Accelerating data-driven discovery in type 1 diabetes: an informatics-based approach

Type 1 diabetes (T1D) is a lifelong chronic disease characterized by the absolute or near-absolute loss of insulin. For affected individuals, management of T1D is an unremitting challenge that involves constant blood glucose monitoring and lifelong administration and titration of exogeneous insulin. Unfortunately, findings from decades of research have not yet comprehensively translated into substantially improved health outcomes, suggesting that limitations inherent in the use of small patient samples and traditional analytical methods have curbed discovery of actionable disease insights. Understanding and addressing ongoing worsened health outcomes in T1D -- as well as particular vulnerabilities experienced by subgroups of individuals impacted by the disease -- requires actualization of a research paradigm that potentiates and advocates for analyses of complex T1D data at scale. The work described in this dissertation aims to accelerate translational research potential through iterative, data-intensive approaches that holistically integrate three interfacing domains: health informatics, clinical medicine, and biomedical research. Using approaches firmly rooted in explainable artificial intelligence, we successfully demonstrate the following: (1) development of a computational phenotyping methodology that enables automated diabetes/T1D case identification in heterogeneous, nationwide electronic health record (EHR) data, (2) a data-driven, contrast pattern mining approach for discovery of clinically significant phenotypic heterogeneity in T1D, and (3) development of a highly scalable pipeline to facilitate diabetes health outcomes research using multi-site EHR data. We also substantiate our commitment to openly disseminating our findings and tools to the research community at large. The totality of this work demonstrates that our informatic innovations are poised to significantly expand current research capabilities by leveraging automated processes that facilitate and enable rapid discovery of disease insights.

Name thesis:degree_name

Ph. D.

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Informatics (MU)

Grantor dc:publisher

University of Missouri--Columbia

Year dc:date.issued

2022

Author dc:creator

• Tallon, Erin M.

Advisor dc:contributor.advisor

• Shyu, Chi-Ren

Language dc:language.iso

eng, English

DOI

https://doi.org/10.32469/10355/94294

OAI identifier oai:identifier

oai:mospace.umsystem.edu:10355/94294