The relationship between lifetime trauma exposure and psychosis in a multi-country case-control study in Africa

Abstract

Background:

Exposure to traumatic events is a known risk factor for psychosis. Additionally, psychosis may be a risk factor for exposure to traumatic events. There are little data on the relationship between traumatic events and psychosis in sub-Saharan Africa, particularly in large, cross-country samples using the same instrument.

Methods:

In a case-control study, 21,606 adults were recruited with psychosis (cases) and 21,329 adults without any history of psychosis (controls) in Ethiopia, Kenya, South Africa, and Uganda from 2018 to 2023 (n = 42,935). Lifetime trauma exposure was assessed using the Life Events Checklist-5. Regression models included the: i) prevalence of any trauma exposure; ii) cumulative burden of trauma exposure; and iii) the odds of exposure to specific trauma types. Analyses were run by case-control status for the full sample and within each country; trauma types endorsed by cases and controls were further stratified by sex.

Results:

There was a modest increased odds of trauma among cases compared with controls. Cases had higher odds of reporting exposure to ≥1 trauma and ≥3 trauma types (adjusted odds ratio (AOR) = 1.23, 95 % CI: 1.18–1.28 and AOR = 1.19, 95 % CI: 1.15–1.23, respectively). The trauma types with the highest odds were sexual violence (AOR = 1.99, 95 % CI: 1.86–2.14), physical violence (AOR = 1.69, 95 % CI: 1.62–1.76), and network trauma (causing injury, harm, or death to someone else) (AOR = 1.52, 95 % CI: 1.38–1.67). Similar trends were seen within each country. Sexual violence and physical violence were most endorsed by female cases and male cases, respectively. Network trauma was most endorsed by male cases and particularly from South Africa.

Conclusion:

People in eastern and southern Africa report significant exposure to trauma with a slightly higher prevalence among individuals with psychosis. Special attention should be paid to potential trauma exposure including interpersonal violence when providing treatment for this population.

1. Background

Exposure to traumatic events is a known risk factor for many mental health conditions, including posttraumatic stress disorder (PTSD) and psychosis (Kessler et al., 2017; Varese et al., 2012). Numerous studies have found that prior experiences of trauma, particularly during childhood, are associated with the development of psychosis (Varese et al., 2012; Woolway et al., 2022), and that the association between trauma exposure and psychosis may be dose-respondent, with additional traumas further increasing the likelihood of psychosis (Matheson et al., 2013; Read et al., 2005; Shevlin et al., 2007). In addition, multiple studies have found that people with psychosis have higher rates of exposure to physical assault and sexual assault than the general population (Mauritz et al., 2013).

Trauma exposure, defined as exposure to actual or threatened death, serious injury, or sexual violence (American Psychiatric Association, 2013), is heterogeneous and consists of many different types of traumatic events that can occur across the lifespan, such as being in a natural disaster or being in a war zone. Sexual, physical, and psychological violence (Mercy et al., 2017), in particular, have been associated with increased psychopathology (Dworkin et al., 2017; Kessler et al., 2010). Population-representative studies have found differences in patterns of victimization by gender in that men and boys have a higher risk of exposure to physical violence and women and girls to sexual violence (Benjet et al., 2018; Finkelhor et al., 2013), which is consistent with findings from populations with psychosis (Khalifeh and Dean, 2010). Several theories have been proposed to explain why experiencing traumatic events can increase the risk of developing psychosis, including the stress-vulnerability model. This model posits that some people have a biological vulnerability to psychosis. It suggests that while people can tolerate a certain level of stress, such as a traumatic event, once a certain level of stress is surpassed, either through repeated exposure or through the intensity of a stressor, people become more susceptible to developing psychosis (Zubin and Spring 1977).

In addition, there are theories for why people with psychosis may be at higher risk for exposure to traumatic events. Theorists suggest that cognitive impairment, symptoms of psychosis (i.e., positive and negative symptoms), and/or social and physical isolation that can sometimes accompany psychosis can raise the likelihood of being targeted by a perpetrator (de Vries et al., 2019). In other situations, such as natural disasters and wars, it has been documented that people with psychosis have sometimes been left behind because of their inability to flee (Murthy and Lakshminarayana, 2006).

There is a paucity of data on the relationship between traumatic events and psychosis in sub-Saharan Africa, with the majority of studies having taken place in Europe, North America, and Australasia. Of two seminal meta-analyses, one on trauma exposure as a risk factor for developing psychosis and one on psychosis as a risk factor for exposure to trauma, only two of the 68 included studies took place in a low- or middle-income country (LMIC), and of those, only one was from Africa (de Vries et al., 2019; Varese et al., 2012). Of the literature from LMICs in Africa, research from South Africa found that the odds of developing schizophrenia were much higher in people who reported exposure to trauma in childhood, and particularly to traumas related to physical and emotional abuse (Mall et al., 2020). Additional studies in South Africa have focused on how childhood trauma is associated with pre-morbid adjustment (Kilian et al., 2017), poorer cognitive performance (Kilian et al., 2018), and abnormalities in brain function (Asmal et al., 2019) in those with schizophrenia compared to those without it. In qualitative research in South Africa, people with psychosis described traumatic events of familial disruptions in childhood, being bullied, experiencing violence from peers and people in the community, and the experience of psychosis itself (Ntlantsana et al., 2024). In qualitative research in Ethiopia which aimed to assess what kinds of events people with severe mental illness considered traumatic, participants described many events that were consistent with Western-defined clinical criteria [Criterion A events in The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5)], such as sexual assault, although some were not, such as forced marriage by abduction, being restrained, and the experience of severe mental illness itself (Ametaj et al., 2021).

A recent scoping review on the types of trauma that adults with pre-existing serious mental health conditions report in LMICs found that interpersonal violence, including sexual and physical assault, was the most reported trauma type (Stevenson et al., 2024). Physical restraint, including shackling, chaining, and human rights violations, emerged as a significant trauma category within interpersonal violence. Overall, the underlying mechanisms linking psychosis and trauma are postulated to be similar across contexts including the impact of childhood and interpersonal traumas; however, the specific experiences of trauma are more contextual including what is considered traumatic, how common different trauma types are, and how each relates to psychosis.

Most of the studies in LMICs, including those in Africa, have relied on small samples and used various trauma measures or qualitative methods that limit comparability. This analysis seeks to address these gaps by examining lifetime trauma exposure among people with psychosis in a much larger sample across four countries in Africa, and by using the same instrument that allows for comparison across Ethiopia, Kenya, South Africa, and Uganda.

Using data from a case-control study of 42,935 participants with a diagnosis of psychosis (cases) (n = 21,606) and those without (controls) (n = 21,329) in Ethiopia, Kenya, South Africa, and Uganda, the current study aimed to assess the differences between cases and controls in the: 1) prevalence of any trauma exposure; 2) cumulative burden of trauma exposure; and 3) patterns of trauma types across the four countries. Differences in the patterns of trauma types were further examined between cases and controls by sex (male cases vs. male controls and female cases vs. female controls). Our hypotheses were that compared to controls, people with psychosis would be more likely to report: 1) exposure to any type of traumatic event, in particular physical and sexual violence; 2) higher cumulative burden of trauma exposure; and 3) general trends would be common across the samples from Ethiopia, Kenya, South Africa, and Uganda, but there would be cross-country differences in the prevalence of and types of trauma exposure because of each country’s different histories and cultures. We expected that the trauma type analyses by case-control status and by country (and further stratified by sex) would drive some of the differences we would find in the full sample.

