIV transmission in the context of armed conflict in the Tigray Region, Northern Ethiopia: a dynamic risk equation model Authors: Abdelkadir Muzey Mohammed, Habtu Alemayehu Atsbeha, Yohannes Yirga Kefela, Woldegebriel Assefa Woldegerima, Kiros Tedla Gebrehiwot Published: 2026 ABSTRACT The two-year armed conflict in the Tigray Region of Northern Ethiopia (2020–2022) created a humanitarian crisis marked by widespread sexual violence, healthcare disruption, and displacement of health personnel. However, reliable epidemiological evidence on the contribution of conflict-related sexual violence to HIV transmission remains scarce. This study addresses this gap by quantifying the impact of conflict-related rape on HIV incidence and prevalence in Tigray. A dynamic risk equation model was developed to capture HIV transmission during conflict. In this framework, conflict intensity influences key transmission drivers, including rape prevalence, number of perpetrators, trauma severity, reduced post-exposure prophylaxis (PEP) use, and healthcare service degradation. Data from the Central Statistics Agency of Ethiopia, UNAIDS regional estimates, World Health Organization, Ethiopian Public Health Institute, and relevant literature were used. Uncertainty analyses were conducted, complemented by global sensitivity analysis and numerical simulations. This model-based study indicates that mass rape could cause a median of 122 (interquartile range, IQR: 13–681) new HIV infections, indicating conflict-related mass rape could cause a median increase of 13.8% (IQR: 1.9%–62.5%), while the median increase in prevalence could be 0.002% (IQR: 0.00027%–0.0089%). These results suggest the increase in incidence due to rape is epidemiologically notable though the effect on population prevalence is small, reflecting that escalating conflict intensity characterized by increased rape frequency, declining health system capacity, and reduced PEP coverage further amplifies HIV transmission. Integrated strategies, including prevention of sexual violence, timely PEP access, and mobile health services, are essential to mitigate HIV transmission. Humanitarian agencies and conflicting parties should act on both public health and human rights grounds. Introduction Sexual violence in conflict has been widely documented in Burundi, Sierra Leone, Rwanda, the Democratic Republic of Congo, Liberia, Sudan, and Uganda1, and most recently in Tigray, Northern Ethiopia2. Armed conflict is often accompanied by widespread sexual violence, displacement, food insecurity, and collapse of health services3–5. Sub-Saharan Africa, which bears the highest HIV burden, is also the region most affected by armed conflicts6. Consequently, conflicts are recognized as accelerators of HIV spread, contributing to increased incidence and prevalence7,8. The Tigray war (2020–2022) exemplifies these dynamics. Reports confirmed widespread rape and other forms of sexual violence were perpetrated against children and older persons, with ages ranging from nine to 60 years old9,10. Community surveys indicated that 43% of women experienced sexual and gender-based violence (SGBV) and nearly 8% were raped, while fewer than 10% accessed support services11. Estimates suggest that 100,000–120,000 women were raped12,13, with rates exceeding those documented in Sierra Leone, Uganda, and Ukraine2. Patterns of conflict-related sexual violence in Tigray included consecutive, individual, and multiple-perpetrator rapes, involving up to fifteen perpetrators from different armed forces and allied militias2,4,9,10. Many women and girls experienced gang rape by three to nine soldiers, with gang rapes accounting for 68.4% of cases2. According to Insight Insecurity, among 106 surveyed survivors, 67% reported a single rape, while 33% experienced mass rape involving multiple perpetrators; of these 33 cases, 76% involved multiple perpetrators and 24% involved a single perpetrator4. At the same time, the health system collapsed: 70% of hospitals, 83% of health centers, and over 700 health posts became non-functional14. Presence of armed forces, unsafe transport, drug shortages, and traditional barriers restricted access to care3,9,15. The United Nations Population Fund16 reported that only 1% of facilities provided rape management services, leaving over 90% of survivors without timely post-exposure prophylaxis. As a result, many women developed untreated injuries, infections, and elevated HIV risk. Fisseha et al.2 found 2.7% of survivors contracted HIV, while regional prevalence rose from 1.4% pre-war to 3%17. More generally, conflict event data and qualitative reports from sources including the Armed Conflict Location & Event Data Project18, Human Rights Watch19, the Office for the Coordination of Humanitarian Affairs20, and the International Human Rights Commission Experts Ethiopia & United Nations9 highlight that the intensity of the conflict during 2021–2022 can reasonably be represented on a normalized scale from 0 to 1, with peak periods of violence in 2021 corresponding to values approaching 0.8, reflecting maximum disruption of health services21. Armed conflict increases HIV transmission risk through multiple interlinked pathways: (1) a direct increase in sexual violence, particularly mass rape with multiple perpetrators; (2) the disruption of health systems, leading to the collapse of HIV prevention, testing, and treatment services; (3) reduced access to PEP for survivors; and (4) increased population displacement and mobility, which can accelerate transmission. This study focuses on integrating the first three pathways through a conflict-intensity-dependent modeling framework. This study’s findings are relevant to national and international strategies, including Ethiopia’s National HIV/AIDS Strategic Plan and UN Security Council resolutions on conflict-related sexual violence (e.g., UNSCR 1820, 1888), which emphasize the need for integrated prevention and response in humanitarian settings. Despite the scale of violence and its consequences on the Tigray health system, reliable epidemiological evidence on SGBV during conflict remains scarce2,4,9. Specifically, no quantitative estimates exist for the number of HIV infections directly attributable to conflict-related rape in Tigray, representing a critical knowledge gap for humanitarian response and public health planning. This is in line with Tewabe et al.22, who reported that no rigorous empirical studies have documented the extent of SGBV and its impact on HIV epidemics in Northern Ethiopia, necessitating the use of mathematical models23–26. Mathematical modeling provides a vital alternative to estimate the impact of SGBV on the HIV epidemic where direct data are unavailable; for instance, see the works of Mwangi & Luboobi26, Anema et al.23, and Supervie et al.24. Anema et al.23 and Supervie et al.24 developed a static risk equation model to quantify the HIV infections attributable to conflict-related mass rape in sub-Saharan African countries. Building on these foundations, the present study develops a dynamic risk equation model that extends earlier static models. The key innovations include: (1) a cumulative exposure risk from multiple perpetrators, (2) a dynamic transmission probability \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r(c)$$\end{document} that is a function of conflict intensity, and (3) the integration of trauma-related risks and the synergistic effects of health system collapse and PEP decline. This approach aims to quantify how mass rape during the Tigray conflict contributed to HIV incidence and prevalence, offering structured insights for humanitarian response and public health planning. Methods Conceptual framework of the study The conceptual framework, illustrated in Fig. 1, maps the causal pathways from conflict intensity to the risk of HIV transmission. This framework is built on the central premise that conflict is the primary exogenous driver, acting as a “force multiplier” that cascades through demographic, behavioral, and systemic vulnerabilities to produce adverse health outcomes. The model architecture reveals a complex, nonlinear system where the whole is greater than the sum of its parts, and its value lies in its ability to disaggregate a seemingly intractable problem into measurable, interacting components. Conflict intensity as the core driver At the apex of the framework is Conflict Intensity (c). The model conceptualizes conflict not as a binary state but as a continuous variable. This is a critical advancement, as it allows for the calibration of the model across different conflict phases—from low-intensity instability to full-scale war. The diagram emanating from this central node signifies the pervasive, often chaotic, and nonlinear way conflict disrupts every facet of society. It does not simply increase the number of risk factors but fundamentally alters their nature and interrelationships. The dynamic nature of our model is based on the conflict intensity, which influences several factors to vary. Pathways of vulnerability The framework deconstructs the overall risk equation, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I = X r P \beta _r$$\end{document}, into four distinct pathways, highlighting the multistage nature of the risk: Demographic Exposure(Xr): This component represents the scale of the population at risk. The framework correctly distinguishes between the baseline susceptible population (women and girls aged 5–49) and the conflict-induced behavioral risk—the rape proportion (r). A strength of this model is that it forces the analyst to consider both the size of the vulnerable group and the intensity of the specific violence they face, rather than conflating the two.The Assailant Risk Pathway(P): A key innovation