Date on Master's Thesis/Doctoral Dissertation
Industrial Engineering, PhD
Operations research in health services; trauma center location; patient safety; bi-objective optimization; simheuristic; on-scene injury assessment
Trauma continues to be the leading cause of death and disability in the US for people aged 44 and under, making it a major public health problem. The geographical maldistribution of Trauma Centers (TCs), and the resulting higher access time to the nearest TC, has been shown to impact trauma patient safety and increase disability or mortality. State governments often design a trauma network to provide prompt and definitive care to their citizens. However, this process is mainly manual and experience-based and often leads to a suboptimal network in terms of patient safety and resource utilization. This dissertation fills important voids in this domain and adds much-needed realism to develop insights that trauma decision-makers can use to design their trauma network. In this dissertation, we develop multiple optimization-based trauma network design approaches focusing minimizing mistriages and, in some cases, ensuring equity in care among regions. To mimic trauma care in practice, several realistic features are considered in our approach, which include the consideration of: (i) both severely and non-severely injured trauma patients and associated mistriages, (ii) intermediate trauma centers (ITCs) along with major trauma centers (MTCs), (iii) three dominant criteria for destination determination, and (iv) mistriages in on-scene clinical assessment of injuries. Our first contribution (Chapter 2) proposes the Trauma Center Location Problem (TCLP) that determines the optimal number and location of major trauma centers (MTCs) to improve patient safety. The bi-objective optimization model for TCLP explicitly considers both types of patients (severe and non-severe) and associated mistriages (specifically, system-related under- and over-triages) as a surrogate for patient safety. These mistriages are estimated using our proposed notional tasking algorithm that attempts to mimic the EMS on-scene decision of destination hospital and transportation mode. We develop a heuristic based on Particle Swarm Optimization framework to efficiently solve realistic problem sizes. We illustrate our approach using 2012 data from the state of OH and show that an optimized network for the state could achieve 31.5% improvement in patient safety compared to the 2012 network with the addition of just one MTC; redistribution of the 21 MTCs in the 2012 network led to a 30.4% improvement. Our second contribution (Chapter 3) introduces a Nested Trauma Network Design Problem (NTNDP), which is a nested multi-level, multi-customer, multi-transportation, multi-criteria, capacitated model. The NTNDP model has a bi-objective of maximizing the weighted sum of equity and effectiveness in patient safety. The proposed model includes intermediate trauma centers (TCs) that have been established in many US states to serve as feeder centers to major TCs. The model also incorporates three criteria used by EMS for destination determination; i.e., patient/family choice, closest facility, and protocol. Our proposed ‘3-phase’ approach efficiently solves the resulting MIP model by first solving a relaxed version of the model, then a Constraint Satisfaction Problem, and a modified version of the original optimization problem (if needed). A comprehensive experimental study is conducted to determine the sensitivity of the solutions to various system parameters. A case study is presented using 2019 data from the state of OH that shows more than 30% improvement in the patient safety objective. In our third contribution (Chapter 4), we introduce Trauma Network Design Problem considering Assessment-related Mistriages (TNDP-AM), where we explicitly consider mistriages in on-scene assessment of patient injuries by the EMS. The TNDP-AM model determines the number and location of major trauma centers to maximize patient safety. We model assessment-related mistriages using the Bernoulli random variable and propose a Simheuristic approach that integrates Monte Carlo Simulation with a genetic algorithm (GA) to solve the problem efficiently. Our findings indicate that the trauma network is susceptible to assessment-related mistriages; specifically, higher mistriages in assessing severe patients may lead to a 799% decrease in patient safety and potential clustering of MTCs near high trauma incidence rates. There are several implications of our findings to practice. State trauma decision-makers can use our approaches to not only better manage limited financial resources, but also understand the impact of changes in operational parameters on network performance. The design of training programs for EMS providers to build standardization in decision-making is another advantage.
Hirpara, Sagarkumar Dhirubhai, "Models and algorithms for trauma network design." (2022). Electronic Theses and Dissertations. Paper 4027.
Retrieved from https://ir.library.louisville.edu/etd/4027