Can non-fatal burden estimates from the Global Burden of Disease study be used locally? An investigation using models of stroke and diabetes for South Africa

• Main Authors: Victoria Pillay-van Wyk (Author), Rifqah Abeeda Roomaney (Author), Mweete Debra Nglazi (Author), Oluwatoyin Folashade Awotiwon (Author), Judith M Katzenellenbogen (Author), Tracy Glass (Author), Janetta Debora Joubert (Author), Debbie Bradshaw (Author)
• Format: Book
• Published: Taylor & Francis Group, 2021-01-01T00:00:00Z.
• Subjects: article
• Online Access: Connect to this object online.
• Notes: 1654-9880; 10.1080/16549716.2020.1856471
• Keywords: prevalence-based years of life lived with disability; stroke; diabetes; disease modelling; non-fatal burden; Public aspects of medicine; RA1-1270
• Source: Global Health Action, Vol 14, Iss 1 (2021)

Background: The Global Burden of Disease (GBD) approach estimates disease burden by combining fatal (years of life lost) and non-fatal burden prevalence-based years of life lived with disability (PYLDs) estimates. Although South Africa has data to estimate mortality, prevalence data to estimate non-fatal burden are sparse. PYLD estimates from the GBD study for South Africa can potentially be used. However, there is a divergence in mortality estimates for South Africa between the second South African National Burden of Disease (SANBD2) and 2013 GBD studies. Objective: We investigated the feasibility of utilising GBD PYLD estimates for stroke and diabetes by exploring different disease modelling scenarios. Method: DisMod II software-generated South African stroke and diabetes PYLDs for 2010 from models using local epidemiological parameters and demographic data for people 20-79 years old. We investigated the impact on PYLD estimates of 1) differences in the cause-of-death envelope, 2) differences in the cause-specific mortality estimates (increase/decrease by 15% for stroke and 30% for diabetes), and 3) difference using local disease parameters compared to other country or region parameters. Differences were expressed as ratios, average ratios and ratio ranges. Results: Using the GBD cause-of-death envelope (16% more deaths than SANBD2) and holding other parameters constant yielded age-specific ratios of PYLDs for stroke and diabetes ranging between 0.89 and 1.07 (average 0.98) for males. Similar results were observed for females. A 15% change in age-specific stroke mortality showed little difference in the ratio comparison of PYLDs (range 0.98-1.02) while a 30% change in age-specific diabetes mortality resulted in a ratio range of 0.96-1.07 for PYLDs depending on age. Conclusion: This study showed that GBD non-fatal burden estimates (PYLDs) can be used for stroke and diabetes non-fatal burden in the SANBD2 study.