is the explicit modeling of the assailant. The cumulative exposure probability, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P = 1 - (1 - P_r)^{n_r}$$\end{document}, moves beyond a simple binary (assailant is HIV-positive or not) to incorporate the number of assailants (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}). This is a nuanced acknowledgment of the brutal reality of conflict-related sexual violence, which often involves multiple perpetrators (gang rape). By including \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}, the model captures the exponential increase in risk that arises from such events, a nuance often missing in standard epidemiological models.The Systemic Failure Pathway(E, ϕ , H): This is where the framework depth is most apparent. It posits that conflict degrades the very systems designed to protect individuals, thereby magnifying the biological risk of transmission. The three mediating factors—trauma-related risk (E), PEP coverage (ϕ), and health system functionality (H)—are modeled as decaying exponentials of conflict intensity. This formulation reflects a grim reality: as conflict intensifies, survivors are more likely to experience severe trauma (which can increase biological susceptibility), less likely to access post-exposure prophylaxis (PEP), and will find the health system itself in a state of collapse. The transmission probability function, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r = \tau + \sigma E(1 - \phi )(1 - H)$$\end{document}, is therefore a product of failures, demonstrating that the highest risk occurs when these three factors converge.Fig. 1Conceptual Framework: Dynamic Risk Equation Model for HIV Transmission through Conflict-Related Sexual Violence. The diagram illustrates the pathways through which conflict intensity (c) influences HIV transmission risk. Conflict intensity affects multiple parameters: rape proportion (r), number of assailants (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}), trauma-related exposure (E), PEP coverage (ϕ), and health system functionality (H). These factors feed into the cumulative exposure probability (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P = 1 - (1 - P_r)^{n_r}$$\end{document}), where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document} is the probability that a single assailant is HIV-positive, and the dynamic transmission probability (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r = \tau + \sigma E(1 - \phi )(1 - H)$$\end{document}). The final estimated conflict-related HIV infections are calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I = X r P \beta _r$$\end{document}, where X represents the susceptible female population aged 5–49 years. Demographic Exposure(Xr): This component represents the scale of the population at risk. The framework correctly distinguishes between the baseline susceptible population (women and girls aged 5–49) and the conflict-induced behavioral risk—the rape proportion (r). A strength of this model is that it forces the analyst to consider both the size of the vulnerable group and the intensity of the specific violence they face, rather than conflating the two. Demographic Exposure(Xr): This component represents the scale of the population at risk. The framework correctly distinguishes between the baseline susceptible population (women and girls aged 5–49) and the conflict-induced behavioral risk—the rape proportion (r). A strength of this model is that it forces the analyst to consider both the size of the vulnerable group and the intensity of the specific violence they face, rather than conflating the two. The Assailant Risk Pathway(P): A key innovation is the explicit modeling of the assailant. The cumulative exposure probability, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P = 1 - (1 - P_r)^{n_r}$$\end{document}, moves beyond a simple binary (assailant is HIV-positive or not) to incorporate the number of assailants (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}). This is a nuanced acknowledgment of the brutal reality of conflict-related sexual violence, which often involves multiple perpetrators (gang rape). By including \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}, the model captures the exponential increase in risk that arises from such events, a nuance often missing in standard epidemiological models. The Assailant Risk Pathway(P): A key innovation is the explicit modeling of the assailant. The cumulative exposure probability, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P = 1 - (1 - P_r)^{n_r}$$\end{document}, moves beyond a simple binary (assailant is HIV-positive or not) to incorporate the number of assailants (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}). This is a nuanced acknowledgment of the brutal reality of conflict-related sexual violence, which often involves multiple perpetrators (gang rape). By including \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}, the model captures the exponential increase in risk that arises from such events, a nuance often missing in standard epidemiological models. The Systemic Failure Pathway(E, ϕ , H): This is where the framework depth is most apparent. It posits that conflict degrades the very systems designed to protect individuals, thereby magnifying the biological risk of transmission. The three mediating factors—trauma-related risk (E), PEP coverage (ϕ), and health system functionality (H)—are modeled as decaying exponentials of conflict intensity. This formulation reflects a grim reality: as conflict intensifies, survivors are more likely to experience severe trauma (which can increase biological susceptibility), less likely to access post-exposure prophylaxis (PEP), and will find the health system itself in a state of collapse. The transmission probability function, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r = \tau + \sigma E(1 - \phi )(1 - H)$$\end{document}, is therefore a product of failures, demonstrating that the highest risk occurs when these three factors converge. The Systemic Failure Pathway(E, ϕ , H): This is where the framework depth is most apparent. It posits that conflict degrades the very systems designed to protect individuals, thereby magnifying the biological risk of transmission. The three mediating factors—trauma-related risk (E), PEP coverage (ϕ), and health system functionality (H)—are modeled as decaying exponentials of conflict intensity. This formulation reflects a grim reality: as conflict intensifies, survivors are more likely to experience severe trauma (which can increase biological susceptibility), less likely to access post-exposure prophylaxis (PEP), and will find the health system itself in a state of collapse. The transmission probability function, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r = \tau + \sigma E(1 - \phi )(1 - H)$$\end{document}, is therefore a product of failures, demonstrating that the highest risk occurs when these three factors converge. Conceptual Framework: Dynamic Risk Equation Model for HIV Transmission through Conflict-Related Sexual Violence. The diagram illustrates the pathways through which conflict intensity (c) influences HIV transmission risk. Conflict intensity affects multiple parameters: rape proportion (r), number of assailants (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}), trauma-related exposure (E), PEP coverage (ϕ), and health system functionality (H). These factors feed into the cumulative exposure probability (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P = 1 - (1 - P_r)^{n_r}$$\end{document}), where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document} is the probability that a single assailant is HIV-positive, and the dynamic transmission probability (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r = \tau + \sigma E(1 - \phi )(1 - H)$$\end{document}). The final estimated conflict-related HIV infections are calculated as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I = X r P \beta _r$$\end{document}, where X represents the susceptible female population aged 5–49 years. Key assumptions The conceptual framework is operationalized through five key assumptions that govern the mathematical relationships between conflict intensity and HIV transmission risk. These assumptions reflect both the theoretical underpinnings of the model and the practical constraints of estimating HIV incidence in complex humanitarian emergencies. Simultaneous Exponential Decay: Conflict intensity (c) influences all downstream parameters simultaneously through exponential decay functions. This assumption captures the escalating and nonlinear nature of violence, where small increases in conflict intensity can produce disproportionately large disruptions to social, behavioral, and health systems. The exponential functional form is chosen for its parsimony and its ability to represent accelerating degradation as conflict intensifies.Multiplicative Cumulative Risk: The cumulative risk from multiple perpetrators is treated as multiplicative rather than additive. This reflects the compounding probability of exposure to an HIV-positive assailant when a survivor is raped by multiple perpetrators. Specifically, the cumulative risk is expressed as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P = 1 - (1 - P_r)^{n_r}$$\end{document}, where the multiplicative structure captures the exponential increase in risk associated with gang rape, a common feature of conflict-related sexual violence.Synergistic System Failures: Health system degradation and declines in post-exposure prophylaxis (PEP) availability act synergistically to increase transmission risk beyond their individual effects. This synergy is operationalized in the transmission probability