Super Modelers: 麻豆入口 Faculty Researchers Highlight the Important Uses of Epidemiological Mega Tool
Scholars in Nursing, Engineering, Law and Mathematics explore how mathematical modeling helps predict, prevent and prepare for disease outbreaks
We often hear reference to 鈥渢he models鈥� on newscasts during anticipated severe weather events like hurricanes and snowstorms, referencing the range of likely potential scenarios and impacts the events could bring.
Other important uses of models鈥攖he byproducts of complex mathematical processes鈥攆requently fly under public radar. But in branches of public health science such as epidemiology, there is perhaps no more important method being used to quietly and consistently ensure human health in the face of disease.
"Our role in epidemiology is to understand how diseases spread so we can protect health and save lives," said Monika Pogorzelska-Maziarz, PhD, MPH, CIC, FAPIC, FSHEA, FACE, professor of Nursing at 麻豆入口鈥檚 M. Louise Fitzpatrick College of Nursing. 鈥淚nfectious disease modeling is the number-one tool that we use to try to predict how an epidemic or pandemic will spread.鈥�
And numerous researchers at 麻豆入口 are making their own unique contributions to those efforts.
Predicting The Future
Modeling is a process by which a real-world problem鈥攊ncluding all of its important variables鈥攃an be described and analyzed in a mathematical structure, like an equation. Modeling is used across a wide range of industries and can serve multiple functions, one of which is the ability to forecast future events.
It is an important predictive tool in epidemiology because it is highly customizable and allows for experimentation without harm to humans.
鈥淸Modeling] gives researchers the ability to ask questions without the ethical implications of a real-world study,鈥� said Kaitlyn Muller, PhD, an associate professor of Mathematics and Statistics who works in epidemiological modeling.
Modeling typically begins with the basic principles of epidemiology, such as variables related to distribution and population. These serve as the initial parameters for the model. Additional parameters are then added to account for various scenarios, which further refine the model equation. Unlike statistical modeling, data isn鈥檛 introduced until the end鈥攚hen it is used to validate the model and assess how well it matches real-world outcomes.
鈥淪tatistical modeling starts in the opposite direction, by beginning with already existing data and trying to create an equation that fits it,鈥� Dr. Muller said.
Global-scale disease modeling often relies heavily on statistics. During the COVID-19 pandemic, Dr. Muller noted that the influential model released from London in early spring 2020鈥攐ne that shifted the course for both Great Britain and the United States鈥攚as statistical in nature. It projected how the virus could spread worldwide, growing more complex as new information became available.
On smaller scales鈥攕uch as Dr. Muller鈥檚 work which modeled the potential localized spread of COVID-19鈥攖he math looks a little different.
鈥淭he model we did was very mechanistic and based on those epidemiological principles,鈥� she said. 鈥淲e have this interplay of parameters like symptomatic or asymptomatic, whether you鈥檝e had the disease before, dormancy period or contagious period, human interaction, susceptibility and even suggested policy measures like isolation or no isolation. All of these can be pieces of the model itself.鈥�
鈥淎nd that can inform the response, telling us what resources we will need,鈥� added Dr. Pogorzelska-Maziarz.
This type of work is not unique to novel large-scale global pandemics. There is continuous modeling being done by researchers for more localized disease epidemics across the world, each with its unique set of complications, such as accounting for pathogen-transmitting organisms known as vectors.
Pathogen-spreading vectors like mosquitos add an additional layer when modeling the spread of certain diseases.
鈥淰ectors like mosquitos, for instance, change things,鈥� Dr. Muller said. 鈥淵ou are not just modeling person-to-person interaction, but also vector-to-person and vector-to-vector. If an uninfected mosquito bites a person who has malaria or West Nile virus, that mosquito becomes infected. It鈥檚 going both ways.鈥�
Not only is every disease different, but every place experiencing the disease is as well, environmentally and culturally. For instance, burial practices are taken into account for Ebola modeling, because cultures may differ in how they interact with their deceased, changing how the disease spreads. With leishmaniasis, a parasitic disease, Dr. Muller says modeling has shown using mosquito nettings has potential to greatly reduce the spread, because the disease is prevalent in many places where people live near their livestock.
