South Africa’s health data does not live in one place, and the model your Department of Health actually relies on for its official HIV numbers — Thembisa, built at the University of Cape Town — has been the basis for South Africa’s official UNAIDS estimates every year since 2017. A public health dissertation that only cites Stats SA is missing most of the picture. Eight sources cover the ground a South African public health master’s or honours study is likely to need: modelled estimates, laboratory surveillance, a mortality tracker, a facility scorecard, a household survey and one open microdata archive.
Eight public health data sources, at a glance
| Source | Custodian | What it gives you | Update cycle | Access |
|---|---|---|---|---|
| Thembisa model | University of Cape Town, with UNAIDS and the National Department of Health | Modelled HIV and TB prevalence, incidence, ART coverage and mortality, national and provincial | New version roughly annually (version 5.0 is the current working-paper series) | thembisa.org — downloads and working papers, free |
| GERMS-SA | National Institute for Communicable Diseases (NICD) | Laboratory-confirmed surveillance of antimicrobial resistance and selected invasive pathogens | Ongoing, with periodic surveillance reports | nicd.ac.za, free summary reports |
| Notifiable Medical Conditions (NMC) surveillance | NICD | Case counts for every statutorily notifiable disease, by province | Monthly report | nicd.ac.za, free |
| National Cancer Registry | NICD / National Health Laboratory Service | Pathology-confirmed cancer incidence by type, age and province | Annual, with a multi-year reporting lag typical of cancer registries | nicd.ac.za, free |
| Report on Weekly Deaths in South Africa | South African Medical Research Council, Burden of Disease Research Unit | Near-real-time all-cause mortality and excess-deaths tracking | Weekly | samrc.ac.za, free |
| District Health Barometer | Health Systems Trust | Facility and programme performance indicators — coverage, waiting times, staffing — by district | Annual | hst.org.za, free |
| General Household Survey (health module) | Stats SA | Self-reported health-facility use, satisfaction and health insurance cover, household level | Annual | statssa.gov.za, free; microdata via SuperWEB2 |
| uMkhanyakude HDSS and the ALPHA Network | Africa Health Research Institute, distributed via DataFirst | Longitudinal individual-level health and demographic surveillance microdata, rural KwaZulu-Natal | Historical panel, 2000–2016 for the South African site | datafirst.uct.ac.za, registration required, free |
Read the update-cycle column before you commit to a source. A weekly mortality tracker and a cancer registry with a multi-year lag answer very different research questions, and mismatching them against your timeline is the fastest way to lose a term.
1. The Thembisa model, for HIV and TB estimates you cannot collect yourself
No South African postgraduate is going to run a national HIV prevalence survey. Thembisa exists so nobody has to: it is a mathematical model of the South African HIV epidemic, built and maintained at the University of Cape Town in collaboration with UNAIDS and the National Department of Health, and since 2017 it has been the source behind South Africa’s official UNAIDS estimates. The model has recently been extended to project tuberculosis alongside HIV, which matters because the two epidemics are so entangled here that a TB dissertation without an HIV-adjusted denominator is incomplete.
Cite the working paper for the version you use — the series runs from Thembisa 1.0 through the current 5.0 — not a summary blog post, because the assumptions and outputs change between versions and an examiner who knows the model will ask which one.
Who it suits
A dissertation that needs a national or provincial HIV or TB estimate as context or as a denominator, rather than primary data collection. Combine it with a district-level source below if your unit of analysis is smaller than a province.
2. GERMS-SA, for antimicrobial resistance and hospital-acquired infection
The NICD runs GERMS-SA as its laboratory-based surveillance system for selected invasive pathogens and antimicrobial resistance, drawing on confirmed laboratory results rather than clinical suspicion. For a dissertation on infection control, antibiotic stewardship or resistant organisms in a South African facility, this is the national baseline your own facility-level findings should be measured against.
Who it suits
Nursing, medical microbiology, clinical pharmacology and infection-control topics that need a resistance benchmark rather than raw case counts.
3. Notifiable Medical Conditions surveillance, for outbreak and trend questions
Every disease that South African law requires a clinician to notify — from measles to cholera to viral haemorrhagic fevers — is compiled by the NICD into a monthly Notifiable Medical Conditions report, broken down by province. It sits alongside disease-specific outputs such as the NICD’s weekly measles and rubella situation reports and its respiratory pathogens surveillance, which are useful if your topic tracks a single disease through a season rather than the full notifiable list.
Who it suits
Epidemiology-flavoured dissertations tracking a specific notifiable disease’s trend, seasonality or provincial distribution over a defined period.
4. The National Cancer Registry, for oncology and screening research
Cancer incidence in South Africa is reconstructed from pathology laboratory reports rather than a population registry that captures every case, which the National Cancer Registry states plainly and which your methodology chapter should state too. It is run by the NICD in partnership with the National Health Laboratory Service and is the only national source for cancer incidence by type, age band and province.
Who it suits
Oncology nursing, screening-programme evaluation and health-promotion dissertations that need an incidence baseline. Expect a multi-year lag between diagnosis and the figures appearing in a published edition, and say so rather than treating the newest available year as current.
5. The Report on Weekly Deaths, for mortality and excess-death questions
The South African Medical Research Council’s Burden of Disease Research Unit tracks all-cause mortality using civil registration data and publishes a Report on Weekly Deaths in South Africa on a rolling weekly basis — the most recent edition at the time of writing dated 25 August 2026. It is the tool that let researchers quantify excess mortality during the COVID-19 pandemic, and it remains the fastest way to see whether deaths in a province or age band are running above or below the seasonal norm right now.