function \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r = \tau + \sigma E(1 - \phi )(1 - H)$$\end{document}, where the multiplicative interaction of (1 - ϕ ) and (1 - H) ensures that the combined effect of PEP decline and health system collapse exceeds the sum of their independent contributions.Parameter Independence for Tractability: While we acknowledge that real-world parameters are interdependent–for instance, trauma severity may influence health-seeking behavior, and health system functionality affects PEP access–we treat them as independent for computational tractability. This assumption allows for the calibration of the model using available empirical data and simplifies the mathematical formulation while preserving the core theoretical relationships.Linear Aggregation of Risk Factors: The model aggregates risk factors linearly in the main equation (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I = X \cdot r \cdot P \cdot \beta _r$$\end{document}). This is a conservative approach that may underestimate synergistic effects across the primary components (demographic exposure, assailant risk, and systemic failure). However, it provides a clear, interpretable baseline for estimation and allows for transparency in attributing changes in outcomes to specific input parameters.These assumptions collectively enable the translation of the conceptual framework into a quantifiable model while acknowledging the inherent complexities and data limitations of estimating HIV transmission in conflict settings. Simultaneous Exponential Decay: Conflict intensity (c) influences all downstream parameters simultaneously through exponential decay functions. This assumption captures the escalating and nonlinear nature of violence, where small increases in conflict intensity can produce disproportionately large disruptions to social, behavioral, and health systems. The exponential functional form is chosen for its parsimony and its ability to represent accelerating degradation as conflict intensifies. Simultaneous Exponential Decay: Conflict intensity (c) influences all downstream parameters simultaneously through exponential decay functions. This assumption captures the escalating and nonlinear nature of violence, where small increases in conflict intensity can produce disproportionately large disruptions to social, behavioral, and health systems. The exponential functional form is chosen for its parsimony and its ability to represent accelerating degradation as conflict intensifies. Multiplicative Cumulative Risk: The cumulative risk from multiple perpetrators is treated as multiplicative rather than additive. This reflects the compounding probability of exposure to an HIV-positive assailant when a survivor is raped by multiple perpetrators. Specifically, the cumulative risk is expressed as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P = 1 - (1 - P_r)^{n_r}$$\end{document}, where the multiplicative structure captures the exponential increase in risk associated with gang rape, a common feature of conflict-related sexual violence. Multiplicative Cumulative Risk: The cumulative risk from multiple perpetrators is treated as multiplicative rather than additive. This reflects the compounding probability of exposure to an HIV-positive assailant when a survivor is raped by multiple perpetrators. Specifically, the cumulative risk is expressed as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P = 1 - (1 - P_r)^{n_r}$$\end{document}, where the multiplicative structure captures the exponential increase in risk associated with gang rape, a common feature of conflict-related sexual violence. Synergistic System Failures: Health system degradation and declines in post-exposure prophylaxis (PEP) availability act synergistically to increase transmission risk beyond their individual effects. This synergy is operationalized in the transmission probability function \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r = \tau + \sigma E(1 - \phi )(1 - H)$$\end{document}, where the multiplicative interaction of (1 - ϕ ) and (1 - H) ensures that the combined effect of PEP decline and health system collapse exceeds the sum of their independent contributions. Synergistic System Failures: Health system degradation and declines in post-exposure prophylaxis (PEP) availability act synergistically to increase transmission risk beyond their individual effects. This synergy is operationalized in the transmission probability function \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r = \tau + \sigma E(1 - \phi )(1 - H)$$\end{document}, where the multiplicative interaction of (1 - ϕ ) and (1 - H) ensures that the combined effect of PEP decline and health system collapse exceeds the sum of their independent contributions. Parameter Independence for Tractability: While we acknowledge that real-world parameters are interdependent–for instance, trauma severity may influence health-seeking behavior, and health system functionality affects PEP access–we treat them as independent for computational tractability. This assumption allows for the calibration of the model using available empirical data and simplifies the mathematical formulation while preserving the core theoretical relationships. Parameter Independence for Tractability: While we acknowledge that real-world parameters are interdependent–for instance, trauma severity may influence health-seeking behavior, and health system functionality affects PEP access–we treat them as independent for computational tractability. This assumption allows for the calibration of the model using available empirical data and simplifies the mathematical formulation while preserving the core theoretical relationships. Linear Aggregation of Risk Factors: The model aggregates risk factors linearly in the main equation (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I = X \cdot r \cdot P \cdot \beta _r$$\end{document}). This is a conservative approach that may underestimate synergistic effects across the primary components (demographic exposure, assailant risk, and systemic failure). However, it provides a clear, interpretable baseline for estimation and allows for transparency in attributing changes in outcomes to specific input parameters. Linear Aggregation of Risk Factors: The model aggregates risk factors linearly in the main equation (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I = X \cdot r \cdot P \cdot \beta _r$$\end{document}). This is a conservative approach that may underestimate synergistic effects across the primary components (demographic exposure, assailant risk, and systemic failure). However, it provides a clear, interpretable baseline for estimation and allows for transparency in attributing changes in outcomes to specific input parameters. Model formulation This study builds upon the seminal works of Anema et al23. and Supervie et al24., which developed static risk equation models to estimate HIV infections arising from mass rape in sub-Saharan African countries. In their framework, the number of new infections is a function of the proportion of women raped, HIV prevalence among assailants, per-act transmission probability, and the at-risk female population aged 15–49. We extend this earlier work by developing a dynamic model where conflict-related drivers of HIV infection are explicitly linked to conflict intensity (c). The model estimating the number of new HIV infections from conflict-related rape is:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} I = X \cdot r \cdot P \cdot \beta _r \end{aligned}$$\end{document}where:I = Number of new HIV infections in females aged ≥ 5 years due to mass rapeX = Susceptible women aged 5–49 years who could be infected owing to mass raper = Proportion of the female population aged 5–49 years who could be rapedP = Cumulative risk of exposure to HIV-positive perpetrators\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document} = Number of assailants who could rape a woman per time unitc = Conflict intensity\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document} = Dynamic transmission probability of HIV transmission per act of rape per yearIn a conflict setting where the risk of HIV transmission from sexual violence varies with changing conflict intensity, the use of dynamic transmission probability is essential for accurately modeling incidence27,28. Thus, unlike static models, the probability of HIV transmission \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document} due to conflict-related rape is determined by the interaction of combined risk factors, with components interacting multiplicatively, reflecting the compounded risk in real-world settings. In this context, “dynamic” refers to the model’s ability to capture how HIV transmission risk evolves with conflict intensity (c), which acts as a proxy for the temporal progression of the conflict. While not a time-dependent differential equation model, this formulation allows key drivers–such as rape prevalence, PEP access, and health system functionality–to change as the conflict intensifies or de-escalates, providing a dynamic representation of risk. Formally, the average transmission probability per act of rape per year is expressed as:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \beta _r = \left[ \tau + \sigma E \left( 1 - \phi \right) \left( 1 - H\right) \right] \end{aligned}$$\end{document}Here, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E = E_{\min }e^{\gamma c}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$E_{\min }=E(0)=0$$\end{document} represents trauma-related exposure factors, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\phi = \phi _{\max }e^{-\rho c}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\phi _{\max }=\phi (0.8)$$\end{document} accounts