This is a repeated cross-sectional study conducted in April 2024 [now referred to as Panel 3], that followed the first and second nationally representative panels conducted in September–October 2021 (Panel 1: n = 3,402) and May–June 2022 (Panel 2: n = 3,459) respectively (12, 13). In Panel 3, face-to-face interviews were conducted with 3,171 respondents in all provinces across South Africa (Figure 1).

Respondents provided written informed consent as well as the study obtained ethical approval from the Human Research Ethics Committee (Non-Medical) of the University of the Witwatersrand, South Africa (H21/06/36). Data collection followed the same collection procedures as previously outlined in Panel 1 (12) and Panel 2 (13).

All data were captured in real time using tablets and managed centrally through a secure database system. Respondent household demographic information (age, gender, marital status, employment, education level and household assets) was collected. In this study, a household asset score was subsequently computed in alignment with the Demographic and Health Surveys household questionnaire and used as an indicator of socioeconomic status (SES). This involved counting all key functioning household amenities (e.g., refrigerator, washing machine, television, computer, etc.). In this repeated cross-sectional study, tertiles of the household asset score were additionally calculated and used as a measure of economic differentiation (18–20). Information related to respondent provincial and community size (metropolitan, city/town, rural/village) were additionally recorded.

To assess the respondent’s level of mental health (depression or anxiety), the Patient Health Questionnaire (PHQ-9) (21) and the Generalized Anxiety Disorder (GAD-7) scale was utilized (22). These scales include either 9 (PHQ-9) or 7 (GAD-7) questions, with responses rated on a four-point Likert scale ranging from 0 (“Not at all”) to 3 (“Nearly every day”). Depression levels were classified into five categories: minimal, mild, moderate, moderately severe, and severe, based on PHQ-9 scores of 0–4, 5–9, 10–14, 15–19, and 20–27, respectively (21). Anxiety levels were classified into four categories: minimal, mild, moderate and severe, based on GAD-7 scores of 0–4, 5–9, 10–14 and15–21, respectively (22). Binary depression or anxiety was classified as a PHQ-9 or GAD-7 score of 10 or higher, respectively (21, 22). Additionally, to assess adverse childhood experiences (ACEs) of each respondent, a 12-item ACE questionnaire was used, which collected an individuals’ retrospective report of specific adversities encountered during the first 18 years of life (23). Exposure to ACEs was defined through 12 types of experiences grouped into three categories: emotional and/or physical abuse, sexual abuse, and household dysfunction (23). An overall ACE score was subsequently calculated based on the number of affirmative responses. The ACE score was further classified into three exposure groups, namely, 0 (no exposure), 1–3 (intermediate exposure), and 4–12 (high exposure) based on previous ACE scoring (24).

To evaluate the overall health of the respondent, individuals were asked a series of health-related questions regarding known chronic conditions (heart attack, stroke, high cholesterol, diabetes, overweight/obesity, HIV/AIDS, asthma/chronic obstructive pulmonary disease, joint/muscle problems (e.g., arthritis, gout), tuberculosis, cancer, liver disease, mental health issues (e.g., depression, anxiety, bipolar disorder), chronic kidney disease, and hypertension/high blood pressure). As multimorbidity is commonly defined as the presence of multiple, often chronic, health conditions that coexist, an overall multimorbidity score was calculated based on the total number of chronic conditions a respondent indicated, including mental health. The multimorbidity score was further categorized into three groups, namely, group 1 (0–1 conditions), group 2 (2 conditions), and group 3 (≥3 conditions).

Health literacy was measured using the HLS-EU-Q47, which includes three distinct groups of questions covering healthcare, disease prevention, and health promotion (9). The HLS-EU-Q47 uses a four-point Likert scale, ranging from 1 (“very difficult”) to 4 (“very easy”). Respondents who answered “I do not know” to any of the 47 questions were assigned a score of zero. The overall score was subsequently converted into a scale of 50 and further categorized into four levels, namely, 0–25 (inadequate), >25–33 (problematic), >33–42 (sufficient), and >42–50 (excellent). The HLS-EU-Q47 instrument has not been validated in a South African context, however, in our population, the Cronbach’s Alpha was calculated at 0.972 which is indicative of excellent internal consistency.

