What is a good model for assessing the effect of testing and contact tracing on R0?

I'm imagining it would have parameters perhaps like:

  • testing frequency
  • testing result delay
  • time
  • viral load (varies with time)
  • testing false negatives (probably varies with viral load)
  • path through environment (varies with viral load - i.e. if very sick they self isolate)
  • number of people who intersect that path (perhaps weighted by distance and duration)
  • emission filtration (aka masks)
  • reception filtration (aka masks)
  • percentage of the population tested on a regular frequency (or more complex arrangements if you wish to build that in)
  • percentage of people who go into isolation if they test positive or are requested to do so as a precautionary measure because of the contact tracing conclusions
  • contact tracing density - maybe only 50% of the population is part of the network
  • how deep in the contact tracing web it is recommended to self isolate (perhaps varying with testing frequency because it spreads in time)
  • the growth rate in time (which depends on a lot of the parameters above)
  • some sums or integrals over time and when the testing occurred
  • something something default parameters (some of which can be determined by the original R0 and doubling time of covid-19)
  • something something initial conditions
  • maybe including the percentage of the population already infected or recovered
  • maybe something something pooled tests or randomized population testing
  • maybe something something contact graph structure
  • if very ambitious adding things like packages or perhaps other animals that may be able to catch and transmit it

Ideally it would be able to answer questions like: How much does it decrease R0 if you test 90% of people every 10 days with a false negative rate of 15%, an isolation compliance rate of 85%, a contact tracing web that is 50% dense, and a proactive quarantine over the contact tracing web that is two contacts deep?

New to LessWrong?

New Answer
New Comment
3 comments, sorted by Click to highlight new comments since: Today at 1:59 PM

I'd honestly use twitter and facebook to ask around for epidemiologists if this has decision leverage for donors. There are people working on probabilistic agent models (instead of simplistic quant SEIR like models) but they mostly don't post on the public internet directly about the work.

Wish I'd watched this before. Very good insight into the perils of making models.