| The purpose of this study is to use data mining methods to investigate the physician decisions specifically in treatment of osteomyelitis. The contribution of this dissertation is in industrial mathematics, as it relates to the industry of healthcare and physician decision making.;In Chapters III--V, we review the existing mathematical methods in data mining and statistics, including predictive modeling, time series, and text mining; these methods have not been regularly used in the healthcare industry. Preprocessing is an essential aspect of outcomes research. Dealing with multiple data sources is essential. We demonstrate the needed preprocessing in Chapters VIII--X. Our data contain multiple observations for each patient. We convert this dataset from a one-to-many to a one-to-one observation for each patient; we developed the necessary SAS coding required to perform the preprocessing steps. In other words, we need to have a one-to-many relationship between patients and their procedures (instead of many-to-many). We also show how disparate datasets such as inpatients, outpatients and RX datasets can be merged to examine the relationship of antibiotics to disease treatment.;One innovation here is the use of text mining to define a patient severity index that is far superior to the indices currently defined via logistic regression. Other important innovations in investigating the data include finding the switches of medication and comparing the date of switch with the date of procedures. We are able to find the total number of medication switches and the switched medications. This method can be use to find the first, second or third switch. We could study these medications switched at deeper levels, but this is not necessary in our study, especially with limited access to data. We also create a model to forecast the cost of hospitalization for patients with osteomyelitis.;Using MedStat data in Chapter IX, we have shown that physicians do not use proper antibiotics if antibiotics are used at all, resulting in unnecessary amputations and amputations performed sequentially on patients with osteomyelitis. Other conclusions discovered include the result that physicians assume amputation is the primary treatment for Osteomyelitis. Injection of antibiotics was performed on only a small portion of patients with Osteomyelitis. In many cases, infection has recurred and amputation was performed more than once.;There have been different sources of data used in this dissertation. Two primary data sets have been used in this study; the Nationwide Inpatient Sample (NIS) and Thomson MedStat MarketScan data. We used online sources to obtain background information about the disease and its treatment in the literature. An innovated method was used to capture the information from the web and cluster or filter them to find the most relevant information, using SAS Text Miner.;In this project we found Text Miner in Enterprise Guide a great tool to get data summaries from similar data sets. |
