Introduction
On March 11, 2020, the World Health Organization declared COVID-19 a global pandemic, abruptly dismantling the structural fabric of higher education and healthcare training worldwide. Within weeks, clinical placements were suspended, campuses shuttered, and millions of health science students confronted an existential paradox: trained to safeguard public health yet rendered powerless by an invisible threat. This confluence of educational disruption, social isolation, and vicarious trauma created conditions uniquely hostile to psychological wellbeing.
Prior to the pandemic, health students already exhibited mental health burden exceeding that of the general student population—with depression prevalence reported at 27–40% and burnout rates above 50% in pre-pandemic systematic reviews.1–3 Against this vulnerable baseline, the pandemic did not merely add stress; it systematically amplified pre-existing vulnerabilities through mechanisms that remain incompletely characterized.
Longitudinal evidence accumulating since 2020 consistently demonstrates that healthcare trainees are among the hardest-hit populations in the pandemic mental health crisis. A meta-analysis of 69 studies reported a pooled prevalence of anxiety of 41.1% and depression of 33.8% among university students globally during COVID-19 (Luo et al 2020), with health-specific samples exhibiting markedly elevated rates. Critically, these cross-sectional snapshots fail to capture the trajectory of deterioration—an essential dimension for understanding whether distress is acute and transient or progressive and potentially lethal.
Suicide risk, the most severe outcome of psychiatric deterioration, has received disproportionately little longitudinal attention in this population, despite evidence from national registries in multiple countries indicating increased rates of self-harm ideation among young adults during the pandemic.4–7 This evidential gap carries urgent clinical implications, given that health students are both at elevated personal risk and future providers of crisis mental health care.
A critical but under-investigated moderator of pandemic mental health outcomes is prior trauma exposure—specifically, a history of violence. Adverse childhood experiences (ACEs) and interpersonal violence histories are robustly associated with increased psychiatric vulnerability and sensitize the hypothalamic-pituitary-adrenal (HPA) axis, rendering individuals biologically primed for amplified stress responses under subsequent adversity.8–11 The pandemic, characterized by loss of control, social isolation, and heightened mortality salience, constitutes precisely such an adversity—a “second hit” that may interact multiplicatively, rather than additively, with trauma histories to elevate suicide risk. Despite this theoretically compelling mechanism, no longitudinal study among health science students has formally tested the Time × Violence interaction as a predictor of suicide risk trajectory.12–14 This represents a substantive gap in the evidence base that limits both theoretical understanding and clinical risk stratification.
The present study addresses this gap using a prospective two-wave longitudinal design spanning the pre-pandemic baseline (January 2020) and pandemic peak period (January 2022) across 1401 health science students representing 11 disciplines. We employ validated psychometric instruments (DASS-21; SBQ-R), clinically meaningful effect measures (Number Needed to Harm; Risk Difference), and advanced longitudinal statistical modelling (Generalized Estimating Equations) to: (1) quantify pandemic-attributable deterioration in depression, anxiety, stress, and suicide risk; (2) identify independent demographic and clinical predictors of high suicide risk; and (3) formally test the “trauma multiplier” hypothesis—that pandemic conditions disproportionately escalate suicide risk among students with prior violence exposure. These findings are positioned to inform evidence-based, trauma-informed mental health policy for health education systems globally as they prepare for future public health crises.
Methods
Study Design and Setting
A prospective, two-wave longitudinal cohort study was conducted at a government-accredited health sciences institution in Indonesia. The study conformed to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines for observational longitudinal research (see Supplementary File 1). This study was reviewed and approved by the Health Research Ethics Committee, Faculty of Nursing, Universitas Airlangga, Indonesia (Ethical Approval Number: 3223-KEPK).
The study was conducted in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all participants aged ≥18 years; for participants aged <18 years, parental or legal guardian consent and participant assent were obtained, as applicable.
Participants
Eligibility criteria were: active enrollment in a health science diploma or bachelor’s degree program, provision of informed consent, and completion of baseline assessment in January 2020. Students on extended leave, exchange programs, or with pre-existing diagnosed psychiatric conditions requiring pharmacotherapy were excluded to minimize confounding. Of 1623 eligible students who completed baseline assessment, 1401 (86.3%) completed the follow-up survey in January 2022.
