Comparison of different approaches to incidence prediction based on simple interpolation techniques

Area of Science:

• Epidemiology
• Biostatistics
• Cancer Research

Background:

• Disease incidence prediction is crucial for public health planning.
• Evaluating the reliability of different prediction methods is essential for accurate forecasting.
• Interpolation techniques are commonly used but require careful assessment.

Purpose of the Study:

• To compare the reliability of three disease incidence prediction methods.
• To assess the performance of prediction intervals and estimators using simulation.
• To identify the most dependable method for epidemiological forecasting.

Main Methods:

• Compared three interpolation-based disease incidence prediction methods.
• Assumed Poisson distribution for one method and normal distribution for two others.
• Validated methods using ex post predictions for cancer sites in Finland and a simulation study.

Main Results:

• The prediction method using a Poisson distribution assumption proved most reliable.
• This method demonstrated the smallest coverage error for prediction intervals.
• Simulation results indicated superior precision in its prediction interval estimation.

Conclusions:

• The Poisson-based disease incidence prediction method offers superior reliability.
• This approach is recommended for accurate epidemiological forecasting, particularly for cancer sites.
• Careful assessment of prediction intervals is vital for evaluating method reliability.