Simulation Modeling to Inform Public Health Policy and Prevention
Restricted (Penn State Only)
- Author:
- Zafarnejad, Rey
- Graduate Program:
- Industrial Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- April 26, 2023
- Committee Members:
- Steven Landry, Program Head/Chair
Paul Griffin, Chair & Dissertation Advisor
David Vanness, Outside Unit & Field Member
Steven Landry, Major Field Member
Qiushi Chen, Major Field Member - Keywords:
- Public Health
Healthcare Engineering
Agent-based Modeling
Compartmental Modeling
Big Data Analysis
Cost-effectiveness Analysis
Simulation Optimization
COVID-19
Chronic Kidney Disease (CKD)
Opioid Use Disorder (OUD) - Abstract:
- We explore the utilization of healthcare engineering approaches, including agent-based modeling, epidemiological modeling, microsimulation, cost-effectiveness analysis, and optimization as means of informing policy makers of the impact, efficiency and effectiveness of preventive interventions in three distinct areas: COVID-19, chronic kidney disease (CKD), and opioid overdose use disorders (SUD). We develop an agent-based model for the spread of SARS-CoV-2 quanta in a closed classroom environment that extends traditional transmission models that assume uniform mixing through air recirculation. Several non-pharmaceutical interventions were evaluated including distancing requirements, surveillance testing, and updating ventilation systems. More frequent but short classes, contact tracing, and ventilation were effective in reducing the harm caused by the infection by up to 70%. We then develop a joint compartmental SEIRS-SIRS model for the co-circulation of SARS-CoV-2 and influenza and estimate the impact of interventions and such as vaccination, quarantine, and public education. The VENSIM simulation of the model resulted in R0 = 7.5, which is higher than what was reported for COVID-19 infection alone. Vaccination against COVID-19 dramatically slowed its spread and the co-infection of both diseases significantly, while other types of interventions had a limited impact on the co-dynamics of the diseases. We next developed a longitudinal estimate based on cumulative sum chart called CUSUMGFR to identify patients at risk for development of end stage kidney disease (ESKD) based on glomerular filtration rate (eGFR) ≤ 60 ml/min/1.73 m2. Using electronic health records containing 69 million adults, 5,410 patients (ESKD Group) as well as 85,699 patients (Normal Group) were selected to evaluate intrinsic eGFR variation. The CUSUMGFR statistic, with specificity 0.90 and sensitivity 0.88, identified patients who eventually progressed to ESKD on average 791 days prior to their actual diagnosis date. Early identification of these patients could lead to specific clinical pathways and therapeutic measures aimed at reducing the cost and the incidence of ESKD. In addition, the simple and explainable statistic can help involve primary care to mitigate shortages in nephrologists in the US. We then used a micro-simulation approach to estimate the cost-effectiveness of CUSUMGFR based on the incremental cost-effectiveness per disability-adjusted life years (ICER/DALYs) under a variety of scenarios including universal testing, condition-based testing (diabetes and hypertension), and age-based testing. The CUSUMGFR was cost-effective in all cases and could potentially assist in proving evidence-based guidelines for testing for CKD, since currently there are none officially available. We then developed a simulation-optimization framework for the allocation and inventory management of smart vending machines (VM) as destigmatized venues to engage with hard-to-reach, underserved populations of people who inject drugs. We considered three families of items including overdose reversal products (Naloxone/Narcan®), supply contamination tests (fentanyl test strips or FTS) and safe-injection equipment (injection kits including needles and syringes and other injection paraphernalia). For the case where there is no regulation preventing the stocking of injection kits, we estimated that stocking approximately 10% of the VM available capacity by Naloxone kits, 20% by FTS kits, and the remaining 70% by injection kits will minimize the total disease burden (a 25% reduction in comparison to no vending machine policy) associated with four main complications of opioid use disorder (OUD): HIV/HCV, non-fatal OD and OD deaths. In the end, the findings of this research can help inform policy makers and provide guidance as to which preventive policies would be most effective for various public health challenges, and can further be used to assist in developing preventive protocols and framing proactive guidelines.