Date of Award

January 2020

Document Type

Thesis

Degree Name

Medical Doctor (MD)

Department

Medicine

First Advisor

Charles S. Dela Cruz

Abstract

Fevers are extremely common in hospitalized patients, particularly in those admitted to the intensive care unit (ICU). Up to 50% of fevers in the ICU are non-infectious in origin. Yet, the development of a fever commonly prompts extensive, costly and invasive infectious workups. Patients with persistent fevers are usually treated with antimicrobials long after infectious workups prove unrevealing. While this is viewed as a ‘safe’ approach, inappropriate antibiotic use has shown to be harmful to patients – even effecting an increase in mortality – in addition to contributing to the spread of antibiotic resistance. The purpose of this study is to develop an automated model based on the analysis of fever curves for prompt identification of fever etiology and clinical management optimization. We begin by identifying fever patterns of non-infectious causes and developing mathematical algorithms to distinguish from infectious ones. We demonstrate that dexmedetomidine exposure and hematologic malignancies can generate prolonged fevers with curve patterns that can be differentiated from those caused by bacteremia. We quantify these differences using basic measures of variance and other algorithms, including: area under the curve, sample entropy and detrended fluctuation analysis. These studies can facilitate the development of algorithms amenable to their incorporation into electronic medical records for an automatized recognition of non-infectious fevers. The goal is to personalize therapies with early disease-specific treatments and to reduce unnecessary antimicrobial use.

Comments

This thesis is restricted to Yale network users only. It will be made publicly available on 09/10/2022

Share

COinS