Bayesian multistate modelling of incomplete chronic disease burden data.

Area of Science:

• Epidemiology
• Biostatistics
• Public Health Modeling

Background:

• Multistate lifetable models are crucial for assessing public health interventions but often lack complete incidence and case fatality data.
• Existing methods struggle with incomplete datasets, limiting accurate long-term health impact predictions.

Purpose of the Study:

• To develop and present Bayesian continuous-time multistate models for estimating disease transition rates from incomplete data.
• To provide an accessible R package for implementing these advanced statistical models.
• To extend existing methodologies for age-specific trends over calendar time.

Main Methods:

• Utilized Bayesian continuous-time multistate models to estimate transition rates between disease states.
• Employed flexible modeling techniques like splines and hierarchical models to relate rates across ages and areas.
• Incorporated incomplete data, including incidence, prevalence, and mortality, from sources like the Global Burden of Disease study.

Main Results:

• Successfully estimated case fatality rates for multiple diseases in English city regions using incomplete data.
• Demonstrated the model's ability to handle age-specific trends and varying data sources.
• Provided a formal statistical framework with transparent assumptions for rate estimation.

Conclusions:

• The developed Bayesian models offer a robust approach to estimating disease transition rates with incomplete data.
• The accessible R package facilitates the application of these models in public health research and policy.
• These estimates are valuable for informing health impact models and understanding disease burden across diverse populations.