wirewirewirewire / pandoa

The app informs you, if you had contact with an infected patient.
GNU General Public License v3.0
65 stars 22 forks source link

Risk Score Calculation #13

Open Hans-rooom opened 4 years ago

Hans-rooom commented 4 years ago

We need to calculate a Riskscore.

For a better understanding I have here a short description of the procedure by means of a current example:

My brother was just here in my office. If he and I would both use Pandoa we would now know via GPS that we probably had contact.

If he has been infected only a few days ago, he would develop first symptoms in 10 days. He would probably not do a test or only very late, because he would hardly develop serious symptoms (age 35 years). So if he reports in 14-20 days, I would not get any information before then. By the time I am tested again, the other contacts will receive information much too late. With exponential growth this is all much too slow. Until then there would be millions of cases.

Our platform should therefore calculate the personal risk. This results from your own current condition/status (symptoms, quarantine, risk areas) and the risk of linked contacts (or the contacts determined via GPS). This risk then flows dynamically into the individual risks of 1st degree, 2nd degree contacts, etc.

So I could see daily how high my pot. risk is. The authorities and researchers would also see this data in the same way. So all decisions and predictions can be made much more accurately and above all MUCH HERE.

I see this as the most elementary and important approach of our application. Speed is the only right solution with exponential distribution!

See: https://github.com/wirewirewirewire/pandoa/issues/1 for more reasons why and on what databasis.

EltarieI commented 4 years ago

Das heißt es soll in echtzeit überprüft werden wer/wieviele leute in der Nähe sind und dem entsprechend das pers. Risiko angepasst werden?

So you want to check in real-time who and how many ppl are near your and adjust your personal risc?

Hans-rooom commented 4 years ago

würde ich nicht in echtzeit machen. Reicht es auszuwerten und ggf. mehrmals täglich zu berechnen. sonst bekommen wir einfach zu viel Last aufs System.

EltarieI commented 4 years ago

Würde es reichen eine einfache Zahl anzugegen, also pro person welche tägl. <10 personen trifft +1 und für <10 +2 (um das risiko der anderen Person mit einzurechnen) und dann nur eine einordnung wo man sich unter allen anderen Personen, welche auch die App verwenden, befindet. also sowas wie "sie sind unter den top10%" um somit dann auch eine gewisse einordnung zu bieten

Hans-rooom commented 4 years ago

wäre auf jeden Fall ein Ansatz und für den Prototypen wahrscheinlich die einfachste Lösung.

IDEAL wäre auf Basis der täglichen Selbsteinschätzung aber etwas in der ART:

Gesundheitszustand (ggf. mit smilie) : gut (+0%), mittel (+5%), schlecht(+15%)

Quarantänestatus:

keine Quarantäne wegen aufenthalt in Risikogebiet (welches, wann zurück gekehrt, ggf. wie lange dort gewesen) (+10%) wegen corona fall im umkreis (welcher Kontakt - ID angeben wenn möglich, welche beziehung zum kontakt, wie häufig kontakt zum Kontakt gehabt: kaum, mittel, häufig, intensiv) (+20%)) wegen corona verdacht (+40%) corona festgestellt (=100%) im Krankenhaus (100%) intensiv im Krankenhaus (100%) Infektionsstatus: Unbekannt, Positiv, Negativ (SIEHE Angaben zu TEST)

Test durchgeführt (DATUM) Positiv getestet (DATUM) Negativ getestet (DATUM)

Sonstige Symptome ggf. über Icons (sollten in die gewichtung mit einfließen):

halsschmerzen (+10%) fieber (+15%) schnupfen (-20%) bauchschmerzen (0%) husten (TROCKEN) (+20%) husten (Produktiv) (-5%) kurzatmig/atemnot (+30%) Herzprobleme (+5%) usw.

br1sc commented 4 years ago

imho my potential risk aka "pandemic footprint index" should be calculated locally on local data. if really infected, my last contacts will be alarmed via server, in other cases, only intensional data is shared if i want the person i meet to trust my infection related state. "show your health state" to a person you want or need to meet physically.

E3V3A commented 4 years ago

Hi Guys! This sound, good but English please. Github doesn't have a translation button yet... :( )

I also think a risk score is a great idea.