Determinants of depression severity among university students in KwaZulu-Natal, South Africa: socio-economic, food security and disaster-related predictors

Abstract

Background

Mental health is increasingly recognised as a major public health concern among university students, particularly in low- and middle-income settings where financial insecurity, food access constraints, and environmental disruptions may exacerbate vulnerability. However, limited empirical evidence from South Africa has examined how these structural stressors jointly influence depression severity.

Methods

A cross-sectional online survey was administered to registered students at a university in KwaZulu-Natal (n = 445). Findings reflect associations within a single institution and should not be interpreted as representative of all university students in KwaZulu-Natal or South Africa. Depression severity was measured using the Patient Health Questionnaire-9 (PHQ-9) and categorised into five ordered levels (none to severe). Socio-demographic, economic, food security, and disaster exposure variables were assessed. An ordered probit model with marginal effects was employed to identify determinants of depression severity.

Results

Approximately 59.8% of students reported moderate-to-severe depressive symptoms. Female students exhibited a 7.4% higher probability of severe depression relative to males. Greater dietary diversity was associated with a 2.4% lower probability of severe symptoms, while reliance on less stable funding sources increased risk by 1.1%. Exposure to the July 2021 unrest and April 2022 floods raised the likelihood of severe depression by 3.5% and 3.4%, respectively. Students with advanced age showed a modest protective effect. Other covariates were not statistically significant.

Conclusion

Depression severity among students is associated with gender, food security, financial stability, and exposure to socio-environmental shocks. These findings suggest that university mental health strategies should extend beyond clinical services to include food, financial, and disaster-resilience support. Addressing these determinants aligns with Sustainable Development Goals (SDGs) related to health, food security, education, and climate resilience.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28214-2.

Introduction

Mental health is associated with depression, which is one of the leading contributors to the global burden of disease and a major cause of disability among young adults worldwide [40]. The World Health Organization (WHO) identifies depressive disorders as one of the primary drivers of years lived with disability, with onset frequently occurring during late adolescence and early adulthood, the period when many individuals enter higher education. Adams et al. [1] noted that university students are particularly vulnerable due to the convergence of academic pressure, financial strain, social transition, and reduced family support, all of which may heighten psychological stress. This is corroborated by Kituyi & Kyeyune [27] who indicated that mental health-related cases among university students in Africa are increasing and stem from issues such as depression, addictions, cyberbullying, further compounded by social media. Several studies have reported that between 20 and 40% of university students experience clinically significant depressive symptoms, often accompanied by impaired academic performance and reduced quality of life [8, 17, 20, 21]. Similarly, the World Mental Health–International College Student (WMH-ICS) initiative documented substantial levels of depression and related disorders across tertiary institutions globally, highlighting student depression as an urgent public health concern [5, 16].

Although much of the literature originates from high-income countries, students in low- and middle-income settings may face additional structural stressors that compound depression risks. Socio-economic disadvantage, food insecurity, unstable income sources, and exposure to community-level shocks can intensify psychological vulnerability and limit access to coping resources [28]. Kirkbride et al. [26] stated that social determinant of depression perspective, depression cannot be understood solely as an individual clinical condition but must be situated within broader material and environmental contexts that shape daily life and opportunity structures. Within South Africa, depression challenges among university students are increasingly recognized but remain under-researched relative to need [22]. National and institutional studies suggest that depression, anxiety, and psychological distress are prevalent among tertiary students, often linked to financial hardship, academic exclusion risks, and limited access to counselling services [6, 23]. These pressures are particularly salient for students from historically disadvantaged backgrounds who rely heavily on bursary schemes and household remittances to sustain their studies [6]. Financial uncertainty has been shown to function as both an economic constraint and a chronic psychosocial stressor, contributing to persistent worry, reduced concentration, and emotional exhaustion [29].

Among these intersecting socio-economic constraints, food insecurity emerges as a particularly salient and underexamined pathway through which material deprivation translates directly into psychological distress among students [11]. Food insecurity represents another critical but frequently overlooked determinant of student well-being [11], whilst this is of national and global priority according to the SDGs. Beyond its nutritional implications, inadequate or uncertain access to food is associated with heightened stress, stigma, and diminished cognitive functioning. Empirical evidence consistently links food insecurity with depressive symptoms among university populations. For example, Brown et al. [10] found that food-insecure students in New York had significantly higher odds of depression than their food-secure peers. Comparable patterns have been documented in South Africa, where Wagner et al. [38] reported high levels of food insecurity among university students and significant associations with poorer depression outcomes. These findings suggest that diet quality and food access are closely intertwined with psychological well-being, particularly in contexts marked by economic precarity.

In addition to chronic socio-economic stressors, acute environmental and social disruptions may further compromise depression. Exposure to disasters, civil unrest, or infrastructure breakdowns has been associated with elevated risks of depression and psychological distress, especially among populations already facing material disadvantage [7]. KwaZulu-Natal (KZN) in the past five years has experienced significant community-level shocks, including the July 2021 social unrest, which disrupted local economic activity and damaged infrastructure in the province [25] and the April 2022 floods that overwhelmed critical infrastructure and caused widespread losses [13]. Authors have urged that it was not confined to political unrest but the reality of poverty amplification due to unemployment amongst youth, including graduates who face many socioeconomic pressures pre- and post-university life [14].

