App for COVID-19 contact tracing

Overview

This is a pro­posal for an app that would al­low users to log/​track their in­ter­ac­tions, en­abling faster and more effec­tive con­tact trac­ing in the event that they con­tract COVID-19. Cur­rent meth­ods for con­tact trac­ing are man­ual and rely on hu­man mem­ory, so this tool could greatly im­prove our abil­ity to track the spread of the dis­ease and com­mu­ni­cate vi­tal in­for­ma­tion to peo­ple at risk. In par­tic­u­lar, I think that build­ing this tool could be an effec­tive use of time for EAs, es­pe­cially those with web/​mo­bile de­vel­op­ment ex­pe­rience.

There are a few things I’d love to re­ceive help on:

  • Does this seem worth do­ing? In par­tic­u­lar, are there difficul­ties or draw­backs that I may have missed?

  • There is an open ques­tions sec­tion. If any­body knows the an­swers to those, I would ap­pre­ci­ate it.

  • If you are a web/​mo­bile de­vel­oper and want to con­tribute, or know any­one who is, please let me know.

Motivation

This can help re­duce the spread of COVID-19 im­me­di­ately.

The cur­rent plan that I be­lieve most gov­ern­ments are im­ple­ment­ing is to im­pose strong shut­down and shelter-in type mea­sures to sup­press the dis­ease un­til case num­bers diminish. At that point, they will at­tempt to re­open things while main­tain­ing low case num­bers un­til a vac­cine is de­vel­oped, likely in ~18 months. Con­tact trac­ing is one of the main tools in use to keep the num­ber of cases low.

Ad­di­tion­ally, con­tact trac­ing is pri­mar­ily an in­for­ma­tion and com­mu­ni­ca­tion prob­lem, which seems like some­thing that tech­nol­ogy can help with a lot. There­fore I think it’s a very promis­ing and tractable op­por­tu­nity for those with pro­gram­ming ex­pe­rience to con­tribute, and the EA com­mu­nity has many peo­ple with that back­ground. Most other ways to help with COVID-19 re­quire med­i­cal train­ing or ex­pe­rience in man­u­fac­tur­ing med­i­cal de­vices such as ven­tila­tors.

This could be helpful with fu­ture pan­demics.

Un­for­tu­nately, COVID-19 will likely not be the last pan­demic hu­man­ity faces, and de­spite the po­ten­tial to have a hor­rific death toll, it is still at a lower scale than some of the catas­tro­phes EA has been con­cerned with. My hope is that this will help in the long term by:

  1. Im­prov­ing in­fras­truc­ture for con­tact trac­ing, which seems likely to be use­ful for many in­fec­tious dis­eases, nat­u­ral or syn­thetic. If this had ex­isted be­fore­hand, I think we would be in a bet­ter place.

  2. En­abling us to col­lect data on dis­ease spread and use that to un­der­stand it bet­ter, as­sum­ing we can do this in a way that re­spects pri­vacy.

What is con­tact trac­ing?

Con­tact trac­ing is a com­mon pub­lic health tool, typ­i­cally con­sist­ing of three steps, start­ing when a pa­tient is di­ag­nosed with a com­mu­ni­ca­ble dis­ease:

  1. In­ter­view the pa­tient to iden­tify all peo­ple they re­cently in­ter­acted with (go­ing back 2-3 weeks in the case of COVID-19)

  2. Call all of these con­tacts and ask them to iso­late them­selves.

  3. Mon­i­tor all of them in case they were in­fected with the dis­ease.

For more in­for­ma­tion, see the WHO site or the Wikipe­dia page.

Min­i­mum Vi­able Product

The pro­posal is to make an app that would record con­tacts in real time to en­able eas­ier, faster, and higher re­call con­tact trac­ing once some­body is sick. Since I think it is quite ur­gent to get this out quickly, I will de­scribe what I think is the min­i­mum vi­able product (MVP) be­low and then de­scribe ad­di­tional fea­tures af­ter.

The MVP would be a web ap­pli­ca­tion that does the fol­low­ing:

  1. Prompt a user each day to man­u­ally record ev­ery­one they have in­ter­acted with that day (likely from a list of FaceBook friends or phone con­tacts). They can also list peo­ple they in­ter­act with ev­ery day to avoid con­stantly up­dat­ing.

