Kinsa, a company that sells smart thermometers, has a dashboard that shows which regions of the US have an unusually high number of fevers. They have previously used these methods to track regional flu trends in the US. (FitBit has done something similar.)

I wrote a post here describing my attempt to turn their data into a rough estimate of the total number of coronavirus infections in the United States. Something similar could be done for smaller regions.

[ Parent Question — LessWrong Coronavirus Agenda ]

How can we estimate how many people are C19 infected in an area?

by Elizabeth 1 min read18th Mar 20204 comments

21


My favorite dashboard is Our World in Data: it has time series and maps and lets me choose which countries to graph. But it’s only showing confirmed cases, and doesn’t show any unit smaller than a country. Plus on the day I’m writing this, the US was missing yesterday’s day. Plague Plus is trying to do true prevalence estimates, but I don’t like their methodology, in part because it’s the same for every country.

Can I do better than this? Is something out there that shows city level data, with time series? If I must use confirmed cases, how can I translate those numbers into true prevalence estimates, given region-specific conditions?

New Answer
Ask Related Question
New Comment

4 Answers

Kinsa, a company that sells smart thermometers, has a dashboard that shows which regions of the US have an unusually high number of fevers. They have previously used these methods to track regional flu trends in the US. (FitBit has done something similar.)

I wrote a post here describing my attempt to turn their data into a rough estimate of the total number of coronavirus infections in the United States. Something similar could be done for smaller regions.

There's currently a Foretold community attempting to answer this question here, using both general Guesstimate models and human judgement taking into account the nuances of each country. We've hired some superforecasters from Good Judgement who will start working on it in a few days.


(Tangential: as part of the Epidemic Forecasting project at FHI we are feeding this data into GLEAM, which is a global SEIR model running on high-performance computers, based on a database of millions of airline and commute connections. The model also tries to factor in in seasonality, air traffic reductions, and effectiveness of various containment measures.)

The Johns Hopkins Center for Systems Science in Engineering has time series data at the state and province level for some countries (US, China, Canada, Australia). They used to have county-level data for the US but no longer provide it. Unfortunately the case numbers are only confirmed + presumptive positive, so it's not everything you're asking for, but it seems like it gets close.


https://github.com/CSSEGISandData/COVID-19/tree/master/csse_covid_19_data/csse_covid_19_time_series

A few theoretical data sources that are pretty geographically specific:

  • Hospital patient care records
  • 911 calls
  • Patient records from 911 calls
  • 311 calls
  • 911 detailed questions and answers

In the United States, some more concrete thoughts :

  • 911 calls can be aggregated by asking my day job to integrate with your database.
  • Detailed Q and A in 911 can be aggregated by the Q and A software vendor's partners (hi!) or themselves.
  • Patient data from 911 is aggregated sporadically at the state level (NEMSIS) and my employer
  • Hospital data is aggregated by ESSENCE (CDC and state level) with an unknown delay
  • There are also other sporadic aggregators of hospital data.

If this is too much advertising, you can edit or reject this comment.

To summarize, your question is somewhat technically feasible. It's just a lot of work.

1 Related Questions

Parent Question