Aaron Miller, PhD, is an assistant professor of epidemiology in the University of Iowa College of Public Health. His research interests include data mining and machine learning, disease surveillance, epidemiological and applied statistical methods, and infectious disease modeling.
He is a member of the UI’s computational epidemiology research group, also known as CompEpi, led by principal investigator Alberto Segre, professor and chair of computer science. The group also includes Dan Sewell, assistant professor of biostatistics, along with faculty from computer science, internal medicine, and pharmacy. CompEpi is part of the CDC’s Modeling Infectious Diseases (MInD) in Healthcare Network.
Miller answered a few questions about how computational tools are used to model, simulate, visualize, and understand the spread of disease, including COVID-19.
Q: In simple terms, what is infectious disease modeling?
A: Infectious disease modeling is the use of mathematical, statistical, and computational methods to represent the transmission of infectious diseases across individuals and populations. Models can be used to represent individual actions of people coming in contact with one another or different elements of the environment (e.g., door knobs, patient beds), and the resulting transmission of disease.
After modeling individual actions, a variety of different aggregate outcomes can be evaluated, such as disease incidence, mortality, or recovery. Thus, infectious disease modeling allows us to understand disease transmission and the resulting consequences of an outbreak across a range of different scenarios and population types.
Q: How is this research used?
A: This research is used both to better understand the factors contributing to disease transmission and to forecast how transmission can be expected to occur under different scenarios. Often the complexity of a particular situation makes it challenging to identify the factors contributing to disease transmission. For example, infections that occur among patients within a healthcare setting can be driven by the actions of multiple types of healthcare workers, other patients, the environment, medications or treatments, etc. Using data from infections that occurred in a particular institution, an infectious disease model might then be used to assess how hand hygiene among healthcare workers contributed to an outbreak.
Alternatively, models can be developed to represent or simulate a real-world settings and can then be used to “play out” the stages of an outbreak. This not only allows us to estimate how an outbreak might be expected to unfold, but also to analyze how different potential interventions (e.g., social distancing or self-quarantine) might impact the overall outbreak. For example, we might develop a model of a particular healthcare environment, then use this model to evaluate the impact of healthcare workers choosing to go to work versus remaining at home if they become ill.
Q: What are some of the current modeling projects the University of Iowa CompEpi group is involved with?
A: Currently, our CompEpi group is funded through the CDC’s MInD-Healthcare Network to conduct research modeling infectious diseases in healthcare settings. Our work primarily involves the modeling of common healthcare-associated infections, such as C. difficile, MRSA, and surgical site infections, along with various prevention mechanisms.
One unique aspect of the Iowa CompEpi group is that we frequently use mobile devices, remote sensors, and other novel data sources (e.g., social media data) to develop more accurate disease models by better measuring disease parameters and population dynamics.
Given the COVID-19 pandemic, our group, and others in the MInD Network, have begun to adapt our models to the unique features of COVID-19. This will allow us to evaluate how the pandemic may be expected to impact transmission within the healthcare environment.
Q: What are some of the challenges of modeling a disease outbreak, especially in its early stages?
A: The biggest challenge with modeling a disease outbreak is what is known as “parameterizing” the model; this is the process of selecting, or estimating, the values that describe the particular disease and setting of interest. For example, disease transmissibility rates, mortality or recovery rates, infection duration, and contact rates between individuals are all types of parameter values that must be specified before modeling an outbreak scenario. An infectious disease model is only as good as the values used to parameterize it, and more accurate parameter values will produce more accurate model estimates.
Models have numerous, sometimes hundreds, of different individual parameter values that may need to be specified or estimated prior to modeling an outbreak. Consequently, modeling groups devote a significant amount of effort to identifying the individual parameter values that go into a model and to better understanding how the choice of different parameter values may impact the resulting model estimates.
At the early stages of an outbreak, disease parameters are often unknown and estimates may fluctuate widely as new information becomes available. For example, with the COVID-19 outbreak, most of these values are still largely unknown and estimates vary widely across different settings where testing has been more/less widely utilized. Indeed, it may be years before researchers fully understand the specific parameters that define the current pandemic.
Thus, at the start of an outbreak, models must be constructed using a range of plausible parameters, and researchers may be required to evaluate outcomes across a wide range of possible scenarios using limited information to define these scenarios. However, infectious disease modeling is often the best way for researchers to assess an outbreak, and possible interventions, in real-time before exact information is available.
Q: What advice do you have for students who may be interested in pursuing a career in epidemiology?
A: Students interested in epidemiology who would like to pursue a career in infectious disease modeling should take coursework in both infectious diseases and quantitative disciplines. Infectious disease modeling requires a range of interdisciplinary methods that include computer science, statistics, mathematics, geographical information systems, and even economics and engineering.