Attrition analysis using independent-samples t-tests and chi-square tests confirmed no significant baseline differences between completers and non-completers (all p > 0.05), indicating that loss to follow-up was non-differential and unlikely to introduce selection bias.
Data Collection Procedures
Data were collected via standardized, self-administered electronic questionnaires administered through a secure institutional platform (Google Forms, encrypted). Identical procedures were employed at both time points. Research assistants trained in survey administration provided standardized instructions and were available to clarify item wording.
To minimize social desirability bias, students were explicitly informed that responses were anonymous and would not affect academic standing.
Measures
Mental Health Outcomes
The Depression Anxiety Stress Scale-21 (DASS-21) is a 21-item psychometrically validated instrument measuring three subscales: depression (items 3, 5, 10, 13, 16, 17, 21), anxiety (items 2, 4, 7, 9, 15, 19, 20), and stress (items 1, 6, 8, 11, 12, 14, 18). Each item is rated on a 4-point Likert scale (0–3); subscale scores are doubled for comparison with DASS-42 normative data. Binary classification employed validated cutoffs: depression ≥10, anxiety ≥8, stress ≥15 as “abnormal” (mild to extremely severe).
Cronbach’s α for this sample: depression = 0.89, anxiety = 0.85, stress = 0.87.
Suicide Risk
The Suicide Behaviors Questionnaire–Revised (SBQ-R) is a four-item instrument assessing suicidal ideation, suicide attempt history, frequency of ideation, and disclosure of suicidal intent. The established binary cutoff of SBQ-R ≥ 8 was used to classify participants as high suicide risk, consistent with validated thresholds in community and student samples. Cronbach’s α in this sample = 0.79.
History of Violence
Prior violence exposure was assessed using a single structured item: “Have you ever experienced physical, emotional, or sexual violence at any point in your life?” (Yes/No), supplemented by a validated 5-item Violence Exposure Subscale adapted from the Adverse Childhood Experiences (ACE) questionnaire. This approach was selected for feasibility within the student population context.
Statistical Analysis
All analyses were conducted using R v4.3.1 (R Core Team, 2023) with packages lme4, geepack, and pROC (see Supplementary Statistical Code for the full annotated R script). Statistical significance was set at α = 0.05 (two-tailed).
Descriptive Statistics
Continuous variables are reported as means ± standard deviations (SD) or medians with interquartile ranges (IQR) based on distributional assessment (Shapiro–Wilk test). Categorical variables are presented as frequencies and percentages.
Longitudinal Change
McNemar’s test was used to compare paired binary mental health outcomes (pre/during pandemic). Paired odds ratios (OR), 95% confidence intervals (CIs), Risk Differences (RD), and Number Needed to Harm (NNH) were calculated as clinically meaningful effect measures.
Subgroup Analyses
Interaction terms were tested using logistic regression stratified by gender, age group, academic level, and residence type. Heterogeneity in OR was quantified using the Breslow-Day test.
Multivariable Logistic Regression
Predictors of high suicide risk (SBQ-R ≥ 8) during the pandemic were identified through backward stepwise selection (entry criterion p < 0.10; retention criterion p < 0.05), adjusting for age, gender, academic level, and residence. Model calibration was assessed by the Hosmer-Lemeshow test; discrimination by area under the receiver operating characteristic curve (AUC-ROC).
Generalized Estimating Equations (GEE)
To account for within-subject correlation across the two time points, a GEE model with exchangeable working correlation structure was fitted to the binary suicide risk outcome. The critical Time × History of Violence interaction term was formally tested to evaluate the “trauma multiplier” hypothesis. Model fit was assessed using the Quasi-likelihood under the Independence Model Criterion (QIC).
Sensitivity Analyses
Complete-case analysis (n = 1357) was compared with the primary multiple imputation model (m = 20; MCMC convergence diagnostics are presented in Supplementary Figures S2 and S3). Analyses were repeated in adult-only subsamples (≥18 years, n = 1124) to assess robustness independent of the minor subgroup.
Results
Participant Flow and Baseline Characteristics
Of the 1623 health students who completed the baseline assessment (January 2020), 1401 completed the follow-up survey (January 2022), yielding a retention rate of 86.3%. Attrition analysis revealed no significant baseline differences between completers and non-completers (p > 0.05), minimizing attrition bias. The final analytical sample was predominantly female (88.0%) and aged 18–21 years (80.2%).
