[Evaluation of estimation of prevalence ratio using bayesian log-binomial regression model].

Area of Science:

• Biostatistics
• Epidemiology
• Public Health

Background:

• Accurate estimation of prevalence ratios (PRs) is crucial for understanding health-related behaviors.
• Log-binomial regression is a common method, but can suffer from convergence issues.
• Bayesian approaches offer an alternative for estimating PRs, potentially overcoming limitations of conventional methods.

Purpose of the Study:

• To evaluate the estimation of prevalence ratio (PR) using a Bayesian log-binomial regression model.
• To assess the application of this model in estimating the PR of medical care-seeking for infant diarrhea.
• To compare the performance of Bayesian log-binomial regression with conventional log-binomial regression.

Main Methods:

• Utilized Bayesian log-binomial regression model in Openbugs software.
• Estimated the PR of medical care-seeking related to caregivers' recognition of infant diarrhea risk signs.
• Compared point and interval estimations, and model convergence between Bayesian and conventional log-binomial models, adjusting for covariates like education and distance.

Main Results:

• Caregivers' recognition of infant diarrhea risk signs was significantly associated with a 13% increase in medical care-seeking.
• All three Bayesian log-binomial regression models converged, yielding PR estimates between 1.128 and 1.132.
• Conventional log-binomial models showed convergence for simpler models, but the most complex model (model 3) encountered misconvergence, requiring the COPY method.

Conclusions:

• Bayesian log-binomial regression effectively estimates prevalence ratios with reduced risk of misconvergence.
• The Bayesian approach demonstrates advantages in application compared to conventional log-binomial regression, particularly with complex models.
• Findings support the use of Bayesian log-binomial regression for robust estimation of health-related prevalence ratios.