Leveraging Digital Twins for Clinical Pathways: Exploring Arthroplasty Registry and Clinical Database
Master thesis

View/ Open
Date
2024-06-03Metadata
Show full item recordCollections
- Master theses [252]
Abstract
This thesis investigates the integration of digital twins into clinical pathwaysto improve predictive modeling of patient outcomes. Within the Digital TwinProject we have analyzed two different data sources; Data from the NorwegianArthroplasty Register (NAR) and the Medical Information Mart for IntensiveCare (MIMIC) databases. The primary objective was to explore the possibilitiesof improving patient outcomes through, among other things, better predictions.The research employs design and data science methodologies to explorethe feasibility of creating digital twins using clinical data. Two distinct datasets were analyzed, resulting in different types of pathways. The clinical re-gister contains simplified patient data, the pathways are straightforward andcontain one main distinguishing point, the occurrence of complications. TheMIMIC-IV database contained detailed patient data which resulted in detailedand complicated clinical pathways. Thus, there was no one generalized ap-plicable clinical pathway in which a digital twin could be integrated. Thestudy therefore evaluated the extent of available data, the potential to connectexternal data sets, and the flow of data within each database. The study foundthat the MIMIC-IV database was more suitable for the development of digitaltwins. The study concluded that while NAR provides valuable insights intopatient safety and outcomes, it lacks detailed, time-stamped data that wouldallow the creation and integration of digital twins into a pathway.As a proof-of-concept, basic machine learning pipelines were implementedto demonstrate the capability of digital twins to make real-time predictionsand update clinical pathways. The findings illustrate how event logs couldinitialize digital twins in MIMIC-IV, investigate suitable predictive methods,and visualize results. This is only one type of modeling; there are numerouspossibilities to explore and integrate. This opens up the possibility for futureresearch in various clinical settings.This thesis contributes to the field by providing insight into the practicalapplication of digital twins in healthcare, specifically in the context of hiparthroplasty