A notable vulnerable subgroup comprised minors aged <18 years (n = 277, 19.8%). Over half were enrolled in diploma programs (56.8%), and 27.3% resided in boarding houses during the pandemic. The sample represented 11 distinct health science disciplines (Table 1 and see Supplementary Table S1 for extended baseline demographic and clinical characteristics).Table 1Demographic and Clinical Characteristics of the Study Cohort (n = 1401)CharacteristicCategoryn%p-value†GenderMale16812.0–Female123388.0Age (years)< 18 (Minors)27719.8–18–21112480.2Academic LevelDiploma79656.8–Bachelor’s Degree60543.2ResidenceWith Family101972.7–Boarding House38227.3History of ViolenceYes18713.3–No121486.7Baseline High Suicide Risk (SBQ-R ≥ 8)Yes16912.1–No123287.9Note: †Attrition analysis p-value (completers vs non-completers).Abbreviation: SBQ-R, Suicide Behaviors Questionnaire–Revised.
Longitudinal Changes in Mental Health Outcomes
Paired longitudinal analysis revealed statistically significant and clinically meaningful deterioration across all mental health domains (Table 2 and see Supplementary Table S2 for extended longitudinal mental health outcomes and sensitivity metrics). Anxiety demonstrated the most severe pandemic-attributable deterioration, with the proportion of students with abnormal anxiety symptoms rising 15.9 percentage points from 48.5% to 64.4% (paired OR = 1.73, 95% CI: 1.56–1.92, p < 0.001; NNH = 7), indicating that for every seven students exposed to pandemic conditions, one additional student developed clinically significant anxiety. Depression prevalence increased from 35.0% to 44.2% (NNH = 11) and stress from 25.0% to 29.6% (NNH = 22).
Suicide risk escalated from 12.1% to 15.4% (paired OR = 1.33, 95% CI: 1.16–1.52, p < 0.001; NNH = 30).Table 2Longitudinal Changes in Mental Health Status Before and During the Pandemic (n = 1401)DomainTime PointNormal, n (%)Abnormal, n (%)McNemar χ2p-valuePaired OR (95% CI)RD (%)NNHDepressionBefore911 (65.0)490 (35.0)58.42<0.0011.44 (1.31–1.58)+9.211During782 (55.8)619 (44.2)–––––AnxietyBefore721 (51.5)680 (48.5)127.85<0.0011.73 (1.56–1.92)+15.97During499 (35.6)902 (64.4)–––––StressBefore1051 (75.0)350 (25.0)24.67<0.0011.26 (1.15–1.38)+4.622During986 (70.4)415 (29.6)–––––Suicide Risk (SBQ-R ≥ 8)Before1232 (87.9)169 (12.1)18.34<0.0011.33 (1.16–1.52)+3.330During1185 (84.6)216 (15.4)–––––Note: Abnormal = Mild to Very Severe (DASS-21 validated cutoffs).Abbreviation: OR, Odds Ratio; CI, Confidence Interval; RD, Risk Difference; NNH, Number Needed to Harm.
Subgroup Analysis and Effect Modification
Stratified analyses revealed significant heterogeneity in the magnitude of mental health deterioration across demographic strata. Female students exhibited significantly higher odds of depression deterioration compared to males (OR = 1.52 vs 1.18; p for interaction = 0.032). Students residing in boarding houses faced higher anxiety escalation risk (OR = 1.89, 95% CI: 1.62–2.20) compared to those living with family (OR = 1.65, 95% CI: 1.46–1.86).
The minor subgroup (<18 years) demonstrated heightened vulnerability across all domains, particularly anxiety (OR = 1.89, 95% CI: 1.61–2.22; Figure 1 and see Supplementary Figures S1 and S2 for extended subgroup analyses across all mental health domains). Figure 1Subgroup-specific odds ratios for depression among health science students during the pandemic. This forest plot presents the odds ratios (OR) with 95% confidence intervals (CI) for depression across key demographic and socioeconomic subgroups within the longitudinal cohort of health science students.
The analysis is stratified by gender (male vs female), age (<18 years vs ≥18 years), academic level (Diploma vs Bachelor’s Degree), and residence type (living with family vs boarding house). The vertical dashed line at OR = 1.00 represents the null effect; data points to the right of this line indicate an increased risk of depression during the pandemic. The overall pooled estimate is displayed at the top.
