Primary Research Area: Kiel’s primary area of research focuses on the mathematical and statistical modeling of infectious diseases, with an emphasis on improving how transmission dynamics are understood, predicted, and controlled. His earlier work has included agent-based modeling and systems of ordinary differential equations (ODEs) to estimate mortality and prevalence rates across a range of healthcare and community environments. Currently, his research centers on Bayesian methodological innovation for infectious disease modeling. In particular, he is developing new applications of the Delayed Rejection Adaptive Metropolis (DRAM) algorithm (see paper 2 and paper 2 in ongoing work) that allow for the stepwise estimation of multiple sets of unknown parameters within complex transmission models. This approach provides a more flexible and realistic framework for handling uncertainty, heterogeneous populations, and incomplete data—common challenges in infectious disease research.
Kiel is also exploring how Bayesian computation and AI-driven learning techniques can be integrated into epidemiologic modeling workflows. This includes ongoing work examining how structural and behavioral factors, such as substance use, influence infection risk and transmission patterns in vulnerable populations. Importantly, these methodological advances are designed not only for infectious disease applications but also for broader public health challenges and emerging health trends where uncertainty, complex systems, and heterogeneous risk patterns play a central role. Related work can be found in his published and in-progress studies.
Primary Research Goal: To develop and apply Bayesian innovations in infectious disease modeling that improve predictive accuracy, better represent real-world transmission dynamics, and ultimately contribute to improved patient outcomes, stronger public health decision-making, and more effective responses to evolving health trends.
Objectives:
Optimize prevention and intervention strategies for infectious disease transmission in diverse healthcare settings, including urban safety-net hospitals and long-term care facilities
Improve clinical and population-level outcomes for patients affected by infectious diseases
Advance understanding of transmission mechanisms in heterogeneous and high-risk populations
Develop more reliable and data-informed models for predicting transmission rates, mortality, and disease prevalence
Integrate Bayesian computation, machine learning, and mechanistic modeling to enhance infectious disease forecasting and intervention planning
Extend these modeling frameworks to address broader public health problems and trends, including health disparities, behavioral risk patterns, and emerging population health threats
More detailed information about Kiel’s previous projects—including work on human contact networks, transmission heterogeneity, and current research at the intersection of machine learning and population health—is provided below.
Past Research Projects
1) Title: "An Agent-Based Model to Assess the Impact of Shared Staff and Occupancy Rates on Infectious Disease Burden in Nursing Homes"
Authorship: Primary Author.
Funding Source: Centers for Disease Control (CDC)
Summary: The COVID-19 pandemic revealed that infection could spread from one nursing home to another through asymptomatic staff members working in both nursing homes. Nevertheless, the transmission dynamics of infection both within and between nursing homes have not been entirely studied for how various levels of shared staff differ in their impact on disease incidence, prevalence, and mortality. Our project aimed to provide more quantitative details about these impacts and determine if there are any significant trends regarding the transmission of infectious diseases within the nursing home with different levels of shared staff. To do this, we created an agent-based model (ABM) to simulate COVID-19 outbreaks in two artificial nursing homes set to various levels of shared staff (0%, 5%, 10%, 15%, 20%, and 30%) and recorded the results in terms of prevalence, incidence, and mortality for each nursing home.
Key Highlights:
The model simulations indicate that reducing the percentage of shared staff below 5% plays a significant role in controlling the spread of infection from one nursing home to another through personal protective equipment usage, rapid testing, and vaccination .
As the percentage of shared staff increases to more than 30%, basic prevention measures (PPE use, vaccination, and isolation) become less effective, and the prevalence of infection reaches an endemic state in both nursing homes.
Estimated individual reproduction numbers suggest that certain staff and residents are more susceptible to becoming super spreaders of an infection once you start employing a 15% shared staff level.
Effectiveness of infection control and basic preventive measures (PPE use, vaccination, and isolation) decreases as the level of staff sharing between nursing homes increases, leading to an endemic state in both nursing homes.
Impact on incidence, prevalence, and mortality rates from shared staff is higher when the nursing home has a higher contact rate.
