Client side symptom classification algorithm

[cankisagun]

This issue builds on Issue #34

Given 40% of Covid19 carriers have only mild symptoms and may not have tested, it's important to identify interactions with these individuals as well.

In order to reduce the data load on SafeTrace API, we propose to run a classification on self reported symptoms (with output high risk symptoms, medium risk symptoms and low risk symptoms) on the client side and pass user status to SafeTrace API

Next step: Determine client side algorithm based on issue #5
Requirements: The algorithm needs to be written in Javascript to run on the client-side

[kistn010]

This issue builds on Issue #34

Given 40% of Covid19 carriers have only mild symptoms and may not have tested, it's important to identify interactions with these individuals as well.

Comment Updated
Data from Wuhan reveal that nearly 80% of infections were spread by patients with mild symptoms or none at all. Citation.

In order to reduce the data load on SafeTrace API, we propose to run a classification on self reported symptoms (with output high risk symptoms, medium risk symptoms and low risk symptoms) on the client side and pass user status to SafeTrace API

Next step: Determine client side algorithm based on issue #5

[cankisagun]

@codeanticode this seems is similar to your research. We are trying to classify user reported symptom (fever >38 / 100, Cough, sore throat, shortness of breath) based on risk (high risk symptoms, low .. etc.)
If you have any feedback we'd be very happy to incorporate into our thinking
Cheers

[kistn010]

@cankisagun Apologies. Other obligations prevented me from posting some content. Hopefully things will slow down this weekend and I'll be able to contribute. To be clear, you're looking for hoping to categorize a user's risk based on his/her signs/symptoms (SS), correct?

[cankisagun]

@kistn010, yes that's the idea.
@codeanticode is also working on a similar algorithm to classify the risk of symptoms.

The goal of this is to report to individuals not only diagnosed patients around them but also users who show high risk symptoms

[kistn010]

Understood. The majority of people who contract the virus experience only mild symptoms (if any at all). Thus, focusing on signs/symptoms may fail to differentiate. I think greater gain would come from stratifying users based on their exposure (to the virus/infected person). This is a primary issue given the amount of asymptomatic transmission that's occurring (sp., people worldwide are transmitting the disease without even knowing they're infected). If individuals know the level of risk associated with a particular exposure, it should be easier for them to quarantine more quickly/responsibly (despite being asymptomatic).

…

On Sat, Apr 4, 2020 at 2:55 PM cankisagun ***@***.***> wrote: @kistn010 <https://github.com/kistn010>, yes that's the idea. @codeanticode <https://github.com/codeanticode> is also working on a similar algorithm to classify the risk of symptoms. The goal of this is to report to individuals not only diagnosed patients around them but also users who show high risk symptoms — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#35 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKPYK7UHWI7OKLGQAAF56ETRK6GC3ANCNFSM4LXBURJQ> .

-- *Brett M Kistner* *Doctor of Pharmacy* *University of Minnesota* *P (309) 360 8672* *E [email protected] <[email protected]>*

[codeanticode]

@cankisagun sorry for the slow reply. I have worked on some diagnosis/prognosis prediction models for Ebola with similar sign & symptom data as the input. Logistic regression models are a common choice for kind of prediction tasks.