Patterns of depression prevalence and antidepressant use in South Africa, 2002-2024: a system dynamics modelling perspective

Using a system dynamics model calibrated to national survey data, this study reveals that while depression prevalence in South Africa has remained relatively stable from 2002 to 2024, antidepressant use remains low overall with a stark disparity between the private and public health sectors.

The Gist

Imagine South Africa's population as a massive, bustling city with millions of residents. For years, city planners (the researchers) have been trying to understand a hidden problem: depression. They know it's a major issue, but the data they have is like a jigsaw puzzle with missing pieces and some pieces from different puzzles mixed in. Some surveys say depression is rare; others say it's everywhere. Some count people who are currently sad, while others ask people to remember if they were ever sad years ago (which is hard to do accurately).

To solve this puzzle, the researchers built a digital "City Simulator" (a system dynamics model). Instead of just looking at snapshots, this simulator runs a movie of the city's life from 1985 to 2024, tracking how people move in and out of different emotional states.

Here is what the simulation revealed, explained simply:

1. The "Weather" of Depression

The researchers found that the "weather" of depression in South Africa has been surprisingly stable over the last 20 years.

• The Trend: About 5% of adults are currently experiencing depression. It's like a steady, low-hanging cloud that hasn't cleared up, though it got a little heavier during the COVID-19 pandemic (a temporary storm) before settling back down.
• Who is most affected? The "cloud" hangs lower over women and older adults (60+). Men and younger people have a slightly clearer sky, but they aren't immune.

2. The "Amnesia" Problem (Lifetime vs. Current)

This is the most surprising part of the study.

• The Old View: Past surveys asked people, "Have you ever been depressed?" Most people said "No" or "Maybe once." This led to the belief that only about 10-15% of people are ever at risk. It was like thinking only a few people in the city have ever caught a cold.
• The New View: The simulator looked at the pattern of how depression comes and goes. It found that depression is often like a fickle visitor that people forget to mention. Because people forget past episodes (especially if they weren't severe or didn't get treatment), the old surveys underestimated the problem.
• The Reality: The model suggests that nearly 70% of South African adults will experience at least one episode of depression in their lifetime. It's not a rare disease affecting a small group; it's a common human experience that affects the vast majority of the population at some point.

3. The "Medicine Cabinet" Gap

Even though the problem is huge, the solution is hard to find for most people.

• The Private Sector (The VIP Lounge): People with private health insurance are like VIPs. About 11% of them are taking antidepressants. They have easy access to doctors and pharmacies.
• The Public Sector (The General Crowd): For the rest of the population (90% of the country), access is extremely difficult. Only 0.9% of people in the public system are on medication.
• The Bottleneck: Imagine a line of 100 people who just developed depression. In 2024, only 12 of them managed to start treatment. The system is overwhelmed, and the "gatekeepers" (doctors) are too few to let everyone in.

4. Why the Gap Exists

The study highlights a few reasons why the "VIPs" get help and the "General Crowd" doesn't:

• Rules of the Road: In South Africa, antidepressants are tightly controlled "scheduled" drugs. Only doctors can prescribe them, and they need to be very careful. But most public clinics are run by nurses, who often can't prescribe these drugs. It's like having a fire truck that only a specific driver can operate, but that driver is rarely at the station.
• Stigma and Gender: Men are less likely to ask for help than women, partly due to cultural norms. It's like men being taught to fix a flat tire themselves rather than calling for a tow truck, even when the car is broken.
• The "Chronic" Misconception: The study suggests that because depression often happens to people who aren't in a "high-risk" group (it can happen to anyone), we can't just focus on treating a small group of "repeat offenders." We need to build a system that can handle a flood of new cases from the general population.

The Bottom Line

The researchers used their "City Simulator" to tell us two main things:

1. Depression is more common than we thought: It's not just a problem for a few "sick" people; it's a widespread experience that touches most of us eventually.
2. The treatment system is broken: While the rich get the medicine they need, the poor are left waiting. The gap between the private and public sectors is a chasm.

The Takeaway: To fix this, South Africa needs to stop treating depression like a rare disease affecting a few "high-risk" people and start treating it like a common public health issue. They need to train more nurses to prescribe medication, reduce the stigma (especially for men), and build a system that can handle the fact that depression is a frequent visitor in almost everyone's life.

Technical Summary

1. Problem Statement

Depression is the leading cause of disability in South Africa, yet data regarding its prevalence and the utilization of antidepressants are fragmented, inconsistent, and often rely on screening instruments that may overestimate true prevalence. There is a lack of longitudinal data to distinguish between transient depressive episodes and chronic recurrence, leading to uncertainty about the lifetime risk of depression and the "at-risk" population size. Furthermore, there is a significant gap in understanding treatment access inequalities between the public and private health sectors. The study aims to synthesize disparate data sources to estimate trends, inequalities, and the plausibility of different hypotheses regarding depression heterogeneity (i.e., whether depression is concentrated in a small high-risk group or distributed across the population).

2. Methodology

The authors developed a deterministic system dynamics model of the South African population aged 15 and older, calibrated using a Bayesian framework.

