In Tumbes, to northern region the north coast of Peru, the outbreak of dengue and chikungunya is currently increasing in magnitude and severity. As in many other regions of the country the budget constrains and the limited human resources available does not allows to achieve high coverage rates nor interrupt transmission properly, so we do need to plan of interventions in a more effective way. So my question is which strategy, model or algorithm might be the best to prioritize which communities intervene first in order to maximize the effectiveness of our interventions (standard WHO integrated control intervention including reservoirs management + larvicidal + aerial spraying) if we have a team that only can covered 10% of the regions households per week? Thanks ahead for your mostly appreciated feedback on this matter.
For further reference about Tumbes demographics and geography please refer to http://www.wikiwand.com/en/Tumbes_Region
Can you tell us more about what is different between the communities?
Do you know anything about R_0?
If R_0 is close to 1, then a small change in R_0 has a big impact on epidemic size. However if R_0 is much larger than 1 then a small change in R_0 has almost no effect on the epidemic size.
Widespread interventions would be expected to have a small impact on R_0 everywhere, but focused interventions would be expected to have a large impact on R_0 in those places where they happen.
Based on your question, it looks like R_0 is well above 1. In that case, focused interventions where they are most likely to be effective would make more sense. But obviously this is challenging since it means letting the epidemic spread unaffected in other areas. Since it looks like the best option is more focused interventions, I would expect those focused interventions to be most effective in areas with larger populations.
However - please take note: I am not very experienced with vector-borne diseases. So there may be subtleties that I am not considering. What I am suggesting is a general rule, and may need to be modified in light of the particular circumstances.
Thanks for your feedback. I understand that with better data we can do far better planing, but that is key element of the problem, we don't have better data. The national epidemiological surveillance system in Peru basically only counts weekly cases at each of the reporting units level (which range from health post to regional hospitals), so we can clean the data a bit and estimate prevalence rates at reporting units and then aggregate them to get the regional weekly prevalences. We are currently analyzing NASA data (from TRMM and MERRA products) to characterize climate suitability for vector-borne diseases at the national level so we did a secondary analysis focusing our scope only in Tumbes (see the graph for further reference, it shows the distribution of health centers in Tumbes as well as the overlapping grids of the yearly total precipitation estimates in 2014 according to TRMM). The issue with this analysis is that the grids of data are too wide to offer enough variability to analyze climate suitability at each of the Tumbes reporting units. We have only 11 grids of weather data within the Tumbes area and 44 health centers reporting units. We will deal with such limitation and do some cluster analysis but meanwhile, the problem of maximizing the effectiveness of the current vector control interventions remains at the top of our priorities. If you know a way to estimate the R0 with such limited data in an outbreak context for Chikungunya and Dengue that will be very helpful. Regardless we still need to come up with a better strategy to distribute or interventions in a more effective and scientific way. Our current advice to the regional officers is to prioritize those areas in the region with the highest weekly prevalences for both diseases (because both are transmitted by the same vector: Aedes aegyti) and every week update such analysis to do the same exercise. This way it is expected that regardless of the effectiveness of the integral intervention deployed we might maximize its impact and interrupt transmission around those reporting units rather than over disperse our efforts. I trust that there is a better way to do so we are eager to learn for you what might be such strategy...
Thanks Sergei! I'll prepare an English version of my database without identifiers to share it with you. Just give me until tomorrow to make it as intuitive as possible.
- Badges
- Science topic