鈥淚f you really want to understand what's happening,鈥� she said, 鈥測ou do have to dig down deeper and look at those things.鈥�
Disease Modeling Incorporates More Than Just Math
Peter Muller, PhD, associate professor of Mathematics and Statistics, and his student researchers are doing that digging. His research group is currently working on models to find the 鈥渟weet spot percentage鈥� of how many people in Haiti need to be vaccinated, and at what dosage, against human papillomavirus (HPV) to have the disease greatly diminished鈥攐r better, eradicated鈥攊n the country.
His students chose Haiti out of a desire to focus on a country without a national vaccine program or subsidized vaccination strategies. Additionally, Dr. Peter Muller says, in underserved countries, it is harder to obtain data, 鈥渟o hypotheticals like these are what we have to go on.鈥�
Ultimately, they hope to be able to validate their models using a technique that looks at how the model performs against data from a country with similar parameters.
In the modeling, students are accounting for different variables to home in on that 鈥渟weet spot.鈥�
鈥淚 have them read through medical literature and even some psychology,鈥� Dr. Peter Muller said. 鈥淭hey are learning how to parse through data from the Centers for Disease Control and Prevention and also nongovernmental organizations, and how to compare apples and oranges, so to speak, when data is reported differently.
鈥淲e may not be trained in those subjects to the extent that someone in that discipline is, but we are to the extent that we can interpret it into a mathematical model,鈥� added Dr. Peter Muller, who has training himself in medical imaging. 鈥淥ur citations are not all just math papers.鈥�
Ultimately, once the parameters are refined into a working, validated iteration of a model, it can then be used to determine aspects like cost effectiveness, or methods for driving the vaccination percentage to be as high as it needs to be. One of those methods in this specific case, he says, is to address the stigma surrounding the virus, framing vaccination as a preventative for certain cancers HPV can lead to, rather than for the virus itself.
鈥淜nowing that there is a path for potentially increasing the acceptance of getting a vaccine can then play a role in further iterations of the model itself,鈥� he said.
Disease Doesn鈥檛 Travel; People Do
While some researchers working in disease modeling focus more on the disease itself, others, like Chenfeng Xiong, PhD, assistant professor of Civil and Environmental Engineering, look at it from the lens of human mobility.
鈥淒isease doesn鈥檛 necessarily move, but people do,鈥� Dr. Xiong said. 鈥淲e have to rely on a very accurate prediction of human mobility if we want to stop the spread of a disease before it gets out of control.鈥�
Dr. Xiong is leading a research effort aimed at understanding, and ultimately predicting, the spread of infectious diseases by focusing on that key variable of human movement, which has become increasingly important in the connected world of the 21st century.
Disease exposure during the first COVID-19 pandemic wave in Lagos State, Nigeria, estimated by Dr. Xiong’s large-scale integrated human mobility and epidemiological model.
Dr. Xiong鈥檚 work in this field began more than five years ago when he created a COVID-19 dashboard while working at the University of Maryland during the height of the pandemic. Using anonymized mobile device location data from over 100 million monthly active samples, his dashboard tracked human mobility between counties across the United States. Metrics from the dashboard revealed that increased human mobility positively correlated to additional spread of the disease.
This COVID-19 dashboard laid the groundwork for Dr. Xiong鈥檚 research at 麻豆入口, where he is in the fourth year of a five-year project funded by the National Institutes of Health (NIH) to develop a large-scale model designed to predict human movement and determine how that movement contributes to the spread of infectious diseases. This model tracks location data over time and learns to predict where people will move and when, even anticipating what mode of travel they will use and what routes they will take.
