Predicting and assessing the adaptability of healthcare facilities to emerging challenges, using simulation-powered predictive analytics and reinforcement learning
Yehuda Kalay, Professor
Architecture
Applications for Spring 2025 are closed for this project.
The performance of healthcare facilities is an important issue due to the high value of the services they provide on one hand, and the high cost of constructing and operating them on the other. Various disciplines have offered measures to assess such performance, focusing on different indicators. The healthcare industry offers people-centered measures, focusing on quality of care as well as staff and patient satisfaction. Operations Research (OR) offers operational and cost measures, focusing on optimizing the flow of activities performed in the facility. Architecture and Engineering offer building-oriented measures, focusing on optimizing the physical environments in which people perform their activities. All these measures are based on evidence gleaned from past experiences, drawing conclusions from post-occupancy studies and surveys that examine the correlations between decisions and their outcomes.
But how well are healthcare facilities prepared to handle new, emerging challenges, such as pandemics, earthquakes, mass-casualty events? How quickly can they adapt to new situations? At what cost (financial and human)?
The hypothesis underlying this research is that the overall performance of healthcare facilities (like many other types of facilities) is the product of three measures: (1) clinical outcomes and staff/patients’ satisfaction, (2) operational efficiency, and (3) space utilization. These measures are not independent of one another: often improving one will affect others—negatively or positively. Decisions made separately in one part of the facility may directly or indirectly impact other parts. Measures taken to resolve one conflict at the present may create other conflicts down the line.
The research aims to assess the adaptability of healthcare facilities to handle emerging needs in short (days), medium (months), or long (years) time frames. It does so by integrating and trading off the effects of modifying operational procedures, staff/patient satisfaction measures, and space utilizations intended to address emerging challenges. It uses forward-looking simulation and AI-based reasoning to predict and assess the effectiveness and cost of alternative courses of action.
The Cardiac Catheterization Lab (CCL) at St. Bernardine Medical Center (SBMC) in southern California has been chosen as case study for the research. It is a complex and dynamic facility that serves a large population in all matters related to cardiac care. At any given moment, decisions must be made concerning the allocation of resources (spaces, people, activities) in a manner that will maximize operational efficiency, space utilization, and staff and patient satisfaction. Such actions are taken simultaneously by multiple actors located in different parts of the facility, who are typically not aware of the actions, or even needs, of other actors. A preliminary study of the CCL was conducted to understand the prevailing spatial, workflow and staffing conditions at SBMC. It comprises the system’s knowledge base, from which future activities and states of occupancy can be projected by means of digital event-based simulation.
We are developing an abstracted digital model of SBMC’s Cath Lab, a Discrete Event Simulation engine that generates alternative future occupancies and activities scenarios, and a Reinforcement Learning (RL) system for analysis and evaluation of the simulated alternatives.
We are looking for 1-2 URAP students who will help develop the simulation engine and RL assessment.
Role: • Collaborative coding, troubleshooting, and applying various Reinforcement Learning models to simulate the hospital's daily operations and decision making.
• Visualizing and reporting results.
• Reading, understanding, summarizing, and extracting key actionable insights and models from leading-edge papers in the broad fields of machine learning.
• URAP students who participate in the project are expected to be self-motivated, able to suggest development directions as well as critique such directions, and learn new methods as needed.
Qualifications: Applicants should possess skills in at least one of the following, and strong interest in the others:
• Knowledge of Search, Graph, Reinforcement Learning and other machine learning models (related courses such as CS188, CS189, CS186).
• Experience with coding, especially related to applying machine learning models (preferably in Python).
• Interest and motivation to read research papers in a variety of fields describing AI / ML models.
• Data visualization and reporting (e.g. Pandas, Matplotlib).
Day-to-day supervisor for this project: Professor Davide Schaumann
Hours: to be negotiated
Off-Campus Research Site: Once a week Zoom meetings, lasting 1-2 hours.
Engineering, Design & Technologies Digital Humanities and Data Science