Skip to main content
NCBI home page
PLOS Medicine logoLink to PLOS Medicine
. 2025 Sep 26;22(9):e1004765. doi: 10.1371/journal.pmed.1004765

Incidence and prognostic factors of self-harm and subsequent unnatural death in South Africa: A cohort study

Veronika Whitesell Skrivankova 1, Stephan Rabie 2, Mpho Tlali 3, Naomi Folb 4,5, Chido Chinogurei 3, Sarah Bennett 4, Aimee Wesso 4, Yann Ruffieux 1, Morna Cornell 3, Soraya Seedat 6, Matthias Egger 3,7,8, Mary-Ann Davies 3, Gary Maartens 5, John Joska 2, Andreas D Haas 1,3,*
Editor: Charlotte Hanlon9
PMCID: PMC12503312  PMID: 41004534

Abstract

Background

Self-harm is a major public health concern globally and in South Africa. Individuals with a history of self-harm are at increased risk of unnatural death, including suicide and fatal accidents. This study investigates the incidence and predictors of self-harm and its role as a predictor for subsequent unnatural death.

Methods and findings

We conducted a cohort study using insurance claims and vital registration data from beneficiaries of South African medical insurance schemes (2011–2022), aged 10 years or older. We estimated the cause-specific cumulative incidences of healthcare encounters for intentional self-harm (International Classification of Diseases 10th Revision [ICD-10] codes X60-X84) and unnatural deaths (ICD-10 codes V01-Y98), using the Aalen–Johansen method. We assessed predictors of both outcomes using Cox regression. We followed 1,356,119 beneficiaries (median age 33 years, 52.2% female) for a median of 3 years, during which 7,510 (0.6%) had a healthcare encounter for self-harm. The 5-year cumulative incidence of self-harm ranged from 0.2% in males aged 10–14 to 2.1% in females aged 15–24. Sex, age, and mental disorders were strong predictors for self-harm, while HIV was a modest predictor. Among individuals who survived a self-harm event, the five-year cumulative incidence of subsequent unnatural death was 3.43% (95% CI [2.38, 4.76]) for males and 0.77% (95% CI [0.48, 1.19]) for females. Non-fatal self-harm was a strong predictor of subsequent unnatural death in both males (hazard ratio [HR] 7.03, 95% CI [5.27, 9.39]) and females (HR 4.63, 95% CI [3.00, 7.15]). The study’s main limitations include potential under-ascertainment of self-harm incidence due to reliance on routine data, and the unavailability of the exact cause of death, preventing analysis of suicide.

Conclusion

Self-harm is common among beneficiaries of South African private medical insurance, with the highest risk in young females and individuals with mental disorders. These groups may benefit from targeted screening and early intervention. Non-fatal self-harm was a strong predictor of subsequent unnatural death, underscoring the need for suicide-specific brief interventions for individuals presenting with self-harm.

Author summary

Why was this study done?

  • People who have harmed themselves in the past are more likely to do so again and are at higher risk of dying by suicide or other unnatural means.

  • With the right support from healthcare providers, self-harm can often be prevented.

  • Although self-harm is common in African countries, there is limited information to guide how health services can best support people affected by it.

What did the researchers do and find?

  • In a cohort of over 1.35 million members of South African medical insurance schemes, each year, around 2 in every 1,000 females and 1 in every 1,000 males accessed healthcare following self-harm.

  • The risk of self-harm is especially high among adolescents and young adults aged 15–24, and among people with mental health or substance use problems.

  • Individuals with a history of self-harm were at much greater risk of dying from unnatural causes, such as suicide, in the years following the self-harm event. The risk of dying from unnatural causes was particularly high among men.

What do these findings mean?

  • South Africa is facing a high burden of self-harm. Young adults, females, and individuals with mental health or substance use conditions are at increased risk and represent key populations for targeted screening and early intervention.

  • Individuals presenting for self-harm, especially males, are at high risk of dying from unnatural causes and may benefit from suicide-specific brief interventions, such as safety planning, motivational techniques, and structured follow-up care.

  • The main limitations of the study are that we may have missed some cases of self-harm because we relied on routine medical records, and that we could not examine suicide mortality because the exact cause of death was not available.

In a cohort study using registry data, Veronika Whitesell Skrivankova and colleagues investigate the incidence and prognostic factors of self-harm and subsequent unnatural death in South Africa.

Introduction

Intentional self-harm (hereafter referred to as self-harm) includes a spectrum of behaviors in which individuals deliberately harm themselves. The literature distinguishes three types of self-harm behaviors: non-suicidal self-injury, defined as deliberate, directly injurious behavior (e.g., cutting, or burning) without any intent to die; suicide attempts, defined as non-fatal potentially self-injurious behaviors with at least some intent to die; and suicide, defined as fatal self-directed injurious behavior with intent to die [1,2].

Self-harm contributes substantially to the global disease burden and places considerable strain on health systems, especially in low- and middle-income countries (LMICs) like South Africa, where self-harm rates are high [3,4] and resources for prevention and care are scarce [5,6]. The crude suicide rate in South Africa in 2021 was estimated at 35.4 per 100,000 population among males, 9.9 among females, placing the country among those with the highest rates globally [7]. Lifetime prevalence of suicide attempts in South Africa is 3.8% among females and 1.8% among males [8]. Non-suicidal self-injury, which predominantly affects adolescents and young adults [1,2], is also highly prevalent with lifetime estimates ranging from 33% to 56% [9,10].

Self-harm cases presenting to emergency departments in South Africa are predominantly episodes of self-poisoning without evidence of suicidal intent or suicide attempts. At hospitals in Cape Town and KwaZulu-Natal, poisoning, mainly with medication, accounted for 72%–80% of presentations, while highly lethal methods such as hanging or laceration were less common (6%–14%). Across all methods, most patients (66%–78%) reported no intent to die [11,12]. While non-suicidal self-injury is common among adolescents in South Africa, it is infrequently seen in hospital or emergency settings [11,12].

Self-harm behaviors are often recurrent and may escalate in severity [2,13]. Individuals with a history of self-harm are at increased risk of unnatural death including suicide and accidental death [14–17]. Non-suicidal self-injury has consistently been shown to predict suicidal ideation and suicide attempts [18]. Intentional self-poisoning, regardless of intent, is a strong predictor subsequent suicide [14]. The risk of recurrent suicidal behavior after a suicide attempt is well documented, with 16% reattempting and 3% dying by suicide within one year [13,15]. Beyond suicide, individuals with a self-harm history also face increased risks of accidental, alcohol-related, and drug-related mortality [16,17].

The World Health Organization (WHO) recommends early identification of individuals at risk of self-harm and suicide, and management, support, and monitoring of those with a history of self-harm as key effective suicide prevention strategies [19]. The organization urges countries to implement evidence-based strategies tailored to local contexts and informed by local data [19], to achieve the international target of reducing suicide mortality by one third from 2015 levels by 2030 [19,20]. However, current data on self-harm in South Africa are scarce, primarily derived from small, cross-sectional, single-center studies without control groups [11,12,21–23]. The existing studies have not estimated the incidence or predictors of self-harm and subsequent mortality. Such estimates are essential to support early identification of at-risk groups and to inform the development of evidence-based strategies for managing individuals presenting with self-harm, tailored to their risk of subsequent unnatural death.

To address these gaps, we analyzed longitudinal data from over 1 million beneficiaries of South African medical insurance schemes. We aimed to estimate self-harm incidence, identify its predictors, and evaluate non-fatal self-harm as a predictor of subsequent unnatural death. Our analyses were not designed to identify causal (etiologic) risk factors. Instead, we aimed to identify individual-level characteristics that serve as risk markers, indicating increased future self-harm or mortality risk, without necessarily implying causality [24].

Methods

Study design and participants

We conducted a cohort study using longitudinal reimbursement claims data from a South African medical insurance scheme (hereafter referred to as the medical scheme) and a private sector HIV disease management program (hereafter referred to as the HIV program). The design and objectives of this study align with prognostic factor research [24], as defined in the PROGnosis RESearch Strategy (PROGRESS) framework [25]. We refer to prognostic factors as ‘predictors’ throughout the manuscript, as study a general population of medical insurance beneficiaries, not individuals with a specific disease. We conducted two primary analyses: (1) the self-harm analysis, investigating the incidence and predictors of intentional self-harm, and (2) the mortality analysis, studying non-fatal self-harm as a predictor of subsequent unnatural death. We included individuals aged 10 years or older who were covered by the medical scheme (1 Jan 2011–30 Jun 2020) or enrolled in the HIV program (1 Jan 2011–1 Oct 2022) (Fig 1). We excluded HIV program members also included in the medical scheme cohort, and individuals with unknown sex. From mortality analysis, we excluded individuals not linkable to the vital registration system. We conducted analyses according to a pre-specified protocol (S1 Appendix), and list protocol deviations in S2 Appendix. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 STROBE Checklist). The Human Research Ethics Committee of the University of Cape Town (HREC REF Number: 84/2006) and the Cantonal Ethics Committee of the Canton of Bern granted (KEK-Nr: 150/14) permission to analyze data.

Fig 1. Study inclusion flowchart.

Fig 1

Open in a new tab

Data sources

We combined data from the medical scheme and HIV program to increase sample size, given the rarity of unnatural death after self-harm. The dataset included sociodemographic characteristics, reimbursement claims for inpatient and outpatient care with International Classification of Diseases 10th Revision (ICD-10) diagnoses, pharmacy claims, and laboratory results. These data were deterministically linked to vital registration records (1 Jan 2011–26 Jan 2021) from the National Population Register using national identifiers. Death registration is estimated to be 93% complete among adults and slightly lower among children [26,27]. Vital registration data included date of death, and cause of death (natural, unnatural, or unknown).

Setting

The medical scheme is open to all South African residents but requires paid membership. Most residents do not have private insurance and rely on public sector services [28]. The HIV program includes individuals with HIV whose medical schemes have contracted the program to manage their HIV care. Medical scheme beneficiaries generally have higher socio-economic status and may face fewer psychosocial stressors of self-harm than individuals without insurance [29]. In addition to medical scheme benefits, all beneficiaries are entitled to Prescribed Minimum Benefits (PMBs), which include essential services. For self-harm, PMBs cover up to 3 days of inpatient care or six outpatient visits per episode, including treatment of injuries, complications, and underlying conditions [30].

Outcomes

In the self-harm analysis, the outcome was an individual’s first observed healthcare encounter for intentional self-harm (ICD-10 codes X60-X84) [31]. Repeated self-harm events were not considered. In the mortality analysis, the outcome was death from unnatural causes (ICD-10 codes V01-Y98), including suicides and fatal accidents.

Predictors

We defined predictors as individual-level characteristics, measured before the outcome among people with a given start point, that indicate increased risk of future self-harm or mortality, without necessarily implying causality [24,32]. In the self-harm analysis, we included sex, age group (10–14, 15–24, 25–39, 40+  years), HIV status, and mental disorder diagnoses as potential predictors. We defined HIV status using antiretroviral therapy codes (excluding prophylaxis), HIV-related ICD-10 diagnoses (B20–24, F02.4, R75, Z21), relevant lab tests (e.g., viral load, CD4), or enrollment in the HIV program. Mental disorders were grouped by ICD-10 codes into organic mental disorders (F00–09), substance use disorders (F10–16, F18–19), psychotic disorders (F20–29), bipolar disorder (F31), depression (F32–34.1), anxiety (F40–48), personality disorders (F60–69), and other mental disorders (F50–59, F70–99). For individuals with both depression and bipolar disorder, only the bipolar diagnosis was considered from the time of diagnosis onward.