Project Files Archive: https://github.com/Corkran1/NH_COVID
Paper Link: https://link.springer.com/article/10.1186/s12879-025-10786-w
2) Title: " Bayesian Inference of Nosocomial Methicillin-Resistant Staphylococcus aureus Transmission Rates in an Urban Safety-Net Hospital"
Authorship: Primary Author.
Funding Source: Centers for Disease Control (CDC)
Summary: Methicillin-resistant Staphylococcus aureus (MRSA) is a strain of Staphylococcus aureus that poses significant challenges in treatment and infection control within healthcare settings. Recent research suggests that the incidence of healthcare-associated MRSA (HA-MRSA) is higher among patients treated in safety-net hospitals compared to those in non-safety-net hospitals. This study aimed to identify HA-MRSA transmission patterns across various nursing units of a safety-net hospital, with the goal of improving patient outcomes and facilitating the implementation of targeted infection control measures. A retrospective analysis was conducted using surveillance data from 2019 to 2023. A compartmental disease model was applied to estimate MRSA transmission rates and basic reproduction number for each nursing unit of an urban, multicenter safety-net hospital before and during the COVID-19 pandemic. Posterior probability distributions for transmission, isolation, and hospital discharge rates were computed using the Delayed Rejection Adaptive Metropolis (DRAM) Bayesian algorithm.
Key Highlights:
Analysis of 187,040 patient records revealed that inpatient nursing units exhibited the highest MRSA transmission rates in three out of the five years studied (2019 to 2023).
Transmission pattern found by our simulations consists of relative calm for a few years, then a spike, and then back to relative calm transmission.
Notable transmission rates were observed in certain inpatient and progressive care units (0.55 per individual per month; 0.018 per individual per day) and the surgical ICU (0.44 per individual per month; 0.015 per individual per day).
In contrast, the Nursery NICU and Medical ICU had the lowest transmission rates. Although MRSA transmission rates significantly declined across all units in 2021.
When comparing this to the regular hospital average transmission rate of 0.30, the mean transmission rates from our study indicate moderate infection control, though occasional spikes highlight areas for improvement.
Project Files Archive: https://github.com/Corkran1/UHKC-MRSA-Transmission-Model-FIies
Link: https://www.journalofhospitalinfection.com/article/S0195-6701(25)00234-8/abstract
3) Title: " Assessing the Impact of Biosecurity Compliance on Farmworker and Livestock Health within a One Health Modeling Framework"
Authorship: Secondary Author.
Funding Source: Centers for Disease Control (CDC)
Summary: The adherence to protocols designed to prevent outbreaks and spread of infectious diseases is known as biosecurity compliance. While several models only focus on animal health, the present work develops a one-health modeling framework to investigate how different degrees of compliance can influence the health of both animals and farmworkers. The model consists of a set of ordinary differential equations representing the spread of substance-susceptible and substance resistant pathogens in the environment and animal population, coupled with a set of stochastic differential equations representing the spread of pathogens in the farm worker population. The next-generation matrix approach is employed to estimate the basic reproduction numbers (R0) under three different hypotheses of transition, transition–reservoir, and reservoir both for substance-susceptible and substance-resistant pathogens, where they all agree on the same threshold value (i.e. R0 >1 implies disease outbreak and R0 <1 results in the disappearance of infection). Using the data from the existing literature, the model is validated and certain parameter values are estimated. Also, the conditions for the existence and stability of disease free and three different endemic equilibria are established based on the R0 expressions. Focusing on transmission of infection within adult dairy cows on a dairy and the risk of spread to farmworkers, we investigate the effects five different biosecurity measures and their combinations: (1) animal movement control and quarantine, (2) disease monitoring and reporting(3) hygiene and disinfection, (4) feeding and watering practices, (5) antimicrobial stewardship, and (6) manure and contamination management.
Key Highlights:
Numerical simulations indicate that compliance with measures associated with host-to-host transmission has the highest impact on the prevention of outbreaks and improving the health of both animals and farmworkers.