2. Methods

2.1. Study countries

This study took place across Ethiopia, Kenya, South Africa, and Uganda. Ethiopia is the second most populous country in East Africa with a population of 132 million people. Its gross domestic product (GDP) per capita is $2800 USD and its Gini co-efficient, a metric of inequality in which a number closer to 100 signifies higher inequality, is 35/100. There are a little more than 100 psychiatrists for the entire country, almost all of whom are located in the capital city, Addis Ababa. Ethiopia has experienced multiple armed, civil, and ethnic conflicts in the past 40 years. Kenya is a coastal country in East Africa and is adjacent to Ethiopia and Uganda. It has a population of 58 million people, its GDP/capita is $1300 USD, and its Gini co-efficient is 38.7/100. There are again just over 100 psychiatrists in Kenya, for whom the majority are based in the capital, Nairobi. Kenya’s history has been largely peaceful, although there have been instances of civil conflict around elections as well as intercommunal armed conflicts related to resource scarcity - livestock, grazing land, and water sources. South Africa is the southernmost country in Africa. It has a population of 60 million people, a GDP/capita of $6,000, and its Gini co-efficient is 63/100, the highest in the world. There are just over 930 psychiatrists in the country. South Africa was governed by apartheid, a system of institutionalized racial discrimination and segregation by the white minority government over non-white South Africans, from 1948 to 1994. Uganda is a landlocked country in East Africa and has a population of 49 million people, its GDP/capita is $1000, and its Gini co-efficient is 42.7/100. There are less than 50 psychiatrists in the country, who also primarily work in the capital city, Kampala. Uganda experienced more than 20 years of violent conflict in the northern region between rebel groups and Ugandan government forces starting in the 1980’s. Ethiopia, Kenya, South Africa, and Uganda are all designated as low- and middle-income countries by the World Bank (World Bank, 2024). [Ethiopian citations (CIA, 2025a; WHO, 2022a); Kenyan citations (CIA, 2025b; WHO, 2022b); South African citations (CIA, 2025c; WHO, 2022c); Ugandan citations (CIA, 2025d; WHO, 2022d).]

2.2. Participants

Data for this study were collected as part of a broader initiative, Neuropsychiatric Genetics of African Populations-Psychosis (NeuroGAP-Psychosis), which aimed to understand the genetic architecture of psychosis. A detailed description of NeuroGAP-Psychosis can be found in the published protocol (Stevenson et al., 2019). In brief, participants were recruited from February 2018 to March 2023. Participants were eligible to participate in the study if they were aged 18 years or older, had decision-making capacity to consent, provided informed consent, and were fluent in one of the languages in which the study was conducted. Potential participants were excluded if they did not meet the criteria above, were under the acute influence of alcohol or drugs, or were hospitalized for management of a substance use disorder.

As part of the NeuroGAP-Psychosis study, cases were included if they had received a previous formal diagnosis from a clinician of psychosis, which was operationalized as schizophrenia, schizoaffective disorder, bipolar disorder, mania not otherwise specified, or psychotic disorder not otherwise specified according to the DSM-5. Conditions were grouped under the umbrella of “psychosis” due to the large genetic overlap and diagnostic instability across these disorders (Cardno and Owen, 2014; Kämpe et al., 2024; Lee et al., 2013; Wood et al., 2021). Controls were included if they did not have any of the aforementioned psychiatric conditions.

Participants were recruited from more than 50 hospitals and community health centers in both rural and urban settings across the four countries. Cases were inpatients at psychiatric hospitals (Ethiopia and South Africa cases only) or outpatients receiving mental health services at health facilities. Controls were recruited from the same clinics or at health facilities in the same catchment areas to reflect the same underlying population as the cases. Within recruitment locations, cases and controls were recruited from the same health facilities if they provided mental health services and other healthcare services. If a recruitment site provided only psychiatric services, controls were recruited from a nearby health facility in the same municipality. Participants were frequency matched during recruitment to enroll a target ratio of 50:50 cases and controls, with an attempt to keep the sex and age distribution of the cases and controls similar.

Trainings were provided to the study staff to administer the structured phenotyping tools including trauma exposure in the same way for both cases and controls. In this secondary analysis of psychosis and trauma, only phenotypic data from NeuroGAP-Psychosis was used.

2.3. Procedures

The NeuroGAP-Psychosis tools were administered in Acholi-Luo, Afan Oromo, Afrikaans, Amharic, English, isiXhosa, Kiswahili, Luganda, Lugbara, and Runyankole. All study tools were translated from English to the nine other languages through a forward and backward translation process. Forward translation was completed by a team member or hired linguist who was bilingual in English and the local language. A separate team member/hired linguist, who was also bilingual in both languages, then back-translated the questions to English. The original source text and the back-translated version were then compared by the study teams at each site. For differences in the versions, discrepancies were reviewed by team members who came to a consensus on the appropriate translation to use.

Informed consent was provided by participants by signing the consent form or, if a participant was illiterate, using their fingerprint to mark and a witness to sign the consent form to ensure the participant had understood the process. All questionnaires were administered in person and read out loud to participants by trained research staff with a minimum of a bachelor’s degree. Ethical clearances to conduct this study were obtained from all participating sites, including the Harvard T.H. Chan School of Public Health (#IRB17-0822). Please see the Declarations section of the manuscript for the full list of ethics committees and approval numbers.

2.4. Measures

2.4.1. Sociodemographic characteristics

For sociodemographic information, participants were asked their age, biological sex at birth, marital status, highest level of education achieved, and living arrangement.

2.4.2. Trauma exposure

The Life Events Checklist for DSM-5 (LEC-5) is a self-report measure used to assess for potentially traumatic events that can result in psychopathology, such as PTSD (Weathers et al., 2013). The LEC-5 is a 17-item measure that lists 16 events that a person may have experienced over the life course (e.g., natural disaster or an unwanted sexual experience) and one additional item to capture any other very stressful event not captured in the previous items. The response options for 15 of the trauma types selected for this analysis were “happened to me” and “doesn’t apply.” For two items that cannot be directly experienced, “sudden violent death” and “sudden accidental death,” the response options were restricted to “witnessed it” or “doesn’t apply.” These response options were chosen for being the best predictors of developing PTSD symptoms (Weis et al., 2022).

The LEC has been previously used in Africa (Duko et al., 2020; Gray et al., 2015; Wadley et al., 2020) and has been shown to adequately capture traumatic events in the NeuroGAP-Psychosis dataset within Ethiopia, Kenya, South Africa, and Uganda (Girma et al., 2022; Kwobah et al., 2022; Morawej et al., 2024; Stevenson et al., 2023).

2.5. Statistical analysis

Traumatic events from the LEC-5 items were categorized into seven types which were drawn from the South African Stress and Health Study (SASH) (Atwoli et al., 2013) as part of the World Mental Health Surveys (WMHS), which utilized the World Health Organization’s Composite International Diagnostic Interview (CIDI) for DSM-IV’s PTSD module (Kessler and Üstün, 2004). A previous confirmatory factor analysis of SASH had grouped the 27 items from the CIDI into seven classes (Atwoli et al., 2013). In subsequent exploratory and confirmatory factor analyses of the LEC-5 in the NeuroGAP-Psychosis sample, the 7-factor model based on SASH was found to be the best fit in Ethiopia, Kenya, South Africa, and Uganda (Girma et al., 2022; Kwobah et al., 2022; Morawej et al., 2024; Stevenson et al., 2023). The seven trauma types were as follows: 1) accidents (natural disaster, fire or explosion, transport accident, serious accident at work, home, or during a recreational activity, exposure to a toxic substance, life-threatening illness or injury); 2) physical violence (being attacked, hit, slapped, kicked, beaten up, being shot, stabbed, threatened with a knife, gun, bomb); 3) sexual violence (rape, attempted rape, made to perform any type of sexual act through force or threat of harm, or other unwanted sexual experience); 4) war (exposure to war-zone in the military or as a civilian or captivity, being kidnapped, abducted, held hostage, prisoner of war; 5) severe suffering (severe human suffering); 6) witnessed death (witnessed an accidental death or a homicide or suicide); and 7) network trauma (causing injury, harm, or death to someone else). An additional open-ended question from the LEC-5, “any other stressful experience,” was excluded from the factor analyses because it could not be grouped with any of the other items, however, it was included here as a category for assessing the total prevalence of any exposure to a traumatic event and cumulative trauma burden.

There was not perfect overlap between the LEC-5 and the SASH groupings due to the phrasing of the different items and because we were mapping 16 items from the LEC-5 to SASH. The LEC-5 item “caused injury, harm, or death to someone else” most closely aligned with the SASH grouping “network events,” which commonly refers to events that occurred in a person’s family or social network, although “caused injury, harm, or death to someone else” reflects only some types of network events. We used the term “network trauma” as the best bridge between the LEC-5 item and the SASH term in the factor analyses in Ethiopia, Kenya, South Africa, and Uganda. Because we were using the names of those trauma types here, we used the same naming convention for consistency.

Exposure to any traumatic event was defined as experiencing at least one traumatic event (≥1) and the highest cumulative trauma burden was defined as exposure to ≥3 trauma types. A sum score was created by calculating the events across the eight trauma types with a score ranging from zero to eight. For any exposure, the sum score was grouped into 0 or ≥1 and cumulative burden was grouped into 0, 1, 2, and ≥3. We operationalized ≥1 traumatic event to align with the WMHS (Benjet et al., 2016) and the cut points to align as closely as possible with WMHS and the schizophrenia and trauma study in South Africa referenced previously (Mall et al., 2020).