In the mortality analysis, we considered a healthcare encounter for non-fatal self-harm as the main predictor. As a secondary predictor, we classified self-harm encounters by lethality of the self-harm method, based on evidence that individuals who use highly lethal methods at their index event are at greatest risk of suicide [33]. High lethality methods included hanging (X70), gas poisoning (X67), jumping (X80–X81), firearms (X72–X75), and drowning (X71). Less lethal methods included drug or chemical poisoning (X60–X66, X68–X69), fire-related injuries (X76–X77), motor vehicle accidents (X82), sharp or blunt object injuries (X78–X79), and other or unspecified methods (X83–X84) [34].

Statistical analysis

Baseline was defined as the latest of medical scheme or HIV program enrollment, the individual’s 10th birthday, or January 1, 2011. Age group, non-fatal self-harm, HIV status, and mental disorders were modeled as time-varying covariates, with individuals considered exposed from the first recorded occurrence. Data preparation and statistical analyses were done using R version 4.3.1 and Stata version 18.

Self-harm analysis.

Individuals were followed from baseline until the earliest of the end of insurance coverage, medical scheme database closure, death, or first observed self-harm encounter. We calculated crude incidence rates per 100,000 person-years by sex and age group with 95% confidence intervals (CIs) based on the Poisson distribution. Using Royston-Parmar flexible parametric survival models [35], we estimated self-harm incidence as a continuous function of age, using a spline with 5 degrees of freedom, stratified by sex, and HIV status. We estimated cause-specific cumulative incidence of self-harm using the Aalen–Johansen estimator, treating death as a competing event and stratifying by age group and sex [36]. Individuals exited the risk set upon first self-harm. We used Cox regression to estimate age- and sex-adjusted and fully adjusted hazard ratios (HRs) for predictors of self-harm, treating death as a censoring event. Fully adjusted models included age, sex, HIV, and mental disorders.

Mortality analysis.

Individuals were followed from baseline until the earliest of the end of insurance coverage, medical scheme database closure, vital registration database closure, or death. We estimated the cumulative incidence of unnatural death using the Aalen–Johansen estimator, stratified by sex and self-harm history, with natural and unknown causes treated as competing events. Using Cox models, we estimated cause-specific age-adjusted and fully adjusted HRs to evaluate non-fatal self-harm as a predictor for subsequent unnatural death, treating natural or unknown causes as censoring events. Age-adjusted models used age as the underlying time scale and were stratified by sex, with separate models fitted for males and females. Fully adjusted models additionally controlled for mental disorders.

Post-hoc analyses.

We conducted the following additional post-hoc analyses: First, we assessed whether the lethality of the method used at the index self-harm event predicted subsequent unnatural death. Second, we used flexible parametric survival models to estimate age-adjusted, time-varying HRs for non-fatal self-harm. Third, we fitted a combined Cox model including both males and females, using age as the time scale and adjusting for sex to estimate association between prior non-fatal self-harm and unnatural death across sexes. In addition, we fitted the same model with an interaction term between sex and prior non-fatal self-harm to formally test for effect modification. Finally, we used descriptive statistics to compare healthcare utilization prior to self-harm encounters with that of matched controls prior to hypothetical encounter dates. For this analysis, each self-harm case was matched to two randomly selected controls who had no self-harm encounter. Matching criteria included the same baseline year, age group (10–14, 15–19, 20–39, 40–59, 60+), and duration of follow-up (<2 years, 2–4 years, 4–6 years, 6–8 years, or 8+  years). The hypothetical event date for controls was set as their baseline date plus the interval from baseline to the self-harm encounter of the matched case. We calculated standardized differences to assess group differences, interpreting values <0.1 as not meaningful, >0.2 as moderate, and >0.5 as large [37]. Further details are provided in Text A in S3 Appendix.

Sensitivity analyses.

In post-hoc sensitivity analyses, we repeated all primary analyses in the medical scheme cohort only, to assess the impact of oversampling individuals with HIV due to the inclusion of HIV program data. To assess generalizability to individuals not linked to the vital registration system, we compared those included versus excluded from the mortality analysis. Self-harm incidence was compared using Cox regression adjusted for age, sex, HIV status, and mental disorders.

Results

Characteristics of the study population

We followed 1,356,119 individuals (Fig 1) (709,073 females, 647,046 males) for a median of 3.3 years (interquartile range [IQR] [1.2, 6.8]). Median age at baseline was 33 years (IQR 21–46). By the end of follow-up, 177,193 (13.1%) had HIV diagnosis and 348,157 (25.7%) had a mental health diagnosis, most commonly anxiety (16.1%) or depression (13.5%) (Table 1).

Table 1. Characteristics of the study population by sex.

Female Male Total
N = 709,073
(52.2%)		 N = 647,046
(47.8%)		 N = 1,356,119 (100.0%)
Characteristics at baseline
 Age, years
  10–14 104,952 (14.8%) 107,055 (16.5%) 212,007 (15.6%)
  15–24 111,327 (15.7%) 95,318 (14.7%) 206,645 (15.2%)
  25–39 237,889 (33.5%) 204,255 (31.6%) 442,144 (32.6%)
  40+ 254,905 (35.9%) 240,418 (37.2%) 495,323 (36.5%)
  Median (IQR) 33 (21, 46) 33 (20, 46) 33 (21, 46)
Characteristics at the end of follow-up
 HIV status
  HIV–negative 606,426 (85.5%) 572,500 (88.5%) 1,178,926 (86.9%)
  HIV–positive 102,647 (14.5%) 74,546 (11.5%) 177,193 (13.1%)
 Any mental health disorders 206,878 (29.2%) 141,279 (21.8%) 348,157 (25.7%)
  Organic mental disorder 5,972 (0.8%) 4,863 (0.8%) 10,835 (0.8%)
  Substance use disorder 2,663 (0.4%) 6,540 (1.0%) 9,203 (0.7%)
  Psychotic disorder 2,795 (0.4%) 2,687 (0.4%) 5,482 (0.4%)
  Bipolar disorder 13,895 (2.0%) 7,933 (1.2%) 21,828 (1.6%)
  Depression 116,173 (16.4%) 66,538 (10.3%) 182,711 (13.5%)
  Anxiety disorder 135,820 (19.2%) 82,761 (12.8%) 218,581 (16.1%)
  Personality disorder 920 (0.1%) 804 (0.1%) 1,724 (0.1%)
  Other mental health disorders 33,601 (4.7%) 34,136 (5.3%) 67,737 (5.0%)
 Self-harm encounters
  Event of undetermined intent 5,255 (0.7%) 7,777 (1.2%) 13,032 (1.0%)
  Intentional self-harm 5,208 (0.7%) 2,302 (0.4%) 7,510 (0.6%)
 Follow-up time, years
  Median (IRQ) 3.4 (1.2,7.0) 3.3 (1.2,6.7) 3.3 (1.2,6.8)
 Linked to National Population Register 651,866 (91.9%) 604,822 (93.5%) 1,256,688 (92.7%)
 Died during follow-up* 17,495 (2.7%) 22,034 (3.6%) 39,529 (3.1%)
  Natural cause* 15,921 (2.4%) 18,349 (3.0%) 34,270 (2.7%)
  Unnatural cause* 923 (0.1%) 2,759 (0.5%) 3,682 (0.3%)
  Unknown cause* 651 (0.1%) 926 (0.2%) 1,577 (0.1%)
Open in a new tab

Data are n (%) unless otherwise stated.

*Among those linked to National Population Register.

Abbreviation: IQR, interquartile range.

Characteristics of individuals with self-harm

Among 7,510 individuals (0.6%) with a self-harm encounter (5,208 females, 2,302 males), most were hospital admissions (80%). Less lethal methods, mainly poisoning, accounted for 97.8% of cases; highly lethal methods were rare (2.2%). Median age at self-harm encounter was 31 years (IQR [20, 40]). Half of the cases (50.4%) had received a mental health diagnosis before the self-harm encounter, with another 23.7% being diagnosed within 7 days after the encounter (Table 2).

Table 2. Characteristics of individuals with healthcare encounter for self-harm.

Female Male Total
N = 5,208
(69.3%)		 N = 2,302
(30.7%)		 N = 7,510
(100.0%)
Self-harm encounter type
 Hospital admission 4,313 (82.8%) 1,699 (73.8%) 6,012 (80.1%)
 Outpatient encounter 895 (17.2%) 603 (26.2%) 1,498 (19.9%)
Self-harm method
 Highly lethal 57 (1.1%) 105 (4.6%) 162 (2.2%)
  Self-poisoning by exposure to gasses 5 (0.1%) 22 (1.0%) 27 (0.4%)
  Self-harm by hanging strangulation and suffocation 16 (0.3%) 30 (1.3%) 46 (0.6%)
  Self-harm by drowning and submersion 7 (0.1%) 1 (0.0%) 8 (0.1%)
  Self-harm by firearms or explosives 17 (0.3%) 38 (1.7%) 55 (0.7%)
  Self-harm by jump. from heights, or before moving objects 12 (0.2%) 14 (0.6%) 26 (0.3%)
 Less lethal 5,151 (98.9%) 2,197 (95.4%) 7,348 (97.8%)
  Self-poisoning by an exposure to substances excl. gasses 4,901 (94.1%) 1,940 (84.3%) 6,841 (91.1%)
   Nonopioid analgesics, antipyretics and antirheumatics 1,062 (21.9%) 296 (15.4%) 1,358 (20.0%)
   Antiepileptic, sedative-hypnotic, antiparkinsonism and other psychotropic drugs 1,098 (22.6%) 379 (19.7%) 1,477 (21.8%)
   Narcotics and psychodysleptics 133 (2.7%) 90 (4.7%) 223 (3.3%)
   Other drugs acting on the autonomic nervous system 139 (2.9%) 55 (2.9%) 194 (2.9%)
   Other and unspecified drugs, medicaments and biological substances 1,988 (40.9%) 765 (39.7%) 2,753 (40.6%)
   Alcohol 88 (1.8%) 102 (5.3%) 190 (2.8%)
   Organic solvents and halogenated hydrocarbons 65 (1.3%) 38 (2.0%) 103 (1.5%)
   Pesticides 107 (2.2%) 94 (4.9%) 201 (3.0%)
   Other and unspecified chemicals and noxious substances 178 (3.7%) 107 (5.6%) 285 (4.2%)
  Self-harm by steam, hot vapors, and hot objects 5 (0.1%) 3 (0.1%) 8 (0.1%)
  Self-harm by sharp object 98 (1.9%) 92 (4.0%) 190 (2.5%)
  Self-harm by blunt object 28 (0.5%) 57 (2.5%) 85 (1.1%)
  Self-harm by smoke, fire and flames 17 (0.3%) 11 (0.5%) 28 (0.4%)
  Self-harm by crashing of motor vehicle 3 (0.1%) 5 (0.2%) 8 (0.1%)
  Self-harm by other or unspecified means 99 (1.9%) 89 (3.9%) 188 (2.5%)
 Time from baseline to first self-harm encounter, years
  Median (IRQ) 2.6 (1.1, 4.8) 2.7 (1.2, 4.8) 2.6 (1.1, 4.8)
Age at self-harm event, years
 10–14 259 (5.0%) 83 (3.6%) 342 (4.6%)
 15–24 1,658 (31.8%) 629 (27.3%) 2,287 (30.5%)
 25–39 2,096 (40.2%) 872 (37.9%) 2,968 (39.5%)
 40+ 1,195 (22.9%) 718 (31.2%) 1,913 (25.5%)
 Median (IQR) 30 (20, 39) 33 (21, 43) 31 (20, 40)
Mental health diagnoses before the self-harm encounter*
 Any mental health disorders 2,652 (50.9%) 1,130 (49.1%) 3,782 (50.4%)
  Organic mental disorder 57 (1.1%) 43 (1.9%) 100 (1.3%)
  Substance use disorder 94 (1.8%) 180 (7.8%) 274 (3.6%)
  Psychotic disorder 66 (1.3%) 40 (1.7%) 106 (1.4%)
  Bipolar disorder 501 (9.6%) 207 (9.0%) 708 (9.4%)
  Depression 1,944 (37.3%) 746 (32.4%) 2,690 (35.8%)
  Anxiety disorder 1,692 (32.5%) 636 (27.6%) 2,328 (31.0%)
  Personality disorder 17 (0.3%) 12 (0.5%) 29 (0.4%)
  Other mental health disorders 451 (8.7%) 235 (10.2%) 686 (9.1%)
New mental health diagnoses within 7 days of self-harm encounter**
 Any mental health disorders 1,297 (24.9%) 485 (21.1%) 1,782 (23.7%)
  Organic mental disorder 35 (0.7%) 27 (1.2%) 62 (0.8%)
  Substance use disorder 67 (1.3%) 95 (4.1%) 162 (2.2%)
  Psychotic disorder 25 (0.5%) 26 (1.1%) 51 (0.7%)
  Bipolar disorder 176 (3.4%) 75 (3.3%) 251 (3.3%)
  Depression 1,461 (28.1%) 521 (22.6%) 1,982 (26.4%)
  Anxiety disorder 321 (6.2%) 124 (5.4%) 445 (5.9%)
  Personality disorder 11 (0.2%) 9 (0.4%) 20 (0.3%)
  Other mental health disorders 25 (0.5%) 15 (0.7%) 40 (0.5%)
Open in a new tab

