COVID-19 Risk Assessment App Idea for Vetting and Discussion

Please note: Th­ese opinions are my own. I’m not an ex­pert, but I have a Q&A be­low with Sukrit Silas, an ex­pert on in­fec­tious dis­ease who sup­ports these ideas and thinks they merit fur­ther dis­cus­sion.

Below I dis­cuss some worst-case sce­nar­ios for COVID-19, and my pre­limi­nary ideas for an app that could help slow its spread. Like any­one, I don’t know how bad things will get; we just have to wait and see. But ob­vi­ously, if you want to be ready for a deadly pan­demic, how­ever un­likely, you have to start well in ad­vance. This is par­tic­u­larly true for de­vel­op­ing an app, which re­quires time and re­sources.

I’m hop­ing for some EA feed­back and com­ments.


Introduction

COVID-19 has a rel­a­tively long in­fec­tious in­cu­ba­tion pe­riod, roughly 5 and up to 14 days, dur­ing which there may be asymp­tomatic trans­mis­sion. This al­lows it to spread more quickly than many other viruses do. It also makes con­tact trac­ing — to keep in­fec­tions un­der con­trol — harder, slower, and less scal­able.

How­ever, it is now easy to get GPS data for in­di­vi­d­ual smart­phone users. In fact, map­ping apps like Google Maps and MapQuest already col­lect this data, by de­fault, for their users.

This GPS data could perform a rapid, au­to­matic, and prob­a­bil­is­tic form of con­tact trac­ing. It could also alert users who might be ex­posed to the es­ti­mated risks as­so­ci­ated with a given area.

The app could also provide in­di­vi­d­u­al­ized recom­men­da­tions, based on risk calcu­la­tions from their shared GPS data. Th­ese recom­men­da­tions would re­duce un­cer­tainty and could help peo­ple bet­ter un­der­stand their own risks, make more in­formed de­ci­sions to avoid per­sonal ex­po­sure, and slow the spread of the virus. Ul­ti­mately, this could save lives.

We could also provide users with a heat map, show­ing the risk lev­els of sur­round­ing ar­eas. This map could be more fine-grained than Johns Hop­kins, while still coarse-grained enough to pro­tect users’ pri­vacy. Cru­cially, users would not be able to iden­tify other users, based on the heat map in­for­ma­tion.

I want to have an app like this, to use my­self. And I want other peo­ple to have it, and use it. My goal with this post is to in­crease the chance that some­one can, and will, build such an app. Do­ing so could have a mea­surable im­pact on slow­ing the spread of COVID-19.


Pri­vacy Focused

China has de­vel­oped its own coro­n­avirus app, but it wor­ried me. I’m a liber­tar­ian, and I started think­ing about whether an al­ter­na­tive pri­vacy-re­spect­ing ver­sion could work. I un­der­stand China’s do­ing ev­ery­thing they can to fight the virus, but the level of pri­vacy vi­o­la­tion is huge. So, to be clear, that is not what I’m propos­ing.

I be­lieve that level of in­va­sive­ness is also com­pletely un­nec­es­sary for the app’s effec­tive­ness. I think the high­est im­pact in­for­ma­tion that in­di­vi­d­u­als could give is a few an­swered ques­tions and their GPS data — which most peo­ple share with many pri­vate com­pa­nies already. For the av­er­age per­son, the ad­di­tional pri­vacy bur­den could be min­i­mal to non-ex­is­tent, if the app is built by some­one who cares about pri­vacy.

The app could be built by a not-for-profit group col­lect­ing only the (limited) in­for­ma­tion that would have the high­est im­pact on the spread of the virus. The­o­ret­i­cally, the app’s database could still be hacked and abused by gov­ern­ments. So the app would need good se­cu­rity, and per­haps even to be blocked in coun­tries where cit­i­zens’ rights are most likely to be com­pro­mised.

This could be bal­anced with con­sid­er­a­tions re­gard­ing how likely it is to pre­vent even worse fu­ture in­cur­sions on your pri­vacy that could fol­low a so­cietal break­down, city or com­mu­nity quaran­tines, a to­tal­i­tar­ian gov­ern­ment app, forced com­pli­ance mea­sures, mil­i­tary in­volve­ment, or other things no one wants.

Of course, there’s still a risk that this app wouldn’t be ac­cepted be­cause of in­cor­rect pat­tern match­ing and as­so­ci­a­tion with China’s app. Quick adop­tion will be a hard challenge, even with­out ad­di­tional PR challenges. But the ac­tual pri­vacy harms seem like they can be sig­nifi­cantly miti­gated.