The frequencies of sociodemographic characteristics, trauma exposure (any, cumulative), and trauma types were calculated for: 1) the full sample; 2) by country; and 3) by case-control status for the full sample and within each country. Further, 4) trauma types by case-control status were stratified by sex for the full sample and by country. Means and standard deviations were calculated for continuous variables and the Student’s t-test was used to evaluate differences. Counts and percentages were computed for categorical variables and the Chi-square test was used to evaluate bivariate differences.

For all regression analyses, we classified the independent variable as the case-control status and the dependent variable as the likelihood of reporting a traumatic event(s). Although “exposure” is typically synonymous with a predictor variable in statistics, we use the phrase “exposure to a trauma type” as part of the outcome variable throughout the manuscript to describe whether a participant experienced one of the trauma types in the LEC-5. Logistic regression was run to calculate odds ratios (OR) for any trauma exposure and each trauma type. Ordinal logistic regression was run to calculate OR for cumulative trauma burden as an ordered categorical variable. Age, sex, and level of education attainment were included to control for potential confounding. In the logistic regressions for trauma types between male cases and male controls and female cases and female controls, only age and education were adjusted for. We additionally conducted the Brant-Wald test to assess the assumption of proportional odds for our ordered logit model (Brant, 1990) and calculated e-values for the primary results of the study to estimate the potential impact of unmeasured confounding (VanderWeele and Ding, 2017).

All analyses were conducted using Stata 18 (StataCorp, 2023).

3. Results

The study sample consisted of 42,953 cases and controls. Eighteen participants were missing the LEC-5 (0.04 %), so the final analytic sample was 42,935 participants. The sample was split almost evenly between cases and controls, with cases making up 50.3 % (n = 21,606) and controls making up 49.7 % (n = 21,329). The mean age of the participants was 36.7 years (SD = 11.7), and more than half of the participants were male (57.3 %). Most participants completed at least some secondary education (70.3 %). Almost a third of the sample was from Ethiopia (30.3 %), followed by 27.9 % from Uganda, 22.3 % from South Africa, and 19.5 % from Kenya. For selected demographics for the total study sample, for case-control status, for each country, and for case-control status for each country, see Table 1.

Table 1. Study demographics.

Total	Study sample						Ethiopia						Kenya						South Africa						Uganda
	All		Cases		Controls		All		Cases		Controls		All		Cases		Controls		All		Cases		Controls		All		Cases		Controls
	42,935		21,606	50.3	21,329	49.7	12,996		6499	50.0	6497	50.0	8353		4414	52.8	3939	47.2	9589		4694	49.0	4895	51.1	11,997		5999	50.0	5998	50.0
Selected variables	Count	%	Count	%	Count	%	Count	%	Count	%	Count	%	Count	%	Count	%	Count	%	Count	%	Count	%	Count	%	Count	%	Count	%	Count	%
Sex at birth
Female	18,314	42.7	9092	49.7	9222	50.4	4518	34.8	2295	50.8	2223	49.2	3916	46.9	2026	51.7	1890	48.3	3473	36.2	1639	34.9	1834	37.5	4518	34.8	3204	53.4	3203	53.4
Male	24,621	57.3	12,107	49.2	12,514	50.8	8478	65.2	4276	50.4	4202	49.6	4437	53.1	2388	53.8	2049	46.2	6116	63.8	3055	65.1	3061	62.5	8478	65.2	2795	46.6	2795	46.6
Age categories (yrs)
18-29	13,000	30.3	6477	30.0	6523	30.6	3402	26.2	1655	25.5	1747	26.9	2925	35.0	1567	35.5	1358	34.5	2469	25.8	1144	24.4	1325	27.1	3402	26.2	1655	25.5	1747	26.9
30-44	19,226	44.8	9756	45.2	9470	44.4	6643	51.1	3355	51.6	3288	50.6	3491	41.8	1853	42.0	1638	41.6	4081	42.6	2036	43.4	2045	41.8	6643	51.1	3355	51.6	3288	50.6
45-59	8932	20.8	4460	20.6	4472	21.0	2565	19.7	1299	20.0	1266	19.5	1564	18.7	789	17.9	775	19.7	2572	26.8	1284	27.4	1288	26.3	2565	19.7	1299	20.0	1266	19.5
60+	1777	4.1	913	4.2	864	4.1	386	3.0	190	2.9	196	3.0	373	4.5	205	4.6	168	4.3	467	4.9	230	4.9	237	4.8	386	3.0	190	2.9	196	3.0
Marital status
Single	19,736	46.0	11,455	53.0	8281	38.8	1642	12.6	755	11.6	887	13.7	3289	39.4	1916	43.4	1373	34.9	6048	63.1	3528	75.2	2520	51.5	4278	35.7	2352	39.2	1926	32.1
Married or cohabitating	16,406	38.2	6115	28.3	10,291	48.3	5011	38.6	2976	45.8	2035	31.3	3682	44.1	1592	36.1	2090	53.1	2408	25.1	590	12.6	1818	37.1	5174	43.1	2098	35.0	3076	51.3
Widowed	1553	3.6	742	3.4	811	3.8	4985	38.4	1740	26.8	3245	50.0	325	3.9	165	3.7	160	4.1	281	2.9	141	3.0	140	2.9	625	5.2	285	4.8	340	5.7
Divorced or separated	5221	12.2	3277	15.2	1944	9.1	1315	10.1	1005	15.5	310	4.8	1057	12.7	741	16.8	316	8.0	841	8.8	425	9.1	416	8.5	1913	16.0	1258	21.0	655	10.9
Level of education
No formal education	1172	2.7	653	3.0	519	2.4	549	4.2	394	6.1	155	2.4	98	1.2	54	1.2	44	1.1	51	0.5	33	0.7	18	0.4	474	4.0	172	2.9	302	5.0
Primary	11,596	27.0	6783	31.4	4813	22.6	3670	28.2	2143	33.0	1527	23.5	2637	31.6	1778	40.3	859	21.8	1206	12.6	750	16.0	456	9.3	4083	34.0	2112	35.2	1971	32.9
Secondary	18,652	43.5	9510	44.0	9142	42.9	4347	33.5	2387	36.7	1960	30.2	2766	33.1	1579	35.8	1187	30.1	6638	69.3	3071	65.6	3567	72.9	4901	40.9	2473	41.2	2428	40.5
University	11,499	26.8	4649	21.5	6850	32.1	4430	34.1	1575	24.2	2855	43.9	2850	34.1	1002	22.7	1848	46.9	1683	17.6	831	17.7	852	17.4	2536	21.1	1241	20.7	1295	21.6
Living arrangement
Alone	5746	13.4	2265	10.5	3481	16.3	1642	12.6	755	11.6	887	13.7	1160	13.9	422	9.6	738	18.7	1497	15.6	488	10.4	1009	20.6	1642	12.6	755	11.6	887	13.7
With parents	13,911	32.4	8923	41.3	4988	23.4	5011	38.6	2976	45.8	2035	31.3	2627	31.5	1900	43.1	727	18.5	3369	35.2	2199	46.9	1170	23.9	5011	38.6	2976	45.8	2035	31.3
With spouse or partner	15,030	35.0	5298	24.5	9732	45.6	4985	38.4	1740	26.8	3245	50.0	3292	39.4	1393	31.6	1899	48.2	2299	24.0	466	9.9	1833	37.5	4985	38.4	1740	26.8	3245	50.0
With friends or other relatives	8041	18.7	4975	23.0	3066	14.4	1315	10.1	1005	15.5	310	4.8	1265	15.2	692	15.7	573	14.6	2281	23.8	1430	30.5	851	17.4	1315	10.1	1005	15.5	310	4.8

Caption: Selected study demographics for the full sample, by country, and by case status. Numbers may not add up due to missing.

The results are organized as follows: 1) any trauma, 2) cumulative trauma, and 3) the three most endorsed trauma types among cases compared to controls: 3a) sexual violence, 3b) physical violence, and 3c) network trauma. Within each section, we first present 1) bivariate comparisons of trauma types between cases and controls, 2) multivariate model results of trauma types between cases and controls, after adjusting for covariates, and 3) differences in both bivariate and multivariate findings across each of the four countries (Ethiopia, Kenya, Uganda, South Africa). For the third section on the three most endorsed trauma types among cases compared to controls, we 4) additionally present differences in bivariate and multivariate results between the two sexes. We include these stratified sex analyses twice, first after the main findings in the total sample and then after presenting differences across each of the four countries.