For all statistical analyses, IBM® SPSS® version 29 (IBM Corporation, Armonk, New York), Stata® version 18.0 (StataCorp, College Station, TX, USA), and GraphPad Prism version 5.03 for Microsoft® Windows (GraphPad Software, San Diego, California, USA) were employed to analyze and visualize the data. Additionally, QGIS (Penn Libraries, Philadelphia, PA) was used to map and scale the geographical location of South Africa (Figure 1).

All statistics were weighted to reflect the most recent census of the South African population aged 18 years and older. The weighting matrix accounted for factors such as age, sex, population group, home language, and provincial distribution to equate for national representativity. In stratified analyses, post-stratified weights were not used as the primary objective was to assess associations within subgroups rather than to produce population level estimates. As such, results reflect weighted estimates within strata based on the survey sample. Proportions across sociodemographics (age, sex, marital status, education level, employment, SES, and urbanicity) and provinces were calculated using crosstabs, with significant differences identified by Chi-square tests and presented as percentages. Univariate and multivariable adjusted binary logistic regressions were performed to assess the odds of probable depression (PHQ-9 score ≥10), probable anxiety (GAD-7 score ≥10), and multimorbidity (chronic conditions excluding mental health) in adulthood, considering either ACEs or ACEs combined with several sociodemographics (age, sex, marital status, education level, employment, SES, and urbanicity) as confounders.

A generalized structural equation model (gSEM) was constructed a priori to examine the relationship between SES, mental health, and whether this relationship was mediated by health literacy and/or multimorbidity. Direct (unmediated), indirect (mediated), and total effects were computed and recorded, with the proportion of the total effect mediated also calculated. Pathways were modified and variables added or removed iteratively, with the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) of each model being compared. The final model was chosen based on its low information criterion values and high theoretical relevance. Direct, indirect, and total effects were estimated using non-linear combination methods.

To prevent multicollinearity in our statistical models, mental health was excluded from the multimorbidity score in cases where it correlated with mental health variables (i.e., any models that included depression, anxiety, or ACEs as dependent or independent variables).

We used the Social Determinants of Health as a theoretical framework, specifically the Fundamental Cause Theory (25), which was particularly relevant for our data analysis; this model aligned well with the construction of the a priori gSEM model, as it accounted for the impact of socio-economic factors (SES) on health outcomes (multimorbidity) while considering psychosocial (mental health) and cognitive-behavioral (health literacy) mediators.

The full characteristics and associations of Panel 1 (12) and Panel 2 (13) have been described elsewhere. In this current sample, a total of 3,171 respondents (female: 50.4%; male: 49.6%) were included for repeated cross-sectional analysis (Table 1). Respondents were predominantly young adults aged 25–34 years (29.6%). The largest proportion of respondents were those who reported a marital status of single (62.9%), employed (47.2%), an education level of graduated high school or equivalent (53.1%), and resided in the Gauteng province [n = 1,087 (27.0%)] (Figure 1). When compared to Panels 1 (2021) and 2 (2022), respondent demographics were closely aligned to both prior panels.

Majority of the respondents scored as minimal risk (56.0%), while those scoring in more severe depressive categories were 22.4% (mild); 14.3% (moderate), 5.6% (moderately severe) and 1.8% (severe) respectively. We calculated an overall probable depression prevalence of 21.7% across South Africa (Table 1; Figure 2) which is lower than that reported in Panel 1 (25.7%) and Panel 2 (26.2%). Contrary to Panel 1 and Panel 2, Eastern Cape province reported the highest prevalence of probable depression (30.1%), and Free State province reported the highest probable anxiety (29.2%) (Figure 2; Supplementary Table S1). Childhood adversity was highest in Limpopo province with a mean score of 2.29 (SD: 2.26) (Supplementary Table S1). Additionally, comparable to Panel 2, the most common type of ACE reported was household dysfunction (40.1%) (Table 1).