Data are n (%) unless otherwise stated.

*Assessed on the day before the self-harm encounter,

**New mental health diagnoses were defined as those received within 7 days after the self-harm encounter, with no prior diagnosis in that category.

Abbreviations: IQR, interquartile range.

Self-harm incidence

The incidence of self-harm encounters per 100,000 person-years was 183 (95% CI [178, 188]) in females, and 90 (95% CI [87, 94]) in males, with variation by age groups (Table A in S3 Appendix). Rates peaked at age 19 in females (522, 95% CI [493, 553]) and at 21 in males (208, 95% CI [191, 226]), then declined (Fig 2A). After 5 years, the cumulative incidence of self-harm ranged from 0.2% in males 10–14 years to 2.1% in females aged 15–24 (Fig 2B and Table B in S3 Appendix). Incidence rates were slightly higher among people with HIV under 40 years of age compared to their counterparts without HIV (Fig A in S3 Appendix).

Fig 2. Incidence of self-harm by age and sex.

Fig 2

Open in a new tab

Panel A presents the incidence per 100,000 person-years as a continuous function of age, for females and males. Panel B shows the Aalen–Johansen estimate of cause-specific cumulative incidence over time since baseline or age-group change, stratified by sex and age group. Individuals were censored at the time of their first documented self-harm event. Repeated self-harm events were not considered.

Predictors of self-harm

In fully adjusted models, younger age and female sex were strong predictors of self-harm: compared to those aged ≥40, individuals aged 15–24 had over 5-fold higher risk (adjusted hazard ratio [aHR] 5.43, 95% CI [5.10, 5.78]), and those aged 25–39 had 2.6-fold higher risk (aHR 2.57, 95% CI [2.42, 2.72]); females had nearly twice the risk of males (aHR 1.79, 95% CI [1.71, 1.88]). HIV status was a modest predictor (aHR 1.19, 95% CI [1.12, 1.26]). All mental disorders were predictors of self-harm, with considerable variation in effect size, with aHRs ranged from 1.20 (95% CI [0.98, 1.47]) for psychotic disorders to 8.70 (95% CI [7.97, 9.52]) for bipolar disorder. Adjusted HRs were higher across all disorders in age- and sex-adjusted models (Fig 3).

Fig 3. Predictors of self-harm.

Fig 3

Open in a new tab

The figure shows age- and sex-adjusted and fully adjusted hazard ratios for predictors of self-harm. Individuals were censored at the time of their first documented self-harm event. Repeated self-harm events were not considered. The error bars represent 95% confidence intervals. Fully adjusted models controlled for age group, sex, HIV, and all mental disorders. * For individuals diagnosed with both depression and bipolar disorder, only the bipolar disorder diagnosis was considered from the date of bipolar diagnosis onwards. Abbreviations: CI, confidence intervals; HR, hazard ratio.

Health care utilization before self-harm

Two-thirds of beneficiaries (69.9%, 5,158/7,379) visited an inpatient or outpatient setting for any reason within 30 days before the self-harm event, including 19.3% who were seen for a mental health condition, and 21.6% who claimed psychiatric medication. All estimates were higher among individuals with self-harm encounter compared to matched controls (Table C in S3 Appendix).

Mortality after non-fatal self-harm

The mortality analysis included 1,256,623 individuals, excluding 99,431 who were not linked to the vital registration system and 65 who had no person-time at risk in the mortality analysis. Among the included individuals, 6,863 had a non-fatal self-harm encounter (4,756 [69.3%] female and 2,107 [30.7%] male), and 1,249,760 had no prior non-fatal self-harm encounter. In the group with a non-fatal self-harm encounter, 68 subsequent unnatural deaths occurred during follow-up (Fig 1). Two years after a non-fatal self-harm encounter, the cumulative incidence of unnatural death was 0.35% (95% CI [0.20, 0.60]) in females and 1.90% (95% CI [1.32, 2.63]) in males, compared with 0.06% (95% CI [0.05, 0.06]) in females and 0.21% (95% CI [0.20, 0.23]) in males without prior self-harm (Fig 4A and Table D in S3 Appendix). The cumulative incidence of unnatural death within two years after a non-fatal self-harm encounter reached 3.84% (95% CI [1.01, 9.89]) in males and 4.26% (95% CI [0.76, 13.00]) in females among those who used highly lethal methods (Fig 4B and Table D in S3 Appendix).

Fig 4. Cumulative incidence of unnatural death by sex and self-harm history.

Fig 4

Open in a new tab

Aalen–Johansen estimate with 95% confidence intervals for cause-specific cumulative incidence of unnatural death over time since baseline or index non-fatal intentional self-harm (NF-SH) encounter, stratified by sex and (A) NF-SH, (B) self-harm method at index encounter. Abbreviations: NF-SH, non-fatal self-harm.

Non-fatal self-harm as predictor for unnatural death

Health care encounters for non-fatal self-harm were strongly predictive of subsequent mortality from unnatural causes in both sexes. Among females, the risk of unnatural death following a non-fatal self-harm encounter was almost five times higher than in females of the same age without a prior non-fatal self-harm encounter (HR 4.63, 95% CI [3.00, 7.15]; Fig 5A). Among males, the risk was seven times higher (HR 7.03, 95% CI [5.27, 9.39]; Fig 5A). In both sexes, HRs were attenuated in fully adjusted models (Fig 5A). Time-varying HRs for non-fatal self-harm as a predictor of unnatural death are shown in Fig B in S3 Appendix. In a combined model including both sexes and adjusting for age and sex, non-fatal self-harm was associated with a 6-fold higher risk of unnatural death (HR 6.01, 95% CI [4.73, 7.65]). The interaction between sex and prior self-harm showed weak statistical evidence for effect modification (p = 0.079).

Fig 5. Hazard ratios for unnatural death comparing individuals with and without a history of non-fatal self-harm by sex.

Fig 5

Open in a new tab

All models were fitted separately for males and females. Age-adjusted estimates control for age using it as the timeline in Cox model. Fully adjusted models additionally control for HIV status, mental health diagnoses. Error bars represent 95% confidence intervals. Panel A shows hazard ratios for a previous non-fatal intentional self-harm (NF-SH) encounter. Panel B shows hazard ratios by method of self-harm (highly lethal vs. less lethal) used during the index self-harm event. Individuals with no prior NF-SH encounter serve as the reference group in both analyses. Abbreviations: NF-SH, non-fatal self-harm; CI, confidence intervals; HR, hazard ratio.

Self-harm method as predictor for unnatural death

Use of highly lethal methods at the index self-harm encounter was a very strong predictor of subsequent unnatural death. Compared with individuals of the same sex and age without prior non-fatal self-harm, the age-adjusted HR was 29.1 (95% CI [7.27, 117.0]) for females and 11.20 (95% CI [3.59, 34.6]) for males (Fig 5B). In both sexes, HRs were attenuated in fully adjusted models (Fig 5B).

Sensitivity analyses

Compared with the medical scheme cohort, the HIV program cohort included a higher proportion of females and fewer individuals younger than 25 years (Table E in S3 Appendix). Main findings remained consistent in sensitivity analyses restricted to the medical scheme cohort, although the strength of associations varied somewhat (Text B and Figs C–F in S3 Appendix). One notable divergence was observed in the analysis of self-harm method as a predictor of unnatural death, where findings were not robust in sensitivity analyses: associations for highly lethal methods were stronger in females but attenuated and not statistically significant in males (Text B and Fig G in S3 Appendix).

Individuals excluded from mortality analysis were younger, had slightly higher HIV prevalence, lower prevalence of depression and anxiety (Text C and Table F in S3 Appendix), and a higher incidence of self-harm (aHR 1.25, 95% CI [1.15, 1.36]) than those included.

Discussion

In this analysis of South African medical insurance data, self-harm incidence was high, predominantly involving self-poisoning, with few cases using highly lethal methods. Key predictors of self-harm included young age, female sex, and mental health diagnoses. Non-fatal self-harm was a strong predictor of subsequent unnatural death in both sexes. As the analyses were prognostic in nature, the identified predictors should not be interpreted as causal but as risk markers to identify individuals at elevated risk.

Self-harm incidence rates in this study (females: 183; males: 90 per 100,000 person-years) were twice as high as 2021 estimates for LMICs (females: 91; males: 42 per 100,000), aligning more closely with rates reported in high-income countries (females: 143; males: 112 per 100,000) [3,38]. Consistent with other studies [39,40], incidence peaked among females aged 15–24. Our findings also align with limited South African research suggesting a higher burden of non-fatal self-harm among females [41], underscoring the substantial risk in young women.

Mental disorders were strong predictors of self-harm. Half of the self-harm cases in our study had a pre-existing mental health diagnosis, and 24% were newly diagnosed within 7 days post-incident, suggesting previously unrecognized illness. The psychiatric morbidity observed in this study is slightly higher than in other LMICs, where fewer than half of those engaging in suicidal behavior had a psychiatric disorder [42]. In contrast, in Western countries, 90% of hospital-presenting self-harm cases had a mental disorder [43]. Growing evidence from LMICs highlights the role of socioeconomic and psychosocial stressors in self-harm [21,22,44], indicating that prevention strategies must address both mental illness and psychosocial factors [40,45,46].

Over two-thirds of individuals who engaged in self-harm had outpatient care contact in the 30 days preceding the episode, and one-fifth received mental healthcare during this period. These findings align with South African data showing that many individuals seek care shortly before suicide [47], underscoring outpatient care as a key setting for prevention. Nurses, counselors, and social workers are well placed to screen for suicidality but require training and supervision to intervene effectively without overburdening the system [48]. Estimates from our study on predictors of self-harm could help inform targeted screening approaches.