The In­cen­tives Align

The most promis­ing thing I want to com­mu­ni­cate about this app is that the in­cen­tives to use it al­ign much bet­ter than one might as­sume.

It’s in your in­ter­est to in­stall the app, if only to see the heat map and mon­i­tor the risk lev­els in your own area. It’s in your in­ter­est to turn on your GPS and keep it on (or con­nect it to the info Google already has for most peo­ple’s phones), be­cause the app couldn’t calcu­late your risk score with­out that. And it’s in your in­ter­est to re­port that you have cold symp­toms — whether you think it’s COVID-19 or not — be­cause then you can see how it changes your risk pro­file, and bet­ter learn what pre­cau­tions to take.

We should also find ways to in­cen­tivize those who already have a con­firmed case of COVID-19 to put that into the app. Also, pre­vent­ing hoaxes (at least early on) is im­por­tant. With any luck, if some­one mak­ing this app can make the right con­nec­tions, they can ask the CDC to give peo­ple with con­firmed cases (or their close con­tact or fam­ily mem­ber) a code for their phone and strongly sug­gest they put in the code to con­firm, and to help ev­ery­one else out. Most peo­ple listen when the gov­ern­ment sug­gests some­thing even if it’s not a re­quire­ment. Th­ese are peo­ple who would also already be sub­ject to CDC con­tact trac­ing, so shar­ing their con­firmed sta­tus /​ GPS is less in­va­sive than what you’re already sub­ject to at that point.

Also, if things got re­ally bad and there were a lot of cases, we could then re­lax the con­straints for need­ing a code be­cause at some point we’d have enough data that hoaxes would have lit­tle effect.


Rea­son­ing About Scenarios

(I’m imag­in­ing one of the worst case sce­nar­ios here, and I’m not mak­ing claims about what’s likely to hap­pen now.)

In some re­ports, many of the lethal and non-lethal cases of COVID-19 seem like they ended up need­ing ex­tended in­ten­sive care that only a hos­pi­tal could provide well—for ex­am­ple, 1-2 weeks of sup­ple­men­tal oxy­gen. If 1323 of the world even­tu­ally get COVID-19, the case fatal­ity ra­tio is 2%, and the trans­mis­sion rates are not re­duced enough be­cause of the roughly 5 and up to 14 day in­fec­tious in­cu­ba­tion pe­riod, many hos­pi­tals would be over­whelmed and the fatal­ities will be con­sid­er­ably higher than they might have been if the spread of the virus had been slower.

In this sce­nario, by par­ti­ci­pat­ing in the app, you’ll get in­for­ma­tion quicker about pos­si­ble ex­po­sure. And, if you do get the virus, you’re more likely to get treated effec­tively at the hos­pi­tal if there are fewer other peo­ple us­ing the same hos­pi­tal re­sources at the same time. In this sce­nario, what helps ev­ery­one also helps you and we all get a slower mov­ing pan­demic where hos­pi­tals don’t get as over­whelmed.

There is also a pos­si­bil­ity in early sce­nar­ios where the virus hasn’t taken hold in a com­mu­nity, if enough peo­ple par­ti­ci­pate in the app, in ad­di­tion to the mea­sures already be­ing im­ple­mented, that the app could not just slow, but help to stop the virus. There lots of caveats here about stop­ping the virus though. It’s the­o­ret­i­cally pos­si­ble in some sce­nar­ios on the mar­gins, but we’re un­likely to be in ex­actly those sce­nar­ios.

This could also help slow down the spread of the virus any­where peo­ple have smart­phones. Even if the virus is con­tained in the United States, in other coun­tries it may not be con­tained.

And, even if the sever­ity of the situ­a­tion in­creases, pro­vid­ing good guidance based on clear in­for­ma­tion also helps to pre­vent panic. Panic makes things worse for ev­ery­one and it takes lives too. The safest world is the one that slows or stops the spread of the virus while con­tin­u­ing to provide nor­mal ser­vices as much as pos­si­ble. There­fore, to save lives, it is im­por­tant both (1) for peo­ple who are at lower lev­els of risk to con­tinue to par­ti­ci­pate and provide nor­mal ser­vices and (2) for peo­ple who learn they are at higher lev­els of risk to stay home and not con­tribute to the spread.

For­tu­nately, we could all help fight the virus even in these sce­nar­ios, sav­ing lives while also ra­tio­nally pro­tect­ing our­selves, just us­ing our cur­rent tech­nol­ogy.