  2. If a user is di­ag­nosed with COVID-19, they are able to no­tify ev­ery­one they have in­ter­acted with for the past three weeks with the click of a but­ton. This no­tifi­ca­tion can be anony­mous and should come up with links to in­struc­tions on how to safely deal with this situ­a­tion.

This would help by mak­ing con­tact trac­ing:

  • Easier: They would only need to click a sin­gle but­ton in­stead of man­u­ally call­ing lots of peo­ple. Since pub­lic health offi­cials may be too over­whelmed to do this prop­erly, this could in­crease the amount of con­tact trac­ing hap­pen­ing sig­nifi­cantly.

  • Higher re­call: I would find it hard to re­mem­ber ev­ery­one I have had con­tact with in the past three weeks, par­tic­u­larly when sick enough to be in the hos­pi­tal. Do­ing this ahead of time should make it more com­plete.

  • Faster: No time is wasted wait­ing on pos­si­bly over­whelmed offi­cials to do the con­tact trac­ing, it can be sent out as soon as a di­ag­no­sis is re­ceived.

Ad­di­tional Features

I think the above will provide some help and re­quire rel­a­tively lit­tle from a user, but there is also the op­por­tu­nity to im­prove this with more fea­tures.

Lo­ca­tion tracking

The most ob­vi­ous is to make this into a phone app and use lo­ca­tion track­ing to de­ter­mine ev­ery­where that peo­ple have been. This is cur­rently be­ing used suc­cess­fully in coun­tries like China, South Korea, and Sin­ga­pore. A less in­ten­sive form of this would just ask the user to man­u­ally add spe­cific stores or crowded lo­ca­tions they have been in and at what time.

The above lo­ca­tion could then be used to make a map of what ar­eas are dan­ger­ous and have had peo­ple with COVID-19 near them, likely ex­piring af­ter three days. South Korea has a map similar to this.

Assess­ing risk level be­yond first-or­der contacts

Tra­di­tional con­tact trac­ing only con­sid­ers first-or­der con­tacts, prob­a­bly for rea­sons of effi­ciency. How­ever, this app could in­form peo­ple if they are con­nected to some­one with COVID-19 in sec­ond de­gree con­nec­tions. Other risk fac­tors could also be con­sid­ered to give an over­all risk rat­ing for a user.

Re­port symp­toms and send less con­fi­dent warn­ings based on that

There is also a short­age of tests in many lo­ca­tions, so many cases are likely go­ing un­di­ag­nosed at first. The app could also query for symp­toms and send out less con­fi­dent warn­ings if re­cent con­tacts have fever/​dry cough or other com­mon in­di­ca­tors of COVID-19.

Gam­ifi­ca­tion types of things to in­crease ad­her­ence to recording

In or­der to make sure that peo­ple re­ally do en­ter their con­tacts each day, we could make it visi­ble to oth­ers how long it has been since they up­dated, send texts if some­one lists them as a con­tact that they didn’t list, etc. I’m sure there are use­ful les­sons to be learned from ex­er­cise/​diet apps that re­quire the user to en­ter all their work­outs or meals as well.

Concerns

There are cer­tainly ways this could go poorly or not be that helpful, and I would love to hear from any­one who has ideas or con­cerns I have not con­sid­ered.

Pri­vacy concerns

Users may have a very strong aver­sion to giv­ing their lo­ca­tion at all times to an un­offi­cial and un­trusted app. This is un­der­stand­able, but I imag­ine there are a num­ber of peo­ple who are will­ing to sac­ri­fice some of their pri­vacy in an emer­gency to pro­tect their health and that of oth­ers. It would definitely be im­por­tant to be very trans­par­ent on what data is be­ing col­lected and make sure the user has opted in to shar­ing that.

Others are already work­ing on this

There are many cases of this be­ing worked on already. South Korea, China, and Sin­ga­pore all have similar cases, mostly run offi­cially by the gov­ern­ment. The WHO is cur­rently work­ing on an offi­cial app that will do some of this.

I think there are a lot of ad­van­tages to hav­ing an offi­cial branded app, and could imag­ine dis­con­tin­u­ing or merg­ing this pro­ject if one of those is suc­cess­ful. How­ever, at the mo­ment, I be­lieve that most peo­ple in Western coun­tries are *not* track­ing their con­tacts.

