Adoption of Machine Learning Framework for Operational Risk Management in Healthcare

Abstract

The healthcare industry confronts high risks because of its complexity. Therefore, recent changes emphasize efficiency through technology and regulation. Clinical governance is adopted by leading organizations to ensure top-quality care, with risk management as a critical component. Pregnancy and childbirth are frequently known as wonderful experiences, but they also can pose unique challenges and risks, particularly in maternity services. Maternal and newborn mortality and morbidity are major concerns, frequently linked to inadequate care during pregnancy and affecting patient satisfaction. This study addresses a gap by exploring probabilistic interactions between pregnancy stages using Machine Learning (ML) algorithms. Bayesian Belief Network (BBN) modeling is used specifically.

Data for this study is sourced from the 2021 maternity patient survey conducted by the NHS in England. The survey covers eight sections with 28 questions, assessing antenatal care, labor, baby's birth, postnatal care, and other sections. Statistical studies were carried out, including reliability tests and descriptive analysis. Continuous numerical data is discretized into two-state, three-state, and mix-state schemes for examination with Bayesian classifier algorithms. The network model is developed using Bayesian search (BS), Peter Clark (PC), and Greedy ThickThinking (GTT) structural learning algorithms.

The mixed-state GTT model achieved the highest accuracy, with an 83% accuracy rate in predicting ‘S5: the staff caring factor’. The study concludes with a sensitivity analysis to have a better grasp of the performance of the chosen model.

Date of Award	16 Dec 2023
Original language	American English
Supervisor	Mecit Simsekler (Supervisor)