for PEP coverage decline, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$H = H_{\max }e^{-\alpha c}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$H_{\max }=H(0.8)$$\end{document} reflects health system degradation, while τ, σ, γ, ρ, and α are scaling constants determined by calibration. For conservative estimation we assume PEP and healthcare services coverage approach 0.2 during the peak conflict period 2021 (at c ≈ 0.8), though PEP availability approached zero and healthcare services were nearly non-functional14–16,20. Thus, we have:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \beta _r = \tau + \sigma E_{\min }e^{\gamma c} \left( 1 - \phi _{\max }e^{-\rho c}\right) \left( 1 - H_{\max }e^{-\alpha c}\right) \end{aligned}$$\end{document}Equation (1) defines the number of new HIV infections as a product of the susceptible population, the rape proportion, cumulative risk, and the dynamic transmission probability. Equation (2) defines the dynamic transmission probability \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document} as a function of the biological per-act risk (τ) and the conflict-driven risk factors. The dynamic transmission probability \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document} is not only a function of the biological per-act risk but also a composite measure that integrates the impact of conflict-driven disruptions. Specifically, it incorporates the degradation of the health system (H) and the decline in PEP coverage (ϕ), both of which are modeled as exponential functions of conflict intensity (c). This formulation allows the model to capture the synergistic effect of increased sexual violence and the simultaneous collapse of essential health services. I = Number of new HIV infections in females aged ≥ 5 years due to mass rape I = Number of new HIV infections in females aged ≥ 5 years due to mass rape X = Susceptible women aged 5–49 years who could be infected owing to mass rape X = Susceptible women aged 5–49 years who could be infected owing to mass rape r = Proportion of the female population aged 5–49 years who could be raped r = Proportion of the female population aged 5–49 years who could be raped P = Cumulative risk of exposure to HIV-positive perpetrators P = Cumulative risk of exposure to HIV-positive perpetrators \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document} = Number of assailants who could rape a woman per time unit \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document} = Number of assailants who could rape a woman per time unit c = Conflict intensity c = Conflict intensity \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document} = Dynamic transmission probability of HIV transmission per act of rape per year \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document} = Dynamic transmission probability of HIV transmission per act of rape per year In sexual violence involving multiple perpetrators, each assailant contributes differently to the survivor’s HIV exposure29,30, and the multiplicative risk from each assailant substantially increases the probability of HIV exposure far beyond that estimated for consensual intercourse. To capture the impact of multi-perpetrator rapes observed during the Tigray conflict, the probability of being raped by an HIV-positive assailant was modeled as a cumulative risk exposure \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P = 1 - (1 - P_r)^{n_r}$$\end{document}, where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document} is the average number of perpetrators raping a woman per year and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document} represents the weighted average HIV prevalence (males aged 15–49) of assailants’ home countries. Data sources and model analysis Estimating HIV transmission risk from conflict-related sexual violence presents profound methodological challenges, particularly in active conflict settings such as Tigray. The collection of primary data during ongoing hostilities is constrained by limited humanitarian access, security risks to researchers and respondents, the displacement of populations, the destruction of health information systems, and the severe trauma experienced by survivors, which complicates ethical data collection2,15,20. Furthermore, conflict-related sexual violence is systematically underreported due to stigma, fear of reprisal, and the collapse of formal reporting mechanisms30,33. Consequently, traditional epidemiological surveillance is largely infeasible, necessitating a reliance on secondary data sources and methodological approaches that can rigorously address these limitations. To address the paucity of data for assessing the consequences of mass rape on HIV incidence and prevalence in Tigray, we employed a multi-pronged strategy. Parameters were estimated based on: (1) available study results and humanitarian reports documenting the pattern of the conflict and conflict-related sexual violence; (2) regional epidemiological data from proximate settings; (3) clinical and biological parameters from established literature; and (4) necessary extrapolation from comparable conflict settings where more robust data exist31–33. Given the inherent uncertainties in these data sources, we implemented a comprehensive approach to uncertainty management: Triangulation: All parameters were triangulated from multiple credible sources to enhance validity and reduce source-specific bias. Primary data sources included: UNAIDS regional HIV estimates36, World Health Organization clinical guidelines38, Ethiopian Public Health Institute epidemiological reports37, peer-reviewed literature on conflict-related sexual violence1,23,24, and humanitarian situation reports documenting the Tigray conflict4,9,16. For conflict-specific parameters such as the number of perpetrators and health system functionality, we drew upon qualitative and quantitative studies conducted during or immediately after the conflict2,14,30.Plausible Ranges: Rather than relying on single point estimates that would falsely imply precision, we specified plausible ranges for each key parameter. This approach acknowledges the uncertainty inherent in conflict-zone data and allows for the systematic exploration of how variation in inputs affects model outputs. Ranges were derived from the minimum and maximum values reported across credible sources, extended where necessary to capture context-specific variation.Monte Carlo Simulation: To propagate parameter uncertainty through the model and characterize the uncertainty in final HIV incidence estimates, we employed Monte Carlo simulation. For each parameter, probability distributions were specified based on the plausible ranges and the nature of the underlying data (uniform distributions were used where only bounds were available; triangular or beta distributions were used where central tendencies could be identified). We conducted 10,000 iterations, randomly sampling from these distributions to generate a distribution of model outcomes.Sensitivity Analysis: One-way and multi-way sensitivity analyses were performed to identify parameters that disproportionately influence model outputs. This allowed us to prioritize data collection efforts for future research and to understand which uncertainties most critically affect policy-relevant conclusions.Model Calibration: The dynamic transmission probability function \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document} was calibrated using empirical bounds derived from established literature. A calibration procedure was implemented to identify scaling constants that minimized the discrepancy between model predictions and observed epidemiological parameters across the range of conflict intensities. Triangulation: All parameters were triangulated from multiple credible sources to enhance validity and reduce source-specific bias. Primary data sources included: UNAIDS regional HIV estimates36, World Health Organization clinical guidelines38, Ethiopian Public Health Institute epidemiological reports37, peer-reviewed literature on conflict-related sexual violence1,23,24, and humanitarian situation reports documenting the Tigray conflict4,9,16. For conflict-specific parameters such as the number of perpetrators and health system functionality, we drew upon qualitative and quantitative studies conducted during or immediately after the conflict2,14,30. Triangulation: All parameters were triangulated from multiple credible sources to enhance validity and reduce source-specific bias. Primary data sources included: UNAIDS regional HIV estimates36, World Health Organization clinical guidelines38, Ethiopian Public Health Institute epidemiological reports37, peer-reviewed literature on conflict-related sexual violence1,23,24, and humanitarian situation reports documenting the Tigray conflict4,9,16. For conflict-specific parameters such as the number of perpetrators and health system functionality, we drew upon qualitative and quantitative studies conducted during or immediately after the conflict2,14,30. Plausible Ranges: Rather than relying on single point estimates that would falsely imply precision, we specified plausible ranges for each key parameter. This approach acknowledges the uncertainty inherent in conflict-zone data and allows for the systematic exploration of how variation in inputs affects model outputs. Ranges were derived from the minimum and maximum values reported across credible sources, extended where necessary to capture context-specific variation. Plausible Ranges: Rather than relying on single point estimates that would falsely imply precision, we specified plausible ranges for each key parameter. This approach acknowledges the uncertainty inherent in conflict-zone data and