Of the apps I am aware of, the WHO app is the only one that seems tar­geted for the US or Western coun­tries. It’s v0.1 does not in­clude con­tact trac­ing and is pri­mar­ily fo­cused on pro­vid­ing in­for­ma­tion. How­ever, it is in­tended to work across lots of cul­tures and might need to make strong com­pro­mises with pri­vacy in or­der to have broad adop­tion and pro­tect its brand (e.g. it would be very bad if peo­ple lost trust in the WHO if its app was poorly re­ceived). It could be benefi­cial to have mul­ti­ple apps with sep­a­rate brands so that peo­ple with differ­ent pri­vacy prefer­ences can use differ­ent apps.

There are some ad­van­tages to de­vel­op­ing mul­ti­ple apps in par­allel. First, many tech pro­jects can fail, be de­layed, or switch fo­cus to new pri­ori­ties. Given the con­se­quences of COVID-19, hav­ing some re­dun­dancy may be a worth­while pre­cau­tion. It could also be good if mul­ti­ple apps try differ­ent ap­proaches and learn les­sons about what works well and what doesn’t. It is always pos­si­ble to stop work­ing on this app and redi­rect users to the offi­cial app if it turns out to be much bet­ter.

Sign up and fol­low-through

Per­haps not enough peo­ple would sign up for this, or once they do sign up they would not con­sis­tently re­port their con­tacts. One thing to note is that even if only a small num­ber of peo­ple sign up, this can still be helpful. Users can send out texts to con­tacts even if those con­tacts don’t have the app. But of course, the more peo­ple that use it, the more of an effect it would have.

Would this cause un­due panic?

A bad ac­tor might list all their con­tacts and then falsely mark them­selves as hav­ing COVID-19. This could then cause many peo­ple to panic and iso­late them­selves or to weaken trust in the warn­ings they re­ceive. Alter­na­tively, even with­out dishon­est re­port­ing, peo­ple may over­re­act to the no­tifi­ca­tions and do more than what is recom­mended. Some of this can be con­trol­led by limit­ing the num­ber of con­tacts and care­ful mon­i­tor­ing, but is still a con­cern.

Can we get this out there?

The Ap­ple and An­droid stores are cur­rently block­ing un­offi­cial COVID-19 apps in or­der to curb mis­in­for­ma­tion about the dis­ease. I pro­posed start­ing with a web app for this rea­son, but it could be very difficult to pub­li­cize it. I think if a web app is show­ing promis­ing re­sults, it might be pos­si­ble to part­ner with an offi­cial or­ga­ni­za­tion to get an app pub­lished, and if not, the web app might be suffi­cient (e.g. run­ning in a browser on a phone).

Open Questions

Is this le­gal with re­gards to pri­vacy/​med­i­cal in­for­ma­tion shar­ing?

If any­one here is a lawyer and has ex­per­tise on HIPAA or other rele­vant reg­u­la­tions, I would love to get ad­vice on this. I’m par­tic­u­larly in­ter­ested in po­ten­tial road­blocks and solu­tions or al­ter­na­tives if you have ideas. I have no le­gal ex­pe­rience and don’t know how much of a blocker shar­ing lo­ca­tion or med­i­cally-re­lated in­for­ma­tion will be even if peo­ple con­sent to it.

How im­por­tant is the lo­ca­tion map­ping as­pect of this?

I think the main con­sid­er­a­tion here is what pro­por­tion of trans­mis­sion is to ran­dom passersby (e.g. cough­ing near some­one on the street or touch­ing the same sur­faces) vs. some­one you know and would be able to record. I am in­ter­ested in see­ing more data on this.

How hard would this be to cre­ate?

I’m a soft­ware en­g­ineer, but I mostly do ma­chine learn­ing and haven’t done much web or phone app work. It seems to me like a rea­son­ably sim­ple app to build, but I might be un­aware of some tech­ni­cal block­ers.

Re­lated In­for­ma­tion

Disclaimer

I apol­o­gize if this post has some er­rors. I did my best to check things, but pri­ori­tized get­ting it out fast for quick feed­back given the ur­gency of the situ­a­tion and how fast things are chang­ing.