Work

Ranking of modified early warning score using a novel pattern matching technique for continuous patient monitoring

Year: 2024
Type: dissertation
Abstract: Remote Patient Monitoring (RPM) applications deploy sensors, fog, edge and cloud technologies to monitor diverse patients' health continuously. However, RPM is challenging, particularly when patients' signs are continually monitored due to difficulties in data stream visualisation, dealing with missing data and generating alerts. The thesis advances a systematic approach for alarm generation, a novel pattern-matching method for vital signs data, ranking of the patterns and sophisticated pre-processing methods as solutions to these problems. The main objective of these contributions is to improve patient care using emerging RPM applications. The thesis's initial contribution (1) involves a thorough examination of the RPM application challenges, primarily on the causes of false-positive alerts during continuous monitoring of patients. Usually, sensors in healthcare are susceptible and produce numerous alerts to identify any possible deterioration of health conditions, which tend to make numerous false positive alarms, leading to alarm fatigue. Based on the identified causes, (2) the thesis proposed an approach to provide a systematic framework for creating effective alarm generators, emphasising physiological data, clinical settings, medical sensor devices, and clinical knowledge. Data integrity is crucial in healthcare, and the widespread use of wearable sensors in RPM has produced significant health data. However, the effectiveness of analytical health data models is compromised by the missing data values, noise, and inconsistencies that frequently appear in health data. The thesis (3) contributes health data pre-processing techniques to manage these obstacles and offers quality data for trustworthy health data analysis. The acquisition of health data presents difficulties for continuous RPM, which could result in transmission errors. The thesis (4) contributes a unique computation pattern-matching method to determine patient risk. Data from a clinical field trial is used to assess the proposed technique, which shows that it can detect and forecast missing patterns comparable to machine learning algorithms. Furthermore, the thesis acknowledges the intricacy of incorporating physiological data obtained from wearable sensors into RPM. The focus is on continuously monitoring vital signs, blood pressure, temperature, heart rate, blood oxygen, and respiratory rate. One of the thesis's significant (5) contributions is the derived continuous vital signs sematic parameters known as Trust, Frequency, Slope, and Trend from the raw physiological RPM data readings. A practical clinical tool for quickly recognising and evaluating a patient's vital signs decline is the Modified Early Warning Score (MEWS), which standardises alerts for clinical intervention. The derived parameters and the standard MEWS support better-informed clinical decisions by improving data representation and enabling automated assessment of prioritised alerts. This thesis underscores the potential for enhancing RPM technology in various healthcare settings. By addressing issues such as false alarms, data quality, and data representation, the research aims to improve patient care and clinical decision-making, aligning with its primary objective.
Author: Arora, Teena
Language: English
Subfield: Surgery
Field: Medicine
Open Access status: green
OpenAlex ID: W7052143866
Cites: 111
Cited by:
Related to: 10