Despite growing recognition of mental distress challenges, important evidence gaps remain on which universities can leverage to improve institutional resilience during chaotic times. Much of the existing literature documents the prevalence and burden of depression and psychological distress among university students rather than modelling their determinants. Global and cross-national estimates provided by the WHO [40], Ibrahim et al. [20], Auerbach et al. [5], Cuijpers et al. [16], and the South African national survey by Bantjes et al. [6], as well as the Sub-Saharan African synthesis by Julius et al. [23], primarily quantify rates of disorder. Where determinants are examined, studies typically focus on single risk factors in isolation, such as financial strain [1, 29] or food insecurity [10, 11, 38], limiting understanding of how multiple socio-economic pressures interact. Although social determinant frameworks emphasise that depression is shaped by broader structural and material [26, 28], empirical applications that jointly integrate these factors remain limited. Moreover, research documenting the socio-economic consequences of community-level shocks, including the 2021 unrest and the 2022 floods in South Africa [13, 25], rarely connects such disruptions to student depression outcomes. Methodologically, most studies [4, 12, 35] rely on binary classifications of depression rather than modelling gradients of severity using ordinal approaches. To our knowledge, no quantitative study in KwaZulu-Natal has simultaneously modelled food security, funding instability, and disaster-related shocks within an ordered response framework. Addressing this gap is essential for identifying the relative importance of intersecting stressors and informing targeted student-support interventions.

This study, therefore examines factors associated with depression severity among university students in KwaZulu-Natal. Specifically, it investigates how socio-demographic characteristics (age, gender, year of study), financial support mechanisms, food security status, and exposure to recent environmental shocks relate to graded levels of depressive symptoms measured using the Patient Health Questionnaire-9. By integrating these dimensions within a unified analytical framework, the study aims to generate evidence to guide policies that strengthen material security, enhance depression support, and promote equitable educational outcomes. In doing so, the research contributes to broader efforts to advance Sustainable Development Goals 3 (Good Health and Well-being), 4 (Quality Education), and 10 (Reduced Inequalities) in higher education settings.

Analytical framework

This study is grounded in the Social Determinants of Mental Health (SDoMH) framework, which posits that psychological well-being is shaped not only by individual characteristics but also by the socio-economic, material, and environmental conditions within which individuals live and study [28]. Rather than conceptualising depression solely as a clinical or intrapersonal disorder, this perspective situates depression outcomes within broader structural contexts that influence exposure to stressors, access to resources, and coping capacity [26, 28]. Empirical evidence consistently demonstrates that poverty, food insecurity, financial instability, and environmental disruptions are associated with elevated risks of psychological distress and depressive symptoms, particularly among young adults navigating transitional life stages such as university education [1, 11, 38]. Drawing on this framework, depression severity among university students is hypothesised to arise from three interrelated domains of influence: socio-demographic, economic, and environmental factors.

First, socio-demographic characteristics, including age, gender, and year of study, may shape vulnerability or resilience to stress. Population-based and university studies report higher levels of depressive and anxiety symptoms among female students and younger cohorts, potentially reflecting gendered social pressures, safety concerns, and adjustment challenges during early stages of university life [5, 6, 20]. Subsequently, empirical findings suggest that the academic disciplines pursued within the university context are factors that influence depressive symptoms and the general well-being of students [39]. Science, Technology, Engineering and Mathematics (STEM) disciplines are also often correlated with higher risk [32]. The students who participated in this study were across the faculties, which also included social sciences.

Second, economic resources which function both as protective factors and as chronic stressors.Although South Africa has national-level policy commitments through the National Student Financial Aid Scheme (NSFAS), serving as a public higher education funding mechanism [30]. This funding model contributes significantly,however, it alone does not eliminate socio-economic vulnerability [14]. Stable and predictable funding mechanisms, such as bursaries or household support, may buffer students from uncertainty and enable them to meet basic needs. Conversely, irregular income sources, debt reliance, or inadequate financial assistance may generate persistent financial strain and psychological distress [1, 29]. Food security is similarly conceptualised as a material determinant of depression. Limited dietary diversity or uncertainty regarding food access has been associated with depression, stress, and reduced academic functioning through both nutritional and psychosocial pathways, including stigma and daily stress related to meeting basic needs [10, 11, 38].

Third, environmental shocks, including the July 2021 civil unrest, which was an unprecedented provincial disaster as per the Disaster Management Act (Act No. 57 of 2002), this disaster transpired during the peak of the Covid 19 pandemic. Following this were the April 2022 KwaZulu-Natal floods, which were declared a national disaster, representing acute stressors that may amplify pre-existing socio-economic vulnerabilities. This study was conducted to capture the compounded effects of consecutive crises, recognizing that students enrolled during 2021 and 2022 were subject to disruptions at different stages of their qualifications. Investigating this period provides critical insight into how successive socio-environmental shocks affect student outcomes.

These events disrupted livelihoods, housing, transport infrastructure, and local economic activity [13, 25]. In student populations, the depression effects of such shocks are expected to operate primarily through indirect economic consequences such as delayed fee payments, housing instability, and income interruptions rather than direct traumatic exposure alone, thereby reinforcing socio-economic pathways to psychological distress [28]. These domains are expected to operate both independently and interactively. For example, the psychological consequences of disaster-related disruptions may be more pronounced among students already experiencing food insecurity or unstable funding. Figure 1 illustrates the hypothesised relationships between socio-demographic, economic, and environmental determinants and depression severity.

Fig. 1.

Conceptual framework illustrating socio-demographic, economic, food security, and environmental determinants of depression severity among university students.

(Source: Author’s own)

Importantly, depression severity is conceptualised as an underlying latent and continuous construct rather than a purely categorical state. The Patient Health Questionnaire-9 (PHQ-9) produces ordered severity classifications (none/minimal, mild, moderate, moderately severe, and severe), which represent progressively higher levels of an unobserved psychological burden. Since these categories reflect rank order without equal intervals, the dependent variable is ordinal rather than continuous. Linear regression would impose unrealistic interval assumptions, while binary models would collapse meaningful distinctions across severity levels. An ordered probit model is therefore theoretically and statistically appropriate, as it assumes a latent continuous propensity toward psychological distress that is segmented by threshold cut-points into observed ordered outcomes. This specification aligns with the SDoMH framework by modelling how socio-economic and environmental exposures shift the probability of students transitioning into progressively more severe depression states. By matching the measurement properties of the PHQ-9 with an ordinal latent-variable framework, the model ensures coherence between theory, data structure, and econometric method.

