Area disease estimation based on sentinel hospital records

Area of Science:

• Epidemiology
• Biostatistics
• Geospatial Analysis

Background:

• Population health estimates often rely on sentinel hospital records, which are prone to bias and uncertainty.
• Traditional estimation methods struggle with biased data, potentially leading to flawed public health conclusions and interventions.

Purpose of the Study:

• To introduce and evaluate the Biased Sample Hospital-based Area Disease Estimation (B-SHADE) technique for generating accurate space-time disease estimates from biased hospital data.
• To address the limitations of existing methods in handling measurement errors and biases in health surveillance data.

Main Methods:

• Developed the B-SHADE technique, a novel approach combining Block Kriging and ratio estimation principles.
• Employed a weighted summation of sentinel hospital records, incorporating space-time information and inter-hospital relationships.
• Validated the technique using hand-foot-mouth disease and fever syndrome incidence data from Shanghai over two years.

Main Results:

• B-SHADE generated unbiased and minimum error variance estimates of area incidence, outperforming mainstream estimators.
• Empirical evaluation demonstrated the technique's superiority in correcting sample bias and improving spatial clustering accuracy.
• The method produced best linear unbiased estimates (BLUE) by enhancing the representativeness of sentinel hospital records.

Conclusions:

• B-SHADE effectively integrates optimal estimation and bias correction, overcoming limitations of Block Kriging and ratio estimators.
• The technique offers flexibility, reducing to established methods under specific data conditions (unbiased samples, no hospital correlation).
• Real-world case studies confirmed the empirical superiority of B-SHADE for disease incidence estimation.