Automating data extraction for epidemiological research

Margadhamane Gokhale, Krishna ORCID: 0000-0002-2167-9572 (2024). Automating data extraction for epidemiological research. University of Birmingham. Ph.D.

Margadhamane Gokhale2024PhD.pdf
Text
Restricted to Repository staff only until 31 December 2028.
Available under License All rights reserved.
Download (4MB) | Request a copy

Abstract

In observational studies utilising Electronic Health Records (EHRs), the transformation of raw, unstructured healthcare data into a structured, 'analysis-ready' format is a critical yet complex process. This transition is crucial for ensuring data accuracy and completeness, facilitating effective research and reliable conclusions. The current state-of-the-art in data extraction is labour-intensive and involves non-standard approaches, which hinders accuracy and research reproducibility. This is because the generation of 'analysis-ready' data is fraught with challenges: such as complexities of study design, biases inherent in EHR databases, and technical complexities. Addressing these challenges requires meticulous planning and execution.

In this thesis, I present a novel conceptual model which outlines a solution to the problem of data extraction for observational research. The thesis outlines the general data extraction process and also presents how the model can be used to extract data for different stakeholder designs. It also describes a novel Extract, Transform, and Load (ETL) based framework which, for the first time, establishes detailed rules and transformations that are involved in the process of data extraction.

The methods developed in this thesis offers many benefits by solving current challenges and expediting the process of data extraction. The novel data extraction framework has been validated multiple times, by answering clinical questions, and resulting in the publication of several peer-reviewed research articles. The wider utility of the framework and future research is also presented.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):