allows for the systematic exploration of how variation in inputs affects model outputs. Ranges were derived from the minimum and maximum values reported across credible sources, extended where necessary to capture context-specific variation. Monte Carlo Simulation: To propagate parameter uncertainty through the model and characterize the uncertainty in final HIV incidence estimates, we employed Monte Carlo simulation. For each parameter, probability distributions were specified based on the plausible ranges and the nature of the underlying data (uniform distributions were used where only bounds were available; triangular or beta distributions were used where central tendencies could be identified). We conducted 10,000 iterations, randomly sampling from these distributions to generate a distribution of model outcomes. Monte Carlo Simulation: To propagate parameter uncertainty through the model and characterize the uncertainty in final HIV incidence estimates, we employed Monte Carlo simulation. For each parameter, probability distributions were specified based on the plausible ranges and the nature of the underlying data (uniform distributions were used where only bounds were available; triangular or beta distributions were used where central tendencies could be identified). We conducted 10,000 iterations, randomly sampling from these distributions to generate a distribution of model outcomes. Sensitivity Analysis: One-way and multi-way sensitivity analyses were performed to identify parameters that disproportionately influence model outputs. This allowed us to prioritize data collection efforts for future research and to understand which uncertainties most critically affect policy-relevant conclusions. Sensitivity Analysis: One-way and multi-way sensitivity analyses were performed to identify parameters that disproportionately influence model outputs. This allowed us to prioritize data collection efforts for future research and to understand which uncertainties most critically affect policy-relevant conclusions. Model Calibration: The dynamic transmission probability function \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document} was calibrated using empirical bounds derived from established literature. A calibration procedure was implemented to identify scaling constants that minimized the discrepancy between model predictions and observed epidemiological parameters across the range of conflict intensities. Model Calibration: The dynamic transmission probability function \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document} was calibrated using empirical bounds derived from established literature. A calibration procedure was implemented to identify scaling constants that minimized the discrepancy between model predictions and observed epidemiological parameters across the range of conflict intensities. Parameter estimation and calibration Parameter estimation drew upon literature and humanitarian reports across several domains: rape proportion (r), HIV prevalence among assailants (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document}), number of perpetrators (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}), per-act transmission probability (τ), trauma-related risk of exposure (E), health system functionality (H), and PEP coverage (ϕ). For population denominators, we used data from the Central Statistics Agency of Ethiopia, applying 2021 population projections conservatively, as the conflict spanned 2020–202235. The susceptible female population aged 5–49 years was estimated based on these population data and standard demographic distributions. The model was further informed by regional UNAIDS data on HIV prevalence in the Horn of Africa36 and reports from the Ethiopian Public Health Institute37. Where direct data were unavailable, we drew upon the methodological approaches and parameter ranges established in peer-reviewed studies of conflict-related HIV transmission in similar settings, adapting these to the specific circumstances of the Tigray conflict23,24,34. The parameter values and their sources are summarized in Table 1.Table 1Ranges of variables and sampled parameters.ParameterDescriptionValueSourceNTotal population6.25 million35XTotal women (5–49 years)2,187,500EstimatedrRaped women (5–49 years)1%–15%23,24pWeighted prevalence of general population0.635%1,36\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document}Assailants prevalencep–8p23,24\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}Number of assailants per unit time1–42,4,9,30\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document}Average probability of HIV transmission0.0028–0.03223,24cConflict intensity0–0.8EstimatedHHealth system coverage0.2–0.92,14ϕPEP coverage0.2–0.838ERape-related exposure risk0.1–1.038 Table 1Ranges of variables and sampled parameters.ParameterDescriptionValueSourceNTotal population6.25 million35XTotal women (5–49 years)2,187,500EstimatedrRaped women (5–49 years)1%–15%23,24pWeighted prevalence of general population0.635%1,36\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document}Assailants prevalencep–8p23,24\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}Number of assailants per unit time1–42,4,9,30\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document}Average probability of HIV transmission0.0028–0.03223,24cConflict intensity0–0.8EstimatedHHealth system coverage0.2–0.92,14ϕPEP coverage0.2–0.838ERape-related exposure risk0.1–1.038 Ranges of variables and sampled parameters. To accurately model the HIV transmission probability \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document} as a function of conflict intensity (c), we calibrated key scaling constants τ, σ, γ, ρ, and α based on epidemiological, healthcare access, and conflict intensity considerations. The Monte Carlo simulation approach was employed to identify optimal values of the parameters that minimize the squared difference between the model output \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document} and the empirically observed bounds23,24. Specifically, 10,000 simulations were run, and for each parameter combination, the squared error was computed; the parameter values with the minimum squared error (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$8.85 \times 10^{-17}$$\end{document}) were selected as the calibrated parameter set: τ = 0.00271, σ = 0.4489, γ = 2.88, ρ = 1.733, and α = 1.88. With these values, equation (3) reduces to:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \beta _r = 0.00271 + 0.4489 E_{\min } e^{2.88c} \left( 1 - \phi _{\max } e^{-1.733c}\right) \left( 1 - H_{\max } e^{-1.88c}\right) \end{aligned}$$\end{document}Evaluation of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r = 0.0028$$\end{document} at c=0 and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r = 0.032142$$\end{document} at c=0.8 shows close agreement with the lower and upper bounds of the empirically observed transmission probability (0.0028–0.032) in established literature23,24. With these estimated values, the model demonstrates an excellent fit and is consistent with Monte Carlo calibration standards31,34, reflecting biological plausibility and comparability to previous estimates in conflict settings. This calibration approach provides a robust framework to incorporate conflict intensity–dependent effects on HIV transmission risk while realistically constraining maximal health intervention coverage under no-conflict baseline conditions. Results The mathematical framework of our risk equation model is detailed in the Methods section. The results below present the model’s output across various parameter scenarios, as detailed in Table 1. Estimated new HIV infections due to conflict-related rape The annual number of rape-related new HIV infections was computed as a product of uninfected women (X), the proportion of the population experiencing sexual violence (r), the cumulative risk of exposure (P), and the dynamic transmission probability \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document}. This methodology allows robust estimation of the population-level impact of conflict-related SGBV on HIV incidence while incorporating Tigray-specific contextual factors.5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} {\left\{ \begin{array}{ll} I = X \cdot r \cdot P \cdot \beta _r \\ P = 1 - (1 - P_r)^{n_r} \\ \beta _r = 0.00271 + 0.4489 E_{\min } e^{2.88c} \left( 1 - \phi _{\max } e^{-1.733c}\right) \left( 1 - H_{\max } e^{-1.88c}\right) \end{array}\right. } \end{aligned}$$\end{document}The estimation in Table 2 specifically investigates how the total population HIV incidence would be affected if 1% to 15% of the female population (aged 5–49 years) per year were raped, the perpetrator’s HIV prevalence was 0.635% to 5.08%, 1 to 4 perpetrators per woman were involved, the trauma-related risk factor was elevated from 0.1 to 1.0, PEP declined from 0.8 to 0.2, and the health system degraded from 0.9 to 0.2.Table 2New HIV infections due to different rape rates in women aged 5–49 years in Tigray, Ethiopia.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document}Xcr\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r(\%)$$\end{document}P\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {new}}$$\end{document}0.0006352,156,8750.100.011.000.6350.006350.003210.0006352,156,8750.200.031.431.2700.018110.004050.0006352,156,8750.300.051.861.9050.035140.0054210.0006352,156,8750.400.072.292.5400.057220.0076660.0006352,156,8750.500.092.723.1750.084020.01091780.0006352,156,8750.600.113.153.8100.115170.01564260.0006352,156,8750.700.133.584.4450.150220.02229360.0006352,156,8750.800.154.005.0800.188660.03151,925 Table 2New HIV infections due to different rape rates in women aged 5–49 years in Tigray, Ethiopia.