Covid-19 screening tool seeking community contributions | Hacker News Reader

Covid-19 screening tool seeking community contributions

46 pointstconley24 comments

Please contribute to a simple mobile friendly COVID-19 screening tool!

github repo with contribution instructions: https://github.com/tconley/screencovid.com

live app in development: https://tconley.github.io/screencovid.com/

This project is based off a phone call I had with Sajung Yun, PhD from Johns Hopkins. He is very much interested in saving lives, and has analyzed the data available to come up with a simple formula for a covid screener. He created a screening tool, but it wasn't a mobile friendly UX. You can see what he's put together at https://www.screencovid.org/. The goal is to get this new tool production ready by Mon 3/23 at which time it'll be available at screencovid.com

You can view Dr Yun's findings and rationale behind the screening at https://github.com/tconley/screencovid.com/raw/master/Presen...

View on HN

You are making presumption that every user is familiar with swiping card UI from Tinder. Also, after some poking around and finding how it actually works, there's no way to reset or change your choice - only hard reload.

I would add simple instructions how to use this tool, e.g. "Swipe left for No, left for YES answer". Also do calculation for user, do not show just numbers. Simple if statement at the end of questionnaire should state "Based on {the algorithm} you should consult a doctor immediately". You must emphasize that there is a margin of error.

Thanks butz, have some community folks working on these enhancements now!

No offense, but the site at screencovid.org seems totally mobile friendly to me, and has a clear ui - check off the symptoms you've got is a known pattern that everybody is familiar with

The GitHub site is a confusing UI (swiping cards away does what?) and it spills off the side of my screen on mobile - I'm not sure why development is needed here when the original site is totally functional

The idea is to improve the UI so it's best of both worlds. we'll have both versions available, thx for the feedback!

Why? seeking fortune or fame...

lol. researcher asked for help and I had enough time sheltering in place playing video games, so helped them out.

Preface IANAD. I suggest that sniffles (rhinorrhea) as well as duration of flu like symptoms (elevated temp, cough) as differentials for your diagnosis tool. Covid-19 is marked by less rhinorrhea and longer duration illness than flu i.e. time before it is severe enough to require hospitalization. You can see the rhinorrhea in the aggregate data in Table 1 https://www.nejm.org/doi/full/10.1056/NEJMoa2002032 You can see the rather long illness period in Figure 2D https://www.nejm.org/doi/full/10.1056/NEJMoa2001316 Flu is typically 4days (3-7 according to CDC https://www.cdc.gov/flu/professionals/acip/clinical.htm) whereas this illness just last and lasts (both anecdotally from the NEJM paper) Apologies if I haven't captured better refs...

One other rather interesting comparison is: https://drive.google.com/file/d/1DqfSnlaW6N3GBc5YKyBOCGPfdqO... page 26. This may help also develop other differentials that your tool can use. Good luck.

Hi, check out https://www.clearstep.health/covid19 for an existing screener that we are deploying with health systems next week!

"Sorry to hear you're concerned about coronavirus (COVID-19)."

Sorry?? Why are you sorry to hear that I'm interested in using your tool?

They mean "Sorry that your circumstances and symptoms are such that you're worried that you may have COVID-19"

Yes, this is what we meant. It does sound a bit weird though, we'll consider changing the wording. Thanks!

I couldn't find a pithy version that was much better. You may want to drop that part altogether--it comes off as a bit of a platitude since you wouldn't be on the site otherwise.

Hi,

I wanted to make a population health tool that would anonymously collect heart rate data from the Apple Watch together with an approximate location. Increase in resting heart rate during sleep is a good sign of infection.

https://github.com/unrelatedlabs/infection-alert

Could not get it publish, because Apple insists on IRB approval.

Any researched want's to get this through IRB? App is done and open source, backend is also built, the data is supposed to be public.

Is there any more documentation on the scoring system?

As I understand it, the weights are just the percentage of COVID+ patients that had each symptom. Fever gets 90.5 points because between 83 and 98% of pts had a fever and (83+98)/2 = 90.5.

The thresholds are back-calculated from recommendations: A 60 year old with a fever and dry cough should see a doctor. Using the point system above, that's 212.9 points, so...that's where the line is drawn.

This feels very adhoc to me, even though the (fractional!) point system makes it seem rigorous.

Furthermore, each of the (binary!) checkboxes seem to linearly add weight to the total value, which seems overly simplified.

Would be great for countries who have givven up on testing the non verry ill ( like netherlands ), to get some idea of how it spreads, (especial if coupled with IP lookup to geo location)

In the symptoms list on the Screening page I don't see "Loss of taste or smell".

Isn't that something that a significant portion of patients have reported?

I have asked the medical folks to look into this.

The ".com" tld, along with no licensing information on github does't make me feel very good about the request for contribution.

Thx for feedback, added MIT license! We have both the .com and .org domains, once we're ready for production they'll show the same material!

I can submit some UX enhancement proposals. Is there a way to do that which makes sense? Github issue with a link?

Yes, pull requests or issues on github are appreciated!

We said that about 2008 crisis as well....ahahaha

Would you please stop posting unsubstantive comments? We ban accounts that do that and you've unfortunately been doing a lot of it.

https://news.ycombinator.com/newsguidelines.html

Reply on news.ycombinator.com