Non-fatal self-harm was a strong predictor of subsequent unnatural death in both sexes. Due to non-proportional hazards, the HRs for non-fatal self-harm should be interpreted as weighted averages of time-varying effects over the follow-up period [45]. We present two adjusted analyses for this predictor which should be interpreted differently based on their adjustment sets. The sex-stratified, age-adjusted analysis indicates the excess risk of unnatural death among individuals with a history of self-harm compared to general population controls of the same age and sex, without adjusting for factors such as mental disorders that form the prognostic profile of people who engage in self-harm. We consider this analysis the most informative, as it captures the overall excess mortality burden associated with self-harm, reflecting the real-world distribution of suicide risk factors faced by affected individuals. This includes a high prevalence of psychiatric illness, affecting almost 80% of our sample, along with a presumably high burden of other psychosocial risk factors, somatic illnesses, and genetic predispositions. In fully adjusted models, additionally accounting for mental disorders, non-fatal self-harm remained a strong predictor, indicating elevated mortality risk even among self-harm survivors without psychiatric conditions. These findings highlight the urgency of implementing indicated interventions, such as safety planning [49–51] and suicide-specific brief therapies [52,53], as recommended by WHO [19], for all individuals with a history of self-harm, including those without mental disorders. Immediate post-discharge care is especially critical, given the high risk of reattempt and suicide during this period [13,15].

Our study population included privately insured individuals with higher socioeconomic status and better access to healthcare than the uninsured population relying on South Africa’s public sector, who may face greater psychosocial stressors [29]. While self-harm and suicide incidence may be expected to be higher in the public sector, South African evidence suggests that socioeconomic status indicators such as low education are not consistently associated with suicidal behavior [8], supporting the generalizability of our findings. Predictors of self-harm are also generally consistent across public- and private-sector settings. In a study conducted in public primary care clinics in deprived neighborhoods in Cape Town, we found high rates of recurrent suicidal behavior—supporting our finding that non-fatal self-harm is an important risk marker—and strong associations between self-harm, female sex, and mental disorders [54]. Notably, HIV was not associated with self-harm in this study, and similarly, our current analysis found only a weak association, in contrast to findings from high-income countries, where HIV is associated with a 4-fold increase in suicide risk [55].

This study relies on routine data, which has inherent limitations. First, intentional self-harm could be underestimated for several reasons: we could not consider recurrent self-harm episodes; providers might be avoiding self-harm codes to reduce stigma; patients may underreport self-harm due to fears of stigmatization or involuntary admission [56]; and individuals who did not seek care after self-harm and those who died before accessing care are not captured. Second, mental health and substance use disorder diagnoses were based on claims data and exclude undiagnosed cases. Depression (13%) and anxiety (16%) appeared well ascertained, but substance use disorders were diagnosed in less than 1%—well below expected rates [57,58]. Estimates for substance use should therefore be interpreted with caution, as they likely reflect a small, potentially selected group with more severe, diagnosed conditions. Third, the exact cause of death was not available, preventing an analysis of suicides after non-fatal self-harm. Instead, we examined deaths from unnatural causes, a composite endpoint that includes suicide, accidents, and homicide. While this is a limitation, it also offers advantages: unnatural deaths are more frequent, improving statistical power, and represent a meaningful outcome for prevention. They are also more reliably recorded than specific causes such as suicide [26,59]. Fourth, we combined data from the medical scheme and HIV program to increase sample size, resulting in oversampling of individuals with HIV, which may have introduced bias. Sensitivity analyses restricted to medical scheme data showed no minimal impact on self-harm incidence, its predictors, or its association with subsequent unnatural death, suggesting that oversampling did not materially affect the findings. However, in males, the association between highly lethal methods and unnatural death was not robust in sensitivity analysis, likely due to small numbers in the highly lethal exposure group. We therefore consider this finding exploratory and recommend replication. Nonetheless, results from the primary analysis are clinically plausible and consistent with a Swedish study [31], supporting their credibility. Fifth, the mortality analysis relied on linked vital registration data, and 7% of beneficiaries who could not be linked were excluded. These individuals differed from those included, which may limit generalizability. However, given that they represented a small proportion of the cohort, the impact on findings is likely minimal. Notably, the excluded group had a higher incidence of self-harm and may have been at greater risk of unnatural death, potentially biasing the association towards the null. We therefore consider our estimates conservative.

In conclusion, this study documents a high incidence of self-harm within South Africa’s private healthcare sector and provides much-needed local data to support the implementation of context-specific strategies prevention strategies. We identify strong predictors of self-harm that can guide targeted screening strategies, which could be implemented in general medical outpatient settings frequently visited in the weeks preceding self-harm—offering a critical opportunity for early detection and proactive management. Our finding that non-fatal self-harm is a strong risk marker for subsequent unnatural death highlights the need for indicated suicide prevention strategies to prevent recurrent self-harm and suicide in clearly defined high-risk populations.

Declarations ethical considerations

The Human Research Ethics Committee of the University of Cape Town, South Africa and the Cantonal Ethics Committee, Bern, Switzerland, granted ethical permission for the analysis. Beneficiaries of the medical scheme or their guardians consented to the use of their data in research.

Declaration of generative AI and AI-assisted technologies in the editing process

During the preparation of this work, we used ChatGPT-4o for proofreading and editing the manuscript to enhance clarity, conciseness, and flow of the text. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

Supporting information

S1 Strobe Checklist. STROBE, Strengthening the Reporting of Observational Studies in Epidemiology.

https://www.strobe-statement.org/.

(PDF)

pmed.1004765.s001.pdf (122.4KB, pdf)
S1 Appendix. Study protocol.

(PDF)

pmed.1004765.s002.pdf (200.9KB, pdf)
S2 Appendix. Protocol deviations.

(PDF)

pmed.1004765.s003.pdf (90.1KB, pdf)
S3 Appendix

Table A. Incidence rates of self-harm encounters per 100,000 person-years, stratified by sex, for different age groups and overall. Table B. Cumulative incidence (%) and 95% confidence intervals of self-harm, stratified by sex, for different age groups and overall. Table C. Healthcare utilization before self-harm events compared with matched controls prior to hypothetical matched event dates. Table D. Cumulative incidence (%) and 95% confidence intervals of unnatural death, stratified by sex and self-harm method, and overall. Table E. Characteristics of the study population by cohort. Table F. Characteristics of individuals included and excluded from the mortality analysis. Fig A. Incidence of self-harm by age, sex, and HIV status. Incidence per 100,000 person-years in general population as a continuous function of age, stratified by sex and HIV status. Individuals were censored at the time of their first documented self-harm event. Repeated self-harm events were not considered. Fig B. Time-varying hazard ratios for unnatural death by sex, comparing individuals with and without prior non-fatal self-harm encounter. Hazard ratios are plotted as a continuous function of analysis time, defined as time since the index self-harm event for individuals with self-harm or baseline for those without self-harm event. Separate models were fitted by sex and adjusted for age. Shaded areas indicate 95% confidence intervals. The dashed horizontal line represents a hazard ratio of 1, indicating no difference in risk. Fig C. Incidence of self-harm by age and sex, restricted to medical scheme data. Incidence per 100,000 person-years in general population as a continuous function of age, stratified by sex. Individuals were censored at the time of their first documented self-harm event. Repeated self-harm events were not considered. Fig D. Cumulative incidence of self-harm by sex and age group, restricted to medical scheme data. Aalen–Johansen estimate with 95% confidence intervals for cause-specific cumulative over time since baseline or age-group change, stratified by sex and age group. Individuals were censored at the time of their first documented self-harm event. Repeated self-harm events were not considered. Fig E. Predictors of self-harm, restricted to medical scheme data. The figure shows age- and sex-adjusted and fully adjusted hazard ratios for predictors of self-harm. Individuals were censored at the time of their first documented self-harm event. Repeated self-harm events were not considered. The error bars represent 95% confidence intervals. Fully adjusted models controlled for age group, sex, HIV, and all mental disorders. * For individuals diagnosed with both depression and bipolar disorder, only the bipolar disorder diagnosis was considered from the date of bipolar diagnosis onwards. Abbreviations: CI, confidence intervals; HR, hazard ratio. Fig F. Hazard ratios for unnatural death comparing individuals with and without a history of non-fatal self-harm by sex, restricted to medical scheme data. All models were fitted separately for males and females. Age-adjusted estimates control for age using it as the timeline in Cox model. Fully adjusted models additionally control for HIV status, mental health diagnoses. Error bars represent 95% confidence intervals. Abbreviations: NF-SH, non-fatal self-harm; CI, confidence intervals; HR, hazard ratio. Fig G. Hazard ratios for self-harm method as a predictor of unnatural death, by sex, restricted to medical scheme data. All models were fitted separately for males and females. Hazard ratios compare individuals who used highly lethal or less lethal methods at their index self-harm encounter with individuals without a history of non-fatal self-harm (NF-SH). Age-adjusted estimates control for age by using it as the timeline in the Cox model. Fully adjusted models additionally control for HIV status and mental health diagnoses. Error bars represent 95% confidence intervals. Abbreviations: NF-SH, non-fatal self-harm; CI, confidence intervals; HR, hazard ratio.

(DOCX)

pmed.1004765.s004.docx (413.3KB, docx)

Abbreviations

aHR

adjusted hazard ratio

CI

confidence intervals

HR

hazard ratio

ICD-10

International Classification of Diseases 10th Revision

IQR

interquartile range

LMICs

low- and middle-income countries

NF-SH

non-fatal self-harm

PMBs

Prescribed Minimum Benefits

STROBE

Strengthening the Reporting of Observational Studies in Epidemiology

WHO

World Health Organization

Data Availability

Data were obtained from the International epidemiology Databases to Evaluate AIDS–Southern Africa (IeDEA-SA). Data cannot be made available online because of legal and ethical restrictions. To request data, readers may contact IeDEA-SA for consideration by filling out the online form available at https://www.iedea-sa.org/contact-us/. The code used for the analysis is available on GitHub (https://doi.org/10.5281/zenodo.17119000).