If COVID-19 Stays with Us

Take a similar sce­nario where the virus also comes around ev­ery year like the com­mon cold, but con­tinues to have a rel­a­tively high fatal­ity rate (how­ever un­likely). Even if we have im­perfect adop­tion of the app, the first year COVID-19 comes around, a lot of peo­ple who didn’t use the app or oth­er­wise pre­pare would be like­lier to get in­fected. The sec­ond year it comes around, the out­break is con­sid­er­ably less be­cause now many of those peo­ple might still have im­mu­nity. And the rest say, “Ok, here we go again” and turn on the app. And the sec­ond time around, it’s more likely to be shorter, or stopped.

Per­haps only a tiny frac­tion of peo­ple use it. In that case, you could think of it more like early Google traf­fic up­dates. Still a lit­tle use­ful. And if it works well, and guidelines tell you to stay home given a level of risk, per­haps it’s like a bad snow day, ex­cept in more lo­cal­ized ar­eas, and more like a snow week-or-two.

Also, for those who start prep­ping or con­sid­er­ing self-quaran­tine for some pe­riod of time, the ques­tion might be asked—what’s your trig­ger for go­ing out into the world again? See­ing the heat map set­tle down nearby you or see­ing your score go down would be very helpful in­for­ma­tion for mak­ing this de­ci­sion.


Ex­am­ple App Questionnaire

* Do you have con­firmed COVID-19? If so, in­clude code. Per­haps op­tions for lab /​ CT di­ag­no­sis.

* Have you been to any of these coun­tries in the past two weeks /​ month?

* Have you been in­formed that you might have been ex­posed? If so, when?

* Have you been told by any au­thor­i­ties to self-iso­late or that you should have self-iso­lated for a given pe­riod? If so, when and for how long? (Maybe to what ex­tent would you say you were able to com­ply dur­ing that pe­riod? 0-100%? Not nec­es­sar­ily your fault, per­haps you were in­formed late /​ or had some other emer­gency.)

* Do you have any up­com­ing travel plans? Are you trav­el­ing through any of the fol­low­ing air­ports?

* Have you had cold or flu symp­toms re­cently? (Even if you’re quite sure it’s not COVID-19) If so, when?

* Are you cur­rently wait­ing on test re­sults for the virus?

You could eas­ily go back to the ques­tion­naire to up­date your an­swers at any time as they change or you need to make a cor­rec­tion. You could get pinged once in a while and asked for up­dates. You could also get pinged if your level of risk in­creases.

I think ask­ing peo­ple on the ques­tion­naire whether they just have cold or flu symp­toms is pos­si­bly re­ally use­ful in­for­ma­tion. For ex­am­ple, right now, if you come down with a cold for the fore­see­able fu­ture in the U.S., it’s very un­likely it’s COVID-19. Lots of the usual stuff is go­ing around. How­ever, let’s say you are liv­ing in an area where the heat map is light­ing up. This gives the app very use­ful in­for­ma­tion about whether you might be ex­posed, not nec­es­sar­ily for di­rect health ad­vice, but for es­ti­mat­ing risk for your­self and oth­ers, con­sid­er­ing where you may have gone out to (even only briefly) re­cently.


Q&A With Sukrit Silas

In 2019, Sukrit Silas was listed in “Forbes 30 Un­der 30: Health­care.” Work­ing in the lab of No­bel lau­re­ate An­drew Fire, Silas dis­cov­ered how bac­te­ria use CRISPR, which is part of their im­mune sys­tem, to learn to at­tack viruses that are made of the ge­netic ma­te­rial RNA. His com­pany, BillionToOne, raised $15M in 2019 for de­vel­op­ing di­ag­nos­tics for beta tha­lassemia and Down syn­drome. He’s cur­rently a post­doc­toral fel­low in the the Lab­o­ra­tory of Prof. Jonathan S. Weiss­man (UCSF).

Q: Do you think this app would be use­ful for slow­ing the spread of a virus?

A: It doesn’t make as much sense to do for some­thing gen­eral, like the flu, be­cause you’re not asymp­tomatic for long be­fore you’re in­fec­tious. But if there is a real threat where the virus ei­ther stays ac­tive on dry sur­faces for a long time, or like with this coro­n­avirus, where there is a long in­fec­tious in­cu­ba­tion pe­riod, this would be very use­ful.

Q: Might an app like this be broadly ap­pli­ca­ble for pub­lic health in the long term?