Particularly, low compliance with this measure results in high prevalence and incidences of infection, whereas good or excellent levels of compliance prevent outbreaks or result in minimal incidences of Salmonella infection in cows.
Further model analysis indicates that full compliance with the other three biosecurity measures does not guarantee the prevention of outbreaks in a dairy farm.
These results are consistent with the local and global sensitivity analyses of the model.
The modeling framework developed in this study can be applied to other zoonotic infections, and it can be used as a guiding tool for optimal resource allocation, leading to the minimization of disease spread both in animal and human populations.
Link:
4) Title: " Comparative Analysis of Machine Learning Models for Predicting Hospital- and Community-Associated Urinary Tract Infections Using Demographic, Hospital, and Socioeconomic Predictors"
Authorship: Secondary Author.
Funding Source: Centers for Disease Control (CDC).
Summary: Urinary tract infections (UTI) are among the most common infections encountered in both community and healthcare settings. Differentiating between community-associated UTI (CA-UTI) and healthcare-associated UTI (HA-UTI) is crucial for understanding their epidemiology, identifying risk factors, and developing appropriate treatment strategies. Machine learning (ML) techniques have shown significant potential in improving the accuracy of predicting these infections, enabling more effective interventions and better patient outcomes. While previous studies have demonstrated the utility of ML models in various healthcare settings, there is still a need for a comparative analysis of different ML approaches, particularly in distinguishing between CA-UTI and HA-UTI and assessing the risk of UTI among hospitalized patients. Using 2019-2023 patient demographics, hospital, and socioeconomic data, this study aims to build, validate, and compare machine learning models—Decision Tree (DT), Neural Network (NN), Logistic Regression (LR), Random Forest (RF), and Extreme Gradient Boosting (XGBoost) to differentiate between the incidences of HA-UTI and CA-UTI. Additionally, it seeks to identify key predictors of UTI using demographic, hospital, and socioeconomic variables.
Key Highlights:
Results: The DT model demonstrated the highest sensitivity, particularly in handling the highly imbalanced data of HAI, with a sensitivity of 87%. LR achieved the best overall accuracy, at 95.9% for HA-UTI and 93.2% for HA-UTI vs. CA-UTI. RF performed best in cross-validation, reaching 99.1% for HA-UTI and 96.2% for HA-UTI vs. CA-UTI. NN showed the highest specificity, at 93.4%, for HA-UTI vs. CA-UTI. The AUC values further supported these findings, ranging from 71.9% for NN to 96% for RF, reflecting the robustness of these models across different annual datasets.
Among patient demographics, hospital, and socioeconomic variables, all models consistently identified the nurse units (e.g., inpatient units and mental health units) as the most significant predictors of UTI. In addition to nurse units, LR and DT identified location (e.g., various clinics and medical centers) as a key predictor. For HA-UTI versus CA-UTI, variations were observed across the years, with patient age, median household income, and gender intermittently emerging as key predictors.
In addition to nurse units, LR and DT identified location (e.g., various clinics and medical centers) as a key predictor. For HA-UTI versus CA-UTI, variations were observed across the years, with patient age, median household income, and gender intermittently emerging as key predictors.
The predictive accuracy of the machine learning models is relatively the same, with some differences in sensitivity and specificity for identifying both HA-UTI vs. CA-UTI and HA-UTI. Nurse units consistently emerge as the most significant predictors across all years.
The importance of all predictors, such as socioeconomic factors and location, varies from year to year, highlighting the need for incorporating those variables in the surveillance systems to optimize the accuracy of predictions.
Current Publications in Progress
1)Title: "Bayesian Inference of Hospital-Acquired MRSA Transmission Among Patients Declaring Substance Use at Admission"
Authorship: Primary Author.