Notably, female students, those aged <18 years, diploma-level students, and those residing in boarding houses exhibited elevated odds of depression, with confidence intervals entirely above unity, underscoring the disproportionate vulnerability of these subgroups to pandemic-related psychological distress. The color coding distinguishes subgroups with statistically significant risk elevation (Orange) from those with non-significant or protective trends (blue).This forest plot illustrates the odds ratios (OR) and 95% confidence intervals (CI) for depression risk across key demographic subgroups. A vertical dashed line at OR = 1.00 indicates the null effect.
Subgroups are stratified by Gender (Male, Female), Age (<18 years, ≥18 years), Academic Level (Diploma, Bachelor’s Degree), and Residence (With Family, Boarding House). Each estimate is represented by a circular point with a horizontal CI line. Statistically significant risk elevations (CI entirely >1.00) are highlighted in orange, while non-significant trends are in blue.
Clinically, the overall pooled OR is approximately 1.45 (95% CI: 1.35–1.60). Female students show OR approximately 1.52 (95% CI: 1.38–1.68), minors <18 years show OR approximately 1.50 (95% CI: 1.28–1.82), and diploma students show OR approximately 1.58 (95% CI: 1.40–1.80), all demonstrating significant vulnerability. Male students show a non-significant trend with OR approximately 1.18 (95% CI: 0.88–1.55), as the confidence interval crosses the null value.
Bachelor’s degree students exhibit OR approximately 1.28 (95% CI: 1.15–1.45). Regarding residence, students living with family show OR approximately 1.32 (95% CI: 1.20–1.48), while those residing in boarding houses demonstrate OR approximately 1.52 (95% CI: 1.30–1.78), indicating elevated risk associated with physical isolation from family support.Forest plot displaying subgroup-specific odds ratios (OR) and 95% confidence intervals (CI) for depression deterioration among health science students during the pandemic, stratified by gender, age, academic level, and residence.Note: OR > 1 indicates increased risk of depression during pandemic.
Predictors of High Suicide Risk: Multivariable Analysis
In the fully adjusted multivariable logistic regression model (Table 3 and see Supplementary Table S2 for additional multivariable regression outputs and model diagnostics), history of violence emerged as the strongest independent predictor of high suicide risk during the pandemic (aOR = 3.30, 95% CI: 1.62–6.73, p < 0.001). Severe anxiety (aOR = 3.33, 95% CI: 2.34–4.74) and moderate-to-severe depression (aOR = 1.93, 95% CI: 1.39–2.68) were similarly robust predictors. Demographic variables (gender, age, academic level, residence) lost statistical significance in the fully adjusted model, indicating that their crude associations with suicide risk were largely mediated by trauma history and acute symptom severity.
Model calibration was satisfactory (Hosmer-Lemeshow p = 0.61) and discrimination was excellent (AUC = 0.85, 95% CI: 0.81–0.89).Table 3Multivariable Logistic Regression Predicting High Suicide Risk (SBQ-R ≥ 8) During the PandemicVariableCategoryUnadjusted OR (95% CI)p-valueAdjusted OR (95% CI)*p-valueVIFHistory of ViolenceNoReference–Reference––Yes4.12 (3.08–5.51)<0.0013.30 (1.62–6.73)<0.0011.31Depression SeverityNormal/MildReference–Reference––Moderate2.34 (1.72–3.18)<0.0011.93 (1.39–2.68)<0.0011.58Severe/Very Severe5.67 (4.12–7.80)<0.0013.15 (2.18–4.55)<0.0011.62Anxiety SeverityNormal/MildReference–Reference––Moderate2.01 (1.48–2.73)<0.0011.67 (1.20–2.32)0.0021.54Severe/Very Severe4.89 (3.58–6.68)<0.0013.33 (2.34–4.74)<0.0011.59GenderMaleReference–Reference––Female1.67 (1.21–2.31)0.0021.34 (0.94–1.91)0.1081.22Age Group≥ 18 yearsReference–Reference––< 18 years1.52 (1.14–2.03)0.0041.18 (0.85–1.64)0.3121.18Notes: *Adjusted for all variables in the table plus academic level and residence. Hosmer-Lemeshow p = 0.61; AUC = 0.85 (95% CI: 0.81–0.89). VIF = Variance Inflation Factor (all <2.0; multicollinearity acceptable).
Bold = primary independent predictors.
