Skip to content

Probability that one individual is infected

Luis Aleixo requested to merge feature/p_one_infected into master

Implementation of the probability of one individual is infected in a given population (@mandrein) - Probability_to_be_infected.pptx.

On the Event data there is a new section for Deterministic exposure or Probabilistic exposure (radio buttons). If the later is checked, three new UI input fields are displayed - Population, New confirmed cases (weekly) and Confidence level needed for the calculation of the new probability. If the

Confidence level (in the infected population monitoring, or "Ascertainment Bias"):

  • High (mandatory population-wide surveillance) - AB=2
  • Medium (recommended population-wide surveillance) - AB=5
  • Low (surveillance only for symptomatic patients) - AB=10

This probability uses the previous CARA probability multiplied by a Binomial distribution that follows the probability of having x infected persons in an event with n people. n is the Total number of occupants, while x is a value that goes from 1 until the number of infected people, if lower than 10:

  • Calculate the P(I) when the number of infection persons is 1 and multiply by the distribution output when x is 1.
  • Calculate the P(I) when the number of infection persons is 2 and multiply by the distribution output when x is 2.
  • ...

The final probability is a sum of all these calculations.

Performance tests:

Before:

[I 220602 09:24:52 web:2239] 200 POST /calculator/report (::1) 6511.08ms
[I 220602 09:24:59 web:2239] 200 POST /calculator/report (::1) 4665.08ms
After:

[I 220602 09:21:35 web:2239] 200 POST /calculator/report (::1) 10708.70ms
[I 220602 09:21:45 web:2239] 200 POST /calculator/report (::1) 9095.44ms

Assuming this scenario, with a total population of 270000 inhabitants, 100 cumulative cases in the past 7 days, and medium confidence level, the probability of on-site transmission, having at least 1 new infection in an event with 5 occupants, is 0.04%.

Screenshot_2022-10-03_at_17.19.42

If there is any short-range interaction, the card displays (in an unordered list) two P(I), one that considers the short-range interactions P(I), and other that doesn't:

Screenshot_2022-10-03_at_17.18.17

@mandrein In this implementation we ask for the 7-day average of new reported cases instead of the 10-days.

New section in calculator:

Screenshot_2022-10-03_at_16.55.58

Note: New activity type was added to this MR through this (merged) MR.

The tests that were added are based on calculations performed with an excel file : test_pi_one_infected.xlsx

  • The P(I) values were extracted from CAiMIRA from three different scenarios - lambda t: 36.; lambda t: 72.; lambda t: 1.2.
  • For each of the previous scenarios, we calculated the respective results for a population of 11, 21 and 31 people, where initially there was only 1 infected person.
Edited by Luis Aleixo

Merge request reports