NJIT
Accurate patient health condition identification and risk prediction from health data
Abstract
dc:description.abstract<p>Online Health communities (OHCs), Electronic Health Records (EHRs) and Claims data contain rich patient information. Accurate identification of patient's health conditions is critical for providing effective and safe health care service and for trustable knowledge discovery. This dissertation proposes several deep-learning based techniques on identifying patient health conditions and predicting adverse drug event (ADE) risks using various health data.</p><p>First, this dissertation presents a novel patient experience mining model to differentiate the data expressing user experiences from the one describing hearsays. Indeed, OHCs contain information with varying degrees of quality. Identifying information that describes patient health experience from OHCs is important for trustable knowledge discovery and effective recommendation. The proposed model holistically captures linguistic features of text, user information and context information to effectively classify patient experience in OHC.</p><p>The second problem discussed in the dissertation is to accurately identify patient medical conditions in an EHR system. This is the basis of precisely documenting patient status, coding for billing, and supporting data-driven clinical decision making. However, patient disease information is often not fully captured in structured EHR systems, but may be documented in unstructured clinical notes. Not all disease mentions in clinical notes actually refer to the patient's condition. The dissertation presents a two-step workflow for identifying patient diseases from clinical notes and a deep-learning model for disease mention classification.</p><p>Last, this dissertation presents proposed techniques for predicting a patient's risk of experiencing an ADE in the target ADE list if being given a target drug using medical claims data. Such risk assessment can help physicians to choose safer treatment among alternatives. A hierarchical neural network model that combines the characteristic of claim codes and encounter time relation is proposed.</p>
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy in Mathematical Sciences - (Ph.D.)
- Discipline thesis:degree_discipline
- Mathematical Sciences
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Shi, Jinhe
- Contributors dc:contributor
-
- Yi Chen
- Dantong Yu
- Dimitri Theodoratos
Subjects
dc:subject × 8Identifiers
dc:identifier.*- Repository record dc:identifier
- https://digitalcommons.njit.edu/dissertations/1909