• Model Structure:

  • Population Stratification: The population is stratified by age (15–90+), sex, and HIV status/stage (negative, positive/undiagnosed, diagnosed/untreated, on ART). Demographic inputs are derived from the Thembisa HIV model.
  • Risk Stratification: Individuals are divided into high-risk (genetic/biochemical factors, adverse childhood experiences, high recurrence risk) and low-risk (single episode, low recurrence) groups.
  • Compartments: Within each stratum, individuals transition between states: Never depressed, Currently depressed (untreated/treated), and Recovered (untreated/treated/previously treated).
  • Incidence Drivers: Depression incidence is modeled as a function of age, sex, HIV status, and specific risk factors: exposure to physical violence, binge drinking prevalence, chronic conditions (arthritis, asthma, IHD, stroke), and excess mortality during the COVID-19 pandemic.
  • Treatment Dynamics: The model simulates health-seeking behavior (sex-dependent), antidepressant initiation rates (increasing over time), treatment efficacy (reduced recurrence and faster recovery), and discontinuation rates (higher in acute phases).

• Calibration & Data Integration:

  • Prevalence Data: Calibrated against nine nationally representative household surveys (2002–2024) using different tools (CIDI, CES-D, PHQ-9). The model adjusts for the ratio of screening tool prevalence to "true" diagnostic prevalence.
  • Longitudinal Data: Used National Income Dynamics Study (NIDS) panel data to calibrate the association between initial and subsequent depression episodes (recurrence risk).
  • Antidepressant Data: Calibrated against 18 data points from surveys/cohorts and 11 data points from drug dispensing records (2008–2023), accounting for heterogeneity between public/private sectors and chronic disease cohorts.
  • Bayesian Approach: Prior distributions were specified for parameters based on literature. Posterior distributions were estimated using Incremental Mixture Importance Sampling (IMIS) to generate means and 95% confidence intervals (CIs).

• Sensitivity Analyses:

  • Model A (Base): 7% high-risk population.
  • Model B: Added a linear time trend for incidence.
  • Model C: Excluded COVID-19 effects.
  • Model D: Assumed a "high-risk only" hypothesis (15% of population at risk, zero incidence in low-risk group).
  • Model E: Intermediate hypothesis (15% high-risk, but non-zero incidence in low-risk group).
  • Evaluation: Models were compared using Bayes factors and the Number Needed to Treat (NNT) to prevent one depression case.

3. Key Contributions

• Data Synthesis: Successfully integrated cross-sectional prevalence surveys, longitudinal panel data, and drug dispensing records into a unified dynamic framework.
• Heterogeneity Testing: Provided empirical evidence against the "small high-risk group" hypothesis. The study demonstrates that models assuming depression is confined to a small, highly recurrent sub-population fail to reproduce observed longitudinal associations in the data.
• Lifetime Risk Estimation: Challenged the conventional view of low lifetime prevalence (10–15%) by estimating a much higher cumulative risk, attributing discrepancies in literature to recall bias in retrospective surveys.
• Inequality Quantification: Quantified the stark disparity in treatment access between South Africa's private and public sectors.

4. Key Results

• Prevalence Trends:
  • Point prevalence of depression slightly declined from 5.1% (2002) to 4.5% (2024), with a transient spike during the COVID-19 period.
  • In 2024, prevalence was higher in women (5.3%) than men (3.6%) and highest in adults aged 60+.
  • Lifetime Prevalence: Estimated at 70.6% (95% CI: 67.8–73.6%) for adults, significantly higher than retrospective survey estimates. This suggests a large portion of the population experiences at least one episode, rather than a small group experiencing chronic recurrence.
• Antidepressant Use:
  • Use increased from 1.0% (2008) to 2.8% (2024).
  • Inequalities: In 2024, 11.0% of the private sector population used antidepressants versus only 0.9% in the public sector. Women were 4.4 times more likely to use antidepressants than men.
  • Treatment Gap: The ratio of new antidepressant initiations to new incident depression cases in 2024 was only 0.12, indicating a massive treatment gap.
• Model Comparison:
  • Model A (Base) provided the best fit to the data.
  • Model D (High-risk only) failed to converge when calibrated to both prevalence and longitudinal data, overestimating recurrence. This supports the conclusion that depression risk is widely distributed, not concentrated in a small sub-population.
  • NNT: The Number Needed to Treat to prevent one case of depression was approximately 11 in the base model, compared to 4.5 in the rejected "high-risk only" model.

5. Significance and Implications

• Policy Shift: The finding that lifetime risk is high (70.6%) and not confined to a small high-risk group implies that strategies focusing solely on treating recurrent cases in a specific sub-population are insufficient. Instead, broad preventive strategies targeting social determinants (violence, poverty, chronic disease) are necessary to reduce population-level incidence.
• Health System Reform: The extreme disparity in treatment access (11% in private vs. 0.9% in public) highlights a critical failure in the public health system. The authors advocate for regulatory changes to allow nurse-initiated antidepressant treatment, as the current requirement for psychiatrist oversight is a major bottleneck given the scarcity of psychiatrists in the public sector.
• Methodological Advancement: The study demonstrates the utility of system dynamics modeling in low-resource settings where routine mental health data is scarce. It provides a template for other low- and middle-income countries to synthesize fragmented data for strategic planning.
• Recall Bias Correction: The study offers a robust explanation for the discrepancy between low retrospective lifetime prevalence estimates and high dynamic model estimates, attributing it to substantial recall bias in cross-sectional surveys, particularly for untreated or less severe episodes.

In conclusion, the paper argues that depression in South Africa is a widespread, population-level issue with a significant treatment gap, necessitating a shift from narrow, high-risk-targeted interventions to broader, systemic approaches that address social determinants and expand access to treatment in the public sector.