Martyn Fyles

Martyn Fyles

Statistics PhD student at the University of Manchester, funded by the Alan Turing Institute. Infectious disease modeller, statistician, data scientist.

Household Branching Process and Contact Tracing Software (Household-Contact-Tracing)

Available at: https://github.com/TTI-modelling/TestingContactModel

During the SARS-CoV-2 pandemic, I led the development of household-contact-tracing - a simulation package designed to provide rapid analysis of advanced contact tracing methods without large, computational overheads. The SARS-CoV-2 epidemic is modelled using a household branching process, which is essentially a modified multi-type branching process typically used to represent the spread of a pathogen through a human population segmented into household. Human social structures, such as households, are important to the spread of respiratory pathogens due to the close proximity of members of the household. Further, interventions such as quarantine could be enacted at the level of households. Contact tracing between members of the household is also considered to be dramatically easier, and faster than the tracing of individuals between households, and as such the dynamics of household structures may play important roles when considering the efficacy of contact tracing interventions, particularly if leveraged appropriately.

The model has been restructured to allow for a more modular framework. Various epidemic processes can be implemented, such as a household-based branching process or an individual-based branching process. These processes model epidemics under different assumption, an spawn household/individuals with various attributed, depending upon what is required of the simulation. For example, if we are investigating lateral flow tests, then individuals will be spawned with required status’ that are required to implement lateral flow testing enabled contact tracing processes. Separate to the epidemic process, a contact tracing process is being propagated over the structure (network) generated by the epidemic process at each time step. Again, contact tracing processes are modular, and can be combined with other processes depending upon which question is being asked. For example, we combine a contact tracing process with a model of testing demand which allows us to understand how contact tracing efficacy decays as the demand for tests exceeds the capacity for everyone to get tested.

So far, we have used the household-contact-tracing simulation framework to:

  • Provide early predictions of the efficacy of contact tracing in the UK
  • Provide scenario modelling of different lockdown relaxations
  • Investigated household-based contact tracing (a variation on standard contact tracing leveraging household structure)
  • Investigated two-step contact tracing (tracing contacts-of-contacts of SARS-CoV-2 cases)
  • Evaluated the utility of out-of-household isolation strategies in preventing household outbreaks, and in protecting vulnerable individuals
  • Provided a rapid evaluation of the use of LFD tests in daily contact testing strategies
  • Investigated the use of LFD tests in border control policies
  • Developed a framework for analyzing the implementation of contact tracing processes when there are limited resources available
  • Provided modelling support to consider how to optimally use limited sequencing resources

We are in process of adding all additional modelling modules to the code that we have released at: INSERT LINK HERE. Further, we are completing additional documentation and unit testing - currently we have the most basic models available for use.