Methodology

Study design and setting

This study employed a quantitative, cross-sectional analytic design to examine socio-demographic, economic, and environmental factors associated with depression severity among university students. An online survey design, which was a web-based questionnaire available on various university platforms. Data were collected during the 2024 academic year at Mangosuthu University of Technology public higher education institution located in KwaZulu-Natal, South Africa. The institution serves a predominantly low- and middle-income student population and has experienced socio-economic disruptions related to the July 2021 civil unrest and the April 2022 floods, making it an appropriate setting for investigating the intersection between structural stressors and student depression.

Study population and sampling

The study population comprised registered undergraduate and postgraduate students enrolled at a public higher education institution in KwaZulu-Natal, South Africa, during the 2024 academic year. Students were eligible to participate if they were aged 18 years or older and had completed at least one full academic year of study. First-year students were excluded from the study to ensure that participants had sufficient exposure to the university environment and its associated academic, financial, and social stressors, which are central to the study’s analytical framework. Participants were recruited using a non-probability convenience sampling approach. An electronic recruitment notice containing a secure link to the online questionnaire was disseminated through institutional communication platforms accessible to students across multiple faculties. Participation was voluntary, and students self-selected into the study after providing informed consent.

While convenience sampling enabled efficient data collection and broad reach within the institutional context, it does not produce a probability-based representative sample. As such, the final analytic sample (n = 445) should be interpreted as indicative of associations within the study population rather than representative of all university students in KwaZulu-Natal or South Africa more broadly. The potential for self-selection bias is acknowledged, particularly if students experiencing psychological distress were more motivated to participate. Nevertheless, this sampling approach is widely used in university mental health research, including large national and international student surveys, due to ethical considerations, feasibility constraints, and the sensitivity of mental health data collection. Despite these limitations, the achieved sample size was assessed as adequate for the analytical model employed. For ordered probit models, a commonly applied rule of thumb recommends a minimum of ten observations per predictor variable; with eleven predictors in the final model, this implies a minimum of approximately 110 observations, which the achieved sample (n = 445) substantially exceeds. Comparable cross-sectional studies of university student mental health in South Africa and internationally have reported samples ranging from approximately 300 to 600 participants, situating the present study within an accepted range. The diversity of participants across academic levels, funding sources, accommodation types, and exposure to socio-environmental shocks strengthens the internal validity of the findings within the institutional context.

Data collection and instruments

Data were collected using a structured, self-administered online questionnaire designed to capture socio-demographic characteristics, economic conditions, food security status, exposure to socio-environmental shocks, and depression severity among university students as shown in Table 1. The questionnaire was administered during the 2024 academic year using REDCap (Research Electronic Data Capture), a secure, web-based platform widely used for academic and health research data collection. The survey instrument was adapted from a previously implemented university student mental health questionnaire used at the University of the Western Cape. Minor contextual modifications were made to ensure relevance to the KwaZulu-Natal setting and to include items relating to exposure to the July 2021 civil unrest and the April 2022 floods. Depression severity was measured using the validated Patient Health Questionnaire-9 (PHQ-9). The full English-language version of the questionnaire is provided as Supplementary Files 1–3.

Table 1.

Description, measurement, and theoretical justification of variables included in the ordered probit regression model

Category	Categories	Justification
Depression severity (PHQ-9)	Categorical dependent variable (0 = none/minimal, 1 = mild, 2 = moderate, 3 = moderately severe, 4 = severe)	Measures level of depressive symptoms among students
Age	Recorded as a continuous variable in completed years	Younger students may experience higher depression due to adaptation challenges and academic stress
Gender (ref: Female)	Binary variable (Whether the student is a Female (reference category) or Male)	Female students are often found to report higher depressive symptoms compared with males in similar studies
Year of Study	Categorical variable (2nd year, Third year, Fourth year, Honours, Advanced Diploma)	Senior students may have developed coping skills and support systems, reducing the risk of depression
Funding Status (ref: NSFAS bursary)	Categorical variable based on primary source of financial support (0 = NSFAS bursary, 1 = Bursary (other than NSFAS), 2 = Employment income, 3 = Parent/guardian support, 4 = Student loan and 5 = Other)	Students with stable bursaries or financial aid are less stressed; a lack of secure funding increases Depression risk
Food Security (HDDS)	Measured using the HDDS, capturing the number of food groups consumed in the previous 24 h. Categorised as: (1 = Low (0–4 food groups), 2 = Medium (5–8 food groups)and 3 = High (9–12 food groups))	Low dietary diversity and food insecurity are linked to psychological distress and depressive symptoms
Influence of Food Prices	Influence of Food Prices on Purchasing Decisions. Measured on a 4-point Likert scale: (1 = Strongly disagree, 2 = Disagree, 3 = Agree and)4 = Strongly agree	High sensitivity to food prices indicates financial strain, which may elevate stress and depression
Accommodation (ref: Living at home)	Categorical variable describing current living arrangement: (0 = Living at home, 1 = Private/rented room or flat, 2 = Residence off-campus, 3 = Residence on-campus and 4 = Other)	Living conditions (on/off campus, at home) affect social support and study environment, influencing depression
Marital Status	Categorical variable: (0 = Living together, 1 = Married and 2 = Single)	Relationship status may provide support or increase stress, impacting depression levels
Flood exposure (April 2022 KwaZulu-Natal floods)	Categorical variable indicating the level of impact the floods had on a student (Severely affected, Moderately affected, Mildly affected and Not Affected at all)	Flood-related disruptions to housing, transport, income, and essential services may increase financial strain, academic instability, and psychological distress
Social unrest exposure (July 2021 unrest)	Binary variable indicating if the social unrest affected the student: (1 = Yes,0 = No)	Civil unrest may disrupt livelihoods, mobility, and financial stability, increasing stress and depressive symptoms
Academic Performance Affected	Binary variable (Yes/No): 1 = Yes (academic performance was negatively affected) and 0 = No	Academic challenges or poor performance can increase perceived failure and depressive mood

Depression severity measurement

Depression severity was measured using the Patient Health Questionnaire-9 (PHQ-9), a validated self-report screening tool that assesses the frequency of nine depressive symptoms over the preceding two weeks. Each item is scored on a four-point Likert scale ranging from 0 (“not at all”) to 3 (“nearly every day”), yielding a total score ranging from 0 to 27. Consistent with established clinical and research guidelines, PHQ-9 scores were categorised into five ordered levels of depression severity: none/minimal (0–4), mild (5–9), moderate (10–14), moderately severe (15–19), and severe (20–27). The PHQ-9 has demonstrated strong psychometric properties across diverse populations, including university students and South African samples, and is widely used in large-scale epidemiological studies of student mental health. Its ordinal structure makes it suitable for modelling graded depression severity rather than binary outcomes.