Who it suits
Any dissertation asking whether mortality has shifted — from a policy change, a disease outbreak, a heatwave or a service disruption — because it is close to real time where most health data is not.
6. The District Health Barometer, for facility and programme performance
Health Systems Trust, an independent non-profit whose stated mission is strengthening health systems in South Africa, compiles the District Health Barometer as its flagship annual publication: facility- and programme-level indicators — antenatal coverage, immunisation rates, staffing ratios, waiting times — broken down to district level from the routine data district health information offices submit. It is the natural comparator for any dissertation set in a specific district or comparing districts against each other, and the routine reporting system that feeds it is the same District Health Information System your facility’s own monthly statistics come from.
Who it suits
Health systems, public administration and nursing management dissertations comparing performance across districts or tracking one district’s indicators over several editions. A dissertation set inside education rather than health faces the same district-versus-national choice; the guide to where education dissertations find their datasets works through it from the schooling side.
7. The General Household Survey, for what people actually experience
Every other source above counts events inside the health system. Stats SA’s General Household Survey asks households directly: which facility they used, how satisfied they were, whether anyone in the household has medical scheme cover. It is the closest thing South Africa has to a population-representative measure of the patient experience, and because Stats SA has run it annually for two decades, it is one of the few sources here that supports a trend analysis over time. The guide to getting South African data through Stats SA and SuperWEB2 covers the account setup and table-building steps this survey shares with every other Stats SA release.
Who it suits
Health-seeking-behaviour, health-financing and patient-satisfaction dissertations that need a population denominator rather than a facility-level count.
8. The uMkhanyakude HDSS and the ALPHA Network, for longitudinal individual-level data
If your question needs to follow the same people over years rather than count events, the Africa Health Research Institute’s health and demographic surveillance system in uMkhanyakude, rural KwaZulu-Natal, is the South African site inside the international ALPHA Network of population HIV and mortality studies. DataFirst distributes the South African panel, covering 2000 to 2016, to registered researchers at no cost, alongside comparable longitudinal sites in Malawi, Zimbabwe and Tanzania that support cross-country comparison.
Who it suits
A dissertation with a genuinely longitudinal question — HIV incidence over time, mortality by household characteristic, migration and health — that a cross-sectional survey cannot answer, and where a facility- or district-level pilot is not the point.
Match the source to your question before you register your topic
Four questions narrow eight sources to the one or two you actually need. Is your question modelled or measured — Thembisa if you need an estimate nobody has directly counted, one of the surveillance systems if you need a confirmed case. Is your unit of analysis a person, a facility, a district or the country — the HDSS panel, the District Health Barometer, and the national surveillance systems respectively answer at those different levels, and mixing levels without saying so is a common examiner objection. Does your timeline tolerate a lag — a weekly mortality report and a cancer registry with a multi-year lag cannot both anchor a one-year study without an explicit justification. And is your access free and immediate, or does it require registration — DataFirst’s HDSS access needs an application, so start that process before you finalise your proposal, not after.
Whichever source anchors your study, the same decisions about test choice follow once you have the data in hand; the guide to choosing the right statistical test for a South African dissertation picks up from exactly this point. If your dissertation sits inside the higher-education system rather than the health system, the companion figures in the roundup of South African postgraduate enrolment statistics follow the same custodian-and-edition discipline. And every figure you quote from any of these eight sources needs a citation naming the custodian and the version or edition you used, in the Harvard variant your faculty sets out in the comparison of Harvard and APA referencing at South African universities.
Build the data chapter once you have chosen your source
Choosing between eight national data systems, each with its own access process, update cycle and citation convention, is exactly the kind of decision that stalls a public health dissertation for weeks. Tesify drafts the data and methods section once you have picked your source, keeps the citation consistent with the custodian and version you name, and flags where a lag or a modelled estimate needs a limitation statement — while the underlying data choice, as your faculty’s research integrity policy requires, stays a decision you make and can defend. Draft your public health data chapter with Tesify, then run it past the similarity check before you submit.
Frequently asked questions
Do I need ethics clearance to use these datasets?
Secondary analysis of de-identified, publicly released data such as the General Household Survey or the District Health Barometer usually needs a lighter ethics review than primary data collection, but every South African university still requires you to apply. Individual-level HDSS data carries stricter data-use agreements set by the custodian. Check with your faculty’s research ethics committee before assuming any dataset is exempt.
Which source should I use if my dissertation is about one hospital or clinic?
Start with your facility’s own routine District Health Information System data, then benchmark it against the District Health Barometer for your district, since the barometer is compiled from the same routine reporting system.
Is Thembisa the same as a national HIV survey?
No. Thembisa is a mathematical model that combines multiple data streams, including survey and surveillance data, into a single set of estimates; it does not itself collect new data from participants. State this distinction explicitly if you use its outputs.
How current is the data in these sources?
It varies widely by design, from the Report on Weekly Deaths, updated weekly, to the National Cancer Registry, which typically lags several years behind diagnosis because of how pathology-based registries are compiled. Check the most recent edition date on the custodian’s own site before you write your data chapter, since the lag can change between editions.
Can I combine two of these sources in one dissertation?
Yes, and it often strengthens a study — using the District Health Barometer for context alongside your own facility data, for instance. State clearly which claims come from which source and at what level of aggregation, since conflating a national estimate with a facility-level finding is a common examiner objection.
What if the source I need does not cover my province or district?
Say so in your limitations rather than substituting a national figure without comment. The District Health Barometer and the General Household Survey both report to provincial and district level for most indicators, which covers most gaps; the HDSS panel is geographically limited to uMkhanyakude by design.