3.1. Any trauma

Cases reported greater exposure to any trauma (≥1 exposure to trauma type) than controls (68.2 % for cases and 63.3 % for controls, p < .001), representing higher odds for exposure to any trauma than controls after adjusting for age, sex, and education [adjusted odds ratio (AOR = 1.23, 95 % CI: 1.18–1.28)].

These trends were consistent with bivariate comparisons and AORs within each country. In Ethiopia, 56.3 % of cases reported exposure to any trauma, compared to 53.8 % of controls (p = .004; AOR = 1.10, 95 % CI: 1.02–1.18 in Ethiopia). In Kenya, 63.8 % of cases reported any exposure to trauma, compared to 57.9 % of controls (p < .001; AOR = 1.21, 95 % CI: 1.10–1.33). In South Africa, 93.6 % of cases reported any exposure to trauma, compared to 92.0 % of controls (p = .003; AOR = 1.21, 95 % CI: 1.04–1.42). Lastly, in Uganda, 64.5 % of cases reported any exposure to trauma, compared to 53.7 % of controls (p < .001; AOR = 1.59, 95 % CI: 1.47–1.71).

For results for any exposure to trauma in the study sample, by case-control status, by country, and by case-control status by country, see Table 2.

Table 2. Prevalence and odds ratios of any exposure to a traumatic event and cumulative trauma burden.

Study sample	All	%	Cases	%	Controls	%
	42,935	100.0	21,606	50.3	21,329	49.7	p-value	OR	95 % CI	AOR	95 % CI
Any trauma exposure	28,248	65.8	14,743	68.2	13,505	63.3
Cumulative no. of trauma types
None	14,687	34.2	6863	31.8	7824	36.7	<0.001	1.20	1.16–1.25	1.19	1.15–1.23
One type	11,980	27.9	6134	28.4	5846	27.4
Two types	7849	18.3	4128	19.1	3721	17.5
Three + types	8419	19.6	4481	20.7	3938	18.5
Ethiopia	All	%	Cases	%	Controls	%
	12,996	100.0	6499	50.0	6497	50.0	p-value	OR	95 % CI	AOR	95 % CI
Any trauma exposure	7157	55.1	3660	56.3	3497	53.8
Cumulative no. of trauma types
None	5839	44.9	2839	43.7	3000	46.2	0.03	1.08	1.01–1.15	1.06	0.99–1.13
One type	3749	28.9	1925	29.6	1824	28.1
Two types	1967	15.1	1013	15.6	954	14.7
Three + types	1441	11.1	722	11.1	719	11.1
Kenya	All	%	Cases	%	Controls	%
	8353	100.0	4414	52.8	3939	47.2	p-value	OR	95 % CI	AOR	95 % CI
Any trauma exposure	5100	61.1	2818	63.8	2282	57.9
Cumulative no. of trauma types
None	3253	38.9	1596	36.2	1657	42.1	<0.001	1.26	1.17–1.37	1.20	1.10–1.30
One type	2876	34.4	1553	35.2	1323	33.6
Two types	1422	17.0	810	18.4	612	15.5
Three + types	802	9.6	455	10.3	347	8.8
South Africa	All	%	Cases	%	Controls	%
	9589	100.0	4694	49.0	4895	51.1	p-value	OR	95 % CI	AOR	95 % CI
Any trauma exposure	8898	92.8	4394	93.6	4504	92.0
Cumulative no. of trauma types
None	691	7.2	300	6.4	391	8.0	<0.001	1.29	1.19–1.39	1.27	1.17–1.36
One type	1993	20.8	895	19.1	1098	22.4
Two types	2445	25.5	1163	24.8	1282	26.2
Three + types	4460	46.5	2336	49.8	2124	43.4
Uganda	All	%	Cases	%	Controls	%
	11997	100.0	5999	50.0	5998	50.0	p-value	OR	95 % CI	AOR	95 % CI
Any trauma exposure	7093	59.1	3871	64.5	3222	53.7
Cumulative no. of trauma types
None	4904	40.9	2128	35.5	2776	46.3	<0.001	1.51	1.41–1.61	1.50	1.41–1.61
One type	3362	28.0	1761	29.4	1601	26.7
Two types	2015	16.8	1142	19.0	873	14.6
Three + types	1716	14.3	968	16.1	748	12.5

Caption: Prevalence and odds ratios of any exposure to a traumatic event and cumulative trauma burden for the study sample, by case-control status, and by country; Adjusted odds ratios were adjusted by age, sex, and education.

Abbreviations: OR = Odds ratio AOR = Adjusted odds ratio CI = Confidence Interval.

3.2. Cumulative trauma burden

Cases reported greater exposure to ≥3 trauma types (the highest cumulative trauma burden) compared to controls in the overall sample (20.7 % vs. 18.5 %, p < .001; AOR = 1.19, 95 % CI: 1.15–1.23).

In Ethiopia, Kenya, South Africa, and Uganda, the prevalence of ≥3 trauma types ranged from a low of 8.8 % of controls in Kenya to a high of almost 50 % of cases in South Africa (11.1 % vs. 11.1 %, p = .03 in Ethiopia; 10.3 % cases vs. 8.8 % controls, p < .001 in Kenya; 49.8 % cases vs. 43.4 % controls, p < .001 in South Africa; and 16.1 % cases vs. 12.5 % controls, p < .001 in Uganda). After adjusting for age, sex, and education, cases reported a cumulative trauma burden of 1.06–1.50 greater odds than controls (AOR = 1.06, 95 % CI: 0.99–1.13 in Ethiopia; AOR = 1.20, 95 % 1.10–1.30 in Kenya; AOR = 1.27, 95 % CI: 1.17–1.36 in South Africa; and AOR = 1.50, 95 % CI: 1.41–1.61 in Uganda). For results for cumulative trauma burden for the study sample, by case-control status, by country, and by case-control status by country, see Table 2.

3.3. Trauma types

The three trauma types with the highest AORs between cases and controls were sexual violence, physical violence, and network trauma. This pattern was the same in each country except for network trauma in Ethiopia. For detailed results of trauma type endorsements for the study sample and by case-control status, see Table 3. To see trauma types for each country and by case-control status by country, see Table 4. For trauma type endorsements by case-control status further stratified by sex for the study sample and by country, see Table 5for bivariate differences and Fig. 1 for a forest plot of the AORs and CIs.

Table 3. Frequencies and odds ratios of trauma types for the full sample and by case-control status.

Study sample	All		Cases		Controls
	42,935		21606	50.3	21329	49.7
Trauma types:	Count	%	Count	%	Count	%	p-value	OR	95 % CI	AOR	95 % CI
Accident (natural disaster, fire or explosion, transport accident, serious accident at work, home, or during a recreational activity, exposure to a toxic substance, life-threatening illness or injury)	14,341	33.5	6991	32.4	7350	34.6	<0.001	0.91	0.87–0.95	0.90	0.87–0.94
Physical Violence (being attacked, hit, slapped, kicked, beaten up, being shot, stabbed, threatened with a knife, gun, bomb)	15,709	36.7	9205	42.7	6504	30.6	<0.001	1.69	1.62–1.76	1.68	1.62–1.75
Sexual violence (rape, attempted rape, made to perform any type of sexual act through force or threat of harm, or other unwanted sexual experience)	4266	10.0	2725	12.6	1541	7.2	<0.001	1.85	1.73–1.98	1.99	1.86–2.14
War (exposure to war-zone in the military or as a civilian or captivity, being kidnapped, abducted, held hostage, prisoner of war)	3286	7.7	1660	7.7	1626	7.6	0.85	1.01	0.94–1.08	0.98	0.91–1.06
Severe Suffering (severe human suffering)	3356	7.8	1951	9.0	1405	6.6	<0.001	1.41	1.31–1.51	1.37	1.27–1.47
Witnessed death (witnessed an accidental death or a homicide or suicide)	10,410	24.3	4554	21.1	5856	27.5	<0.001	0.70	0.67–0.74	0.69	0.66–0.72
Network Trauma (caused injury/harm/death you caused to someone else)	1891	4.4	1146	5.3	745	3.5	<0.001	1.55	1.41–1.70	1.52	1.38–1.67

Caption: Trauma-type endorsements for the full sample and for cases and controls. Adjusted odds ratios were adjusted by age, sex, and education.