Funding Statement

Research reported in this publication was supported by the Swiss National Science Foundation (grant numbers 193381 [AH], and 189498 [ME]; https://www.snf.ch/en) and the US National Institutes of Health (the National Institute of Allergy and Infectious Diseases, the Eunice Kennedy Shriver National Institute of Child Health and Human Development, the National Cancer Institute, the National Institute of Mental Health, the National Institute on Drug Abuse, the National Heart, Lung, and Blood Institute, the National Institute on Alcohol Abuse and Alcoholism, the National Institute of Diabetes and Digestive and Kidney Diseases, and the Fogarty International Center) (grant number U01AI069924 [MD]; https://www.nih.gov). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

  • 1.Nock MK. Self-injury. Annu Rev Clin Psychol. 2010;6:339–63. doi: 10.1146/annurev.clinpsy.121208.131258 [DOI] [PubMed] [Google Scholar]
  • 2.Hooley JM, Fox KR, Boccagno C. Nonsuicidal self-injury: diagnostic challenges and current perspectives. Neuropsychiatr Dis Treat. 2020;16:101–12. doi: 10.2147/NDT.S198806 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Ferrari AJ, Santomauro DF, Aali A, Abate YH, Abbafati C, Abbastabar H, et al.; GBD 2021 Diseases and Injuries Collaborators. Global incidence, prevalence, years lived with disability (YLDs), disability-adjusted life-years (DALYs), and healthy life expectancy (HALE) for 371 diseases and injuries in 204 countries and territories and 811 subnational locations, 1990-2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet. 2024;403(10440):2133–61. doi: 10.1016/S0140-6736(24)00757-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.World Health Organization. Suicide worldwide in 2019: global health estimate. 2021. [Google Scholar]
  • 5.Sorsdahl K, Petersen I, Myers B, Zingela Z, Lund C, van der Westhuizen C. A reflection of the current status of the mental healthcare system in South Africa. SSM Mental Health. 2023;4:100247. doi: 10.1016/j.ssmmh.2023.100247 [DOI] [Google Scholar]
  • 6.Ruffieux Y, Efthimiou O, Van den Heuvel LL, Joska JA, Cornell M, Seedat S, et al. The treatment gap for mental disorders in adults enrolled in HIV treatment programmes in South Africa: a cohort study using linked electronic health records. Epidemiol Psychiatr Sci. 2021;30:e37. doi: 10.1017/S2045796021000196 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.World Health Organization. Suicide rates. 2025 [cited 4 Apr 2025]. Available from: https://www.who.int/data/gho/data/themes/mental-health/suicide-rates
  • 8.Joe S, Stein DJ, Seedat S, Herman A, Williams DR. Non-fatal suicidal behavior among South Africans : results from the South Africa Stress and Health Study. Soc Psychiatry Psychiatr Epidemiol. 2008;43(6):454–61. doi: 10.1007/s00127-008-0348-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Naidoo S. The prevalence, nature, and functions of non-suicidal self-injury (NSSI) in a South African student sample. S Afr J Educ. 2019;39: 1–10. doi: 10.15700/saje.v39n3a1697 [DOI] [Google Scholar]
  • 10.Collings SJ, Valjee SR. A multi-mediation analysis of the association between adverse childhood experiences and non-suicidal self-injury among South African adolescents. Int J Environ Res Public Health. 2024;21(9):1221. doi: 10.3390/ijerph21091221 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Pieterse D, Hoare J, Louw K-A, Breet E, Henry M, Lewis I, et al. Methods of deliberate self-harm in a tertiary hospital in South Africa. S Afr J Psychiatr. 2020;26:1399. doi: 10.4102/sajpsychiatry.v26i0.1399 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Ani JO, Ross AJ, Campbell LM. A review of patients presenting to accident and emergency department with deliberate self-harm, KwaZulu-Natal, South Africa. Afr J Prim Health Care Fam Med. 2017;9(1):e1–7. doi: 10.4102/phcfm.v9i1.1234 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.de la Torre-Luque A, Pemau A, Ayad-Ahmed W, Borges G, Fernandez-Sevillano J, Garrido-Torres N, et al. Risk of suicide attempt repetition after an index attempt: a systematic review and meta-analysis. Gen Hosp Psychiatry. 2023;81:51–6. doi: 10.1016/j.genhosppsych.2023.01.007 [DOI] [PubMed] [Google Scholar]
  • 14.Finkelstein Y, Macdonald EM, Hollands S, Sivilotti MLA, Hutson JR, Mamdani MM, et al. Risk of suicide following deliberate self-poisoning. JAMA Psychiatry. 2015;72(6):570–5. doi: 10.1001/jamapsychiatry.2014.3188 [DOI] [PubMed] [Google Scholar]
  • 15.Demesmaeker A, Chazard E, Hoang A, Vaiva G, Amad A. Suicide mortality after a nonfatal suicide attempt: a systematic review and meta-analysis. Aust N Z J Psychiatry. 2022;56(6):603–16. doi: 10.1177/00048674211043455 [DOI] [PubMed] [Google Scholar]
  • 16.Carr MJ, Ashcroft DM, Kontopantelis E, While D, Awenat Y, Cooper J, et al. Premature death among primary care patients with a history of self-harm. Ann Fam Med. 2017;15(3):246–54. doi: 10.1370/afm.2054 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Morgan C, Webb RT, Carr MJ, Kontopantelis E, Chew-Graham CA, Kapur N, et al. Self-harm in a primary care cohort of older people: incidence, clinical management, and risk of suicide and other causes of death. Lancet Psychiatry. 2018;5(11):905–12. doi: 10.1016/S2215-0366(18)30348-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Hamza CA, Stewart SL, Willoughby T. Examining the link between nonsuicidal self-injury and suicidal behavior: a review of the literature and an integrated model. Clin Psychol Rev. 2012;32(6):482–95. doi: 10.1016/j.cpr.2012.05.003 [DOI] [PubMed] [Google Scholar]
  • 19.World Health Organization. LIVE LIFE: an implementation guide for suicide prevention in countries. Geneva: WHO; 2021. [Google Scholar]
  • 20.World Health Organization. Comprehensive mental health action plan 2013-2030. Geneva: World Health Organization; 2021. [Google Scholar]
  • 21.Raubenheimer L, Jenkins L. An evaluation of factors underlying suicide attempts in patients presenting at George Hospital emergency centre. South African Fam Pract. 2015;57(2):93–9. doi: 10.1080/20786190.2014.976944 [DOI] [Google Scholar]
  • 22.van Zyl PJJ, Bantjes J, Breet E, Lewis I. Motives for deliberate self-harm in a South African tertiary hospital. S Afr J Psychiatr. 2021;27:1524. doi: 10.4102/sajpsychiatry.v27i0.1524 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Du Toit EH, Kruger JM, Swiegers SM, Van der Merwe M, Calitz FJW, Philane L, et al. The profile analysis of attempted-suicide patients referred to Pelonomi Hospital for psychological evaluation and treatment from 1 May 2005 to 30 April 2006. S Afr J Psych. 2008;14(1):6. doi: 10.4102/sajpsychiatry.v14i1.40 [DOI] [Google Scholar]
  • 24.Riley RD, Hayden JA, Steyerberg EW, Moons KGM, Abrams K, Kyzas PA, et al. Prognosis Research Strategy (PROGRESS) 2: prognostic factor research. PLoS Med. 2013;10(2):e1001380. doi: 10.1371/journal.pmed.1001380 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Hemingway H, Croft P, Perel P, Hayden JA, Abrams K, Timmis A, et al. Prognosis research strategy (PROGRESS) 1: a framework for researching clinical outcomes. BMJ. 2013;346:e5595. doi: 10.1136/bmj.e5595 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Bradshaw D, Joubert J, Maqungo M, Nannan N, Funani N, Laubscher R, et al. South African national cause-of-death validation project: methodology and description of a national sample of verbal autopsies; 2020. Available from: https://www.samrc.ac.za/sites/default/files/attachments/2022-08/NationalCause-of-deathValidationReport.pdf
  • 27.Dorrington R, Bradshaw D, Laubscher R, Nannan N. Rapid mortality surveillance report 2019 & 2020. 2021 Nov. Available from: https://www.samrc.ac.za/sites/default/files/attachments/2022-08/Rapid%20Mortality%20Surveillance%20Report%202019%262020.pdf
  • 28.StatsSA. General Household Survey (GHS), 2023; 2024 [cited 9 Jun 2025]. Available from: https://www.gov.za/news/media-statements/statssa-publishes-general-household-survey-23-may-2024
  • 29.Rammila D. Evaluating the potential impact of National Health Insurance on medical scheme members’ rights to have access to health-care services in South Africa. Law democr Dev. 2023;27:360–91. doi: 10.17159/2077-4907/2023/ldd.v27.14 [DOI] [Google Scholar]
  • 30.Council for Medical Schemes. CMScript 7 of 2024: Focus on Attempted Suicide. 4 Sep 2024 [cited 9 Jun 2025]. Available from: https://www.medicalschemes.co.za/cmscript-attempted-suicide/
  • 31.World Health Organization (WHO). International statistical classification of diseases and related health problems 10th revision. 2016 [cited 25 Aug 2019]. Available from: https://icd.who.int/browse10/2016/en
  • 32.Kraemer HC, Kazdin AE, Offord DR, Kessler RC, Jensen PS, Kupfer DJ. Coming to terms with the terms of risk. Arch Gen Psychiatry. 1997;54(4):337–43. doi: 10.1001/archpsyc.1997.01830160065009 [DOI] [PubMed] [Google Scholar]
  • 33.Runeson B, Tidemalm D, Dahlin M, Lichtenstein P, Långström N. Method of attempted suicide as predictor of subsequent successful suicide: national long term cohort study. BMJ. 2010;341:c3222. doi: 10.1136/bmj.c3222 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Cai Z, Junus A, Chang Q, Yip PSF. The lethality of suicide methods: a systematic review and meta-analysis. J Affect Disord. 2022;300:121–9. doi: 10.1016/j.jad.2021.12.054 [DOI] [PubMed] [Google Scholar]
  • 35.Royston P, Parmar MKB. Flexible parametric proportional‐hazards and proportional‐odds models for censored survival data, with application to prognostic modelling and estimation of treatment effects. Stat Med. 2002;21(15):2175–97. doi: 10.1002/sim.1203 [DOI] [PubMed] [Google Scholar]
  • 36.Lau B, Cole SR, Gange SJ. Competing risk regression models for epidemiologic data. Am J Epidemiol. 2009;170(2):244–56. doi: 10.1093/aje/kwp107 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28(25):3083–107. doi: 10.1002/sim.3697 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Institute for Health Metrics and Evaluation. GHDx—GBD results tool; 2021 [cited 26 Jun 2025]. Available from: https://vizhub.healthdata.org/gbd-results/
  • 39.Carr MJ, Ashcroft DM, Kontopantelis E, Awenat Y, Cooper J, Chew-Graham C, et al. The epidemiology of self-harm in a UK-wide primary care patient cohort, 2001-2013. BMC Psychiatry. 2016;16:53. doi: 10.1186/s12888-016-0753-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Knipe D, Padmanathan P, Newton-Howes G, Chan LF, Kapur N. Suicide and self-harm. Lancet. 2022;399(10338):1903–16. doi: 10.1016/S0140-6736(22)00173-8 [DOI] [PubMed] [Google Scholar]
  • 41.Naidoo S, Gathiram P, Schlebusch L. Effectiveness of a Buddy intervention support programme for suicidal behaviour in a primary care setting. South African Fam Pract. 2014;56(5):263–70. doi: 10.1080/20786190.2014.980159 [DOI] [Google Scholar]
  • 42.Knipe D, Williams AJ, Hannam-Swain S, Upton S, Brown K, Bandara P, et al. Psychiatric morbidity and suicidal behaviour in low- and middle-income countries: a systematic review and meta-analysis. PLoS Med. 2019;16(10):e1002905. doi: 10.1371/journal.pmed.1002905 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Hawton K, Saunders K, Topiwala A, Haw C. Psychiatric disorders in patients presenting to hospital following self-harm: a systematic review. J Affect Disord. 2013;151(3):821–30. doi: 10.1016/j.jad.2013.08.020 [DOI] [PubMed] [Google Scholar]
  • 44.Iemmi V, Bantjes J, Coast E, Channer K, Leone T, McDaid D, et al. Suicide and poverty in low-income and middle-income countries: a systematic review. Lancet Psychiatry. 2016;3(8):774–83. doi: 10.1016/s2215-0366(16)30066-9 [DOI] [PubMed] [Google Scholar]
  • 45.Phillips MR, Cheng HG. The changing global face of suicide. Lancet. 2012;379(9834):2318–9. doi: 10.1016/S0140-6736(12)60913-1 [DOI] [PubMed] [Google Scholar]
  • 46.Phillips MR. Rethinking the role of mental illness in suicide. Am J Psychiatry. 2010;167(7):731–3. doi: 10.1176/appi.ajp.2010.10040589 [DOI] [PubMed] [Google Scholar]
  • 47.Kootbodien T, Bantjes J, Joska J, Asmal L, Chiliza B, Stallones L, et al. Healthcare utilization 12 months prior to fatal and non-fatal suicidal behaviour in Cape Town, South Africa. Arch Suicide Res. 2024;28(1):216–30. doi: 10.1080/13811118.2022.2152767 [DOI] [PubMed] [Google Scholar]
  • 48.Jacobs Y, Myers B, van der Westhuizen C, Brooke-Sumner C, Sorsdahl K. Task sharing or task dumping: counsellors experiences of delivering a psychosocial intervention for mental health problems in South Africa. Community Ment Health J. 2021;57(6):1082–93. doi: 10.1007/s10597-020-00734-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Nuij C, van Ballegooijen W, de Beurs D, Juniar D, Erlangsen A, Portzky G, et al. Safety planning-type interventions for suicide prevention: meta-analysis. Br J Psychiatry. 2021;219(2):419–26. doi: 10.1192/bjp.2021.50 [DOI] [PubMed] [Google Scholar]
  • 50.Stanley B, Brown GK. Safety planning intervention: a brief intervention to mitigate suicide risk. Cognit Behav Pract. 2012;19(2):256–64. doi: 10.1016/j.cbpra.2011.01.001 [DOI] [Google Scholar]
  • 51.Bryan CJ, Mintz J, Clemans TA, Leeson B, Burch TS, Williams SR, et al. Effect of crisis response planning vs. contracts for safety on suicide risk in U.S. Army Soldiers: a randomized clinical trial. J Affect Disord. 2017;212:64–72. doi: 10.1016/j.jad.2017.01.028 [DOI] [PubMed] [Google Scholar]
  • 52.Stanley B, Brodsky B, Monahan M. Brief and ultra-brief suicide-specific interventions. Focus (Am Psychiatr Publ). 2023;21(2):129–36. doi: 10.1176/appi.focus.20220083 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Gysin-Maillart A, Schwab S, Soravia L, Megert M, Michel K. A novel brief therapy for patients who attempt suicide: a 24-months follow-up randomized controlled study of the Attempted Suicide Short Intervention Program (ASSIP). PLoS Med. 2016;13(3):e1001968. doi: 10.1371/journal.pmed.1001968 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Tlali M, Whitesell V, van den Heuvel L, Orrell C, Seedat S, Prozesky H, et al. Suicidality among adults living with and without HIV in Cape Town, South Africa. In: International Conference on HIV Treatment, Pathogenesis, and Prevention Research (INTEREST), Cotonou, Benin. 2024. [Google Scholar]
  • 55.Haas AD, Ruffieux Y, Tlali M, Rabie S, Egger M. Is suicide risk 100-fold higher in people with HIV? A critical appraisal of a meta-analysis. Gen Psychiatr. 2025;38(4):e102210. doi: 10.1136/gpsych-2025-102210 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Bantjes J, Nel A, Louw K-A, Frenkel L, Benjamin E, Lewis I. “This place is making me more depressed”: the organisation of care for suicide attempters in a South African hospital. J Health Psychol. 2017;22(11):1434–46. doi: 10.1177/1359105316628744 [DOI] [PubMed] [Google Scholar]
  • 57.Herman AA, Stein DJ, Seedat S, Heeringa SG, Moomal H, Williams DR. The South African Stress and Health (SASH) study: 12-month and lifetime prevalence of common mental disorders. S Afr Med J. 2009;99(5 Pt 2):339–44. [PMC free article] [PubMed] [Google Scholar]
  • 58.Goldstone D, Bantjes J, Nel D, Stanbridge J, Lewis I. Alcohol use predicts emergency psychiatric unit admission for non-fatal suicidal behaviour in the Western Cape (South Africa): a case-control study. Int J Psychiatry Clin Pract. 2020;24(2):163–72. doi: 10.1080/13651501.2019.1711419 [DOI] [PubMed] [Google Scholar]
  • 59.Burrows S, Laflamme L. Assessment of accuracy of suicide mortality surveillance data in South Africa: investigation in an urban setting. Crisis. 2007;28(2):74–81. doi: 10.1027/0227-5910.28.2.74 [DOI] [PubMed] [Google Scholar]

