DOI
https://doi.org/10.25772/KXKJ-3V50
Defense Date
2008
Document Type
Thesis
Degree Name
Master of Science
Department
Mathematical Sciences
First Advisor
James Mays
Abstract
The objective of this study was to find a regression procedure that can better explain the relationship between patterns of antibiotic use and proportions of bacterial resistance. The sample for the study is comprised of 44 University Health System Consortium (UHC) member hospitals, and the data for antibiotic use and proportions of resistance are from the years 2002 to 2005. The hospitals are spread across the Northeast, South, Southwest, Midwest, and Northwest regions of the USA. Based on statistical analysis, MRSA continues to have the highest proportion of resistance among the bacteria examined and has increased significantly since 2002. The antibiotic use in the study was measured in indices called diversity indices. There were six such measures in the study. The study, first using ordinary least squares regression, did not find one single diversity index that adequately predicted the proportion of resistance. There were also concerns that the diversity indices could be measuring the same thing, and therefore all should not be used in the model. The correlations between the three general diversity indices were strong, positive, and linear. Likewise, the three Gram-negative indices were also positively correlated with one another. Multicollinearity diagnostics also showed that there were serious dependencies among general diversity indices. Given the multicollinearity results and the correlation coefficients for the indices, it can be concluded that all six indices should not be in the same model together. Logistic regression and weighted least squares regression using the logit transformation were also performed, and just like the ordinary least squares results, there was no one single diversity index or a combination of diversity indices that adequately predicted the proportion of resistance.
Rights
© The Author
Is Part Of
VCU University Archives
Is Part Of
VCU Theses and Dissertations
Date of Submission
8-19-2008