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document}Xcr\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r(\%)$$\end{document}P\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _r$$\end{document}\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {new}}$$\end{document}0.0006352,156,8750.100.011.000.6350.006350.003210.0006352,156,8750.200.031.431.2700.018110.004050.0006352,156,8750.300.051.861.9050.035140.0054210.0006352,156,8750.400.072.292.5400.057220.0076660.0006352,156,8750.500.092.723.1750.084020.01091780.0006352,156,8750.600.113.153.8100.115170.01564260.0006352,156,8750.700.133.584.4450.150220.02229360.0006352,156,8750.800.154.005.0800.188660.03151,925 New HIV infections due to different rape rates in women aged 5–49 years in Tigray, Ethiopia. Table 2 reveals that the data are evidently right-skewed, with 1 to 21 new HIV infections in the low estimate; 66 to 426 new HIV infections in the intermediary estimate; and 936 to 1,925 in the high estimate. This shows that incidence due to rape varies substantially from the low estimate and gradually increases to the highest estimate under extreme scenarios.Table 3Statistical summary of estimated new infections.DescriptionQ1MedianQ3IQRMeanValues13122681668 (13–681)444.75 Table 3Statistical summary of estimated new infections.DescriptionQ1MedianQ3IQRMeanValues13122681668 (13–681)444.75 Statistical summary of estimated new infections. The first quartile (Q1 = 13) represents the 25th percentile of the estimated new infections, indicating that in one quarter of the modeled scenarios–typically those with lower rape proportions, fewer perpetrators, and less intense conflict–the number of new infections is 13 or fewer. The third quartile (Q3 = 681) represents the 75th percentile, showing that in three quarters of scenarios, the estimate is below 681. The wide interquartile range (IQR = 668) highlights the substantial uncertainty inherent in the estimates, driven by the range of plausible values for key parameters such as rape proportion, number of perpetrators, and conflict intensity (Table 3). The mean in this dataset (444.75) is a value pulled upward by the extremely high estimate of 1,925, highlighting that it may not reflect the typical or expected number of new infections39. The right-skewed distribution of estimates (Table 2) indicates that the potential for a large number of new infections is driven by the co-occurrence of high-risk conditions–specifically a high proportion of the population experiencing rape, involvement of multiple perpetrators, and intense conflict that cripples the health system–rather than being a typical outcome. This is in line with Hoaglin et al.40, who reported that the median estimate is often more appropriate than the mean, especially in public health data characterized by large variability. Our intermediary estimate is, therefore, most likely to be representative of the actual situation of the conflict in Tigray, as it was also used by Anema et al.23 as well as Supervie et al.24. Consequently, our estimation shows that mass rape could cause a median of 122 (IQR: 13–681) new HIV infections. This highlights that the median of new HIV infections due to mass rape could be 2 per 100,000 individuals of the Tigray population per year. Furthermore, the result shows that widespread rape could cause about 31 new HIV infections per 100,000 individuals of the Tigray population in the most extreme scenarios where: 15% of the female population aged 5–49 years were raped, HIV prevalence among perpetrators was 8 times the weighted prevalence of the assailants’ country population prevalence, the average number of assailants per woman per year was 4, conflict intensity was 0.8, trauma-related exposure was maximum 1.0, PEP decline was maximum 0.8, health system degradation was maximum 0.9, and the HIV transmission rate was highest at 4 times the average high rate. This high-end scenario corresponds to the most severe documented conditions, which in Tigray (2021) was characterized by widespread violence and near-total health system collapse. Recent empirical studies from Tigray have documented wartime HIV prevalence of 1.2% among tested individuals42 and poor HIV outcomes increasing from 28.6% pre-war to 54.5% during active conflict43, providing contextual validation for our modeled estimates. Increase in incidence and prevalence The percentage increase in HIV incidence due to conflict-related rape was calculated following the framework proposed by Supervie et al.24, adapted to account for all new HIV infections in the population through all transmission paths. To contextualize the burden attributable to sexual violence within the broader epidemic, the incidence increase is computed as the proportion of new HIV infections attributable to conflict-related rape relative to the population-level total number of new HIV infections from all transmission pathways, aligning with WHO recommendations for comprehensive HIV epidemic modeling28,38. This formulation provides a clear quantification of the impact of sexual violence on population-level HIV dynamics while maintaining alignment with prior modeling studies23,24.6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} {\left\{ \begin{array}{ll} \text {Incidence Increase (\%)} = \dfrac{I}{I_{\text {baseline}}} \times 100 \\ \text {Absolute Prevalence Increase (\%)} = \dfrac{I}{\text {Total Population}} \times 100 \end{array}\right. } \end{aligned}$$\end{document}where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {baseline}}$$\end{document} represents new infections of the total population through all transmission pathways.Table 4HIV and population estimates for Tigray in 2021.DescriptionTotal PopulationI\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {baseline}}$$\end{document}PLHIVIncidence Increase (%)Absolute Prevalence Increase (%)Values6,250,00012288649,14113.80.002Source35Estimated3737EstimatedEstimated Table 4HIV and population estimates for Tigray in 2021.DescriptionTotal PopulationI\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$I_{\text {baseline}}$$\end{document}PLHIVIncidence Increase (%)Absolute Prevalence Increase (%)Values6,250,00012288649,14113.80.002Source35Estimated3737EstimatedEstimated HIV and population estimates for Tigray in 2021. Fig. 2Estimated impact of conflict-related rape on HIV incidence and prevalence. (a) Increase in HIV incidence showing median 13.8% (IQR: 1.9%–62.5%) with 95% bootstrap confidence interval ranging from 1.24% to 105.64%. (b) Increase in HIV prevalence showing median 0.002% (IQR: 0.00027%–0.0089%) with 95% bootstrap confidence interval ranging from 0.00018% to 0.015%. The right-skewed distributions reflect substantial uncertainty, with occasional extreme values suggesting potential for explosive increases under certain conditions. The small absolute prevalence increase reflects the large population denominator and existing pool of people living with HIV. Estimated impact of conflict-related rape on HIV incidence and prevalence. (a) Increase in HIV incidence showing median 13.8% (IQR: 1.9%–62.5%) with 95% bootstrap confidence interval ranging from 1.24% to 105.64%. (b) Increase in HIV prevalence showing median 0.002% (IQR: 0.00027%–0.0089%) with 95% bootstrap confidence interval ranging from 0.00018% to 0.015%. The right-skewed distributions reflect substantial uncertainty, with occasional extreme values suggesting potential for explosive increases under certain conditions. The small absolute prevalence increase reflects the large population denominator and existing pool of people living with HIV. The results presented in Table 4 and Fig. 2 indicate that the estimated percentage increase in HIV incidence due to conflict-related rape was a median of 13.8% (IQR: 1.9%–62.5%; 95% bootstrap CI: 1.24%–105.64%), indicating a notable relative increase in new infections due to rape compared to the baseline incidence of 886 cases/year. The percentage increase in absolute prevalence was a median of 0.002% (IQR: 0.00027%–0.00886%; 95% bootstrap CI: 0.00018%–0.015%), demonstrating that conflict-related mass rape has a notable impact on HIV incidence and prevalence. Even though the absolute effect on population prevalence remains small, the incidence effect is epidemiologically important, given that even modest increases can destabilize HIV control efforts28. Our estimated incidence increase of 13.8% is consistent with a systematic review confirming that armed conflicts consistently disrupt HIV care and increase transmission risk across Sub-Saharan Africa45. Furthermore, conflict-related sexual violence has been recognized as a driver of infectious disease outbreaks, including HIV, in humanitarian settings44. Global sensitivity analysis and numerical simulation To enhance the robustness and reliability of our findings, we conducted a comprehensive global sensitivity analysis and numerical simulations against different parameter scenarios. Global sensitivity analysis Global sensitivity analysis was employed to evaluate how variations in key model parameters influence the overall model outcome, providing complementary insights for disease spread, transmission, control, and intervention planning34. To fully explore the input space and take into account the simultaneous variation of input parameters, we used the concept of elasticity, which is given by:7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{aligned} \varepsilon _{Q}^{p} = \frac{\partial Q}{\partial p} \times \frac{p}{Q} \end{aligned}$$\end{document}where the quantity Q is the key model outcome I and p is a model parameter/variable. This sensitivity index quantifies the relative impact of parameter p on Q and indicates whether a small relative increase in parameter p will increase or decrease