Decision Letter 0

Suzanne De Bruijn

6 Feb 2025

Dear Dr Haas,

Thank you for submitting your manuscript entitled "Intentional Self-Harm and Unnatural Death in South Africa: A Cohort Study" for consideration by PLOS Medicine.

Your manuscript has now been evaluated by the PLOS Medicine editorial staff as well as by an academic editor with relevant expertise and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by Feb 10 2025 11:59PM.

Login to Editorial Manager here: https://www.editorialmanager.com/pmedicine

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Feel free to email us at plosmedicine@plos.org if you have any queries relating to your submission.

Kind regards,

Suzanne

Suzanne De Bruijn, PhD

Associate Editor

PLOS Medicine

Decision Letter 1

Suzanne De Bruijn

30 May 2025

Dear Dr Haas,

Many thanks for submitting your manuscript "Intentional Self-Harm and Unnatural Death in South Africa: A Cohort Study" (PMEDICINE-D-25-00406R1) to PLOS Medicine. The paper has been reviewed by two subject experts and a statistician; their comments are included below and can also be accessed here: [LINK]

As you will see, the reviewers found the paper interesting and felt that the dataset was very important. However, the raised a large number of questions and felt that further clarification was needed regarding aspects of the methodology and analysis. After discussing the paper with the editorial team and an academic editor with relevant expertise, I'm pleased to invite you to revise the paper in response to the reviewers' comments. Please note that the manuscript will require substantive revisions to address the reviewers’ and editors’ comments, and we plan to send the revised paper to some or all of the original reviewers. At this stage, cannot provide any guarantees regarding publication.

In addition to these revisions, you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests shortly.

When you upload your revision, please include a point-by-point response that addresses all of the reviewer and editorial points, indicating the changes made in the manuscript and either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please also be sure to check the general editorial comments at the end of this letter and include these in your point-by-point response. When you resubmit your paper, please include a clean version of the paper as the main article file and a version with changes tracked as a marked-up manuscript. It may also be helpful to check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper.

We ask that you submit your revision by June 20th. However, if this deadline is not feasible, please contact me by email, and we can discuss a suitable alternative.

Don't hesitate to contact me directly with any questions (sbruijn@plos.org).

Kind regards,

Heather

Heather Van Epps, PhD

Consulting Editor

[on behalf of]

Suzanne De Bruijn, PhD

Associate Editor

PLOS Medicine

sbruijn@plos.org

-----------------------------------------------------------

Comments from the academic editor:

1. Intentional self-harm is not necessarily suicidal so should not be considered synonymous with a “suicide attempt” (e.g. in the abstract) or suicidal behaviour (background). The authors need to engage with the intentional self-harm literature in the background and explain how it relates to suicidal intentions and attempts in this context. For example, cutting in high-income countries is described as ‘intentional self-harm’ but is conceptualised as a maladaptive coping strategy for emotional regulation and rarely suicidal. In the findings from this paper, cutting seems to be uncommon as a means of intentional self-harm so the authors can make the argument that is more closely linked to suicidal intention.

2. How complete is the coverage of the national vital registration database?

3. It is necessary to present characteristics of the two cohorts separately. The inclusion of a cohort based on HIV status will affect the findings – recommend that the stratified findings are presented in a supplementary file.

4. Please change ‘completed suicide’ to ‘death by suicide’ in line with recommended language to reduce stigma.

-----------------------------------------------------------

Comments from the reviewers:

Reviewer #1 (statistics):

Thanks for the opportunity to read your manuscript. My role is statistical reviewer, so I have focused on the design, data, and analysis that are presented. I have put general comments first, followed by questions relevant to a specific section of the manuscript (with a page/paragraph reference).

This manuscript presents the results of a cohort study investigating incidence and risk factors of self-harm and unnatural deaths in South Africa. Data come from routinely collected sources, a private HIV management program, medical insurance scheme, and a national vital registration database. People aged 10 or over covered the insurance scheme between 2011 and 2020 were included. The main outcomes were the first coded healthcare record for intentional self-harm, and 'unnatural deaths'. A range of covariates including demographics and health status were available. For the mortality data, the primary exposure of interest was a healthcare record for non-fatal self-harm. Healthcare utilisation was compared between exposed and non-exposed, with matching on age and length of follow-up. Overall rates by age and sex were estimated from Poisson regression, and flexible parametric survival models were used for incidence of self-harm (with death as competing risk), including age as covariate and stratified by sex and HIV status. The approach to the analysis is sound (the FSPM is a good choice). I found this to be an interesting and well-written manuscript with well-presented results. I had some requests for clarification about the analysis and data sources, my primary concern was about the interpretation of the multivariable models.

1. It would help in understanding the coverage and limitations if some additional details about the datasets and linkage were provided. In particular, the eligibility and coverage of the medical insurance scheme, HIV disease management program, and vital registration scheme (particularly what information is available from this database). What method of data linkage (deterministic, SLK, probabilistic) was used and on what variables? Why were two databases (insurance + HIV management) used?

2. A limitation of some of the results (e.g. fig 2) with a multivariable model is that the effect estimates have different interpretations depending on the causal model that is assumed ('table 2 fallacy' https://doi.org/10.1093/aje/kws412). Essentially the estimates will be a mix of 'direct-effects' and 'total-effects' which have different interpretations. I would consider either a) proposing a causal model so that the estimates can be interpreted correctly, b) a + concentrating on previous self-harm exposure (in mortality analysis), c )or looking at the data more descriptively.

3. Is it possible that some participants might be recurrent cases, with the data recording intentional self-harm occurring before the study inception date? Effectively, the study has no look-back period, which is concerning as because the length of lookback period affects incidence estimates (e.g. https://doi.org/10.1016/j.jclinepi.2005.12.013)

4. P4, paragraph 3. Can the codes for 'unnatural deaths' be included?

5. P4, paragraph 4. Were the insurance and management program databases the source for these covariates?

6. P4, paragraph 5. Presumably those who died during the first healthcare encounter would have no or very little follow-up data?

7. Was non-fatal self-harm exposure included a fixed covariate, or was this a time-varying covariate?

8. Table S3. With your sample size, even a trivial difference will be 'significant', I would consider putting in an estimate of standardised difference instead of the p-value.

Reviewer #2:

Thank you for the opportunity to review this informative manuscript. The authors present a valuable and rare dataset from South Africa, offering insights into the risk factors for self-harm and its subsequent mortality outcomes in a non-Western setting. While the study could make potentially important contributions to the literature, particularly in under-researched contexts, I have several major concerns related to the methodological clarity, study design, analystical approaches, and conbining two studies in one manuscript.

Comments on the overall study

1. This manuscript presents valuable data with the potential to significantly inform suicide prevention efforts in South Africa and similar contexts. However, the combination of two complex studies, unclear definitions, missing methodological details, and insufficient adjustment for key covariates limit its current clarity and rigor. I would recommend revising the manuscript to either (1) more clearly separate and expand upon the two components, or (2) divide it into two distinct papers with improved methodology and presentation. [Editorial note: we prefer that you do not divide the manuscript into two independent papers]

2. This manuscript combines two conceptually distinct yet related studies: one that investigates risk factors for first-time self-harm in the general population ("self-harm study"), and another that examines the association between self-harm and subsequent unnatural death ("mortality study"). While these topics are connected, they require different study designs, assumptions, and analyses. Presenting both studies in a single paper leads to conceptual and methodological confusion. A clearer distinction in structure and explanation between these two studies is necessary, and it may be more appropriate to report them in separate papers.

