Modelling and optimizing socio-technical operations in healthcare using the FRAM and reinforcement learning

Salehi, Vahid (2022) Modelling and optimizing socio-technical operations in healthcare using the FRAM and reinforcement learning. Doctoral (PhD) thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (3MB)

Abstract

This PhD research work is intended to model, analyze, and optimize socio-technical operations in healthcare using a systemic approach and reinforcement learning. An extensive literature review is presented, and the main knowledge gaps related to modelling and optimizing socio-technical operations in healthcare are clearly outlined and addressed in this research work. Introduction: Hospital to home transition processes of frail older adults include a set of actions for frail people who are discharged from hospital to their home in the community. The transition process exhibits dynamic interactions between technology, humans, organizations, and the environment. The non-linear dependencies among these influential parameters complicate the understanding of the transition process and the mechanism of modelling its operations. Objectives: The objectives of this research work are (a) To identify the strengths and shortcomings of the FRAM in modelling complex socio-technical systems; (b) To develop a comprehensive model of the hospital-to-home transition process for frail patients; (c) To capture and visualize different characteristics of variability in the transition process; (d) To monitor frail patients’ transitions from hospital to home; (e) To identify challenges of the transition process; and (f) To explore functional pathways to identify transition processes with the highest quality of care and services for frail older people. Methodology: This research work uses the Functional Resonance Analysis Method (FRAM) to study and model the complexity of the transition process. A complementary tool for the FRAM (DynaFRAM) is also used to characterize functional and system variability in order to identify the challenges of successful transition processes. Additionally, this research employs the reinforcement learning technique to explore the functional transition model generated by the FRAM to investigate a basic method to optimize the transition process for frail people. Results and discussion: The results of this research work show that FRAM-generated models can serve as a basis in further analyses regarding complexity, safety, and risk management. The results also indicate that the DynaFRAM tool helps monitor patients’ hospital-to-home transitions and characterize different types of variability in functional and system outputs. A comprehensive model¹ of the transition process was built using the FRAM. It includes a library of 38 functions classified in five categories. The outcomes of using the DynaFRAM for monitoring patients’ transitions revealed functions with significant variability. The variability observed in the outputs of these functions could be challenging as the variability of a function can reinforce the variability of down-stream functions and affect the performance of the entire transition process. Finally, the results of coupling the FRAM to reinforcement learning help evaluate the system performance in terms of accumulated action value achieved by an artificial agent during functional pathways. Conclusion: In light of the FRAM, the complexity of the transition process can be visualized and understood better. The application of the DynaFRAM helps enhance the situation awareness of frail patients through providing healthcare providers with where a patient is and what they need during the transition process. Coupling the FRAM and reinforcement learning would benefit the healthcare system by providing guidance on how to provide the best care to frail patients in the light of various circumstances. ¹The transition model is called comprehensive as it includes the perspectives of healthcare professionals, patients, and caregivers. It also involves pre-discharge and post-discharge processes.

Actions (login required)

View Item