It is impossible to exactly capture the trajectory of real-world outbreaks from epidemiological models. However, model predictions can be useful if they produce reasonable approximations of causal outcomes and can be investigated for effects of mechanistic changes in the underlying process. But, models are built with various structural and parameterized assumptions (2016 review), and it is important to understand how those decisions affect the predictions. For this project, we have selected three tutorials that attempt to apply the basic models for disease transmission to simple datasets in order to inform policy decisions. The main purpose of these tutorials is to gain insight into the structures and the assumptions of the models and to provide example tools, datasets, and learning resources that are useful in guiding potential data science projects.
In order to run these tutorials, users will need a resource for running python-based Jupyter Notebooks. The free, default option will be to use Google Colaboratory (“Colab”) environment, which should be available through the Google Apps @ Illinois account. During the tutorial, students will create their own copies of the notebooks, read through the instructions, and run the executable cells. On the provided sample data sets, the execution steps should finish fairly quick. There are three different tutorials provided for this project. The first explains the basic SIR model and attempts to simulate the consequences of different social distancing policies. The second tutorial expands the model by adding the (E)xposed partition for the population and builds an interactive widget for simulating a campus outbreak. The final tutorial runs a simulation based on population health statistics to assess the predicted strain on hospital resources. Please fill out the [cloud access request form] if you would like Google Cloud Platform access to modify the Jupyter notebooks for larger simulations beyond the capabilities of the free Google Colab accounts.
There are many possibilities to extend beyond these tutorials and create a data science project. Projects might focus on ways to improve upon the basic models by incorporating more model features, such as states for being vaccinated, deceased, or asymptomatic, or transitions representing reinfection. Alternatively, a project might want to use additional networking data types when constructing the model, such class schedules, transportation patterns, or social networks. The project could focus on applying these modeling methods to up-to-date, local datasets, policies, and hospital resources or to past outbreaks such as H1N1, MERS, or Ebola. The STEM software project is a well-supported research tool that supports many of the modeling variations with datasets from different past outbreaks. Another direction to pursue could be creating a live, interactive dashboard for others to follow the course of the pandemic. Or finally, a project might be focused on building models that simultaneously capture both disease transmission and economic impact which can then guide potential policy interventions.