Supplementary Table S2 presents the prevalence of mental health and multimorbidity among respondents stratified by socio-demographic determinants. In line with Panels 1 and 2, the prevalence of probable depression and anxiety was higher in women (≥14.0%); widowed, divorced, or separated (≥17.7%) and/or with only a basic level of education, i.e., primary school and/or partial secondary school (≥18.3%). Comparable to Panel 1, probable depression was highest in those who were unemployed (25.3%). Respondents in Panel 3 also showed a comparable higher prevalence of probable depression and high ACE exposure in those respondents with an SES score in the lowest tertile (≥19.2%). In contrast to Panels 1 and 2, probable depression was higher in middle aged adults (45–54 years: 25.4%), probable anxiety was highest in those aged 35–44 years (16.9%) and those in the highest ACE exposure groups were men (17.9%); widowed, divorced or separated (20.0%); in the age range of 25–34 years (18.7%), currently retired (20.0%). Multimorbidity of 3 or more morbidities across the socio-demographics was comparable to Panel 2 (65 + years: 23.4%; women: 8.6%; widowed, divorced or separated: 17.7%; retired: 21.1%).

We repeated the univariate and multivariable adjusted binary logistic regressions (Supplementary Table S3) as seen in both Panels 1 and 2 to further corroborate the odds of having either probable depression or anxiety with higher levels of ACE exposure (model 1), or having ACE exposure, independent of socio-demographic characteristics (model 2). We confirmed the results determined in both panels showing that the likelihood of having probable depression (OR, 1.123 [95% CI 1.123; 1.124]) or probable anxiety (OR, 1.122 [95% CI 1.121; 1.122]) increases with each standard deviation increase in the ACE score (p < 0.001), independent of several socio-demographic determinants. We also reassessed the odds of having a higher number of morbidities with higher levels of ACE exposure (Supplementary Table S4) and confirmed that, respondents who reported having 0–1 morbidities were 80% less likely (OR, 0.796 [95% CI 0.796; 0.796]) to have experienced ACEs (p < 0.001) while, the likelihood of having 2 morbidities or 3 or more morbidities increased by 19% (OR, 1.193 [95% CI 1.193; 1.194]) or 21% (OR, 1.205 [95% CI 1.205; 1.206]) respectively, with each standard deviation increase in the ACE score (p < 0.001).

To assess the impact of SES, health literacy, multimorbidity, and mental health (Figure 3; Table 2), a gSEM was constructed a priori. In this model, SES was specified to influence both the outcomes (depression and multimorbidity) and the mediators (childhood adversity, anxiety, and health literacy), with childhood adversity and health literacy also modeled as influencing anxiety. The results revealed significant direct effects, specifically, SES on depression (p < 0.001), and ACE on multimorbidity (p < 0.001). The relationship between SES and depression was partially mediated by health literacy (4.2%) and ACE exposure (4.3%). In contrast, the impact of SES on multimorbidity was fully mediated by ACE (23.7%) and depression (54.7%). Furthermore, ACE was found to have a significant indirect and total effect on depression (p < 0.001), fully mediated by multimorbidity (75.3%) and anxiety (89.8%). Additionally, the effect of ACE on multimorbidity was partially mediated by depression (15.3%). These findings therefore highlight the complex interrelationships among these variables.

This study highlights the complex interplay between SES, childhood adversity, health literacy, and mental and physical health outcomes. The findings suggest that improving health literacy and addressing childhood adversity may help mitigate the mental health and multimorbidity burdens associated with socioeconomic disadvantage. The mediating roles of depression, anxiety, ACE and multimorbidity underscores the need for integrated care models that address both mental and physical health. While the cross-sectional design limits causal inference, the results support targeted, upstream interventions and policies aimed at the social determinants of health (i.e., SES, health literacy) to improve long-term health outcomes.

To conclude, the study reveals that one in five South Africans experience depressive symptoms, with notable differences in the prevalence of mental health across the nine provinces. The study further highlights childhood adversity, particularly household dysfunction, as a significant factor linked to mental health issues and increased multimorbidity. Additionally, health literacy was found to mediate the relationship between SES and depression, suggesting that lower health literacy may heighten vulnerability to mental health challenges. These findings underscore the urgent need for targeted interventions to address childhood adversity, improve health literacy, and enhance mental health resources across regions to reduce the burden of mental health in South Africa.