Following from last week’s blog post, Pete and Sam and I decided that in the framework of intervention space we’re exploring(namely,“what if interventions were conducted extremely well”), we should proceed with annual interventions and no outbreaks/importation(the latter using the assumption that our highly effective interventions would be equally effective everywhere). However, there were still a few parameters we were interested in exploring, namely:
Heterogeneous biting risk;
Correlated ITN usage, in which some portion of the population never uses a net irrespective of intervention coverage;
Health-seeking rate in the ACT intervention. Here,“coverage” refers to someone’s probability of seeking care at all given a clinical case, while the“health seeking rate” refers to their per-day probability of seeking care, given that they’re going to seek care eventually. So far, we’ve been setting this probability to 0.15.
Heterogeneous Biting
Some preliminary data from Burkina implies that biting risk is certainly not homogeneous, and that an exponential is perhaps an acceptable distribution to describe risk in the population:
With that in mind, I reran all the simulations from last time, but with an exponential biting risk(lambda=1). Results are below-- homogeneous risk is always on the left, heterogeneous on the right:
ITNs:
IRS:
ACT:
Adding biting heterogeneity adds extensive residual transmission to the model. I personally am more comfortable with this framework than homogeneous transmission, and think it will be important to include even under highly effective intervention schemas.
Correlated Net Usage
NB: These simulations are in the heterogeneous biting framework, but results should be similar(if attenuated) under homogeneous biting.
For visibility, I’m zooming in on our Cameroonian site, Martae. As we increase the proportion of the population who never receive or use nets, we see a considerable weakening of net impact:
I don’t think we need to get as extreme as 50, or even 20%, non-usage, but including some small proportion of non-users would be reasonable in our scenarios, to represent travelers or ethnically/religiously isolated groups.
ACT Health-Seeking Rate
NB: Same stipulations and setting as for correlated net usage.
Here’s Martae again, varying both the probability of seeking care at all(“coverage”) and the per-day probability of seeking care after symptom onset(“health-seeking rate”, represented by subplots):
Varying this parameter has a smaller impact, probably because everyone infected at a given coverage does seek care eventually. We see a bit of a“drop in the elbow” for faster health-seeking rates, but the overall message stays the same. I’m happy sticking with our 0.15.
Feedback Needed
My main questions for the group(especially Pete, Sam, Jaline) are:
Do we keep heterogeneous biting? If yes, are we happy with the way I’ve parameterized it?
What proportion of our population should never use nets? I’d like the answer to be“more than zero”, but otherwise am open to suggestions.
What should the ACT health-seeking rate be? In the absence of strong objections, I’d like to stick with 0.15.
Once these questions are answered, I’m happy to start a full run of all interventions and their interactions for these sites.
Heterogeneous Biting
Correlated Net Usage
ACT Health-Seeking Rate
Feedback Needed