Q. The sign of the index indicates the direction of effect, where positive means increasing the parameter increases the output, while negative means the opposite. The sensitivity of new HIV infections due to conflict-related sexual violence among women (I) was examined against the parameters rape proportion (r), HIV prevalence among perpetrators (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document}), number of perpetrators per victim (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}), conflict intensity (c), trauma-related risk of exposure (E), post-exposure prophylaxis coverage (ϕ), and health system degradation (H).Table 5Sensitivity of new infection due to mass rape.ParameterValuesSensitivity Index \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varepsilon _{I}^{p}$$\end{document}r0.051.000\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document}0.0190.973\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}1.860.982c0.301.640E0.0240.080ϕ0.5250.930H0.491.070 Table 5Sensitivity of new infection due to mass rape.ParameterValuesSensitivity Index \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varepsilon _{I}^{p}$$\end{document}r0.051.000\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document}0.0190.973\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}1.860.982c0.301.640E0.0240.080ϕ0.5250.930H0.491.070 Sensitivity of new infection due to mass rape. The sensitivity indices reveal that all parameters show a positive relationship with I, indicating that increases in any of these parameters lead to higher numbers of new infections. Among the parameters, conflict intensity (c) and health system degradation (H) exhibited the highest sensitivity, because they affect multiple downstream pathways simultaneously. Conflict intensity influences the rape proportion, the number of perpetrators, and the transmission probability, while health system degradation directly impacts access to all services, including PEP, highlighting the strong influence of war conditions on HIV transmission through increased sexual violence, disrupted social structures, and decline of healthcare systems. Rape proportion (r), number of perpetrators (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}), and perpetrators’ HIV prevalence (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document}) were also highly sensitive, reflecting the direct impact of sexual violence prevalence and the likelihood that the perpetrator affects HIV infection burden. Post-exposure prophylaxis coverage (ϕ) was also sensitive, indicating the importance of timely post-exposure interventions in preventing HIV infections. Trauma-related exposure (E) had lower but still positive sensitivity, indicating secondary contributions to overall risk (Table 5). Numerical simulations To complement the global sensitivity analysis and further understand how HIV transmission behaves under different conflict conditions, we performed numerical simulations. These simulations show how new HIV infections respond when key factors such as the proportion of women raped, the HIV prevalence among perpetrators, the number of assailants per survivor, and the overall intensity of conflict are varied over realistic ranges. Figure 3 presents the impact of key parameters on new HIV infections by demonstrating how the trajectories of new HIV infections behave as each of these primary conflict drivers varies: rape proportion, assailant HIV prevalence, number of assailants that could rape a woman, and conflict intensity.Fig. 3Impact of conflict drivers on new HIV infections. (a) Rape proportion (r). (b) Assailant HIV prevalence (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document}). (c) Number of assailants (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}). (d) Conflict intensity (c). Impact of conflict drivers on new HIV infections. (a) Rape proportion (r). (b) Assailant HIV prevalence (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P_r$$\end{document}). (c) Number of assailants (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$n_r$$\end{document}). (d) Conflict intensity (c). Figure 3(a) demonstrates that new infections increase in direct proportion to the rape proportion; as the proportion of women subjected to rape increases across its range, new infections rise from low levels to substantially higher magnitudes, indicating that every incremental increase in exposure yields a corresponding rise in HIV infections with no evidence of saturation1. The impact of assailant HIV prevalence is presented in Figure 3(b), which reveals a more concerning pattern: as the prevalence among perpetrators rises, new infections grow at an accelerating rate, where higher perpetrator prevalence results in a proportional increase in infections with cumulative risk amplifying the effect. Even marginal increases in perpetrator HIV prevalence can precipitate substantially larger surges in infections, reflecting the compounded risk confronting survivors exposed to multiple infected assailants23,24. Figure 3(c) further shows that the greatest increase in risk occurs when transitioning from a single assailant to multiple assailants, with additional perpetrators beyond the first contributing progressively smaller incremental increases in risk. Each supplementary assailant elevates cumulative risk, thereby generating a non-linear increase in infection risk, a pattern consistent with the high frequency of gang rape documented in Tigray where the majority of cases involved multiple perpetrators2,41. Finally, Figure 3(d) demonstrates that conflict intensity exerts a critical threshold effect: at lower levels of conflict, infections remain relatively contained, yet beyond a definable point, infections rise sharply and rapidly. As conflict intensity escalates across its range, new infections increase sharply due to the synergistic effects of heightened rape frequency, progressive health system collapse, and diminished PEP access. This behavior reflects the compounded and simultaneous breakdown of social order, health infrastructure, and protective services as armed conflict intensifies14. Figure 4 examines how intermediate mechanisms translate conflict conditions into actual HIV infections. These mechanisms include cumulative exposure risk, trauma-related susceptibility, health system functionality, and PEP coverage.Fig. 4Relationship between mediating mechanisms and new HIV infections. (a) Cumulative exposure probability (P). (b) Trauma-related exposure (E). (c) Health system functionality (H). (d) PEP coverage (ϕ). Relationship between mediating mechanisms and new HIV infections. (a) Cumulative exposure probability (P). (b) Trauma-related exposure (E). (c) Health system functionality (H). (d) PEP coverage (ϕ). Figure 4(a) demonstrates that cumulative exposure, which is driven jointly by the HIV prevalence among assailants and the number of perpetrators, amplifies infection risk substantially beyond what would be anticipated from simple additive effects. Each additional exposure multiplicatively increases the overall probability of infection, underscoring the compounded nature of risk in multi-perpetrator sexual violence23. Figure 4(b) illustrates the role of trauma-related susceptibility: elevated trauma exposure increases vulnerability to HIV transmission through biological mechanisms including mucosal damage and compromised barrier integrity. While trauma increases susceptibility, the protective benefit of reducing trauma exhibits diminishing returns at extreme levels due to physiological constraints; nevertheless, baseline transmission risk persists even when trauma is minimized27,29. Figure 4(c) reveals a critical pattern for health system functionality: as PEP coverage declines across its range, new infections increase substantially, highlighting the protective effect of timely post-exposure prophylaxis. Small reductions in health system capacity produce moderate increases in infections; however, near-total collapse precipitates catastrophic surges in infections, reflecting the simultaneous loss of testing, treatment, prevention, and support services. This reality was witnessed in Tigray where the majority of hospitals became non-functional15,45. Figure 4(d) demonstrates that reduced health system functionality leads to elevated infection rates due to constrained access to testing, treatment, and supportive care. Infections remain relatively low when coverage is maintained above a critical threshold, but declining below this level triggers rapid escalation, underscoring that PEP programs must sustain sufficient coverage to remain effective, particularly in conflict settings where access is already compromised16,38. Taken together, these simulations reveal a consistent and coherent picture: conflict intensity emerges as the most powerful overarching driver of new HIV infections, as it simultaneously exacerbates structural vulnerabilities (health system collapse and reduced PEP access), behavioral risks (increased rape proportion and number of assailants), and biological susceptibility (trauma-related exposure). Health system degradation operates as a proximate mechanism through which conflict intensity amplifies transmission risk, while rape proportion, number of assailants, assailant HIV prevalence, cumulative exposure, trauma, and PEP coverage all contribute to shaping the magnitude and velocity of infection spread. Critically, several of these factors exhibit threshold behavior, meaning that early intervention before conflict escalates, before health systems collapse, and before PEP coverage drops below critical levels can yield disproportionately large reductions in HIV burden. Integrated strategies that simultaneously address rape prevention, health system protection, and