Establishing disease models for the Belgian national burden of disease study: challenges and perspectives

Abstract The Belgian national burden of disease study (BeBOD) was launched in 2022, and currently provides disease burden estimates for 38 key diseases. The calculation of the non-fatal component of the disease burden, the Years Lived with Disability (YLD), follows a stepwise approach that aims to establish priority outcomes; quantify prevalence “best estimates”; establish disease models; and perform expert evaluation of methods and results. The disease models visualize the relationship between the different health states associated with a given outcome. Health states include the different acute and chronic stages of the outcome (including complications), which may be stratified in different severity levels (e.g., mild, moderate, severe). Disease models used in burden of disease studies primarily aim to document the considered health states, and do not aim at a representation of the complete clinical picture of the condition. The disease models instead help in understanding how the number of cases for each health state is calculated. Models typically start with one “parent node”, which contains all cases. This parent node then gives rise to multiple “child nodes”, with the terminal child nodes representing the individual health states. In BeBOD, disease models and severity distributions are adapted from the Global Burden of Disease (GBD) study, and complemented with local data where possible. This process ensures consistency with available disability weights, while still allowing for local data to feed into the calculations. In this presentation, we will demonstrate the BeBOD process for establishing disease models based on a number of archetypical examples, which represent varying challenges linked to disease model complexity, clarity and completeness of the GBD disease model documentation, and availability of local data. Future perspectives will be discussed, including the improved leveraging of local clinical data, and the implementation of tailored active data collection.