2024_A Novel Risk-Prediction Development and Modelling for the Coronavirus Disease 2019 (COVID-19) Screening Using Expert Agreement and Machine Learning

Bibliographic Details
Format: General Document
_version_ 1860798329713065984
building INTELEK Repository
collection Online Access
collectionurl https://intelek.unisza.edu.my/intelek/pages/search.php?search=!collection3
copyright Copyright©PWB2025
country Malaysia
date 2024-09-11 14:08
format General Document
id 17234
institution UniSZA
originalfilename 17234_195aa1fd7575163.pdf
person Dr. Mohd Salami Bin Ibrahim
recordtype oai_dc
resourceurl https://intelek.unisza.edu.my/intelek/pages/view.php?ref=17234
sourcemedia Server storage
Scanned document
spelling 17234 https://intelek.unisza.edu.my/intelek/pages/view.php?ref=17234 https://intelek.unisza.edu.my/intelek/pages/search.php?search=!collection3 General Document Malaysia Library Staff (Top Management) Library Staff (Management) Library Staff (Support) Terengganu Faculty of Medicine English application/pdf 1.7 Microsoft® Word for Microsoft 365 Server storage Scanned document UniSZA Private Access UniSZA Copyright©PWB2025 367 Covid-19 Machine Learning Dissertations, Academic Epidemiological Modelling UniSZA Public Health Surveillance Artificial Intelligence Dr. Mohd Salami Bin Ibrahim Risk Prediction Model Expert Agreement Screening Algorithm Predictive Analytics Clinical Decision Support COVID-19 (Disease) — Diagnosis — Statistical methods COVID-19 (Disease) — Risk Assessment — Mathematical Models Machine Learning — Medical Applications Artificial Intelligence — Health Risk Prediction Medical Screening — Predictive Models 2024_A Novel Risk-Prediction Development and Modelling for the Coronavirus Disease 2019 (COVID-19) Screening Using Expert Agreement and Machine Learning During the COVID-19 pandemic in 2020, a comprehensive web-based screening strategy was established worldwide, leveraging epidemiological links, symptoms, and comorbidities. Similarly, Universiti Sultan Zainal Abidin (UniSZA) established the COVID-19 Health Risk Assessment and Self-evaluation (CHaSe) tool to screen at-risk staff and students. However, the conventional screening approaches faced fundamental limitations due to the lack of comprehensive data and the evolving understanding of COVID-19, particularly within university environments with limited resources. These gaps highlighted the need for innovative and tailored screening models. Reflecting on this problem, the objective of this study is to retrospectively analyse the CHaSe project data to identify the significant COVID-19 predictors, determine medical experts' agreement on screening risk categories, and validate a final prediction model. This study is a three-phase retrospective research using secondary data. Phase 1 involved systematic cataloguing of the global COVID-19 database to identify all predictors of COVID-19. Subsequently, a Principal Component Analysis (PCA) was conducted on data from 200 CHaSe pilot participants to identify predictors that explain most of the dataset variation. Then, Linear Discriminant Analysis (LDA) from 1113 randomly selected CHaSe real mass screening data was conducted to determine significant predictors for modelling prediction. In Phase 2, a total of 11 significant predictors, with 1536 possible combinations, were included. A novel dashboard-based rating instrument was developed and validated to ascertain statistical agreement risk categories for all combinations among purposively selected 11 medical experts. Phase 3 involved transforming the ratings all 11 medical experts provided into a training dataset, subsequently modelled as a multiple linear regression via machine learning. The final prediction model was turned into a classifier and validated via a statistical agreement analysis with medical experts. Phase 1 identified 52 COVID-19 predictors, with PCA revealing 10 variables contributing to 63.4% of the data variation. LDA verified 13 significant predictors for predicting COVID-19 risk categories. In Phase 2, the ordinal-weighted Agreement Coefficient 2 (AC2) reached 0.81 (95% CI [0.79, 0.82], P-value < 0.001), indicating substantial agreement among medical experts on COVID-19 risk categories. This study developed two models in Phase 3, one with and one without the close contact variable. These models were merged to create the final classifier model using experts-selected cutoff points. This model achieved an ordinal-weighted AC2 of 0.89 (95% CI [0.87, 0.90], P-value < 0.001), statistically benchmarked as an almost perfect agreement with all Phase 2 medical experts. The novel approach allows a prediction model incorporating experts' critical judgments. These pioneering inventions fuel potentially more genuine insight into modelling prediction in future medical practice. 2024-09-11 14:08 uuid:85D1746A-CFDC-4F9C-AE8F-433D9C816B22 17234_195aa1fd7575163.pdf Thesis
