Gao, Xiaotian
(2015)
Analysis of kidney volume and functional outcomes using survival and classification tree models.
Master's Thesis, University of Pittsburgh.
(Unpublished)
Abstract
Tree models have been widely used for clustering problems in areas like evidence-based decision-making, machine learning and data mining. The inherent properties of tree models, e.g. recursively dividing sample space, make it more flexible and superior in situation of nonlinear classifications and complex sample structure. In this thesis, they would be applied to the data from the Consortium for Radiologic Imaging Studies of Chronic Kidney Disease (CRISP) to explore the association between Total Kidney Volume (TKV) and Chronic Kidney Disease (CKD) stage 3.
In this current study, multivariable Cox survival models were used to adjust for baseline confounders and assess the relationship between TKV and time to CKD Stage 3. The same questions, were also analyzed using survival tree models, identifying the combination of variables associated with similar survival, and thus facilitating the identification of high and low risk sets. Variations of the tree modeling approach were employed to maximize model fit and generalizability, including pruning and bagging.
Classification tree models, and the same variations (pruning and bragging) were also fit to the development of the dichotomous outcome of CKD Stage 3 by a fixed time point. Receiver operator characteristic (ROC) curves with and without cross-validation are presented, and associated classification statistics (sensitivity, specificity and area under the curve) are calculated to characterize the prognostic ability of the tree models. Findings are then compared to the standard logistic model.
Both tree models and regression models agreed on the significance of baseline total kidney volume and estimated glomerular filtration rate in predicting CKD prognosis. Cutoff values were also determined.
From the public health significance perspective, these cutoffs could be advisory to actual clinical decision and prognostics of CKD. Comparing with current continuously GFR monitoring for CKD progress, two baseline predictors measured in the early phase of the disease, makes early interventions more practical.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
28 September 2015 |
| Date Type: |
Publication |
| Defense Date: |
3 June 2015 |
| Approval Date: |
28 September 2015 |
| Submission Date: |
3 June 2015 |
| Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
| Number of Pages: |
51 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
School of Public Health > Biostatistics |
| Degree: |
MS - Master of Science |
| Thesis Type: |
Master's Thesis |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
Regression Trees; Chronic Kidney Disease |
| Date Deposited: |
28 Sep 2015 16:54 |
| Last Modified: |
15 Nov 2016 14:28 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/25329 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[img]](http://d-scholarship.pitt.edu/style/images/fileicons/application_pdf.png)
![[feed]](/images/feed-icon-32x32.png){kind=icon}