What further sets his work apart is its ability not just to monitor these patterns but also to depict them dynamically颅鈥斅璦nticipating outbreaks, identifying potential epicenters and suggesting targeted interventions based on mobility data. Dr. Xiong鈥檚 recent work has even looked at how severe weather events like hurricanes changed that flow of human travel during the pandemic.
To illustrate the importance of the model for prevention, Dr. Xiong describes a major airport in the United States. 鈥淥ur model can track the movement of an infected person, even if they travel on a plane,鈥� he said. 鈥淚f it shows that a major airport is at high risk of receiving imported cases of a disease, like seasonal flu, we could recommend quick interventions to that airport to reduce the chance of disease spreading from travelers to the wider community.鈥�
Although a model cannot stop people from traveling, it can alert health officials as to where an epicenter might occur next, so that prevention measures can be instituted before a potential outbreak.
While Dr. Xiong鈥檚 research into the spread of disease began in response to COVID-19, its scope has since broadened. He is hopeful that his modeling can expand to cover other diseases, including measles, seasonal flu and even tuberculosis. He is also exploring how it can be utilized to track diseases that appear more commonly outside the US, with a demonstration model already up and running in Nigeria in partnership with the Institute of Human Virology.
Models can help inform direct action in places like hospitals, where contagious diseases can spread quickly.
From Equation to Action
The modeling of all these hypotheticals is a key driver in ultimate policy and action plans.
When a disease outbreak occurs, whether a regional epidemic or potential global pandemic, response to that outbreak often begins with models from a country鈥檚 high-level agencies. A bigger outbreak means more information considered from more sources.
For a global pandemic, models trying to calculate how contagious a disease is likely to be would usually start with a partnership between an umbrella organization like the CDC or another country鈥檚 equivalent, and a collaborator. For COVID-19, that was the National Center for Immunization and Respiratory Diseases.
鈥淚n the US, the information goes up the chain at CDC and the Department of Health and Human Services, and it informs the specific recommendations they put out,鈥� said聽Ana Santos Rutschman, JD, LLM,聽professor of Law in 麻豆入口鈥檚 Charles Widger School of Law and director of the聽Health Innovation Lab. 鈥淭he World Health Organization also makes its own calculations using similar models and collaborating with country-level public health agencies.鈥�
A more localized outbreak within the United States would also begin with modeling done by those country-level agencies, like the CDC.
鈥淭heir leadership may then issue general recommendations, but in the US quite a lot of the policymaking in this area will be done at state level,鈥� Professor Santos Rutschman said.
State-level public health authorities then typically recommend a policy or guidance that aligns closely with that of the CDC. These state authorities both do their own modeling and consider those from researchers at universities like 麻豆入口.
Models can also help inform direct action within a more specific sector, like hospitals. Dr. Pogorzelska-Maziarz works frequently with statistical modeling to understand how infectious diseases spread within a hospital setting. Those models are crucial determinants of hospitals鈥� response to both disease outbreaks and routine care.
鈥淲e can turn that data into action,鈥� she said. 鈥淚nfectious disease modeling can show patterns in hospital-acquired infections, staffing workflows, resource use and more, helping teams make changes to routine practices like staffing, room design and infection prevention protocols. Infection control models also provide data for building a business case for hospital resources, which directly impact patients.鈥�
That, she says, is an example that underscores the importance of interdisciplinary collaboration in disease modeling鈥攍ike that of 麻豆入口 researchers and countless other universities, agencies and institutions working behind the scenes to try to predict in order to prepare.
鈥淎nd when it comes to diseases, preparedness means that you're not聽辞苍濒测听prepared, but always vigilant,鈥� said Dr. Pogorzelska-Maziarz. 鈥淚t鈥檚 a never-ending process. We tend to hear about disease modeling only when a new outbreak occurs, but epidemiologists analyze data every day to anticipate risks, strengthen preparedness and protect patients.鈥