spellingShingle 2024_A Novel Risk-Prediction Development and Modelling for the Coronavirus Disease 2019 (COVID-19) Screening Using Expert Agreement and Machine Learning
state Terengganu
subject Dissertations, Academic
COVID-19 (Disease) — Diagnosis — Statistical methods
COVID-19 (Disease) — Risk Assessment — Mathematical Models
Machine Learning — Medical Applications
Artificial Intelligence — Health Risk Prediction
Medical Screening — Predictive Models
summary During the COVID-19 pandemic in 2020, a comprehensive web-based screening strategy was established worldwide, leveraging epidemiological links, symptoms, and comorbidities. Similarly, Universiti Sultan Zainal Abidin (UniSZA) established the COVID-19 Health Risk Assessment and Self-evaluation (CHaSe) tool to screen at-risk staff and students. However, the conventional screening approaches faced fundamental limitations due to the lack of comprehensive data and the evolving understanding of COVID-19, particularly within university environments with limited resources. These gaps highlighted the need for innovative and tailored screening models. Reflecting on this problem, the objective of this study is to retrospectively analyse the CHaSe project data to identify the significant COVID-19 predictors, determine medical experts' agreement on screening risk categories, and validate a final prediction model. This study is a three-phase retrospective research using secondary data. Phase 1 involved systematic cataloguing of the global COVID-19 database to identify all predictors of COVID-19. Subsequently, a Principal Component Analysis (PCA) was conducted on data from 200 CHaSe pilot participants to identify predictors that explain most of the dataset variation. Then, Linear Discriminant Analysis (LDA) from 1113 randomly selected CHaSe real mass screening data was conducted to determine significant predictors for modelling prediction. In Phase 2, a total of 11 significant predictors, with 1536 possible combinations, were included. A novel dashboard-based rating instrument was developed and validated to ascertain statistical agreement risk categories for all combinations among purposively selected 11 medical experts. Phase 3 involved transforming the ratings all 11 medical experts provided into a training dataset, subsequently modelled as a multiple linear regression via machine learning. The final prediction model was turned into a classifier and validated via a statistical agreement analysis with medical experts. Phase 1 identified 52 COVID-19 predictors, with PCA revealing 10 variables contributing to 63.4% of the data variation. LDA verified 13 significant predictors for predicting COVID-19 risk categories. In Phase 2, the ordinal-weighted Agreement Coefficient 2 (AC2) reached 0.81 (95% CI [0.79, 0.82], P-value < 0.001), indicating substantial agreement among medical experts on COVID-19 risk categories. This study developed two models in Phase 3, one with and one without the close contact variable. These models were merged to create the final classifier model using experts-selected cutoff points. This model achieved an ordinal-weighted AC2 of 0.89 (95% CI [0.87, 0.90], P-value < 0.001), statistically benchmarked as an almost perfect agreement with all Phase 2 medical experts. The novel approach allows a prediction model incorporating experts' critical judgments. These pioneering inventions fuel potentially more genuine insight into modelling prediction in future medical practice.
title 2024_A Novel Risk-Prediction Development and Modelling for the Coronavirus Disease 2019 (COVID-19) Screening Using Expert Agreement and Machine Learning
title_full 2024_A Novel Risk-Prediction Development and Modelling for the Coronavirus Disease 2019 (COVID-19) Screening Using Expert Agreement and Machine Learning
title_fullStr 2024_A Novel Risk-Prediction Development and Modelling for the Coronavirus Disease 2019 (COVID-19) Screening Using Expert Agreement and Machine Learning
title_full_unstemmed 2024_A Novel Risk-Prediction Development and Modelling for the Coronavirus Disease 2019 (COVID-19) Screening Using Expert Agreement and Machine Learning
title_short 2024_A Novel Risk-Prediction Development and Modelling for the Coronavirus Disease 2019 (COVID-19) Screening Using Expert Agreement and Machine Learning
title_sort 2024_a novel risk-prediction development and modelling for the coronavirus disease 2019 (covid-19) screening using expert agreement and machine learning