Editor's Note: We are thrilled to add Eva Reid to our editorial staff as a regular columnist. In this first article, she explores modeling health data and how that impacts the geospatial community in the coming months with vaccines and further disease tracking. Welcome to the DirectionsMag family, Eva!
Any geographer or geospatial professional knows that “It’s more than just a map.” Dr. Este Geraghty, chief medical officer and health solutions director at ESRI, said the same in our conversation recently about GIS, public health, and a new agent-based model developed by Epistemix, called FRED.
The geographic scientific community has understood that space, place, and health are related for at least 160 years. Some would argue that it’s been much longer than that and goes back to ancient Greece. Regardless, the relationship between place and health has led to many innovations in health care and health research. GIS is an effective tool to understand this relationship, and to support decision-making in public health.
GIS has been used in modern public health research for over 20 years to map patterns of resources and disease, but also to monitor, assess, and predict health outcomes — everything from keeping tabs on outbreaks of disease to assessing potential infections and disease risks and predict sudden emergencies. As technology has advanced, so has the opportunity for improved public health surveillance, understanding virus transmission, and distribution of protective equipment and medication.
GIS and epidemiology
GIS was used extensively to understand recent H1N1 and Ebola outbreaks, and has been used to track other communicable diseases, including Zika, West Nile virus, and measles. Much emphasis is placed on understanding the spatial distribution of the disease epidemiology and answering questions like:
- How quickly is the virus spreading and where is it going?
- How quickly are local and regional hospital resources being depleted?
- What medications are available? Where are they?
Statistical models and simulations are important tools in the epidemiologist’s toolkit to predict the possibility and severity of disease outbreak. These tools are a main source of obtaining and analyzing information for determining the type and intensity of disease intervention. GIS-supported modelling is also immensely helpful, providing place-based “what-if” scenarios to explain disease spread and helping to prioritize how and where vaccines are distributed.
One of the challenges with including GIS modelling in managing a pandemic is that some epidemiologists don’t touch GIS every day. GIS capacity in local health departments is generally “moderate to low,” according to Dr. Geraghty. Not for lack of interest, says Dr. Geraghty, but GIS often ends up in an IT group within the health department (or in another agency completely) and GIS is “not IT, it’s informatics.”
Another challenge often mentioned in healthcare research is the limitation of available population data. Health planning models may have to rely on data that have quality or completeness issues, or have to be aggregated in order to provide individual privacy. Census data, for example, are collected at the individual level but aggregated into various geographies. These data are used in a variety of ways, including identifying populations of need, where to support infrastructure investment, where to promote public programs, and how to encourage investment in specific areas. As good and useful as they are, there are some difficulties with aggregate data and making decisions about human beings using these data. These include:
- Aggregated data provide privacy but don’t allow for understanding individual situations.
- All variables are not always available at all aggregation levels.
- Data are relatively static; some variables may be as old as the last population enumeration.
- Finding environmental risk data is difficult.
Demographic and socioeconomic spatial data also are limited in that they do not always describe patients’ behavior, for example behavior in seeking appropriate services. Enter agent-based modelling or ABM. ABM is a method of analysis used to simulate macro- and micro-level health system behavior. This methodology has been used in a variety of areas of study including transportation, infrastructure investment, and public health.
Epistemix is a new company based on the research of Dr. Donald Burke and John Grefenstette. With their colleagues at the University of Pittsburgh, they developed the Framework for Reconstructing Epidemiological Dynamics as an agent-based modeling system to predict the progression of disease outbreaks. While the system is not new, the partnership between ESRI and Epistemix is a little over a year old, and seeks to leverage the power of geography and behavioral modeling to understand and manage the spread of disease.
A timely connection
Originally, the researchers were looking at applying the system to the spread of measles, but Drs. Geraghty, Burke, and the team quickly realized that FRED was ideal for mapping the spread of COVID-19, and understanding issues of herd immunity, vaccine distribution, and the implications of human behavior in disease management. Understanding where data gaps exist can lead to better, more complete decision-making.
“We are in such a state of uncertainty,” said Dr. Geraghty. “Anything that has scientific methodologies and can give us actionable insights, even if imperfect, I think is incredibly valuable because we need some level of certainty and clarity to make decisions.”
What we know, and where we go from here
The CDC has already determined how many doses of the vaccine likely are needed in each state, based on a variety of factors. As we move into the next phase of the COVID-19 event, it won’t be enough to just map vaccine distribution locations and potential populations of need. We also need to understand the surrounding environment(s), potential risks, and how populations negotiate these factors. Questions raised include:
- Those facilities that have cold storage capabilities – are they near populations that can actually get to them? If they are, do people’s behaviors match?
- Are populations at risk likely to take advantage of services?
Dr. Geraghty pointed out that, in addition to some of these general issues around disease spread and appropriately locating vaccine distribution sites, there are also particular public health challenges when dealing with vaccination, including some degree of vaccine hesitancy in communities of color. How do these concerns get incorporated into models? Can we actually model all of the reasons for hesitancy? It seems that Epistemix’s FRED model might be the way.
Dr. Geraghty said that one of the great benefits Epistemix adds to the conversation is the ability to look at a concept like herd immunity geographically and understand that different communities have different levels of interaction and movement; we might find that we can achieve herd immunity with lower vaccination rates in some areas. This is yet to be determined, but the possibility is there, and could lessen concerns like vaccine hesitancy and whether there are enough distribution sites with appropriate chilling facilities.
It will be interesting to see how health departments work through the coming challenges, but collaborations like ESRI and Epistemix and the inclusion of the FRED modeling system may play a significant role in how the public health community combats this current global health crisis.
Alboaneen, D.; Pranggono, B.; Alshammari, D.; Alqahtani, N.; Alyaffer, R. Predicting the Epidemiological Outbreak of the Coronavirus Disease 2019 (COVID-19) in Saudi Arabia. Int. J. Environ. Res. Public Health 2020, 17, 4568. https://www.mdpi.com/752142
Jennifer Badham, Edmund Chattoe-Brown, Nigel Gilbert, Zaid Chalabi, Frank Kee, Ruth F. Hunter, Developing agent-based models of complex health behaviour, Health & Place, Volume 54, 2018, Pages 170-177, https://doi.org/10.1016/j.healthplace.2018.08.022.
Cassidy, R., Singh, N.S., Schiratti, PR. et al. Mathematical modelling for health systems research: a systematic review of system dynamics and agent-based models. BMC Health Serv Res 19, 845 (2019). https://doi.org/10.1186/s12913-019-4627-7