INDIANAPOLIS—George Mohler, Ph.D., a professor with the Department of Computer and Information Sciences has received a National Science Foundation grant of more than $149K to develop a new point-processes based algorithm for modeling and forecasting emerging infectious diseases.
Mohler, who is a point-processes expert, says this type of algorithm is better suited for emerging events when there is less data available and more randomness occurring.
Check out this interview with Mohler for a more in-depth look at how point-processes work and how the models can be used by local governments to make informed public health decisions.
Description of the video:
Our new algorithms for threat detection grant is about modeling infectious diseases using models that are called point processes. And common models used during the COVID-19 pandemic like compartmental models kind of need a large group of people to be infected and they're only really applicable uhm when the numbers get really bad. These point processes are more suitable for when you have an emerging epidemic where there's small event counts and there's a lot of randomness involved and what these models, these point process models that we hope to develop can do is detect emerging epidemics and assess the probability that those will turn into a pandemic. Every day we're discovering new diseases and there's an email list and a database that's compiled of all these different diseases that people reporting and what we would like to know is you know what is the risk that this will turn into the next COVID-19 pandemic. Another thing we've seen is you know we're not, we're not randomly testing you know at the levels maybe we could be or testing everybody for diseases. We're doing you know contact tracing based on very limited data and so you have to kind of fill in missing data, you know who who was infecting two and a social network we might not know that for sure but these models can try to fill in that missing information to kind of give it a better picture of an existing pandemic or assess the risk of future ones. People are using these models to estimate the reproduction number of the virus you know so for example, we know that the reproduction number of the delta variant is much higher than the previous variants we've seen which is concerning because that means you spread it to more people if you're infected on average so these models can be used for inferring that those sorts of parameters of a of a pandemic they will also you know be able to be used for forecasting. So you know local governments can make decisions on you know when to open or closed schools or went to require mass so these models can be used to inform kind of public health decisions for this pandemic but also for future ones.
Mohler’s research over the next three years will use COVID-19 data collected by the state and county governments, along with historical datasets (measles, SARS, MERS) and emerging data coming from the Program for Monitoring Emerging Diseases (ProMED). It will be coupled with census data and google mobility data to model how the local spread of diseases is linked to variations in geography, demographics and human behavior.
“Receiving this grant is a way to continue to grow my research program on modeling contagion and infectious diseases. It will allow me to support a PhD student to work on this research with me. It will also allow me to go to conferences to present this work as we conduct the research, to spread the work more broadly. With this support from the NSF program on Algorithms for Threat Detection, we will be able to tackle pressing problems that otherwise we wouldn’t have the resources to solve,” said Mohler.
The methods developed with this project will not be limited to epidemiology applications. The models will be applicable to social media, criminology, seismology and other areas of study.