search

mrc-ide / covid-sim

This is the COVID-19 CovidSim microsimulation model developed by the MRC Centre for Global Infectious Disease Analysis hosted at Imperial College, London.
GNU General Public License v3.0
1.23k stars 256 forks source link

Addition of human dimensons #178

Closed MrHash closed 4 years ago

MrHash commented 4 years ago

In the interest of developing accurate predictive models for billions of humans in very diverse groups, a feature for adding modelling dimensions in order to more accurately predict infection and morbidity risks such as follows would be very useful.

  1. Age
  2. Gender
  3. BMI
  4. Pre-existing conditions
  5. Lifestyle (e.g. smoking, exercise)
  6. Size & type of household (house, apartment, nursing home)
  7. Population density
  8. Vocation type
  9. Diet and prophylaxis

There may be other useful dimensions. Without finer detail on risk profiles across such different groups, the potential influence of such models on global policy can cause unexpected side effects such as increasing morbidity risk on vulnerable groups where previously risk was low, or causing knee-jerk blanket responses which can disrupt the global economy.

weshinsley commented 4 years ago

I assume you've read our model enough to know it includes spatial population density, age, household size, and a degree of workplace assignment (explicit school types, pupils and staff). The other things on your list may be interesting if you can find the data, and have a solid understanding of the way they affect covid transmission and susceptibility.

It looks you are wanting to start your own model, which is fine. But once again, our issue tracker is not a helpful place to discuss that.

MrHash commented 4 years ago

ok great do you have any sources for your model specifications? thanks

On Thu, 7 May 2020, 14:00 Wes Hinsley, notifications@github.com wrote:

I assume you've read our model enough to know it includes spatial population density, age, household size, and a degree of workplace assignment (explicit school types, pupils and staff). The other things on your list may be interesting if you can find the data, and have a solid understanding of the way they affect covid transmission and susceptibility.

It looks you are wanting to start your own model, which is fine. But once again, our issue tracker is not a helpful place to discuss that.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mrc-ide/covid-sim/issues/178#issuecomment-625438731, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAC7ODNSEDOLCFXIMM4AKOTRQMAORANCNFSM4M3RNU3Q .

weshinsley commented 4 years ago

Depending on what you have in mind by "model specifications", you'll need to start with basic classical epidemiology, starting at what SIR models are, and the effects of spatial heterogeneity. From that, the papers on the readme tell you how the covid-sim implements those things. Many of our other reports besides no. 9 are our best estimates for parameterising the classical models with what we currently know about covid, for which we have a team who collect data from international governments and health-care statistics sites. If you mean sources of data, for spatial population density lookup worldpop, UNWPP for age distributions, Eurostat for households/schools/employment inputs.

MrHash commented 4 years ago

It seems that an unexpected side effect that was not considered in the model has shown that individuals who are exposed are infecting everyone in their household or nursing home because of the inability to separate adequately. Will you be updating your models to account for such effects?

weshinsley commented 4 years ago

Please, I beg you, read the things I am telling you to read. This is an absolutely fundamental household behaviour, entirely expected and modelled.

(Care homes not explicitly modelled at the moment in this simulation)

MrHash commented 4 years ago

I'm reading thanks. Could you point me to the section of the code that models this specific behaviour?

weshinsley commented 4 years ago

Lookup CalcHouseInf

MrHash commented 4 years ago

Great. One more question.

I also noticed the DoMassVacc parameter amongst others relating to vaccination in the CovidSim at https://github.com/mrc-ide/covid-sim/blob/030c35001755fd33a6eaad038c72ff8541de410f/src/CovidSim.cpp#L1347 and in the SetupModel

Could you please point me to the reading material or specification where RNA vaccinations are available for human use and can be modeled based on existing data? Otherwise i'd appreciate an explanation as to how vaccination effects can be modeled without any data from human applications. thanks

dlaydon commented 4 years ago

We are not currently modelling any vaccination, either mass or reactive, because there is no available vaccine for Covid-19. We have left this in the model code (i.e. DoMassVac is turned off by default, as you can see in the line you have highlighted) as it may prove useful if a vaccine is developed.

MrHash commented 4 years ago

thanks for elaborating. could you indicate the source of the algorithm for the implementation as it appears, even though it is disabled by default. thanks

On Thu, 7 May 2020, 16:36 dlaydon, notifications@github.com wrote:

We are not currently modelling any vaccination, either mass or reactive, because there is no available vaccine for Covid-19. We have left this in the model code (i.e. DoMassVac is turned off by default, as you can see in the line you have highlighted) as it may prove useful if a vaccine is developed.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mrc-ide/covid-sim/issues/178#issuecomment-625510102, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAC7ODIYOTIUVRHD7PCKB2TRQMSXLANCNFSM4M3RNU3Q .

bbolker commented 4 years ago

Not clear what "algorithm" you're looking for, but you can always search the code for whatever strings you like, e.g. here: https://github.com/mrc-ide/covid-sim/blob/master/src/Update.cpp#L1237

MrHash commented 4 years ago

Thanks, i can see the code. In case you are unclear, when modeling the real world the code follows an "algorithm" that is presumed to be defined clearly and reflects real world data. Can you provide the source of the reasoning behind the code specifically relating to vaccinations? Or is it purely speculative?

On Fri, 8 May 2020, 19:38 Ben Bolker, notifications@github.com wrote:

Not clear what "algorithm" you're looking for, but you can always search the code for whatever strings you like, e.g. here: https://github.com/mrc-ide/covid-sim/blob/master/src/Update.cpp#L1237

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/mrc-ide/covid-sim/issues/178#issuecomment-626076278, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAC7ODIOWR2YKVHAPHGW2HTRQSQZLANCNFSM4M3RNU3Q .

bbolker commented 4 years ago

I don't know if the implementation of vaccination in this simulation code was ever used for public/peer-reviewed papers: this pubmed search%20AND%20vaccination) doesn't find anything obvious. As previously commented, the code in here will be purely speculative until such time as there is a vaccine, when we can start specifying plausible ranges of efficacy, availability, side effects, etc. etc.. If you have specific questions about the implementation of vaccination in the code, someone could try to answer them, but a general "where does this come from?" is hard to answer (and low priority) at this point.

MrHash commented 4 years ago

Ok thanks, you have answered my question. I was simply trying to establish what basis there would be for modelling vaccination efficacy in the current model, and it appears to be speculative.