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

Model-based Decision Support for Sepsis Endotypes

Zhang, Li Ang (2018) Model-based Decision Support for Sepsis Endotypes. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

Download (20MB) | Preview


Sepsis is a high mortality syndrome characterized by organ dysfunction due to a severe and dysregulated acute inflammatory response to infection. Research into therapies for this syndrome has historically ended in failure, which has largely been attributed to the elevated levels of subject heterogeneity. What may have been previously attributed to variability in sepsis may be due to mechanistic differences between patients. Endotypes are distinct subtypes of disease, where underlying causes such as mechanistic or pathway related differences manifest into phenotypes of disease.

The lack of mechanistic understanding of immune mediator dynamics and the responses they trigger necessitates a mathematical modeling approach to analyze its complexities. A transfer function model is proposed to describe and cluster the dynamics of key inflammatory mediators. Five sepsis endotypes were discovered and revealed motifs of overwhelming inflammation, various levels of immunosuppression, sustained inflammation, and immunodeficiency. An accurate clinical tool was proposed to classify subjects into endotypes using six-hour trajectories of clinical data.

A physiological ordinary differential equation model of sepsis is proposed that characterizes the interactions of inflammatory signaling molecules, neutrophils, and macrophages across the bone, blood, and tissue compartments of the body. This model used to generate individual subject fits against human sepsis data. Population-level parameter analysis implicated macrophage cell death and cytokine half- dynamics in endotype-level differences.

Several proof-of-concept statistical models were introduced to demonstrate that it is possible to estimate the pre-hospital time of sepsis subjects and to quantify their sepsis-induced systemic tissue damage. A nearest-neighbor-based method was verified against animal and human data and revealed that identifying infection time-zero of sepsis patients can be quickly estimated with high accuracy using commonly measured clinical features. A logistic regression ensemble model demonstrated revealed early organ dysfunction were significant contributors to systemic damage and mortality. Knowledge of time-zero and systemic damage levels, in combination with an endotype classifier, provides clinicians with a clear depiction of where a subject is located on their sepsis trajectory. Such a decision support system enables therapy timing, early organ support, and targeted therapies to guide personalized treatment and shift patients towards better outcomes in sepsis.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Zhang, Li AngzhangL@pitt.eduzhangL0000-0001-9468-2513
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairParker, Robert S.rparker@pitt.edurparker
Committee MemberBanerjee, Ipsitaipb1@pitt.eduipb1
Committee MemberClermont, Gillescler@pitt.educler
Committee MemberSwigon, Davidswigon@pitt.eduswigon
Date: 25 January 2018
Date Type: Publication
Defense Date: 25 August 2017
Approval Date: 25 January 2018
Submission Date: 28 November 2017
Access Restriction: 1 year -- Restrict access to University of Pittsburgh for a period of 1 year.
Number of Pages: 151
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Chemical Engineering
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: mathematical modeling, sepsis, cytokines, ordinary differential equations, machine learning, statistical analysis
Date Deposited: 25 Jan 2018 13:34
Last Modified: 25 Jan 2019 06:15


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item