Longitudinal Suicide Risk Trajectories: GEE Model
The GEE model (Table 4) confirmed a significant main effect of time on suicide risk (β = 0.312, p < 0.001; aOR = 1.37, 95% CI: 1.15–1.63). The Time × History of Violence interaction term was statistically significant (β = 0.423, p = 0.002; aOR = 1.53, 95% CI: 1.17–1.99), formally confirming the “trauma multiplier” hypothesis: students with prior violence exposure experienced a disproportionately steeper trajectory of suicide risk escalation during the pandemic compared to non-exposed peers, over and above the main effects of time and violence exposure individually. The predicted probabilities of high suicide risk before and during the pandemic, stratified by history of violence, are presented in Figure 2 (see Supplementary Figures S3 and S4 for granular visualization of longitudinal psychological trajectories across different student cohorts).Table 4Generalized Estimating Equations (GEE) Model for Longitudinal Suicide Risk TrajectoriesPredictorβ CoefficientSEWald χ2p-valueaOR (95% CI)Intercept−2.8470.234147.82<0.001–Time (During vs Before)0.3120.08912.31<0.0011.37 (1.15–1.63)History of Violence (Yes vs No)1.1940.15658.47<0.0013.30 (2.43–4.48)DASS-21 Total Score0.0910.01082.81<0.0011.10 (1.08–1.12)Time × History of Violence Interaction0.4230.1349.970.0021.53 (1.17–1.99)Notes: Working correlation structure: Exchangeable.
QIC = 2847.3. Bold row indicates the primary interaction effect confirming the trauma multiplier hypothesis.Abbreviations: SE, Standard Error; aOR, Adjusted Odds Ratio.Figure 2Predicted probability of high suicide risk before and during the pandemic, stratified by violence exposure. This line graph illustrates the longitudinal change in the predicted probability of high suicide risk among health science students, comparing two critical exposure groups: those with no history of violence exposure (blue line) and those with documented violence exposure (Orange line).
Data points represent marginal means derived from multivariate logistic regression models, with error bars denoting 95% confidence intervals. The analysis reveals a stark divergence in trajectories: students with violence exposure demonstrated a substantially elevated baseline risk before the pandemic (approximately 30%), which escalated dramatically during the pandemic to over 40%. In contrast, students without violence exposure maintained a consistently low and stable risk profile (approximately 10%) across both time periods.
This pronounced interaction effect exemplifies the “trauma multiplier effect”, wherein pre-existing trauma synergistically amplifies the psychological impact of pandemic-related stressors, resulting in a disproportionately higher suicide risk among the most vulnerable subset of the cohort.A line graph showing predicted probability of high suicide risk before and during pandemic by violence exposure.
Sensitivity Analyses
Complete-case analysis (n = 1357) yielded effect estimates within 5% of the primary multiple imputation model (aOR for violence: 3.21, 95% CI: 1.57–6.58), confirming robustness to missing data assumptions. When analyses were restricted to adult participants (≥18 years, n = 1124), the association between violence history and suicide risk remained virtually unchanged (aOR = 3.18, 95% CI: 1.54–6.45, p < 0.001), indicating that the primary association was not materially influenced by the inclusion of participants younger than 18 years.
Discussion
This prospective longitudinal cohort study of 1401 health science students demonstrates that the COVID-19 pandemic was associated with a significant, clinically meaningful deterioration in mental health across all assessed domains, with anxiety exhibiting the most severe increase (NNH = 7). Crucially, we provide the first formal longitudinal evidence of a “trauma multiplier” effect: prior violence exposure more than tripled baseline suicide risk (aOR = 3.30) and, through a statistically significant interaction with pandemic exposure, produced a disproportionately steep suicide risk escalation trajectory in trauma-exposed students (interaction aOR = 1.53). These findings extend and deepen prior cross-sectional literature by quantifying the longitudinal interplay between pre-existing trauma vulnerability and acute societal crisis.15–17
The disproportionate escalation of anxiety (NNH = 7 vs NNH = 11 for depression) is consistent with the acute, uncertainty-driven nature of pandemic stress, which maps directly onto the cognitive threat-appraisal mechanisms central to anxiety disorders. Prolonged uncertainty regarding academic continuity, clinical exposure, and occupational safety are established anxiety-inducing stressors specific to health students.18–20 The observed increase in anxiety prevalence corresponded to an absolute risk difference of 15.9 percentage points, equivalent to an NNH of approximately 7 when calculated from the observed pre-pandemic and pandemic-period risks. Because this was an observational before-and-during-pandemic comparison without a counterfactual control condition, the NNH should be interpreted as a descriptive measure of the observed risk difference rather than as definitive evidence of a causal effect attributable solely to the pandemic.