Abbreviations: OR = Odds ratio AOR = Adjusted odds ratio CI = Confidence Interval.

Table 4. Frequencies and odds ratios of trauma types for each country and by case-control status.

Ethiopia	All		Cases		Controls
	12,996		6499	50.0	6497	50.0
Trauma types:	Count	%	Count	%	Count	%	p-value	OR	95 % CI	AOR	95 % CI
Accident	3430	26.5	1632	25.2	1798	27.7	0.002	0.88	0.82–0.95	0.88	0.81–0.95
Physical Violence	3303	25.5	2197	33.9	1106	17.0	<0.001	2.50	2.30–2.71	2.46	2.26–2.68
Sexual violence	629	4.9	464	7.2	165	2.5	<0.001	2.96	2.47–3.55	3.26	2.70–3.93
War	854	6.6	377	5.8	477	7.3	<0.001	0.78	0.68–0.90	0.75	0.65–0.86
Severe Suffering	484	3.7	222	3.4	262	4.0	0.07	0.84	0.70–1.01	0.83	0.69–1.01
Witnessed death	2911	22.4	1104	17.0	1807	27.8	<0.001	0.53	0.49–0.58	0.52	0.48–0.57
Network Trauma	373	2.9	167	2.6	206	3.2	0.04	0.81	0.66–0.99	0.78	0.63–0.96
Kenya	All		Cases		Controls
	8353		4414	52.8	3939	47.2
Trauma types:	Count	%	Count	%	Count	%	p-value	OR	95 % CI	AOR	95 % CI
Accident	2034	24.4	968	22.0	1066	27.2	<0.001	0.75	0.68–0.83	0.73	0.66–0.81
Physical Violence	2198	26.4	1451	32.9	747	19.0	<0.001	2.09	1.89–2.31	1.98	1.78–2.20
Sexual violence	387	4.6	257	5.8	130	3.3	<0.001	1.81	1.46–2.24	2.18	1.73–2.74
War	207	2.5	102	2.3	105	2.7	0.29	0.86	0.65–1.14	0.92	0.69–1.23
Severe Suffering	397	4.8	262	5.9	135	3.4	<0.001	1.78	1.44–2.20	1.44	1.15–1.79
Witnessed death	2201	26.4	1114	25.3	1087	27.7	0.01	0.88	0.80–0.98	0.81	0.74–0.90
Network Trauma	216	2.6	167	3.8	49	1.2	<0.001	3.12	2.26–4.30	3.02	2.17–4.21
South Africa	All		Cases		Controls
	9589		4694	49.0	4895	51.1
Trauma types:	Count	%	Count	%	Count	%	p-value	OR	95 % CI	AOR	95 % CI
Accident	5437	56.8	2692	57.4	2745	56.3	0.282	1.05	0.96–1.13	1.03	0.95–1.11
Physical Violence	7262	75.9	3751	80.0	3511	71.9	<0.001	1.56	1.42–1.72	1.57	1.42–1.73
Sexual violence	2065	21.6	1245	26.6	820	16.8	<0.001	1.79	1.62–1.98	2.03	1.83–2.26
War	526	5.5	300	6.4	226	4.6	<0.001	1.41	1.18–1.68	1.41	1.18–1.69
Severe Suffering	1820	19.0	1951	9.0	772	15.8	<0.001	1.53	1.38–1.70	1.49	1.34–1.65
Witnessed death	3314	34.6	1388	29.6	1926	39.5	<0.001	0.64	0.59–0.70	0.62	0.57–0.68
Network Trauma	1023	10.7	642	13.7	381	7.8	<0.001	1.88	1.64–2.15	1.81	1.58–2.07
Uganda	All		Cases		Controls
	11997		5999	50.0	5998	50.0
Trauma types:	Count	%	Count	%	Count	%	p-value	OR	95 % CI	AOR	95 % CI
Accident	3440	28.8	1699	28.4	1741	29.2	0.35	0.96	0.89–1.04	0.96	0.89–1.04
Physical Violence	2946	24.6	1806	30.1	1140	19.1	<0.001	1.83	1.68–1.99	1.84	1.69–2.01
Sexual violence	1185	9.9	759	12.7	426	7.1	<0.001	1.89	1.67–2.14	1.92	1.69–2.18
War	1699	14.2	881	14.7	818	13.7	0.12	1.08	0.98–1.20	1.14	1.02–1.27
Severe Suffering	655	5.5	419	7.0	236	4.0	<0.001	1.83	1.55–2.15	1.86	1.58–2.19
Witnessed death	1984	16.6	948	15.8	1036	17.4	0.02	0.89	0.81–0.99	0.90	0.81–0.99
Network Trauma	279	2.3	170	2.8	109	1.8	<0.001	1.57	1.23–2.00	1.57	1.23–2.01

Caption: Trauma-type endorsements within each country and by case-control status within each country. Adjusted odds ratios were adjusted by age, sex, and education.

Abbreviations: OR = Odds ratio AOR = Adjusted odds ratio CI = Confidence Interval.

Table 5. Sex stratified trauma type endorsements by case-controls status.

Full study sample	Cases-Male		Controls-Male			Cases-Female		Controls-Female
Trauma types:	Count	%	Count	%	p-value	Count	%	Count	%	p-value
Accident	4394	35.2	4651	38.5	<0.001	2597	28.6	2699	29.4	0.27
Physical Violence	6016	48.1	4319	35.7	<0.001	3189	35.1	2185	23.8	<0.001
Sexual violence	720	5.8	355	2.9	<0.001	2005	22.1	1186	12.9	<0.001
War	970	7.8	1028	8.5	0.03	690	7.6	598	6.5	0.004
Severe Suffering	1071	8.6	872	7.2	<0.001	880	9.7	533	5.8	<0.001
Witnessed death	2796	22.4	3673	30.4	<0.001	1758	19.4	2183	23.8	<0.001
Network Trauma	907	7.3	612	5.1	<0.001	239	2.6	133	1.4	<0.001
Ethiopia	Cases-Male		Controls-Male			Cases-Female		Controls-Female
Trauma types:	Count	%	Count	%	p-value	Count	%	Count	%	p-value
Accident	1174	27.6	1249	29.8	0.03	458	20.7	549	23.9	0.009
Physical Violence	1581	37.1	818	19.5	<0.001	616	27.8	288	12.6	<0.001
Sexual violence	89	2.1	45	1.1	<0.001	375	16.9	120	5.2	<0.001
War	312	7.3	425	10.1	<0.001	65	2.9	52	2.3	0.16
Severe Suffering	150	3.5	174	4.1	0.13	72	3.2	88	3.8	0.28
Witnessed death	794	18.6	1239	29.5	<0.001	310	14.0	568	24.8	<0.001
Network Trauma	137	3.2	173	4.1	0.03	30	1.4	33	1.4	0.81
Kenya	Cases-Male		Controls-Male			Cases-Female		Controls-Female
Trauma types:	Count	%	Count	%	p-value	Count	%	Count	%	p-value
Accident	594	24.9	629	30.8	<0.001	374	18.5	437	23.2	<0.001
Physical Violence	911	38.2	435	21.3	<0.001	540	26.7	312	16.6	<0.001
Sexual violence	46	1.9	27	1.3	0.11	211	10.4	103	5.5	<0.001
War	66	2.8	77	3.8	0.06	36	1.8	28	1.5	0.47
Severe Suffering	131	5.5	65	3.2	<0.001	131	6.5	70	3.7	<0.001
Witnessed death	610	25.6	607	29.7	0.002	504	24.9	480	25.5	0.68
Network Trauma	131	5.5	41	2.0	<0.001	36	1.8	8	0.4	<0.001
South Africa	Cases-Male		Controls-Male			Cases-Female		Controls-Female
Trauma types:	Count	%	Count	%	p-value	Count	%	Count	%	p-value
Accident	1763	57.8	1855	60.8	0.02	929	56.7	890	48.8	<0.001
Physical Violence	2615	85.6	2466	80.7	<0.001	1136	69.4	1045	57.1	<0.001
Sexual violence	477	15.6	222	7.3	<0.001	768	46.9	598	32.7	<0.001
War	176	5.8	147	4.8	0.09	124	7.6	79	4.3	<0.001
Severe Suffering	606	19.9	536	17.6	0.02	442	27.0	236	12.9	<0.001
Witnessed death	944	31.0	1379	45.2	<0.001	444	27.1	547	30.0	0.06
Network Trauma	548	18.0	327	10.7	<0.001	94	5.7	54	3.0	<0.001
Uganda	Cases-Male		Controls-Male			Cases-Female		Controls-Female
Trauma types:	Count	%	Count	%	p-value	Count	%	Count	%	p-value
Accident	863	30.9	918	33.0	0.10	836	26.2	823	25.8	0.74
Physical Violence	909	32.6	600	21.6	<0.001	897	28.0	540	16.9	<0.001
Sexual violence	108	3.9	61	2.2	<0.001	651	20.3	365	11.4	<0.001
War	416	14.9	379	13.7	0.19	465	14.5	439	13.8	0.38
Severe Suffering	184	6.6	97	3.5	<0.001	235	7.3	139	4.4	<0.001
Witnessed death	448	16.1	448	16.1	0.93	500	15.6	588	18.4	0.003
Network Trauma	91	3.3	71	2.6	0.12	79	2.5	38	1.2	<0.001

Caption: Frequencies of trauma type endorsements stratified by case-control status and by sex for the full sample and by country.