Socio-demographic and economic variables

Socio-demographic information collected included age (in completed years), gender, year of study, marital status, and accommodation type. Economic variables captured students’ primary source of financial support, categorised into bursary funding (NSFAS and non-NSFAS), parental or guardian support, employment income, student loans, and other sources. To assess financial strain related to food affordability, respondents were asked to indicate the extent to which food prices influenced their purchasing decisions using a four-point Likert scale ranging from “strongly disagree” to “strongly agree.”

Food security measurement

Food security status was assessed using the Household Dietary Diversity Score (HDDS), a widely applied proxy indicator of household food access and diet quality. Respondents reported whether they had consumed foods from twelve standard food groups during the previous 24 h. The total HDDS was calculated by summing affirmative responses and subsequently categorised into low (0–4 food groups), medium (5–8 food groups), and high (9–12 food groups) dietary diversity levels. The HDDS has been extensively used in food security research and has demonstrated relevance in assessing dietary adequacy among university student populations.

Exposure to socio-environmental shocks

Exposure to recent socio-environmental shocks was measured through self-reported items capturing the perceived impact of the July 2021 civil unrest and the April 2022 KwaZulu-Natal floods. Exposure to social unrest was captured as a binary variable indicating whether the unrest had affected the respondent’s life in any way (yes/no). Flood exposure was measured using an ordinal scale reflecting the level of impact experienced (not affected, mildly affected, moderately affected, severely affected). These measures were designed to capture indirect disruptions to livelihoods, housing, transport, income, and academic continuity rather than direct traumatic exposure.

Pilot testing and data integrity

Prior to full deployment, the questionnaire was piloted with 20 students to assess clarity, comprehension, and contextual appropriateness. Feedback from the pilot resulted in minor wording adjustments, with no changes to the core constructs or measurement scales. To maintain data integrity, each participant received a unique survey link allowing only one submission, thereby preventing duplicate responses. All responses were collected anonymously, and no personally identifiable information was recorded. The online, self-administered design facilitated broad participation, enhanced privacy when reporting sensitive mental health information, and reduced administrative burden. This approach is consistent with data collection methods employed in national and international university student mental health surveys, including the World Health Organization World Mental Health–International College Student (WMH-ICS) initiative [5] and South African national student mental health studies [6].

Statistical analysis

All data were coded and analysed using Stata version 17. Descriptive statistics (means, standard deviations, frequencies, and %s) were used to summarise socio-demographic characteristics, funding sources, dietary diversity, disaster exposure, and depression severity levels. Multicollinearity among independent variables was checked using Variance Inflation Factors (VIF), with all values below the threshold of 10, indicating no collinearity concerns [33]. Cases with missing values on any variable included in the ordered probit model were excluded via listwise deletion. A review of missing data patterns revealed no systematic differences between complete and incomplete cases, supporting the plausibility of a missing-at-random assumption. The final analytic sample comprised 445 observations. Depression severity, measured as an ordinal variable, was analysed using an ordered probit regression model. This method allowed the estimation of the likelihood of students falling into progressively higher categories of depression severity, conditional n predictor variables. Marginal effects were calculated to interpret changes in probabilities associated with each determinant. Statistical significance was set at p < 0.05, with effects at the 10% level also reported.

Empirical model: ordered probit regression model

Depression severity, as measured by the PHQ-9, produces an ordinal outcome classified into five categories: none/minimal, mild, moderate, moderately severe, and severe depression [3]. Since these categories are ordered but not measured on an interval scale, standard linear regression models would not be appropriate, as they assume continuous outcomes and equidistant intervals between categories [34]. Similarly, binary models such as probit or logit would ignore the multiple thresholds of severity inherent in the PHQ-9 classification [19].

To address this, the analysis employed an ordered probit regression model, which is designed specifically for ordered categorical dependent variables. This model assumes the existence of an unobserved (latent) continuous variable of depression severity,, which is determined by a linear combination of explanatory variables and an error term as shown in Eq. 1:

1

where is a vector of explanatory variables (Table 1), is the vector of parameters, and is an error term assumed to follow a standard normal distribution.

The observed outcome, , is then mapped to discrete categories of depression severity based on threshold parameters (Eq. 2):

2

Category probabilities are given by (Eq. 3):

3

where is the cumulative distribution function of the standard normal distribution, and indexes the depression categories.

The ordered probit model was chosen because it respects the ordinal nature of PHQ-9 outcomes, accounts for the rank-ordering of depression severity and avoids imposing the unrealistic assumption that the difference between categories (e.g., mild vs. moderate) is the same as between others (e.g., moderate vs. severe). Ordered response models are widely applied in health and depression research where outcomes are measured using ranked severity categories. For example, Booysen and Botha [9] employed an ordered probit model to analyse socio-economic determinants of subjective well-being in South Africa, while Contoyannis and Jones [15] and Van Doorslaer and Jones [37] used ordered probit specifications to examine socio-economic gradients in self-reported health status. Similar ordinal latent-variable approaches have also been applied to psychological and mental well-being outcomes, including analyses of mental distress and life satisfaction using ordered probit models [18], and to large-scale university student depression surveys that model symptom severity using ordinal regression frameworks [5]. These precedents support the appropriateness of ordered probit modelling when analysing determinants of depression severity measured on ordered categorical scales.

