Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

A Comparison of Classification Trees to Logistic Regression For Exploring CNS and Systemic Biomarker Predictors of Patient Outcome After Traumatic Brain Injury

Peterson, Madeline (2024) A Comparison of Classification Trees to Logistic Regression For Exploring CNS and Systemic Biomarker Predictors of Patient Outcome After Traumatic Brain Injury. Master's Thesis, University of Pittsburgh. (Unpublished)

[img] PDF
Restricted to University of Pittsburgh users only until 14 May 2026.

Download (1MB) | Request a Copy


There are no approved neuroprotective treatment options for individuals with moderate-to-severe traumatic brain injury (TBI). However, knowledge gaps exist regarding the development and use of early prediction tools that incorporate biomarkers and clinical features reflecting aspects of both central nervous system (CNS) and systemic injury for identifying likely responders to potential neuroprotective treatments. Thus, a secondary analysis was conducted of the ProTECT III multi-center randomized clinical trial utilizing data from N=536 patients who also enrolled in the BioProTECT biosampling protocol. Baseline clinical measures [age, sex, race, Rotterdam CT score, index GCS (iGCS)], assayed systemic injury biomarkers and hormones (serum progesterone, androstenedione, estrone, testosterone, estradiol, and TNFα) and assayed CNS biomarkers (SBDP, S100B, UCHL-1, GFAP) were utilized to predict TBI overall favorable vs. unfavorable outcome using the Glasgow Outcome Scale-Extended (GOS-E) at 6-months and 6-month mortality status. Classification and Regression Tree (CART) models were employed to predict outcome, and findings were compared to logistic regression. Overall, logistic regression outperformed CART modeling; however, both methods identified some consistent baseline predictors. Features common to both GOS-E models included GFAP, age, Rotterdam, iGCS, progesterone, and S100B. Features common to both mortality models included GFAP, age, Rotterdam, and TNFα. The outcomes of this project carry significance for public health as they contribute to an enhanced understanding of biomarkers associated with a patient’s response to TBI and their predictive role in survival and other survivor related outcomes.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Peterson, Madelinemap583@pitt.edumap583
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairCarlson, Jennajnc35@pitt.edujnc35
Committee MemberTang, Lulutang@pitt.edulutang
Committee MemberBarton,
Date: 14 May 2024
Date Type: Publication
Defense Date: 11 April 2024
Approval Date: 14 May 2024
Submission Date: 22 April 2024
Access Restriction: 2 year -- Restrict access to University of Pittsburgh for a period of 2 years.
Number of Pages: 72
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: Traumatic Brain Injury, CART modeling, machine learning, predictive modeling, recovery, logistic regression
Date Deposited: 14 May 2024 19:14
Last Modified: 14 May 2024 19:14


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item