A: Might be a good ex­cuse to build the app, for the next thing… I would definitely build it to be gen­er­al­iz­able though. The threat might not be viral. Pan-An­tibiotic re­sis­tant bac­te­rial in­fec­tions are prob­a­bly go­ing to be a much nas­tier fight. And some­thing like this could re­ally help there.

Q: Do you think show­ing peo­ple a heat map would be use­ful? Some kind of ge­o­graph­i­cal risk es­ti­mate?

A: I think an in­te­gra­tive ge­o­graph­i­cal risk es­ti­mate could be use­ful, and has the ad­van­tage that if you have lots of data (even *bad* data), your pre­dic­tions will still be use­ful.

Q: Where do you see this be­ing most helpful? And what risks do you fore­see?

A: I’m quite in­ter­ested in the po­ten­tial for this to be use­ful for in­fec­tions that ac­tu­ally per­dure, like an­tibiotic re­sis­tant in­fec­tions. I’m gen­er­ally less wor­ried about pri­vacy/​rights is­sues as I am about the pos­si­bil­ity of start­ing a panic/​hurt­ing peo­ple’s liveli­hoods.

Q: Any­thing else you would like peo­ple to know?

A: I think it makes sense to post to EA and gauge in­ter­est/​ex­plore ideas. I just have a few ideas:

(1) I agree with what you say com­pletely: The suc­cess will de­pend on the “fidelity” of the self-re­port­ing more than any­thing else. Espe­cially in places where peo­ple might be afraid of their gov­ern­ments, I’d think a way to re­as­sure peo­ple that their re­sponses will not jeop­ar­dise their safety would be im­por­tant. For ex­am­ple, I trust the LastPass folks, but not Face­book.

(2) Per­haps a way that it would be im­pos­si­ble to cen­trally trace which users are in­fected.

(3) I do think that some­thing like this could be helpful in the long term, in a more gen­eral fight against in­fec­tious dis­ease.

(4) It could be use­ful to get some­one at CDC or some­thing to look at an early pro­to­type and get a sense of tweaks that might be helpful given the ground re­al­ity. I think they’d be in­ter­ested to chat, if a work­ing model were available.


En­vi­sion­ing A Prototype

This is what I imag­ine an im­ple­men­ta­tion as a stan­dalone app might look like.

Name of app

COVID-19 Risk App

What is the app for?

Guidance and GPS based per­sonal risk as­sess­ment for COVID-19

About the app

Peo­ple are get­ting anx­ious about coro­n­avirus. Peo­ple want to know how to un­der­stand their per­sonal risk, how to pro­tect them­selves and pre­pare given their level of risk, and how to help.

This is a COVID-19 risk as­sess­ment app that would calcu­late a risk level for you per­son­ally. To calcu­late your score, in­stall the app and an­swer a ques­tion­naire (anony­mous) about your pos­si­ble ex­po­sure to the virus (cur­rently have it, close con­tacts, when they had it, re­cent travel, etc.) and al­low ac­cess to your GPS lo­ca­tion and your GPS lo­ca­tion his­tory (if available) from Google/​Ap­ple/​Other map­ping apps you already have on your phone from the past 1 month.

The app will show you a heat map with risk lev­els for differ­ent ar­eas nearby. Also, once we have a few days of GPS data, you will get a risk score. You will see the risk score im­me­di­ately if you already have an app with suffi­cient map his­tory on your phone. The app will give you ad­vice about what to do based on your cur­rent level of risk. Your risk level up­dates over time.

When a user up­dates the app that they have the virus (con­firmed with a code from the CDC) and we have the past ~2 weeks of their GPS data, we can quickly and eas­ily get up­dated scores and recom­men­da­tions to self-iso­late or stay home to other users who have shared their GPS data and have some like­li­hood of hav­ing been near an in­fected per­son dur­ing their asymp­tomatic in­fec­tious pe­riod.

The effec­tive­ness of the app for as­sess­ing your risk de­pends on (1) whether you fill in and up­date the ques­tion­naire as ap­pro­pri­ate and (2) how of­ten and for how long you’ve kept your smart­phone near you with your GPS on. Your ques­tion­naire and GPS lo­ca­tion shar­ing are the foun­da­tion of the app, al­low­ing for an im­proved eval­u­a­tion of ex­po­sure and risk with as min­i­mal im­pact on your pri­vacy as pos­si­ble to give you an es­ti­mate of your risk.

The apps also asks you to agree to con­firm that you have the virus on your phone (or ask a con­tact to do so) if you be­come ill and are con­firmed to have the virus.