Fig. 1.

Sex stratified trauma type endorsements by case-control status for the full sample and by country

Caption: Adjusted odds ratios (AORs) and confidence intervals (CIs) of trauma-type endorsements stratified by case-control status and by sex for the full sample and by country. AORs were adjusted by age and education.

3.3.1. Sexual violence

Ten percent of the full study sample endorsed experiencing sexual violence, however, this was higher for cases than controls (12.6 % vs. 7.2 %, p < .001). After adjusting for age, sex, and education, cases were twice as likely to report exposure to sexual violence than controls (AOR = 1.99, 95 % CI: 1.86–2.14). Male cases were more likely to report exposure to sexual violence than male controls (5.8 % vs. 2.9 %, p < .001; AOR = 2.12, 95 % CI: 1.86–2.42). And more than 22 % of female cases endorsed being a victim of sexual violence, which was much higher than reported by female controls (22.1 % vs. 12.9 %, p < .001; AOR = 1.95, 95 % CI: 1.81–2.12).

Similar patterns were found within all four countries. The prevalence of sexual violence among cases in Ethiopia was 7.2 % compared to 2.5 % of controls (p < .001); 5.8 %–3.3 % in Kenya (p < .001); 26.6 %–16.8 % in South Africa (p < .001); and 12.7 %–7.1 % in Uganda (p < .001). After adjusting for potential confounders, cases had 1.9 to 3.3 greater odds of reporting sexual violence than controls (AOR = 3.26, 95 % CI: 2.70–3.93 in Ethiopia; AOR = 2.18, 95 % CI: 1.73–2.74 in Kenya; AOR = 2.03, 95 % CI: 1.83–2.26 in South Africa; and AOR = 1.92, 95 % CI: 1.69–2.18 in Uganda).

Within each country male cases were much more likely to report experiencing sexual violence than male controls (AOR = 2.09, 95 % CI: 1.45–3.02 in Ethiopia; AOR = 1.65, 95 % CI: 1.01–2.71 in Kenya; AOR = 2.40, 95 % CI: 2.02–2.84 in South Africa; and AOR = 1.83, 95 % CI: 1.33–2.52 in Uganda). Female cases reported experiencing sexual violence significantly more than female controls as well. In Ethiopia, 16.9 % of female cases endorsed sexual violence compared to 5.2 % female controls (p < .001; AOR = 3.77, 95 % CI: 3.03–4.69). In Kenya, 10.4 % of female cases endorsed sexual violence compared to 5.5 % of controls (p < .001; AOR = 2.34, 95 % CI: 1.80–3.02). In South Africa, 46.9 % of female cases endorsed sexual violence compared to 32.7 % of female controls (p < .001; AOR = 1.86, 95 % CI: 1.62–2.15). In Uganda, 20.3 % of female cases endorsed sexual violence compared to 11.4 % of female controls (p < .001; AOR = 1.95, 95 % CI: 1.69–2.24).

3.3.2. Physical violence

Physical violence was the most endorsed trauma type in the sample for all participants (36.7 %), although cases reported it significantly more so than controls (42.7 % vs. 30.6 %, p < .001; AOR = 1.69, 95 % CI: 1.62–1.76). More than 48 % of male cases reported experiencing physical violence compared to 35.7 % of male controls (p < .001; AOR = 1.66, 95 % CI: 1.58–1.75). More than 35 % of female cases reported exposure to physical violence compared to 23.8 % of female controls (p < .001; AOR = 1.73, 95 % CI: 1.62–1.85).

Similar endorsement patterns were seen in Ethiopia, Kenya, South Africa, and Uganda. Physical violence was the most reported trauma type for cases in each country (33.9 % cases vs. 17.0 % controls, p < .001 in Ethiopia; 32.9 % vs. 19.0 %, p < .001 in Kenya; 80.0 % vs. 71.9 %, p < .001 in South Africa; 30.1 % vs. 19.1 %, p < .001 in Uganda). This was reflected in the AORs as well (AOR = 2.46, 95 % CI: 2.26–2.68 in Ethiopia; AOR = 1.98, 95 % CI: 1.78–2.20 in Kenya; AOR = 1.57, 95 % CI: 1.42–1.73 in South Africa; and AOR = 1.84, 95 % CI: 1.69–2.01 in Uganda). Sex-stratified analyses between cases and controls within each country mirrored the results of the full study sample.

3.3.3. Network trauma

Network trauma was one of the least prevalent trauma types in the sample for both cases and controls (5.3 % vs. 3.5 %, p < .001). Cases, however, had one and a half greater odds of reporting it than controls (AOR = 1.52, 95 % CI: 1.38–1.67). Both male cases and female cases endorsed network trauma more than their male and female counterparts. More than seven percent (7.3 %) of males cases endorsed network trauma compared to 5.1 % of male controls (p < .001; AOR = 1.45, 95 % CI: 1.31–1.62). Two point six percent of female cases endorsed network trauma compared to 1.4 % of female controls (p < .001; AOR = 1.83, 95 % CI: 1.48–2.27).

Within Kenya, South Africa, and Uganda, cases also reported higher frequencies of network trauma than controls. In Kenya, 3.8 % of cases endorsed network trauma compared to 1.2 % of controls (p < .001; AOR = 3.02, 95 % CI: 2.17–4.21); in South Africa, 13.7 % of cases endorsed it compared to 7.8 % of controls (p < .001; AOR = 1.81, 95 % CI: 1.58–2.07); and in Uganda, 2.8 % of cases endorsed it compared to 1.8 % of controls (p < .001; AOR = 1.57, 95 % CI: 1.23–2.01). In Ethiopia, controls reported a higher exposure to network trauma than cases (3.2 % vs. 2.6 %, p = .04; AOR = 0.78, 95 % CI: 0.63–0.96).

Each country showed low endorsement of network trauma for all subgroups except for male cases in South Africa, of whom 18.0 % endorsed the trauma type. Despite the low prevalence of network trauma in the rest of the sample (a low of 0.4 % for female controls in Kenya to a high of 10.7 % for male controls in South Africa, there were large differences in network trauma by sex. In Kenya and South Africa, male cases had a 1.78–2.72 greater odds of endorsing network trauma than male controls (AOR = 2.72, 95 % CI: 1.89–3.93 in Kenya; AOR = 1.78, 95 % CI: 1.53–2.07 in South Africa). Across Kenya, South Africa, and Uganda, female cases had 2.10–4.56 greater odds of reporting network trauma compared to female controls (AOR = 4.56, 95 % CI: 2.06–10.09 in Kenya; AOR = 2.10, 95 % CI: 1.48–2.97 in South Africa; AOR = 2.10, 95 % CI: 1.42–3.10 in Uganda). Ethiopia was the only country where any control subgroup reported more exposure to network trauma (4.1 % male controls vs. 3.2 % male cases, p = .03; AOR = 0.74, 95 % CI: 0.59–0.94). Female cases and controls in Ethiopia reported the same frequency of exposure to network trauma as each other (1.4 % each, p = .81; AOR = 1.01, 95 % CI: 0.61–1.69).

The e-values associated with all of the primary AORs in this study ranged from 1.11 (CI: 1.00) (for network trauma in Ethiopia between female cases vs. female controls) to 8.59 (CI: 3.54) (for network trauma in Kenya between female cases vs. female controls) with most of the e-values for specific exposure-outcome relationships ranging from 2 to 3 depending on the country.