Although the ordered probit coefficients provide information on the direction of associations, they are not directly interpretable as changes in probabilities. To facilitate substantive interpretation, marginal effects were computed after model estimation. The marginal effect of a variable on the probability of falling into depression category is given by (Eq. 4):

4

where denotes the standard normal probability density function. For continuous variables (e.g., age), marginal effects capture the rate of change in probability associated with a one-unit increase in the variable. For categorical predictors (e.g., gender, funding source, disaster impact), they represent the discrete probability change associated with a shift from the reference category. This two-step approach of estimating the ordered probit model followed by marginal effects. enabled both statistical testing of hypotheses and practical interpretation of how socio-demographic, economic, and environmental factors influenced the likelihood of students being classified in different depression severity categories.

Ethical considerations

The study was conducted in full compliance with ethical standards, following the Declaration of Helsinki, South African Guidelines for Good Clinical Practice, and the South African Medical Research Council (SAMRC) Ethical Guidelines for Research, ensuring participant safety, informed consent, confidentiality, and protection of vulnerable populations.

The study received ethical approval from the Research Ethics Committee (Ethical Clearance Number: RD1/12/2023; Project Registration Number: NSci/03/2023) from the university. Participation was voluntary, and all students provided informed consent before completing the survey. Anonymity and confidentiality were ensured by excluding personally identifying information. Students who scored in the severe category on the PHQ-9 or reported distress were referred to student counselling services for follow-up support.

Results

Descriptive analysis

Table 2 summarises the demographic, socio-economic, and environmental characteristics of the study sample (n = 445) together with the distribution of depression severity measured using the PHQ-9. Based on PHQ-9 categorisation, 9.44% of students reported none to minimal symptoms, 30.79% mild symptoms, and 29.44% moderate symptoms. More severe symptom levels were reported by 18.43% (moderately severe) and 11.91% (severe). Overall, 59.78% of respondents fell within the moderate-to-severe range (PHQ-9 ≥ 10). Comparable levels of psychological distress have been reported among university students internationally and within South Africa [5, 6].

Table 2.

Socio-demographic, economic, food security, and disaster exposure characteristics of the study sample

Variables	Categories	Frequency	%
Depression severity	None to minimal (0–4 PHQ score)	42	9.44
	Mild (5–9 PHQ score)	137	30.79
	Moderate (10–14 PHQ score)	131	29.44
	Moderately severe (15–19 PHQ score)	82	18.43
	Severe (20–27 PHQ score)	53	11.91
Year of study (ref: 2nd year)	2nd year	182	40.90
	third year	163	36.63
	Fourth year	62	13.93
	Honours	2	0.45
	Advanced Diploma	36	8.09
Gender	Female	283	63.60
	Male	162	36.40
Marital status (ref: Single)	Living together	2	0.45
	Married	3	0.67
	Single	440	98.88
Accommodation (ref: Living at home)	Living at home	41	9.21
	Private/rented room or flat	48	10.79
	Residence off-campus	323	72.58
	Residence on-campus	32	7.19
	Other	1	0.22
Influence of prices on the decision to buy food	Strongly disagree	7	1.57
	Disagree	15	3.37
	Agree	139	31.24
	Strongly Agree	284	63.82
Funding (ref: Bursary (nsfas))	Bursary (NSFAS)	324	72.81
	Bursary (other than NSFAS)	29	6.52
	Employment	5	1.12
	Parent/Guardian	50	11.24
	Student loan	8	1.80
	Other	29	6.52
April 2022 floods exposure effect	Severely affected	118	26.52
	Moderately affected	163	36.63
	Mildly affected	104	23.37
	Not Affected at all	60	13.48
HDDS (ref: Low diversity)	Low dietary diversity (0–4)	299	67.19
	Medium dietary diversity (5–8 food groups)	104	23.37
	High dietary diversity (9–12 food groups)	42	9.44
July 2021 social unrest exposure	Yes	354	79.55
	No	91	20.45
Academic performance affected	Yes	192	43.15
	No	253	56.85

		Mean	Std. dev	Min	Max
Age	Continuous	23.34	3.99	18	53

Table 2 further shows that most participants were enrolled in earlier academic years, with 40.90% in the second year and 36.63% in the third year. Smaller proportions were in the fourth year (13.93%) or advanced diploma (8.09%), and only 0.45% were honours students. The sample, therefore, largely represents undergraduate students. Students were predominantly female (63.60%), with males comprising 36.40%. Furthermore, nearly all respondents were single (98.88%), with very small proportions married (0.67%) or living together (0.45%), reflecting the expected relationship status profile of a university-aged population. Most students resided off-campus (72.58%), followed by private rented accommodation (10.79%) and living at home (9.21%). Only 7.19% lived in on-campus residences. These figures indicate that the majority of students relied on private or external housing arrangements rather than university accommodation.

Food-related financial pressure was widely reported. Only 4.94% of students disagreed or strongly disagreed that prices influenced their purchasing decisions, while 95.06% agreed or strongly agreed. This distribution suggests that most students experienced price sensitivity when purchasing food. Similar concerns about food affordability among university students have been reported in previous research on campus food insecurity [11]. Financial support was primarily derived from NSFAS bursaries (72.81%). Smaller proportions relied on parental or guardian support (11.24%), other bursaries (6.52%), or other sources (6.52%), while very few depended on employment (1.12%) or student loans (1.80%). This indicates that most students depended on formal financial aid rather than personal income.

Exposure to the April 2022 floods was common. Only 13.48% reported being unaffected, while the remainder experienced varying levels of impact, including 23.37% mildly affected, 36.63% moderately affected, and 26.52% severely affected. This distribution indicates that the majority of students experienced some degree of disruption related to the floods. A large proportion of students (79.55%) reported being affected by the July 2021 social unrest, while 20.45% reported no impact, indicating widespread exposure within the sample. Dietary diversity levels were predominantly low. Approximately 67.19% of students were classified as having low dietary diversity, 23.37% medium diversity, and only 9.44% high diversity. The HDDS is widely used as an indicator of household food access [36], and the distribution suggests limited dietary variety for many respondents. Nearly half of the respondents (43.15%) indicated that their academic performance had been negatively affected during the previous year, while 56.85% reported no adverse impact. Respondents had a mean age of 23.34 years (SD = 3.99), with ages ranging from 18 to 53 years, indicating a predominantly young adult population with some older or returning students.

