Date on Master's Thesis/Doctoral Dissertation


Document Type

Doctoral Dissertation

Degree Name

Ph. D.



Committee Chair

Cerrito, Patricia

Author's Keywords

Prescription; Medications; Cost analysis


Antibiotics--Administration; Antibiotics--Costs


The purpose of this research study is to examine the use of time series forecasting and text mining to investigate the prescription of antibiotics. The specific objective is to examine the relationship between the total payments, private insurance payments, Medicare payments, Medicaid payments, number of prescriptions and quantity of prescriptions for different antibiotics. Currently, there is no method available to forecast antibiotic prescription costs, so we have adopted several methods that will help health care providers and hospitals to know about the prescription of the antibiotics being prescribed. The payment made for each antibiotic is based upon an average cost and total cost that will include the cost of the antibiotics and insurance payments. It will be beneficial to show health care providers the trends of these antibiotics in terms of the cost analysis. It is also beneficial to make comparisons between several antibiotics in terms of the number of prescriptions and to do further study as to why one antibiotic is prescribed more often than others. We developed time series models that will be used to forecast the prescription practices of the antibiotics. The time series models that we developed for antibiotic prescription are; simple exponential smoothing models, double exponential smoothing model, linear exponential smoothing model. We used exponential models to develop forecasting for antibiotics on which cost increases exponentially. We also developed an autoregressive integrated moving average model for non-stationary data on which the series has no constant mean and variance through time. We developed Generalized Autoregressive Conditional Heteroskedastic Models for volatile variance, and we also incorporated the inflation rate as a model dynamic regressor to see the effect on model forecast. We finally used text mining and clustering to classify the ICD-9 codes into six clusters and make comparisons within each cluster, by plotting the data using kernel density estimation. This project will be beneficial for health care institutions for predicting the trend of the antibiotic prescription, so that further studies can be made why one antibiotic is prescribed more often than others.