Comparison of data-fitting models for schistosomiasis: a case study in Xingzi, China

Area of Science:

• Epidemiology
• Spatial Analysis
• Disease Modelling

Background:

• Epidemiological studies often use Gaussian, binomial, or Poisson models for prevalence data.
• The rationale for model selection is frequently not detailed in scientific literature.

Purpose of the Study:

• To compare Gaussian, binomial, and Poisson models for fitting schistosomiasis risk.
• To identify the most appropriate model for analyzing schistosomiasis prevalence data.
• To map spatial patterns of schistosomiasis risk and identify contributing environmental factors.

Main Methods:

• Utilized parasitological survey data from 36,208 individuals (aged 6-65) across 42 villages in Xingzi county, China.
• Compared Gaussian, binomial, and Poisson regression models to fit schistosomiasis prevalence data.
• Integrated environmental data with parasitological data for spatial risk mapping.

Main Results:

• The Poisson model demonstrated the best fit for the schistosomiasis prevalence data.
• The Poisson model successfully identified environmental risk factors that explain geographical variations in schistosomiasis risk.
• A predictive map of schistosomiasis risk was developed based on identified environmental factors.

Conclusions:

• The Poisson model is the most appropriate for analyzing schistosomiasis prevalence data in this context.
• Environmental factors play a significant role in the spatial distribution of schistosomiasis.
• The predictive risk map is crucial for targeted schistosomiasis control and elimination strategies in China.