biomedical intervention coverage are therefore essential43,45. Overall, Figs. 3 and 4 provide a coherent picture of how structural, behavioral, biological, and health-system factors interact synergistically to drive HIV transmission during armed conflict, with conflict intensity serving as the fundamental force multiplier across all pathways. Discussion The discussion below interprets these model-based estimates. Readers should keep in mind the limitations of the input data and the model structure, which are described in the following section. The impact of mass rape on HIV epidemics in Tigray remains largely unknown due to the difficulty of collecting comprehensive data during the conflict. This study employed a dynamic risk equation model to quantify the contribution of conflict-related rape to HIV transmission, incorporating trauma-related exposure, PEP coverage, and health system accessibility as functions of conflict intensity and cumulative exposure risk. Model parameters were informed by regional epidemiological data from UNAIDS, WHO, EPHI, and relevant literature, and calibrated through 10,000 Monte Carlo simulations to capture uncertainty. The numerical simulations (Figs. 3 and 4) reveal how conflict-driven factors interact to shape HIV transmission. As conflict intensifies, health system functionality and PEP coverage decline. Small initial losses produce moderate infection increases, but near-total collapse leads to disproportionately large surges, reflecting the disruption documented in Tigray where 70% of hospitals became non-functional. Cumulative exposure risk, driven by assailant HIV prevalence and number of perpetrators, amplifies transmission beyond additive effects. The transition from single to multiple assailants produces the largest risk increase, consistent with Tigray’s 68.4% gang rape prevalence. Trauma-related exposure increases susceptibility through biological mechanisms but exhibits diminishing returns at extreme levels. Conflict intensity itself emerges as a critical driver: infections rise steeply beyond a threshold, underscoring that early intervention may yield disproportionately large reductions in HIV burden. Comparing our findings with earlier studies, Anema et al.23 estimated that conflict-related rape in sub-Saharan Africa could lead to a 5–28% increase in HIV incidence, while Supervie et al.24 reported a 0.5–2.0% increase in prevalence in specific conflict zones. Our median estimate of a 13.8% increase in incidence is consistent with these ranges, despite the lower baseline HIV prevalence in Tigray (1.4% pre-war). Recent empirical evidence from Tigray supports our estimates: Weledegebriel et al.43 documented that poor HIV outcomes increased from 28.6% pre-war to 54.5% during active conflict, with ART drug availability dropping by 68.8%. Similarly, Shfare et al.42 reported a wartime HIV prevalence of 1.2% among tested individuals, with 36% STI prevalence and evidence of sexual violence in 19.4% of female STI cases. The modest increase in overall prevalence (0.002%) is explained by the large denominator–the total population (6.25 million) and the existing pool of people living with HIV (approximately 49,141). The additional 122 new infections from mass rape are small relative to this pool. However, the 13.8% relative increase in incidence is epidemiologically notable, representing a sudden surge that can disrupt routine prevention and treatment services during a humanitarian crisis. Sensitivity analysis revealed that the most influential factors driving HIV incidence were conflict intensity, health system coverage, rape proportion, number of assailants, assailant HIV prevalence, PEP coverage, and trauma-related exposure. These findings provide actionable guidance for prioritizing interventions. The study’s methodological framework offers a mechanistic understanding of HIV transmission risk associated with mass rape. The analysis underscores that increasing rape incidences, combined with declining health system capacity and reduced PEP coverage during active conflict, contribute to new HIV infections. This highlights the urgent need for humanitarian agencies to implement targeted measures, including rapid PEP provision, mobile health services, and protective programs. Long-term consequences of armed conflict may exacerbate HIV epidemics post-conflict through health infrastructure collapse and displacement of health workers. Investment priorities should include rebuilding health facilities, strengthening survivor-centered care, and restoring comprehensive HIV prevention programs. Conflict-related sexual violence has been recognized more broadly as a driver of infectious disease outbreaks, including HIV, in humanitarian settings worldwide44. Challenging core model assumptions Several key assumptions warrant explicit consideration, as they directly shape the model’s outputs and the interpretation of its results. First, the model assumes a linear aggregation of risk factors (Assumption 5), which may underestimate the synergistic effect of co-occurring risks (e.g., trauma and lack of PEP). Consequently, the results may represent a conservative estimate of true transmission risk, particularly in settings where multiple vulnerabilities converge. Second, the parameters are treated as independent (Assumption 4), although in reality, conflict intensity simultaneously influences multiple factors. This simplification, while necessary for tractability, may obscure the compounding effects that characterize complex humanitarian emergencies, suggesting that the true uncertainty surrounding the estimates may be even greater than reported. Third, the baseline HIV prevalence among perpetrators is assumed constant, which may have fluctuated during the conflict. If perpetrators came from regions with higher HIV prevalence than our weighted average, our estimates could be conservative; conversely, if the trauma-related risk factor E is overestimated (Assumption 1), our results could be inflated. These directional biases underscore the importance of interpreting the findings with attention to the underlying assumptions. The wide interquartile ranges and broad confidence intervals underscore the substantial uncertainty inherent in estimating outcomes under conflict conditions. This uncertainty should be interpreted not as a weakness of the analysis, but as an honest reflection of the variability in conflict experiences–a variability that is amplified by the very assumptions examined above. For decision-makers, this means that planning should be scenario-based, focusing on interventions (such as ensuring PEP availability) that are effective across a range of plausible outcomes, regardless of how the model’s assumptions are specified. Limitations Despite these insights, the study has limitations. First, the study relied on secondary data with variable quality, as primary data collection was not feasible. The conflict in Tigray was characterized by the destruction of key health institutions, many districts remaining inaccessible, and survivors facing significant stigma and risk of retaliation. Under these constraints, traditional hospital- or community-based data collection was neither operationally feasible nor ethically justifiable. The model does not capture onward HIV transmission through sexual partnerships or mother-to-child transmission, nor does it account for variations in individual behavior, co-infections, or socio-economic factors. Third, the model’s representation of time via conflict intensity (c) is a proxy; it does not capture the precise timing of rape events relative to PEP availability, which could affect transmission risk. Fourth, the assumption of parameter independence may not hold, potentially leading to either over- or under-estimation of risk. These limitations emphasize the need for comprehensive, context-specific data and advanced modeling approaches, such as dynamic compartmental or agent-based models, to better evaluate the long-term impacts and intervention scenarios of conflict-related sexual violence on HIV transmission. Conclusion This model-based study indicates that conflict-related rape during the Tigray war markedly increased HIV incidence by a median of 13.8%, highlighting an urgent public health challenge. While overall prevalence remained minimally affected (median absolute increase of 0.002%), the elevated incidence underscores the critical need for comprehensive strategies combining prevention of sexual violence, timely PEP interventions, and resilient healthcare infrastructure. Strengthening health systems, expanding post-exposure prophylaxis coverage, and providing survivor-centered support are essential to mitigate HIV transmission in conflict-affected populations. For humanitarian agencies and conflicting parties, this means prioritizing: (1) deploying mobile health units to provide PEP and survivor-centered care in hard-to-reach areas; (2) training community health workers in SGBV first-line support and HIV risk assessment; and (3) re-establishing functional health facilities with dedicated services for SGBV survivors as part of the post-conflict recovery plan. Integrating social protection measures with biomedical interventions is crucial for addressing the complex drivers of HIV risk in war-affected settings. This study introduces a novel conflict-intensity-driven dynamic modeling framework that goes beyond static risk equations by capturing the interplay of escalating violence, health system collapse, and cumulative perpetrator exposure. Future research employing more sophisticated modeling techniques, such as dynamic compartmental or agent-based models, is needed to fully capture epidemic dynamics and the potential effects of interventions in such complex humanitarian contexts.