Using Zero-Inflated Poisson Model and Zero-Inflated Negative Binomial Model on Dental Services of Wisconsin, 2014 Data

Professional dental care to ensure optimum oral health of public plays an important role in the public health system. Facing the truth that there has been decline in dental care utilization for decade in 20th century, more and more attention has been paid to the oral health of children from this century. Children from age 0 to 21 years old experience rapid physical and oral development. Investigating the utilization of dental service for these children will provide useful information for the future study of the insurance system. In this thesis, two regression methods will be studied, the Zero-Inflated Poisson model and the Zero-Inflated Negative Binomial model, and compared with the most widely used method in dental health services analysis, the Poisson regression. We will compare the Akaike Information Criterion and Bayesian Information Criterion value to determine the performance of the three models and therefore imply the most optimal model into the Dental service of Wisconsin, 2014 dataset. By comparing the incident risk ratio and odds ratio, we can make prediction for the probability of utilization and number of visits in the future. Future research can be based on this thesis. In this thesis, we demonstrate that Zero-Inflated Negative Binomial model is the most optimal method for this dataset. Contributing variables gender, race, area, age group, and the total enrolled insurance months are the most significant factors in the model. From the incident risk ratio and the odds ratio, combined with the prediction information, we can conclude that children who have a lower probability to use the service, will have a larger number of visits to dentists. In other word, the predictive probability of utilization has a negative relationship with the predictive number of visits.