4. Discussion

This is one of the largest cross-national studies on exposure to traumatic events outside of the WMHS (Kessler et al., 2017), and the largest study of which we are aware on trauma exposure in persons with psychosis. Additionally, it expands the prior focus on childhood trauma preceding psychosis to trauma to consider lifetime traumatic experiences including how they might follow from psychosis. We found that cases endorsed modestly more exposure to any type of trauma, a higher cumulative trauma burden, and higher exposure to sexual violence, physical violence, and network trauma than controls, after adjusting for age, sex, and level of educational attainment. The results confirmed our hypotheses. There were also sex differences in endorsements of trauma types between male cases and male controls and between female cases and female controls. Specifically, female cases reported the most exposure to sexual violence in the sample and had almost twice the odds of reporting it than female controls. In addition, male cases reported the most exposure to physical violence across the whole sample and 1.66 higher adjusted odds than male controls. Male cases endorsed the most exposure to network trauma in the sample and their odds of reporting network trauma was almost 50 % higher than male controls after adjusting for age and education.

We saw similar trends between cases and controls within each country. The findings align with our hypotheses by country as well, with cases reporting a higher cumulative trauma burden than controls except in Ethiopia where controls endorsed a higher cumulative burden than cases.

4.1. Prevalence of trauma exposure

Our results align with the vast research in the field of trauma exposure and psychosis in that cases reported a higher prevalence of any trauma exposure than controls (Maniglio, 2009; Varese et al., 2012). In the full sample, more than 68 % of cases reported ≥1 trauma exposure compared to 63.3 % of controls. This was largely driven by the difference in prevalence of cases reporting exposure to physical violence (42.7 % of cases vs. 30.6 % of controls), which was the most highly endorsed traumatic experience (See section 4.3.2 for further examination of physical violence in the sample.).

Although our results were in line with many other studies, the added odds of reporting exposure to trauma among persons with psychosis (when compared with controls) were smaller than those from high-income countries (Varese et al., 2012). One reason might be that prior research often focuses exclusively on childhood trauma while our study more broadly studied lifetime exposure and did not single out childhood experiences. Additionally, our sampling strategy for controls might have contributed to such lower odds. Because our controls were recruited from outpatient settings and the majority were seeking treatment for health conditions, they were likely less healthy than the general population. Research has shown that chronic physical health conditions are associated with exposure to traumatic events (Scott et al., 2013), which may be reflected in the prevalence of trauma exposure in the controls in our sample, and may have decreased the AORs in trauma exposure between cases and controls. The degree of trauma exposure in our controls across all trauma types may also be considered a strength of our study. Because of the “noise” in our control population, our signal (AOR) was likely lower than it would have been otherwise if we had had healthy controls. That is, exposure to interpersonal violence was that much greater in the cases and drove the results in the overall endorsement of ≥1 and ≥ 3 trauma types. Such findings highlight the importance of interpersonal violence as a specific exposure by participants with psychosis and draw attention to interpersonal violence as a risk factor and/or vulnerability within this population.

4.2. Cumulative trauma burden

Cases reported higher exposure to ≥3 trauma types compared to controls, which supports previous research as well (Grubaugh et al., 2011; Shevlin et al., 2007). While the overall pattern of trauma exposure was similar in cases and controls, cases may have been exposed to a unique range of trauma types due to their diagnoses; for example, patients with psychosis have reported high rates of lifetime trauma exposures in psychiatric settings, including physical assault and sexual assault by staff, police, and other patients in the United States (Frueh et al., 2005). In Uganda, multiple people with schizophrenia, bipolar affective disorder, and other psychoses at two of the same hospitals as our study previously reported being raped during acute illness episodes (Lundberg et al., 2012). Further research in our sample would be needed to assess the nature and setting of trauma exposure among cases, including in relation to the psychiatric setting.

Though our results for cumulative trauma burden were in keeping with the literature in the field, our odds for ≥3 trauma types were lower than similar research in the same countries as our study. In a study on childhood adversity and schizophrenia which took place in two of the same provinces in South Africa as our study, Mall et al. (2020) found an odds of 2.21 (CI: 1.61–3.03) between cases and controls who had experienced ≥3 traumatic events after adjusting for age, sex, education, urbanicity, and locality. While Mall et al. (2020) restricted their study to trauma that occurred in childhood, our study captured lifetime trauma exposure, which could have influenced our different ORs. Mall et al. (2020) also utilized a different measure (the Childhood Trauma Questionnaire), which included 25 items and also collected how often a traumatic event occurred in comparison to the LEC-5’s 17 items, which only captured whether the event ever occurred. The fewer items participants could have endorsed in the LEC-5 and the absence of measuring repeated experiences of the same trauma type likely restricted the number of events reported by participants, potentially limiting the cumulative trauma burden in comparison to the Childhood Trauma Questionnaire.

As prior studies have hypothesized, there may be a dose-response relationship between experiencing multiple traumas and developing psychosis (Mall et al., 2020; Shevlin et al., 2007). Additionally, this cumulative burden likely includes traumas that both precede and follow from the psychosis. However, we do not have data to determine whether multiple traumas occurred before or after the onset of psychosis, and thus this should be interpreted with caution. Differences in cumulative burden were largely due to the significantly higher endorsement of sexual violence, physical violence, and network trauma by cases. For further description of these three trauma types, please see section 4.3 below.

In the Brant-Wald tests, we found the proportional odds model assumption in the ordered logit model was not met. However, we chose to retain the proportional odds rather than use partial proportional odds because it is not uncommon for the former to be violated in large sample sizes (Liu et al., 2023). Additionally, for the reasons mentioned previously regarding alignment with previous studies on cumulative trauma, we felt it was important to keep the ordered logit model.

4.3. Trauma types

4.3.1. Sexual violence

In our full sample, the odds of reporting sexual violence was almost double in cases than controls (AOR = 1.99, 95 % CI: 1.86–2.14). Both male and female cases reported significantly more exposure to sexual violence compared to male and female controls, respectively, though there was notably higher endorsement of sexual violence by female cases than all other subgroups (22.1 % by female cases, 12.9 % by female controls, 5.8 % by male cases, and 2.9 % by male controls).

These findings are in line with the broader literature. As previously described, sexual violence against populations with psychosis can be common, including during psychiatric hospitalization (Frueh et al., 2005). In addition, a study in rural Ethiopia by Ametaj et al. (2021) reported that caregivers and providers of people with severe mental illness (SMI, defined in the article as schizophrenia, bipolar disorder, and severe major depressive disorder) shared that people with SMI were at a higher risk of rape and sexual assault due to additional vulnerability of their illness than others in their community. In separate research in Uganda at one of the hospitals in our study, women with SMI reported a striking difference in sexual violence by non-partners than the general population in the 12 months prior to the study (Lundberg et al., 2015).

In addition, sexual abuse has been shown to be a major risk factor for the onset of psychosis (Morgan and Fisher, 2007; Read et al., 2005; Varese et al., 2012). The aforementioned case-control study of childhood trauma and schizophrenia in South Africa found cases with schizophrenia reported significantly higher exposure to childhood sexual abuse than controls (Mall et al., 2020). Likewise, a national survey in the United Kingdom (UK) found that child sexual abuse was the type of childhood adversity most highly associated with developing definite or probable psychosis; after controlling for depression, child sexual abuse was associated with an odds ratio of 7.4 (95 % CI: 3.6–15.2) with psychosis (Bebbington et al., 2004).

The high endorsement of sexual violence by female cases is consistent with previous findings. Though sexual violence was likely underreported across all groups in our study (Kennedy and Prock, 2018), we expected female participants to report a higher prevalence of sexual violence than male participants because women and girls have reported higher rates of sexual violence than men and boys historically (Barth et al., 2013; Dworkin et al., 2017).

4.3.2. Physical violence

Although physical violence was widely reported across the whole study sample, the adjusted odds of reporting exposure to physical violence was 68 % higher in cases than controls. Both male and female cases reported significantly more exposure to physical violence than male controls and female controls, respectively. While male cases endorsed the most exposure to physical violence (48.1 %), the adjusted odds of physical violence were also driven by the large difference in endorsement by female cases and controls (prevalence of 35.1 % vs. 23.8 %; AOR = 1.73, CI: 1.62–1.85).

