Thread by @abhishekn

🚨New paper thread🚨 How to reopen the economy safely & yet not hurt the economy?

Our empirical model leverages cellphone location data to map city-based social networks & finds one size does not fit all! Answer depends on local context.

full paper: abhishekn.com/files/network-…
👇🏾

1/ Decisions about how to reopen the economy depend on extremely local factors such as how and where people from different age groups intermingle⁠—

and yet our models of disease spread and economic loss don't take this spatial heterogeneity into account. 🤷🏾‍♂️

Can we do better?

2/ Using privacy-preserving models of the city (via replicahq.com) based on cellphone location data of millions of users we quantify physical interactions between people of different ages or from diff. industries.

We find significant heterogeneity across cities!

3/ Core to our approach is the "contact matrix": an accounting of how and where people of different "types" (age-industry pairs for example) meet each other

4/ our empirical model builds on SIR-type analyses of infection transmission from the epidemiology literature to account for structure of social contacts across types -- feed it any contact matrix and it tells you how many people will be sick or will die.

5/ Our epi model is also grounded in newly-emerging medical parameters such as likelihood of infection transmission ("beta"), duration of illness, mortality rates etc that we calculated from recent medical claims/EMR records from covid19researchdatabase.org

6/ Now we can put our model to work! Choose a policy sequence: say "no policy" for 15 days, then "essential only" for 6 days & then "work from home" -- we can construct representative counterfactual contact matrices and feed it to the model.

7/ Moreover, we can keep track of health and economic costs using the medical claims data & occupation-industry based measures of who can work from home and who cant. Every policy generates employment hours lost due to sick & unemployed who cannot work, and also health costs.

8/ We test our model in 3 cities - Chicago, Sacramento & NY. We test policies of #1-Work from Home if possible, #2-Isolate 60+, #3-Alternate Schedules, #4-Essential Only against our baseline of Cautious Reopening (CR) where all is normal except people lower neighborhood contacts

9/ Here is the sample output from our baseline CR policy in Chicago. As you can see, after CR is instituted in Phase 3 (2nd dotted line), many more people in the "blue" region can work, but infections spike at T=120 with an increase in hospitalizations.

10/ We construct "frontiers" that compare the economic and health costs of our key policies by plotting their absolute outcomes as well as relative effectiveness as compared to the baseline CR policy. Here is Chicago: WFH has 67% death rate and only 88% emp. cost compared to CR!

11/ However, these tradeoffs vary a lot across cities. In Sacramento, all policies are roughly equal for deaths ...

12/ ... while in New York, they differ quite a bit, but not as much as Chicago. WFH reduces deaths by 40% in Chicago as compared to the baseline policy, in New York
this number is about 20%. Chicago has many more vulnerable people and policy matters more!

13/ There are many more insights and differences that pop up at a closer inspection of our results and so you should check out the paper (abhishekn.com/files/network-…)

Also: our results may change as this is v.preliminary work & many improvements and additions are in the works!

14/ This has been a very exciting project enabled by a LARGE team (@akbarpour_ @codyfcook #AudeMarzuoli @Simon_Mongey #MatteoSaccarola @Pietro_Tebaldi @shoshievass and #HanbinYang) and other external partners

15/ Comments/suggestions to improve are more than welcome! And keep any eye on reopenmappingproject.com where we will be posting updated results, papers etc and ways to play with our model going forward.