This framing carries direct implications for power calculations in future intervention trials targeting anxiety prevention in this population.
The identification of violence history as the strongest independent predictor of pandemic-era suicide risk (aOR = 3.30) aligns with established neurobiological models of trauma sensitization. Prior adverse experiences—particularly interpersonal violence—induce lasting dysregulation of the HPA axis and amygdala reactivity, rendering the stress response system hyperreactive to subsequent threat stimuli.21–25 The pandemic, characterized by loss of behavioral control, social isolation, and heightened mortality salience, may constitute precisely the “second hit” that transforms latent neurobiological vulnerability into manifest suicidal crisis. The significant Time × Violence interaction term (β = 0.423; aOR = 1.53) provides the first longitudinal statistical confirmation of this mechanistic hypothesis in a health student population, moving beyond correlation to formalize the multiplicative—not merely additive—nature of this risk compounding.
Future studies should distinguish the proposed trauma-multiplier mechanism from a statistical interaction alone by testing whether the observed risk amplification is replicated across crisis contexts and whether modifiable protective factors attenuate this association.
Gender and residential context emerged as significant subgroup moderators of pandemic mental health burden, though their associations with suicide risk were attenuated in the fully adjusted model. The elevated anxiety deterioration observed in female students (OR = 1.52 vs 1.18 in males; interaction p = 0.032) replicates a consistent pattern in the broader COVID-19 mental health literature and likely reflects the intersection of gendered socialization (higher emotional expressivity, help-seeking awareness) with disproportionate domestic burden during lockdowns.26–28 Students in boarding houses—physically isolated from familial social support networks—exhibited nearly double the anxiety escalation risk, underscoring the centrality of social connectedness as a protective buffer against pandemic-induced distress.4,10 These contextual factors should inform targeted outreach priorities within health institutions.2,8
The high cross-domain comorbidity observed during the pandemic—particularly the anxiety-stress correlation of r = 0.85—suggests that these constructs share overlapping pathophysiological pathways under conditions of acute societal threat, potentially constituting a unified “pandemic distress syndrome” rather than independent diagnostic entities. This comorbidity pattern has important implications for intervention design: treatments targeting anxiety and stress simultaneously (eg, unified transdiagnostic protocols) may yield broader symptom relief than domain-specific approaches.6,29,30 The high temporal stability of individual symptom scores (anxiety r = 0.82; depression r = 0.74) further indicates that mental health burden established during the pandemic was not transient but persisted, potentially solidifying into chronic psychopathology risk.
The inclusion of a substantial minor subgroup (n = 277, 19.8%) adds a dimension absent from most health student pandemic studies, which exclusively target adults. This younger cohort exhibited heightened vulnerability across all domains—particularly anxiety (OR = 1.89)—consistent with the neurodevelopmental stage-specificity of stress sensitivity: adolescent brains are in active prefrontal development, rendering emotional regulation less efficient and social isolation more acutely distressing.11,31,32 Sensitivity analyses confirming that primary findings hold when restricted to adults (aOR = 3.18) reassure that the overall cohort results are not driven by adolescent vulnerability, but the minor subgroup findings themselves constitute actionable evidence for age-stratified risk management protocols in health education.
The present findings carry urgent implications for health education policy. The excellent model discrimination (AUC = 0.85) suggests that a brief, two-variable clinical screen—trauma history and current anxiety severity—could identify students at >3-fold elevated suicide risk, enabling resource-efficient, targeted intervention delivery. Integration of routine trauma screening (eg, ACE questionnaire adaptation) into health science program enrollment and annual wellness checks is warranted and feasible given existing institutional frameworks.3,9,13 Beyond institutional screening, the NNH = 7 for anxiety provides a compelling public health argument for universal, low-intensity anxiety prevention interventions (eg, brief mindfulness, psychoeducation) delivered to all health students during crisis periods—analogous to population-level prophylaxis in infectious disease.