Also, this app is for per­sonal recom­men­da­tions for peo­ple who are not oth­er­wise re­ceiv­ing recom­men­da­tions. If lo­cal au­thor­i­ties make other recom­men­da­tions, they take prece­dence.


Why Should I Care?

I think it meets GiveWell’s “Im­por­tance, Tractabil­ity & Ne­glect­ed­ness.”

* Great in scale (it af­fects many peo­ple’s lives, by a great amount)

* Highly solv­able (ad­di­tional re­sources will do a great deal to ad­dress it), and

* Highly ne­glected (few other peo­ple are work­ing on ad­dress­ing the prob­lem).

I ac­knowl­edge that I have limited in­sight about how ne­glected this is, or how many peo­ple are work­ing on this. It’s pos­si­ble that the CDC has a big pro­ject on this and we don’t know. It’s also pos­si­ble they’re re­ally busy right now and new ideas aren’t the pri­or­ity.


What Can I Do Now?

Right now, I’m hop­ing for some EA feed­back and com­ments.

I also think a small, well-funded, ag­ile, and mo­ti­vated group could do this best, who co­or­di­nates with mul­ti­ple large orgs like Google Maps, Johns Hop­kins, CDC, WHO, etc. But I could be con­vinced an ex­ist­ing org with the right set of peo­ple could do this too. In ei­ther case, it could be use­ful to get some­one work­ing on a demo right now be­cause peo­ple (per­haps at the CDC or Google Maps) are more likely to talk to you, and be con­vinced if they can see a work­ing ex­am­ple.

I imag­ine you also wouldn’t want to bother the CDC much when they’ve got im­por­tant things go­ing on right now. Do­ing this as a sep­a­rate group while do­ing some min­i­mal amount of ping­ing one or two peo­ple at the CDC to check would not be too dis­rup­tive. In case the app ends up not work­ing well, at least we didn’t waste any­one’s time much. Ideally, you get peo­ple work­ing with you who are not oth­er­wise work­ing on the prob­lem.

It seems pos­si­ble that Google is already do­ing some­thing like this or think­ing about it, but if not, an EA at Google could ad­vo­cate for this to hap­pen. Google Maps would be very well po­si­tioned to help im­ple­ment a big part of this.

My hope is also that this reaches some­one with con­nec­tions who can do this and form a team. I’m not an ex­pert, but I wrote up a short plan that might be worth skim­ming over quickly—in­clud­ing more pos­si­ble use­ful ques­tions, ex­am­ple recom­men­da­tions, a 1-7 risk scale, and an overview of the peo­ple you would need to form an ini­tial team for the prob­lem. If some­one is in­ter­ested, con­tact me if you think you can con­tribute.


Should I Build My Own Rogue App?

No. My sense from talk­ing to ex­perts is that it would be re­ally easy even for well-in­ten­tioned peo­ple to cause harm. Peo­ple are in­clined to panic in a pan­demic, and in­cor­rect or poorly com­mu­ni­cated in­for­ma­tion could cause peo­ple to act in ways that make things worse. Bad scores, maps, or ad­vice that make peo­ple panic or act too soon (or too late!) would cause more harm than good. You also have to worry about hoaxes.

If you’re con­sid­er­ing build­ing a tool like this, col­lab­o­rat­ing with the health se­cu­rity com­mu­nity is es­sen­tial. You should prob­a­bly be work­ing with some­one who has rele­vant ex­per­tise and con­nec­tions in this space. Ul­ti­mately, I think the only ver­sion that re­ally works with­out likely caus­ing much more harm than good is one that has some sup­port from au­thor­i­ties like the CDC.


Author

Tina White

I’m a PhD Can­di­date at Stan­ford do­ing ap­plied ma­chine learn­ing re­search for aero­dy­namic de­sign. I’ve also done a cou­ple per­sonal web/​mo­bile pro­jects, so I have some sense of what the app might in­volve, and if there was time, I could make a demo. I started go­ing to Stan­ford EA Mee­tups about a year ago be­cause they had con­sis­tent and high qual­ity AI Safety dis­cus­sions.

Web­site: www.tinar­white.com

Email: cr­white@stan­ford.edu


Acknowledgements

I’m grate­ful to Chris Olah and Joshua Zader for writ­ing feed­back, and I ap­pre­ci­ate the com­ments and crit­i­cal feed­back I got from Howie Lem­pel and the Effec­tive Altru­ism Coron­avirus Dis­cus­sion Group.