A multilevel model for continuous time population estimation.

Area of Science:

• Epidemiology
• Biostatistics
• Statistical Modeling

Background:

• Multilist methods are used for population size estimation but typically assume instantaneous list compilation.
• Continuous time lists are common in epidemiology but existing methods have limitations for population size estimation.
• Estimating population size with longitudinal data and identifying factors influencing patient visit intervals are key epidemiological challenges.

Purpose of the Study:

• To propose a novel Bayesian framework for estimating population size from continuously compiled lists.
• To identify patient-specific factors associated with the duration between healthcare visits.
• To address the limitations of existing multilist methods in epidemiological settings with longitudinal data.

Main Methods:

• A Bayesian framework is developed to model interval lengths between patient visits.
• The method handles sparse data, common when patients are observed infrequently.
• The approach integrates population size estimation with the analysis of time-to-event data.

Main Results:

• The proposed method was applied to motivating epidemiological data, demonstrating its applicability.
• A simulation study evaluated the estimator's performance under various conditions.
• The study provides a new statistical tool for continuous time population estimation in epidemiology.

Conclusions:

• The developed Bayesian approach offers a viable alternative for population size estimation using continuous epidemiological data.
• The method effectively models interval lengths and identifies patient factors influencing visit durations.
• Further methodological development is suggested for continuous time population estimation.