Implications for Trauma-Informed Health Professions Education
The findings support the development of trauma-informed health professions education as a framework for strengthening student psychological safety during public health crises. Integrating trauma awareness, early risk identification, peer support, and accessible psychological services into health science curricula may provide a structured institutional response to compounded stress exposure. Future research should investigate whether trauma-informed educational models can reduce psychological distress, improve help-seeking behavior, and mitigate suicide risk among health science students across diverse cultural and educational settings.
Such approaches may contribute to the development of crisis-resilient health professional education systems. Practically, our findings support the implementation of a brief, two-variable clinical screen (prior trauma history + current anxiety severity) as a low-resource, high-yield triage tool. This protocol can be seamlessly integrated into routine university health checks to identify students at >3-fold elevated suicide risk with excellent discriminatory power (AUC = 0.85).
Research Sustainability and Translational Pathway
Beyond its immediate relevance to the COVID-19 pandemic, this study provides a foundation for a sustained research programme on psychological vulnerability and resilience in health professions education. The proposed Trauma Multiplier Framework can be examined as a testable hypothesis in which pre-existing trauma exposure may modify the association between subsequent crisis exposure and adverse mental health outcomes, while social support, adaptive coping, and access to psychological services may represent potentially modifiable protective factors. Three complementary research directions follow from this framework: (1) external validation through multicentre longitudinal studies with repeated assessments across different crisis contexts; (2) prospective evaluation of modifiable protective factors and the development of ethically designed, trauma-informed support interventions; and (3) implementation research examining feasibility, acceptability, equity of access, and mental health outcomes across diverse educational settings.
The present findings establish a rationale for these investigations rather than evidence that the proposed framework or interventions are universally valid or effective. This staged approach may help translate a context-specific longitudinal finding into a progressively validated and context-sensitive evidence base for crisis-resilient health professions education.
Limitations
Several limitations warrant acknowledgment. First, violence history was assessed using a self-report single-item screen rather than a validated multi-item instrument (eg, CTQ, ACE-10), which may underestimate prevalence due to social desirability bias and limits etiological specificity regarding violence subtype. Second, the two-wave design, while prospective, precludes granular trajectory modelling—future studies should employ ≥3 waves to characterize non-linear recovery or chronification patterns.
Third, generalizability is bounded by the single-institution, Indonesian sample; replication in diverse cultural contexts is required before universal policy recommendations can be advanced.1,5,7,14 Fourth, pandemic-era data collection may have introduced response bias due to the emotionally salient survey context. Despite these constraints, this study makes a substantive contribution to global mental health longitudinal evidence and establishes the “trauma multiplier” framework as a theoretically and empirically grounded construct for understanding compounded psychiatric risk in future crisis contexts. Future research should evaluate whether trauma-focused interventions (eg, trauma-informed cognitive behavioral therapy, EMDR) deployed during the early phase of population-level crises can effectively blunt the multiplicative risk amplification documented here.
While this study establishes the Trauma Multiplier effect within an Indonesian cohort, the underlying neurobiological and psychosocial mechanisms are likely universal. Future longitudinal studies in diverse cultural, economic, and high-income settings are explicitly encouraged to validate this framework and refine context-specific trauma-informed interventions.
From a broader perspective, this study aligns with the principles of sustainable development in health education (SDGs 3 and 4). By shifting the paradigm from reactive mental health support to proactive, trauma-informed architectural design, health institutions can build resilient systems capable of withstanding future public health crises, particularly in low- and middle-income countries (LMIC).
Conclusion
COVID-19 pandemic conditions precipitated clinically significant deterioration in mental health and suicide risk among health science students, with anxiety exhibiting the most severe impact (NNH = 7). History of violence emerged as the most potent independent predictor of high suicide risk (aOR = 3.30), operating through a multiplicative—not merely additive—trauma amplification mechanism confirmed by longitudinal GEE modelling. These findings argue compellingly for the integration of universal trauma screening and targeted, trauma-informed mental health interventions into health science education frameworks, positioned as crisis-preparedness infrastructure rather than reactive provisions.
As healthcare systems globally confront the mental health sequelae of compound crises, embedding trauma awareness into the architecture of health professional training is no longer aspirational—it is epidemiologically imperative. The proposed Trauma Multiplier Framework offers a testable foundation for future multicentre longitudinal research, external validation, and the evaluation of trauma-informed strategies to strengthen psychological resilience in health professions education.