Multicollinearity diagnostics

Variance Inflation Factors ranged from 1.02 to 1.30 (mean VIF = 1.12), with all tolerance values exceeding 0.75, indicating no evidence of problematic multicollinearity among explanatory variables. These values are substantially below commonly used thresholds (VIF < 5) [2], suggesting that the regression estimates are stable (Table 3).

Table 3.

Variance Inflation Factors (VIF) for independent variables

Variable	VIF	1/VIF
Funding	1.30	0.769
Year of study	1.28	0.781
Age	1.22	0.820
Accommodation	1.17	0.855
Disaster exposure (July 2021 unrest)	1.13	0.885
Academic performance affected	1.07	0.935
Marital status	1.05	0.952
Disaster exposure (April 2022 floods)	1.05	0.952
Gender	1.04	0.962
HDDS	1.03	0.971
Influence of food prices	1.02	0.980
Mean VIF	1.12

Determinants of depression severity: ordered probit regression

The ordered probit regression results in Table 4 indicate that depression severity among university students is systematically associated with a combination of socio-demographic, economic, food security, and environmental factors. The model was jointly statistically significant (LR χ²(11) = 45.27, p = 0.001), suggesting that the included predictors collectively explain meaningful variation in depression severity. Multicollinearity diagnostics confirmed that coefficient estimates were stable and not distorted by high intercorrelations among explanatory variables.

Table 4.

Ordered probit regression estimates and marginal effects for determinants of depression severity among university students in KwaZulu-Natal

Ordered probit regression analysis				Marginal effects
Depression severity	Coefficient	Std. err	P > z	dy/dx	Std. err	P > z
Year of study	−0.018	0.049	0.709	−0.003	0.008	0.709
Age	−0.023	0.014	0.094*	−0.004	0.002	0.098*
Gender	−0.445	0.107	0.000***	−0.074	0.020	0.000***
Marital status	−0.233	0.337	0.489	−0.036	0.052	0.490
Accommodation	0.101	0.076	0.184	0.015	0.012	0.188
HDDS	−0.156	0.079	0.050**	−0.024	0.012	0.054*
Funding	0.075	0.037	0.043**	0.011	0.006	0.047**
Academic performance affected	−0.020	0.051	0.695	−0.003	0.008	0.695
Influence of prices on the decision to buy food	−0.085	0.075	0.261	−0.013	0.012	0.264
Disaster Exposure: July 2021 unrest (ref: Not affected)	0.254	0.132	0.054*	0.035	0.016	0.036**
Disaster Exposure: April 2022 floods (ref: Not affected)	0.217	0.104	0.037**	0.034	0.017	0.045**
/cut1	−1.761	0.876
/cut2	−0.639	0.874
/cut3	0.165	0.874
/cut4	0.874	0.874

Number of obs = 445 LR chi²(11) = 45.27 Prob > chi² = 0.001 Log likelihood = −649.55885 Pseudo R² = 0.033

Coefficients from the ordered probit model indicate direction and statistical significance of associations with the latent depression severity index. Marginal effects (dy/dx) represent the estimated change in the probability of being classified in the severe depression category (PHQ-9 ≥ 20) associated with a one-unit increase in continuous variables or a discrete change from the reference category for categorical predictors, evaluated at sample means

^*, ^**, ^***Denote significance at the 10%, 5%, and 1% levels respectively

Socio-demographic factors

Age exhibited a negative association with depression severity at the 10% significance level. The marginal effects indicate that each additional year of age was associated with a modest reduction in the probability of being classified in the severe depression category. This pattern suggests that older students may experience slightly lower symptom severity, potentially reflecting greater coping capacity or adaptation to the university environment.

Gender emerged as a strong and robust correlate of depression severity. Male students were significantly less likely than female students to fall into higher severity categories, with a 7.4% point lower probability of severe depressive symptoms. This finding indicates a clear gender gradient in depression severity within the study population, consistent with evidence from university student mental health research showing higher symptom burdens among female students.

Year of study, marital status, and accommodation type were not statistically significant predictors of depression severity in the multivariable model. This suggests that, once economic, food security, and environmental factors were accounted for, academic level and living arrangements did not independently differentiate severity levels within this sample.

Economic and food security factors

Food security, measured through household dietary diversity, was negatively associated with depression severity. Students with higher dietary diversity were less likely to be classified in the severe depression category. The marginal effects indicate that an increase in dietary diversity was associated with a reduction of approximately 2.4% points in the probability of severe depressive symptoms. This finding highlights the relevance of food access and diet quality as correlates of mental health severity rather than merely the presence or absence of depressive symptoms. Funding status was positively associated with depression severity. Students relying on funding sources other than the reference category (NSFAS bursaries) exhibited a higher probability of severe depressive symptoms. Although the magnitude of the marginal effect was relatively small (approximately 1.1%), the association was statistically significant, indicating that financial insecurity or instability may be linked to greater symptom severity among students. The perceived influence of food prices on purchasing decisions was not statistically significant in the multivariable model. This suggests that, after accounting for dietary diversity and funding status, price sensitivity did not independently explain variation in depression severity.

Environmental and disaster-related factors

Exposure to socio-environmental shocks was significantly associated with higher depression severity. Students who reported being affected by the July 2021 civil unrest had a higher probability of being classified in the severe depression category compared with those who reported no impact. Similarly, reported exposure to the April 2022 floods was associated with an increased probability of severe depressive symptoms. These associations are consistent with indirect pathways such as financial disruption, housing instability, transport challenges, and academic interruption; however, the cross-sectional design does not permit causal conclusions about the direction or mechanism of these relationships.

