Bayesian hierarchical models to estimate the force of infection of Helicobacter pylori in Mexico: Evidence from a national survey
Published in 50th Annual meeting of the Society for Epidemiologic Research, 2017
Abstract
Objective
Helicobacter pylori (H. pylori) is one of the most prevalent bacterial infections in the world, present in the stomach of half of the world’s population. The force of infection is defined as the instantaneous per capita rate at which susceptibles acquire infection. The aim of this research is to estimate the agespecific force of infection of H. pylori in Mexico using a novel hierarchical nonlinear Bayesian model by pooling information between Mexican states.
Methods
Data for this study came from national H. pylori seroepidemiology surveys in Mexico. We modeled the number of individuals with H. pylori at a given age in a given state as a binomial random variable. We assumed that the cumulative risk of infection by a given age follows a modified exponential distribution, allowing some fraction of the population to remain uninfected. The cumulative risk of infection was modeled for each state in Mexico and these state-specific cumulative risk curves were shrunk toward the overall national cumulative risk curve using Bayesian hierarchical models. These parameters were used to estimate the force of infection by age in each Mexican state. Models were estimated using Markov chain Monte Carlo (MCMC) methods JAGS.
Results
National H. pylori prevalence estimates plateau at 86.1 % [95% credible interval (CR): 84.2%-88.2%]. The rate of increase of prevalence per year of age is 0.093 [95%CR: 0.084-0.103]. We estimated an average age at infection of the population eventually infected of 12.5 [95%CR: 11.3-13.8] and the age-specific force of infection was highest at birth 0.080 [95%CR: 0.089-0.071] decreasing to zero as age increases.
Conclusion
This study presents the first estimation of the force of infection of H. pylori using seroepidemiologic data. This Bayesian hierarchical model allows estimation of statespecific cumulative risk curves and stabilizes estimation by pooling information between the states, resulting in more realistic estimates.