3. The study excluded individuals who could not be linked to mortality records. However, no comparison is provided between included and excluded individuals. It is essential to assess whether exclusion may have introduced bias. Basic comparisons of self-harm incidence, sociodemographic composition, and the distribution of mental disorders in the included and excluded samples should be presented. It is unclear whether the exclusion was applied to both the self-harm study and mortality study.

4. In both studies, the use of Cox regression assumes proportional hazards, yet there is no report on whether this assumption was tested. Given that mortality risk after self-harm may decline sharply over time, checking and reporting this assumption is essential.

5. While the authors include some variables such as HIV status and psychiatric diagnoses, important sociodemographic and clinical risk factors are missing. These include physical comorbidities and socioeconomic status, which are known to influence both self-harm risk and mortality.

6. The study treats psychiatric conditions in a relatively broad way. However, certain diagnoses—such as ADHD and conduct disorders—are more relevant in youth, whereas depression and substance use may be important in both younger and older adults. Age-specific modelling of psychiatric risk factors would provide more meaningful insights.

Comments on the self-harm study:

7. The self-harm study attempts to identify first-time self-harm events, but no information is given on how prior episodes were ruled out. A standard approach would involve applying a washout period (e.g., 1-2 years) to reduce the likelihood of misclassifying repeat events as first-time episodes. This methodological limitation should be addressed and discussed.

8. Death was treated as a competing event in the self-harm study where it should have instead been treated as a censoring event.

Comments on the mortality study:

9. The mortality study focuses on unnatural death, a key outcome among people with a history of self-harm. However, natural causes could be also major contributors to subsequent mortality, particularly older self-harm patients. It would strengthen the paper to provide data on all-cause mortality, and then further disaggregate by natural vs. unnatural causes.

10. The authors should report the proportion of deaths with unknown (natural vs unnatural causes) or undetermined (suicide vs accident) causes and discuss their implications.

11. Age is a critical modifier in both the risk of self-harm and its mortality consequences. Yet, the study does not present age interactions or age-stratified analyses. Furthermore, while some subgroup analyses were performed, such as by method lethality and sex, no formal tests of interaction were conducted to assess the statistical evidence of effect modifications.

12. In the mortality study, self-harm is modelled as a fixed exposure. However, self-harm occurs at varying times during follow-up and should be treated as a time-varying covariate in survival analyses. Alternatively, a matched cohort design could strengthen causal inferences. However, the control selection is under-explained. For instance, the index date for controls should correspond to the event date of the matched self-harm case, but this is not clearly stated. Furthermore, no information was given regarding how this matching was implemented. Matching variables included age group (e.g., 40-59 years), but such broad categories can introduce residual confounding. Narrower bands (e.g., 5-year groups) should be considered. The inclusion of sex in the matching process would be also relevant and indicated.

13. The final part of the paper describes post-self-harm healthcare contact within specific time frames (e.g., 1 week, 1 month). This is an important but distinct topic from the core analyses, and would be better suited to a separate, focused manuscript.

Reviewer #3:

Thank you for the opportunity to review this manuscript on a topic of clear importance. The methods are quite sound and the findings are striking and far-reaching, with potential for major impact in the field. I appreciated the extremely sounds methods and statistical approach and concise presentation of a variety of findings, but I believe the manuscript would benefit from some clarifications and added detail in a few areas, outlined below.

1. Abstract: The focus on unnatural death (rather than death by suicide) is logical and the ICD codes used to define unnatural death are appropriate. However, a definition of what constitutes unnatural death would strengthen the abstract.

2. I appreciated the acknowledgement of some of the key limitations in the abstract.

3. Introduction: In paragraphs two and three, "Non-fatal suicidal behavior" seems to be used interchangeably with "suicide attempts", which is not the case. Please address.

4. HIV status is considered in the methods and analytic plan but is completely absent from the abstract and introduction. In the prior literature, there is a link between living with HIV and suicide risk. This should be acknowledged and addressed.

5. Add considerations of how the sample of people who had health insurance coverage might differ from the general population.

6. Considerations of social determinants of health (e.g., food insecurity, housing insecurity, race/ethnicity, socioeconomic status) as predictors of unnatural death in this sample are completely absent. Why weren't these considered, and might it be beneficial to explore them in future studies?

7. It appears the second sub-sample were people living with HIV. Given the prior literature highlighting HIV status as a risk factor for suicide, the implications of using this sub-sample should be addressed more thoroughly. Tables S5 and S6 show higher risk among people with HIV but this is not addressed in the Abstract or Results - please add.

8. Did the prevalence of HIV in the final sample reflect the prevalence in the general population? If people living with HIV were over-sampled, how might this have impacted the generalizability of findings?

9. Similarly, comparison of baseline prevalence of common mental disorders in your sample with the prevalence in the general population will give some indication of generalizability.

Any attachments provided with reviews can be seen via the following link: [LINK]

--------------------------------------------------------- ---

General editorial requests:

(Note: not all will apply to your paper, but please check each item carefully)

* We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. Please do not add or remove authors without first discussing this with the handling editor. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

* Please upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

* Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information (web or email address) for obtaining the data. Please note that a study author cannot be the contact person for the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

* We expect all researchers with submissions to PLOS in which author-generated code underpins the findings in the manuscript to make all author-generated code available without restrictions upon publication of the work. In cases where code is central to the manuscript, we may require the code to be made available as a condition of publication. Authors are responsible for ensuring that the code is reusable and well documented. Please make any custom code available, either as part of your data deposition or as a supplementary file. Please add a sentence to your data availability statement regarding any code used in the study, e.g. "The code used in the analysis is available from Github [URL] and archived in Zenodo [DOI link]" Please review our guidelines at https://journals.plos.org/plosmedicine/s/materials-software-and-code-sharing and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse. Because Github depositions can be readily changed or deleted, we encourage you to make a permanent DOI'd copy (e.g. in Zenodo) and provide the URL.

FORMATTING - GENERAL

* Abstract: Please structure your abstract using the PLOS Medicine headings (Background, Methods and Findings, Conclusions). Please combine the Methods and Findings sections into one section.

* At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Ideally each sub-heading should contain 2-3 single sentence, concise bullet points containing the most salient points from your study. In the final bullet point of 'What Do These Findings Mean?', please include the main limitations of the study in non-technical language. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary.

* Please express the main results with 95% CIs as well as p values. When reporting p values please report as p<0.001 and where higher as the exact p value p=0.002, for example. Throughout, suggest reporting statistical information as follows to improve clarity for the reader "22% (95% CI [13%,28%]; p</=)". Please be sure to define all numerical values at first use.

* Please include page numbers and line numbers in the manuscript file. Use continuous line numbers (do not restart the numbering on each page).

* Please cite the reference numbers in square brackets. Citations should precede punctuation.

FIGURES AND TABLES

* Please provide titles and legends for all figures and tables (including those in Supporting Information files).

* Please define all abbreviations used in each figure/table (including those in Supporting Information files).

* Please consider avoiding the use of red and green in order to make your figure more accessible to those with color blindness.

SUPPLEMENTARY MATERIAL

* Please note that supplementary material will be posted as supplied by the authors. Therefore, please amend it according to the relevant comments outlined here.

* Please cite your Supporting Information as outlined here: https://journals.plos.org/plosmedicine/s/supporting-information

REFERENCES

* PLOS uses the numbered citation (citation-sequence) method and first six authors, et al.

* Please ensure that journal name abbreviations match those found in the National Center for Biotechnology Information (NCBI) databases (http://www.ncbi.nlm.nih.gov/nlmcatalog/journals), and are appropriately formatted and capitalised.

* Where website addresses are cited, please include the complete URL and specify the date of access (e.g. [accessed: 12/06/2023]).

* Please also see https://journals.plos.org/plosmedicine/s/submission-guidelines#loc-references for further details on reference formatting.

OBSERVATIONAL STUDIES

* Abstract: Please include the study design, population and setting, number of participants, years during which the study took place (enrollment and follow up), length of follow up, and main outcome measures.

* Please ensure that the study is reported according to the STROBE (or appropriate STOBE extension) guideline (available from: https://www.equator-network.org/reporting-guidelines/strobe) and include the completed STROBE (or STROBE extension) checklist as Supporting Information. Please add the following statement, or similar, to the Methods: "This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist)." When completing the checklist, please use section and paragraph numbers, rather than page numbers.

* For all observational studies, in the manuscript text, please indicate: (1) the specific hypotheses you intended to test, (2) the analytical methods by which you planned to test them, (3) the analyses you actually performed, and (4) when reported analyses differ from those that were planned, transparent explanations for differences that affect the reliability of the study's results. If a reported analysis was performed based on an interesting but unanticipated pattern in the data, please be clear that the analysis was data driven.

* Please state in the Methods section whether the study had a prospective protocol or analysis plan. If a prospective analysis plan (from your funding proposal, IRB or other ethics committee submission, study protocol, or other planning document written before analyzing the data) was used in designing the study, please include the relevant document(s) with your revised manuscript as a Supporting Information file to be published alongside your study and cite it in the Methods section. A legend for this file should be included at the end of your manuscript. If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place. Changes in the analysis, including those made in response to peer review comments, should be identified as such in the Methods section of the paper, with rationale.

Attachment

Submitted filename: PLOS Medicine May 2025.docx

pmed.1004765.s005.docx (18.7KB, docx)

Decision Letter 2

Suzanne De Bruijn

30 Jul 2025

Dear Dr Haas,

Many thanks for submitting your manuscript "Self-Harm and Unnatural Death in South Africa: A Cohort Study" (PMEDICINE-D-25-00406R2) to PLOS Medicine. The paper has been re-reviewed by the subject experts and a statistician; their comments are included below and can also be accessed here: [LINK]

As you will see, the reviewers think the manuscript has improved, but have some remaining concerns. After discussing the paper with the editorial team and an academic editor with relevant expertise, I'm pleased to invite you to revise the paper in response to these last reviewers' comments. Specifically, R#1 comments on the revision being positioned as prognostic, whereas the results are still appropriate for a causality study. R#2 would like you to test for sex-differences, and if these are not present this reviewer suggests to combine the estimates. Please note that we do not expect you to address the comments regarding reporting all-cause mortality numbers as well. We plan to send the revised paper to some or all of the original reviewers, and we cannot provide any guarantees at this stage regarding publication.

In addition to these revisions, you may need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests shortly. If you do not receive a separate email within a few days, please assume that checks have been completed, and no additional changes are required.

When you upload your revision, please include a point-by-point response that addresses all of the reviewer and editorial points, indicating the changes made in the manuscript and either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please also be sure to check the general editorial comments at the end of this letter and include these in your point-by-point response. When you resubmit your paper, please include a clean version of the paper as the main article file and a version with changes tracked as a marked-up manuscript. It may also be helpful to check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper.

We ask that you submit your revision by Aug 20 2025 11:59PM. However, if this deadline is not feasible, please contact me by email, and we can discuss a suitable alternative.

Don't hesitate to contact me directly with any questions (sbruijn@plos.org).

Best regards,

Suzanne

Suzanne De Bruijn, PhD

Associate Editor

PLOS Medicine

sbruijn@plos.org

-----------------------------------------------------------

Comments from the academic editor:

The recommendations regarding the sex differences are valid, and should be addressed.

The suggestion of reporting all-cause mortality separately to unnatural deaths is beyond the scope of this paper.

-----------------------------------------------------------

Comments from the reviewers:

Reviewer #1: Thanks for the revised manuscript and responses to my original review. The updates have clarified most of my original queries.