Academic performance

Self-reported negative effects on academic performance were not significantly associated with depression severity in the adjusted model. This indicates that academic disruption, while prevalent within the sample, did not independently differentiate severity categories once socio-economic and environmental variables were taken into account. Taken together, the results indicate that depression severity among university students is patterned more strongly by structural and material conditions than by academic or residential characteristics. Gender, food security, funding stability, and exposure to recent socio-environmental shocks emerged as the most salient correlates of severity. Importantly, the use of an ordered response model reveals that these factors are associated not only with the presence of depressive symptoms but with shifts toward more severe symptom categories, underscoring the importance of addressing broader socio-economic vulnerabilities in student mental health strategies.

Discussion

This study investigated socio-demographic, economic, food security, and disaster-related determinants of depression severity among university students in KwaZulu-Natal using an ordered probit framework. The findings indicate that depression severity among students is shaped not only by individual characteristics but also by broader structural and environmental stressors. Significant associations were observed for gender, age, dietary diversity, funding stability, and exposure to recent socio-political and climatic shocks, suggesting that student depression is embedded within wider socio-economic conditions.

Consistent with a substantial body of international evidence, female students exhibited higher depression severity compared with males. Gender disparities in depressive symptoms have been widely reported across university settings. Large multi-country surveys coordinated by the World Health Organization World Mental Health initiative show consistently higher prevalence of mood and anxiety disorders among female students [5]. Similar patterns have been documented in South Africa, where national surveys of university students report greater psychological distress among women [6]. These findings suggest that female students may experience disproportionate emotional or social stress burdens within higher education contexts. Age was modestly protective, with older students reporting lower severity. This gradient may reflect greater coping capacity, accumulated life experience, or improved adjustment to academic environments. Comparable trends have been observed in student depression research, where first-year and younger students tend to report higher distress during transitional periods [20].

Economic and food security constraints emerged as important correlates of depression severity. Students with lower dietary diversity were more likely to report severe symptoms, reinforcing growing evidence linking food insecurity to poor depression outcomes. The Society for Nutrition Education and Behaviour highlights that food-insecure college students experience higher levels of stress, anxiety, and depression [11]. From a broader perspective, the social determinants framework emphasizes that unmet basic needs including adequate nutrition which act as chronic stressors that undermine mental well-being [28]. The present results align with this literature and suggest that depression interventions in university settings cannot be separated from food access and affordability concerns. Funding instability was similarly associated with higher severity. Students relying on non-bursary funding sources faced elevated risks of depressive symptoms, indicating the importance of financial security for academic and psychological stability [11]. Financial strain has consistently been identified as a predictor of mental distress among young adults and students [28]. In resource-constrained settings, concerns about tuition, accommodation, and daily living costs may compound academic stress, thereby worsening depression outcomes.

Exposure to environmental and socio-political shocks was also significantly associated with depression severity. Students affected by the July 2021 unrest and April 2022 floods showed higher probabilities of severe depressive symptoms. This finding is consistent with disaster depression research demonstrating increased risks of depression, anxiety, and post-traumatic stress following disruptive events. Large empirical review synthesizing evidence from over 60,000 disaster victims reported consistently elevated psychological morbidity among exposed populations [31]. In the university context, such events may interrupt housing, mobility, income, and educational continuity, thereby potentially increasing vulnerability to psychological distress. Khumalo and Plattner [24] have corroborated the narrative on linking the locus of control with depression, as these pointed out issues are beyond the students' control, especially environmental stressors, as that can instantaneously shift the livelihoods of affected victims.

Importantly, these findings align closely with the United Nations Sustainable Development Goals (SDGs). Depression is explicitly embedded within SDG 3 (Good Health and Well-being), which calls for the promotion of depression and psychosocial well-being [40]. The associations observed between dietary diversity and depression severity highlight the relevance of SDG 2 (Zero Hunger), which emphasizes food security and adequate nutrition. Financial instability and unequal access to educational resources intersect with SDG 4 (Quality Education), as psychological distress may undermine academic performance and retention. Moreover, the observed impacts of flooding and civil unrest underscore the importance of SDG 11 (Sustainable Cities and Communities) and SDG 13 (Climate Action), which seek to reduce vulnerability to environmental and climate-related shocks. Together, these linkages suggest that student depression outcomes are closely tied to broader development challenges rather than solely individual-level factors.

From a policy perspective, the results indicate that university mental health strategies should extend well beyond clinical counselling services, and the specific quantitative findings offer actionable guidance. Given that female students showed a 7.4% higher probability of severe depression than male students, gender-responsive support programmes such as targeted peer mentorship and outreach for female students which warrant prioritisation. The association between higher dietary diversity and a 2.4% reduction in the probability of severe symptoms supports investment in campus food programmes such as subsidised meal schemes and on-campus food pantries. The link between non-NSFAS funding status and a 1.1% higher probability of severe symptoms reinforces the case for expanding stable financial aid coverage and emergency bursary mechanisms. Finally, the 3.5% and 3.4% point increases in the probability of severe depression associated with unrest and flood exposure respectively underscore the need for institutional disaster-response protocols that include mental health support as a core component, including rapid psychosocial services and academic continuity provisions.

Several limitations should be noted. First, the cross-sectional design precludes causal inference; all relationships should be interpreted as associations and temporal or causal direction cannot be established. Second, the convenience sample was drawn from a single institution and does not represent the broader university student population in KwaZulu-Natal or South Africa; findings are therefore context-specific and should not be generalised beyond this setting. Third, participation through an online survey may introduce self-selection bias, particularly if students experiencing higher distress were more likely to respond. Fourth, measures were self-reported and may be subject to recall or reporting error. Future longitudinal and multi-site studies would strengthen causal interpretation and external validity.

Despite these limitations, the study contributes new evidence on the intersection of socio-economic vulnerability, environmental exposure, and depression among university students in South Africa. By situating depression within broader structural determinants and development priorities, the findings highlight opportunities for integrated, multi-sectoral approaches to improving student well-being.