Funding Source: National Research Foundation (NSF)
Summary: The emergence and persistence of infectious diseases in many communities are closely linked with substance use. Behaviors associated with substance use—particularly injection practices, unstable housing, and barriers to routine healthcare—can increase exposure risk, delay treatment, and contribute to the spread of infectious pathogens. At the same time, many individuals with substance use disorders experience underlying health conditions or immune compromise that further elevate susceptibility to infection. This research proposes to advance Bayesian infectious disease modeling frameworks so they better reflect transmission dynamics in high-risk populations, rather than relying on assumptions developed for lower-risk or more homogeneous groups. Our goal is to develop predictive tools that incorporate behavioral, clinical, and structural risk factors to more accurately estimate transmission patterns, identify key recovery and intervention points, and support targeted prevention strategies. These methods will combine Bayesian statistical inference, data-driven learning techniques, and simulation of patient behavior and contact structures.
To demonstrate and validate this framework, methicillin-resistant Staphylococcus aureus (MRSA) will be used as a case study. MRSA is a clinically significant healthcare-associated pathogen with well-documented disparities in burden among vulnerable and socially marginalized populations, including individuals with histories of substance use. By focusing on MRSA transmission within healthcare settings serving high-risk patients, we can evaluate how traditional models underestimate transmission heterogeneity and how enhanced Bayesian approaches improve prediction and intervention planning.
Ultimately, this work aims to produce a flexible modeling framework that can be extended beyond MRSA to other infections disproportionately affecting people with substance use histories, such as hepatitis C and tuberculosis. By improving our ability to model transmission, recovery dynamics, and intervention effects in high-risk populations, this research seeks to inform more effective public health strategies that address both infectious disease spread and the broader health consequences of substance use.
Key Highlight:
MRSA serves as the primary case study for developing and validating Bayesian, data-driven transmission models tailored to high-risk patient populations.
Publication Status: In progress (Accepted publication target date is March 2025)
2)Title: "The Influence of Prior Antibiotic Use on MRSA Transmission Among Patients Declaring Substance Use at Admission."
Authorship: Primary Author
Funding Source: National Research Foundation (NSF)
Study Background: Patients with histories of substance use face elevated risks of healthcare-associated infections due to a combination of medical, behavioral, and structural factors. One factor that may play a major but under-studied role is prior antibiotic exposure, which can alter normal flora, increase antimicrobial resistance, and influence susceptibility to colonization and infection. This project examines how prior antibiotic use affects MRSA (methicillin-resistant Staphylococcus aureus) transmission risk among hospitalized patients reporting substance use. The goal is to understand whether controlling for antibiotic exposure changes which risk factors appear most important for predicting MRSA transmission.
Study Design: Patients are divided into two study populations:
I Control Group – Prior antibiotic exposure is explicitly controlled for in the analysis
II Experimental Group – Antibiotic exposure is not controlled, allowing its influence to operate indirectly through other clinical and behavioral variables. By comparing these groups, we can evaluate how antibiotic exposure shapes the identification of MRSA transmission risk factors.
To identify predictors of MRSA transmission, several logistic regression strategies are applied to both groups:
Simple Logistic Regression: Each predictor is examined individually to assess its unadjusted association with MRSA transmission.
Multiple Logistic Regression: All predictors are included together to estimate their adjusted effects
Additional Methods were used in the Experimental Group include
Stepwise Logistic Regression: An automated selection process identifies a parsimonious set of predictors most strongly associated with transmission.
Hierarchical Logistic Regression: Predictors are entered in conceptually related blocks (e.g., demographics, comorbidities, clinical exposures, behavioral factors) to evaluate how relationships change as additional information is incorporated.
Key Highlight from Paper:
How does controlling for prior antibiotic exposure change which factors appear most strongly associated with MRSA transmission?
Comparing results across modeling approaches and between study groups, this project identifies the following 1) Predictors that remain important regardless of antibiotic exposure, 2) Predictors whose significance depends on whether antibiotic use is factored in, 3) antibiotic exposure may confound or mediate transmission risk.,
Understanding the interaction between antibiotic exposure and transmission risk will allow us 1) to improve infection prevention strategies, 2) better support antimicrobial stewardship efforts, 3) better protect vulnerable patient populations (such as substance users), and 4) better inform policymakers in healthcare settings where patients have substance use histories.
This project can be extended to other infections that disproportionately affect high-risk populations.
Publication Status: In progress (Accepted publication target date is March 2025)