Year: 2023
Type: dissertation
Abstract: Evidence Synthesis is the process of synthesizing information from clinical literature to translate the research findings into patient care and healthcare policy.Throughout the evidence synthesis process, a critical yet challenging step is the quality assessment of clinical studies.Quality in research can be considered through two aspects: methodological quality which concerns how rigorously a research is designed and conducted, and reporting quality which describes how transparently a piece of scientific work is reported as a publication.This thesis explores natural language processing (NLP) approaches to support evidence quality assessment of clinical studies.Specifically, I consider different levels of information granularity used for evidence assessment, and implemented three machine learning developments: (1) Classification of evidence types from clinical publications based on study designs, (2) Classification of sentences from randomized controlled trials (RCTs) with checklist items recommended in reporting guidelines, (3) Extraction of finegrained methodological characteristics from RCTs to assist methodological quality assessment.Applications of these NLP approaches range from assisting authors in checking their manuscripts for compliance with reporting guidelines and supporting journal editors and peer reviewers in assessing papers (pre-publication) to assisting systematic reviewers in synthesizing evidence and meta-researchers in studying research rigor and transparency (post-publication).
Author: Hoang, Khanh Linh
Institution: University of Illinois Urbana-Champaign
Language: English
Subfield: Statistics, Probability and Uncertainty
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7220753960
