Spatio-Temporal Bayesian Models for Malaria Risk Using Survey and Health Facility Routine Data in Rwanda

Area of Science:

• Epidemiology and Public Health
• Geospatial Health
• Biostatistics

Background:

• Malaria poses a significant global health threat, particularly in developing nations, with children under five being highly vulnerable.
• Existing health surveillance often relies on Demographic and Health Survey (DHS) data, which may lack the granularity for real-time, localized malaria elimination strategies.
• Accurate, small-area malaria risk estimates are crucial for effective, tailored interventions.

Purpose of the Study:

• To propose and evaluate a two-step modeling framework integrating survey and routine health data for improved malaria risk incidence estimation.
• To enhance the spatial and temporal quantification of malaria trends at the lowest administrative levels.
• To refine malaria surveillance for children under five in Rwanda.

Main Methods:

• A Bayesian spatio-temporal modeling approach was employed, combining DHS and routine health facility data.
• A two-step process involved fitting a binomial model to survey data, followed by incorporating fitted values into a Poisson model for routine data.
• The model specifically assessed malaria relative risk among children under five in Rwanda.

Main Results:

• Analysis of 2019-2020 DHS data alone indicated higher malaria prevalence in specific regions of Rwanda.
• Integrating routine health facility data identified additional high-risk clusters not apparent from survey data alone.
• The proposed framework successfully enabled the estimation of spatial and temporal trends in malaria relative risk at local levels.

Conclusions:

• Combining DHS and routine health services data offers more precise malaria burden estimates, supporting malaria elimination goals.
• The integrated approach provides a better understanding of subnational malaria relative risk compared to using survey data solely.
• This methodology strengthens active malaria surveillance by leveraging the strengths of both high-quality survey data and small-scale routine data.