Conclusion

This study examined socio-demographic, economic, food security, and disaster-related determinants of depression severity among university students in KwaZulu-Natal using an ordered probit modelling approach. Furthermore, the study captures critical environmental shocks experienced by the province that are of national significance, highlighting how university students are currently affected and how they may be impacted in the future due to global pressures. This research opens up a new dialogue in the country as we observe a progression of national disasters related to climate change and the readiness to support academic discourse from arising stressors impacting human health. The findings demonstrate that depressive symptom severity among students is shaped not only by individual characteristics but also by broader structural vulnerabilities. Gender, age, dietary diversity, funding stability, and exposure to social unrest and flooding were significant correlates of severity, indicating that depression risk is closely linked to financial insecurity, food access constraints, and environmental disruptions.

These results underscore that student depression cannot be addressed solely through clinical or counselling services. Instead, depression outcomes appear embedded within the wider socio-economic and environmental conditions in which students live and study. Interventions that strengthen food security, stabilise financial aid, and improve institutional responses to disaster-related shocks may therefore be critical components of comprehensive student well-being strategies.

From a development perspective, the findings align with several Sustainable Development Goals, particularly SDG 3 (Good Health and Well-being), SDG 2 (Zero Hunger), SDG 4 (Quality Education), and SDGs 11 and 13 (resilience to environmental and climate-related shocks). Improving student mental health thus requires coordinated action across health, education, and social protection systems rather than isolated interventions.

Methodologically, the study contributes evidence from a South African university context using an ordered response framework that preserves the graded nature of PHQ-9 depression severity. This approach provides a more nuanced understanding of how structural stressors shift students into higher symptom categories.

Future research should employ longitudinal designs and broader multi-institutional samples to establish causal pathways and enhance generalisability. Nonetheless, the present findings highlight the importance of integrating psychosocial, economic, and environmental considerations into higher education policies aimed at reducing depression severity and improving student well-being.

Policy implications and recommendations

The findings of this study indicate that depression severity among university students is closely linked to financial insecurity, limited food access, and exposure to environmental disruptions, in addition to demographic characteristics. These patterns suggest that addressing depressive symptoms in higher education settings requires a broader systems-level response rather than relying solely on individual clinical interventions. Several policy and programmatic implications emerge.

Strengthening campus mental health services

Universities should expand routine mental health screening and early identification strategies. The use of brief, validated instruments such as the PHQ-9 during registration periods, academic advising, or student health visits may facilitate early detection of students at risk of severe depressive symptoms. Integrating screening with referral pathways to counselling, peer-support networks, and professional psychological services can improve timely access to care. Increasing the availability of campus-based mental health professionals and tele-mental health options may further reduce barriers to service utilisation.

Improving food security among students

Given the observed association between low dietary diversity and higher depression severity, food access interventions should form part of student support systems. Universities and student affairs departments could implement or strengthen campus food banks, meal voucher schemes, subsidised cafeterias, and emergency food assistance programs. Partnerships with local food suppliers or community organisations may also improve affordable food availability. These initiatives directly support SDG 2 (Zero Hunger) while contributing to improved student well-being.

Enhancing financial stability and bursary support

Financial strain emerged as a significant correlate of depressive symptoms, highlighting the importance of stable funding mechanisms. Policymakers and higher education institutions should prioritise reliable bursary and financial aid programs, including timely disbursement of funds, transparent eligibility processes, and emergency financial relief for students facing unexpected hardship. Expanding financial literacy and budgeting support services may further assist students in managing economic pressures. Such measures align with SDG 4 (Quality Education) by promoting retention and academic success.

Disaster preparedness and resilience planning

The associations between flood and unrest exposure and increased depression severity underscore the need for institutional resilience strategies. Universities located in disaster-prone areas should develop contingency plans that include emergency housing, transport support, academic flexibility, and psychosocial services following disruptive events. Coordinated crisis-response systems and communication channels may mitigate the mental health impacts of environmental and socio-political shocks. These actions support SDG 11 (Sustainable Cities and Communities) and SDG 13 (Climate Action) by strengthening institutional adaptability to external risks.

Integrated, multi-sectoral approaches

Student mental health should be embedded within broader social protection and public health policies rather than treated solely as a clinical issue. Collaboration between university administrations, health services, social development departments, and local governments can facilitate comprehensive support that addresses economic, nutritional, and psychosocial determinants simultaneously. Such integrated strategies are consistent with SDG 3 (Good Health and Well-being) and reflect the interconnections between education, health, and social equity.

Authors’ contributions

T.C.: Conceptualisation; Formal analysis; Methodology; Data curation; Visualization, Writing – original draft; Writing – review & editing. X.M.: Principal investigation; Conceptualization; Methodology; Investigation; Supervision; Writing – review & editing. All authors read and approved the final manuscript and agree to be accountable for all aspects of the work.

Funding

Funded by the Mangosuthu University of Technology through the UCDG8 60469 grant under the Project Registration Number: NSci/03/2023.

Data availability

The data that support the findings of this study are not publicly available due to ethical restrictions and the conditions under which informed consent was obtained, which limit data sharing to protect participant confidentiality. De-identified data may be made available from the corresponding author upon reasonable request and subject to approval by the relevant institutional ethics committee.

Declarations

Ethics approval and consent to participate

Ethical approval for this study was obtained from the Mangosuthu University of Technology Research Ethics Committee (MUT-REC),  (Ethical Clearance Number: RD1/12/2023; Project Registration Number: NSci/03/2023). At the time of approval, the Research Ethics Committee was chaired by Dr Devashan Naidoo, with committee members Dr Joseph Bwapa, Dr Wiseman Nsele, Dr Thulisile Duma, and Dr Annette Mienie, representing different faculties within the institution. All participants provided informed consent electronically prior to participation. Participation was voluntary, anonymity and confidentiality were assured, and no personally identifiable information was collected.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.