This elevated prevalence of physical violence is in keeping with the literature. As mentioned previously, cases may be in unique situations where they are exposed to physical violence by family members and the community because of their condition. In recent research described above in Ethiopia, people with SMI and their caregivers reported that people with SMI commonly experienced physical violence from family members when they refused to be restrained in the course of their illness (Ametaj et al., 2021). In other parts of Ethiopia and in Ghana, people with schizophrenia commonly described family members beating them as punishment or for discipline (Asher et al., 2017; Read et al., 2009); in Ghana and in Uganda, physical assault against people with severe mental health conditions was also reported by pastors and traditional healers as a way to drive away evil spirits, which were considered to be the cause of mental illness (Read et al., 2009; Verity et al., 2021). In addition, as described previously from the United States, people with psychosis experienced physical violence during psychiatric care (Frueh et al., 2005). In separate research in South Africa at one of the same hospitals as our study, researchers found there were 3.8 patient assaults per month amongst inpatients (>80 % with schizophrenia, schizoaffective disorder, and bipolar disorder), with 2.7 assaults/month amongst men and 1 assault/month amongst women (Luckhoff et al., 2013).

In addition, physical assault is commonly reported in the literature on childhood trauma and psychosis. In the study on childhood trauma and schizophrenia in South Africa described previously, cases with schizophrenia reported significantly higher exposure to physical violence than controls, such as being hit so hard by a family member that they had to see a doctor, being hit so hard that it left bruises or marks, or being beaten “with a belt, board, cord, or some hard object” (Mall et al., 2020). Furthermore, Varese et al. (2012)’s meta-analysis of adverse childhood events found that a history of physical assault was associated with almost three times the odds of developing psychosis than controls and a study in the UK found that maternal physical abuse that started prior to age 12 had had the most robust association with adult psychosis of any type of childhood adversity (Fisher et al., 2010).

4.3.3. Network trauma

The adjusted odds of reporting exposure to network trauma (causing injury, harm, or death to someone else) was 52 % higher in cases than controls across the full sample. Though there are studies that have found that people with psychosis are more often perpetrators of violence than the general public (Coid et al., 2016; Whiting et al., 2022), there is also significant research showing that people with psychosis have higher rates of victimization than the general public (de Vries et al., 2019; Maniglio, 2009). Being a perpetrator and a victim are not mutually exclusive, however, and this is not possible to tease out without collecting further information from our participants. Furthermore, network trauma reflected inflicted harm, which may not have been intentional or malicious and may have been accidents or self-defense, thus making it difficult to interpret this trauma type. In our study, the higher endorsement of network trauma was driven primarily by cases from South Africa, and particularly by male cases. Due to stigma of reporting causing harm to others, network trauma was likely underreported in our sample. Given the small number of participants who endorsed this event, the AORs should be interpreted with caution.

It is notable that the most endorsed types of trauma by cases were all interpersonal in nature. Such findings are consistent with the high rates of interpersonal trauma experienced by people with psychosis, especially physical and sexual assault (Belete, 2017; Grubaugh et al., 2011; Mauritz et al., 2013; Zerihun et al., 2021). It has been posited that interpersonal trauma, particularly in childhood, might be related to developing psychosis later in life through a variety of psychological mechanisms including poor attachment, negative schema, dissociative symptoms, subjective loneliness, and impaired social relationships in adulthood (Degnan et al., 2022; Grady et al., 2024; Humphrey et al., 2022; Mauritz et al., 2013; Stain et al., 2013).

4.4. Cross-cultural differences and similarities

Although we saw similar trends in the prevalence, cumulative burden, and types of trauma endorsed between Ethiopia, Kenya, South Africa, and Uganda, we also saw differences in the levels of endorsement by country. Some researchers argue that recognition of potentially traumatic events and subsequent traumatic stress is not simply a reaction to objective experiences, but is shaped by the culture in which one exists and understood through what is deemed “traumatic” in their setting (Chentsova-Dutton and Maercker, 2019), which may influence how participants understood and reported traumatic events in the different countries. However, the similar trends may reflect commonalities across the countries in terms of historical influences and common types of trauma experienced in the region (Abrahams et al., 2014; Fanon, 2008; WHO, 2023).

There were also notable between-country differences of experiencing any trauma between East Africa and South Africa, ranging from a low of 55.1 % in Ethiopia to a high of 92.8 % in South Africa. South Africa’s history of segregation, violence, and state-sponsored oppression under apartheid and its legacy, followed by high rates of crime and socioeconomic disparities, may be reflected in these differences (Bond, 2004; Dubow, 2014; Halvorsen and Kagee, 2010). The high endorsement of trauma exposure could additionally reflect that participants from South Africa were more willing to disclose experiences of trauma. For example, South Africa has an operating nation-wide strategic plan against gender-based violence (GBV) and active campaigns to raise awareness about GBV (Interim Steering Committee of South Africa, 2020; Parliament of the Republic of South Africa, 2024). Such initiatives may increase people’s awareness of their own experiences of trauma and/or destigmatize people speaking openly about their experiences.

National demographic and health surveys (DHS) from Ethiopia, Kenya, and Uganda that collect similar metrics to each other reinforce there are between-country differences in East Africa. Population-representative household surveys with women aged 15–49 found that since age 15, 23 % experienced physical violence and 10 % experienced sexual violence in Ethiopia (Central Statistical Agency Ethiopia, 2017); 34 % experienced physical violence and 13 % experienced sexual violence in Kenya (Kenya National Bureau of Statistics, 2023); and 51 % experienced physical violence and 22 % experienced sexual violence in Uganda (Uganda Bureau of Statistics, 2018). Although the DHS statistics found a mixture of lower and higher prevalences of physical and sexual violence than the female cases in our study, the most striking difference was the endorsement of physical violence from the Ugandan DHS vs. Ugandan female cases in our study (51 % vs. 28.0 % respectively). It is unclear why there would be such a stark difference between the studies.

4.5. Limitations

Several limitations to these findings deserve emphasis. As with many studies that rely on self-report measures, there is the potential for recall bias. For both cases and controls, there is likely underreporting of intimate interpersonal trauma that participants may not feel comfortable sharing due to stigma, such as sexual assault (Kennedy and Prock, 2018). Cases and controls could also have been exposed to the same amount of trauma, but cases may have been more likely to report it. As cases are likely to be more distressed than controls, it is possible they may more readily “access” memories of negative experiences and share them during the study (Baldwin et al., 2019; Dalgleish and Werner-Seidler, 2014).

Prior studies often restrict their samples to first-episode psychosis, where temporality of exposure and outcome can be more clearly defined. In our case, since psychosis was the exposure, we benefited from including individuals with different illness durations. However, since we did not collect granular data on the timing, duration, and recurrence of each trauma type or the timing and duration of psychosis for cases we do not know if the traumatic events reported occurred before or after the onset of psychosis, hence an inability to establish causality. With this study design, we can only measure the magnitude of the associations.

There are potential factors that could have had an influence on the associations seen in this study that are risk factors for or associated with exposure to trauma and psychosis. One of the main challenges in this context is that we do not know as much about risk factors for trauma and psychosis as in some other regions. We do know that family history is a risk factor (Jester et al., 2023) and based on that and the fact that this was not controlled for, unmeasured confounding could explain at least some of the findings. Given the results of the e-values, however, this would be minimal. To the extent which unmeasured confounding could explain the results depends on the context. Furthermore, this paper is important because there is such limited information on trauma and psychosis in Africa and this work is the basis for other studies.

Some types of traumatic events were not likely being captured in the current measure. There may be trauma exposures specific to people with psychosis and to the general population in different regions within Ethiopia, Kenya, South Africa, and Uganda that we are missing, thus underestimating the true counts and interpretations about prevalence, cumulative burden, and types of trauma. For example, as referenced previously, Ametaj et al. (2021) identified locally relevant trauma types not typically captured in trauma exposure indexes such as chaining and marriage by abduction. Furthermore, there may be “stressful” experiences that do not align with the DSM-5’s conventions of traumatic events, but that are significant for people in the course of psychosis (Subramanian et al., 2017; Wang et al., 2021). While a strength of using the same measure is comparison across the four countries, further research should supplement this work with qualitative research that can collect culturally specific trauma types not captured here.

Our study may not be population representative and so inferences should not be made for the general population in these countries. However, our sample is strengthened by its large size of almost 43,000 participants and by the number of recruitment sites within each country.

5. Conclusion

Findings of this cross-country African study underscore the significant cumulative trauma load and exposure to various types of interpersonal violence (e.g., sexual, physical and network trauma) that people with psychosis may face. Disentangling the temporal relationships between trauma exposure and psychosis and developing interventions for underserved and culturally diverse populations is key to providing treatment to such groups going forward.