My remaining concern is that the updates to the manuscript now position the aim of the study to be prognostic, rather than to estimate a causal relationship between the different study covariates and risk of self-harm and unnatural deaths. However, the methods described and the way the results are reported are for an aetiological study rather than a prediction study. If the aim was a predictive model of the study outcomes, the study needs a) a measure of discrimination of the model (e.g. c-statistic), b) level of calibration of predictions from the model, c) a way of measuring or managing model optimism (cross-validation, bootstrapping), and d) a clear vision of the 'use case' of the model, i.e. where it could be implemented and what actions could be taken with predictions from the model are available. The individual coefficients from the model should not be interpreted as effect estimates either (e.g. see Box 2 of van Diepen 2017 which was referenced in the response). The manuscript should report appropriate results of whatever approach is chosen as the aim of the study (descriptive, predictive, causal, see this recent reference summarises this well: https://doi.org/10.1002/sim.10244)

Reviewer #2: I thank the authors for their responses. I have two further comments on the revised paper as outlined below:

1. The authors argued that "The aim of the mortality analysis is to inform suicide prevention efforts. Natural causes of death are not preventable through suicide prevention interventions, and elevated risk may reflect higher comorbidity in this population. To maintain conceptual clarity and focus, we therefore prefer to limit our analysis to unnatural deaths." However, suicide prevention activities that incorporate psychosocial support may have spillover effects on overall health, including physical health. Moreover, a holistic approach to the care of individuals who have self-harmed should address both mental and physical health needs. Importantly, all-cause mortality is a key overarching outcome for this population, and estimating its risk can help contextualise the relative contributions of natural and unnatural causes of death. Therefore, I wonder if the authors could consider providing risk estimates for all-cause and natural-cause mortality, at least in the supplementary materials, if not in the main text, to offer a more comprehensive understanding of mortality risk in this population.

2. In Figure 4, sex-specific analyses were presented. However, the 95% confidence intervals for hazard ratios in males and females overlap substantially, suggesting no statistically meaningful sex difference. It would be worthwhile to formally test for sex differences by including a sex-by-exposure interaction term in the regression models. If no statistical evidence for interaction is found, presenting combined estimates for males and females may be more appropriate.

Reviewer #3: I thank the authors for their careful and thorough attention to the reviewer comments. My comments have now been adequately addressed.

Any attachments provided with reviews can be seen via the following link: [LINK]

--------------------------------------------------------- ---

General editorial requests:

* We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. Please do not add or remove authors without first discussing this with the handling editor. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

* Financial disclosure: Thank you for including your financial enclosure. Please also include URLs from the funders in this statement.

* Ethics statement: Could you please include IRB approval numbers?

FORMATTING - GENERAL

* Thank you for including an Author summary.

-Can you please limit the number of bullet points under 'What Did The Researchers Do And Find' to 3?

-Can you please include limitations in the final bullet point of 'What Do These Findings Mean?

Decision Letter 3

Suzanne De Bruijn

9 Sep 2025

Dear Dr. Haas,

Thank you very much for re-submitting your manuscript "Self-Harm and Unnatural Death in South Africa: A Cohort Study" (PMEDICINE-D-25-00406R3) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by 2 reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

In addition to these revisions, you may need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests shortly. If you do not receive a separate email within a few days, please assume that checks have been completed, and no additional changes are required.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript.

Please note, when your manuscript is accepted, an uncorrected proof of your manuscript will be published online ahead of the final version, unless you've already opted out via the online submission form. If, for any reason, you do not want an earlier version of your manuscript published online or are unsure if you have already indicated as such, please let the journal staff know immediately at plosmedicine@plos.org.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.  

We look forward to receiving the revised manuscript by Sep 16 2025 11:59PM.   

Sincerely,

Suzanne De Bruijn, PhD

Associate Editor 

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

GENERAL EDITORIAL REQUESTS

* Please change your title to: "Self-harm incidence rates and their predictive value for subsequent unnatural death in South Africa: a Cohort Study"

* Please ensure that all abbreviations are defined at first use throughout the text.

* Please confirm that all numbers presented in the abstract are present and identical to numbers presented in the main manuscript text.

GENERAL

* Statistical reporting: Please revise throughout the manuscript, including tables and figures.

- Please report statistical information as follows to improve clarity for the reader ""22% (95% CI [13,28]; p</=)"".

- Please separate upper and lower bounds with commas instead of hyphens as the latter can be confused with reporting of negative values.

- Please repeat statistical definitions (HR, CI etc.) for each set of parentheses."

* Please consider moving some of the methods from the supplements into the main text.

*ABSTRACT: please mention clearly that 2011-2022 were the years for which you have data.

FUNDING STATEMENT

* Thank you for including your funding statement. Could you please include the initials of the people who obtained the grant?

DATA AVAILABILITY

* Thank you for providing a github/Zenodo link in the data availability statement in the manuscript. Can you please ensure the same statement, including the code availability, is also present in the metadata.

FIGURES

* Please consider if moving some of the figures shared in the supplementary information into the main text would aid the reader. Specifically, we would suggest to move Figure A from the supplements to the main text.

OBSERVATIONAL, COHORT, CROSS-SECTIONAL, AND CASE CONTROL STUDIES

*Thank you for including the STROBE checklist, and mentioning this in your methods section. Could you please spell out STROBE? "This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (Appendix S4).""

-When completing the checklist, please use section and paragraph numbers, rather than page numbers.

Comments from Reviewers:

Reviewer #1: Thanks for the revised manuscript and responses to my original review. The explanation given in the responses and the corresponding updates was helpful for me. The updates to the wording to specifically position the manuscript as a 'prognostic factor' study (along with some adjustment of terminology) works well.

Reviewer #2: I have no further comments.

Decision Letter 4

Suzanne De Bruijn

17 Sep 2025

Dear Dr Haas, 

On behalf of my colleagues and the Academic Editor, Charlotte Hanlon, I am pleased to inform you that we have agreed to publish your manuscript "Incidence and Prognostic Factors of Self-Harm and Subsequent Unnatural Death in South Africa: A Cohort Study" (PMEDICINE-D-25-00406R4) in PLOS Medicine.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. Please be aware that it may take several days for you to receive this email; during this time no action is required by you. Once you have received these formatting requests, please note that your manuscript will not be scheduled for publication until you have made the required changes.

In the meantime, please log into Editorial Manager at http://www.editorialmanager.com/pmedicine/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process. 

PRESS

We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with medicinepress@plos.org. If you have not yet opted out of the early version process, we ask that you notify us immediately of any press plans so that we may do so on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

Thank you again for submitting to PLOS Medicine. We look forward to publishing your paper. 

Sincerely, 

Suzanne De Bruijn, PhD 

Associate Editor 

PLOS Medicine

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    S1 Strobe Checklist. STROBE, Strengthening the Reporting of Observational Studies in Epidemiology.

    https://www.strobe-statement.org/.

    (PDF)

    pmed.1004765.s001.pdf (122.4KB, pdf)
    S1 Appendix. Study protocol.

    (PDF)

    pmed.1004765.s002.pdf (200.9KB, pdf)
    S2 Appendix. Protocol deviations.

    (PDF)

    pmed.1004765.s003.pdf (90.1KB, pdf)
    S3 Appendix

    Table A. Incidence rates of self-harm encounters per 100,000 person-years, stratified by sex, for different age groups and overall. Table B. Cumulative incidence (%) and 95% confidence intervals of self-harm, stratified by sex, for different age groups and overall. Table C. Healthcare utilization before self-harm events compared with matched controls prior to hypothetical matched event dates. Table D. Cumulative incidence (%) and 95% confidence intervals of unnatural death, stratified by sex and self-harm method, and overall. Table E. Characteristics of the study population by cohort. Table F. Characteristics of individuals included and excluded from the mortality analysis. Fig A. Incidence of self-harm by age, sex, and HIV status. Incidence per 100,000 person-years in general population as a continuous function of age, stratified by sex and HIV status. Individuals were censored at the time of their first documented self-harm event. Repeated self-harm events were not considered. Fig B. Time-varying hazard ratios for unnatural death by sex, comparing individuals with and without prior non-fatal self-harm encounter. Hazard ratios are plotted as a continuous function of analysis time, defined as time since the index self-harm event for individuals with self-harm or baseline for those without self-harm event. Separate models were fitted by sex and adjusted for age. Shaded areas indicate 95% confidence intervals. The dashed horizontal line represents a hazard ratio of 1, indicating no difference in risk. Fig C. Incidence of self-harm by age and sex, restricted to medical scheme data. Incidence per 100,000 person-years in general population as a continuous function of age, stratified by sex. Individuals were censored at the time of their first documented self-harm event. Repeated self-harm events were not considered. Fig D. Cumulative incidence of self-harm by sex and age group, restricted to medical scheme data. Aalen–Johansen estimate with 95% confidence intervals for cause-specific cumulative over time since baseline or age-group change, stratified by sex and age group. Individuals were censored at the time of their first documented self-harm event. Repeated self-harm events were not considered. Fig E. Predictors of self-harm, restricted to medical scheme data. The figure shows age- and sex-adjusted and fully adjusted hazard ratios for predictors of self-harm. Individuals were censored at the time of their first documented self-harm event. Repeated self-harm events were not considered. The error bars represent 95% confidence intervals. Fully adjusted models controlled for age group, sex, HIV, and all mental disorders. * For individuals diagnosed with both depression and bipolar disorder, only the bipolar disorder diagnosis was considered from the date of bipolar diagnosis onwards. Abbreviations: CI, confidence intervals; HR, hazard ratio. Fig F. Hazard ratios for unnatural death comparing individuals with and without a history of non-fatal self-harm by sex, restricted to medical scheme data. All models were fitted separately for males and females. Age-adjusted estimates control for age using it as the timeline in Cox model. Fully adjusted models additionally control for HIV status, mental health diagnoses. Error bars represent 95% confidence intervals. Abbreviations: NF-SH, non-fatal self-harm; CI, confidence intervals; HR, hazard ratio. Fig G. Hazard ratios for self-harm method as a predictor of unnatural death, by sex, restricted to medical scheme data. All models were fitted separately for males and females. Hazard ratios compare individuals who used highly lethal or less lethal methods at their index self-harm encounter with individuals without a history of non-fatal self-harm (NF-SH). Age-adjusted estimates control for age by using it as the timeline in the Cox model. Fully adjusted models additionally control for HIV status and mental health diagnoses. Error bars represent 95% confidence intervals. Abbreviations: NF-SH, non-fatal self-harm; CI, confidence intervals; HR, hazard ratio.

    (DOCX)

    pmed.1004765.s004.docx (413.3KB, docx)
    Attachment

    Submitted filename: PLOS Medicine May 2025.docx

    pmed.1004765.s005.docx (18.7KB, docx)
    Attachment

    Submitted filename: Responses_to_reviewers_v1.1.docx

    pmed.1004765.s006.docx (145.5KB, docx)
    Attachment

    Submitted filename: Responses_to_reviewers_v1.3.docx

    pmed.1004765.s007.docx (96.7KB, docx)
    Attachment

    Submitted filename: Responses_to_reviewers_v1.0.docx

    pmed.1004765.s008.docx (69.9KB, docx)

    Data Availability Statement

    Data were obtained from the International epidemiology Databases to Evaluate AIDS–Southern Africa (IeDEA-SA). Data cannot be made available online because of legal and ethical restrictions. To request data, readers may contact IeDEA-SA for consideration by filling out the online form available at https://www.iedea-sa.org/contact-us/. The code used for the analysis is available on GitHub (https://doi.org/10.5281/zenodo.17119000).

    Articles from PLOS Medicine are provided here courtesy of PLOS

    RESOURCES