Impending Outbreaks


Table of contents

  1. Anticipating Outbreaks
  2. Graphs

Anticipating Outbreaks

This exploration is focused on "anticipating outbreaks via running percentage changes" of "cumulative observations per $\mathbf{\small{100K}}$ people" w.r.t.with respect to a set of time periods. Why such an approach? Foremost, a mathematical outline which will make it easier to understand the discussion herein.

Let

$\mathcal{C}_{\tau} = \small{100K} \times \Large{\frac{\theta_{\tau}}{\mathcal{P}} }$

wherein

Description
$\tau$Date
$\theta$Observations; $\theta$ is either deaths, positives, tests, or hospitalizations
$\theta_{\tau}$The number of observations by date $\tau$
$\mathcal{P}$The population of the area in question


Then a percentage change on date $\:\tau\:$ w.r.t. $\:\Delta\:$ days ago is

$\text{pc}_{\tau, \Delta} = 100 * \Large{ \frac{C_{\tau} \; - \; C_{\tau - \Delta}}{C_{\tau - \Delta}} }$

noting that

Description
$\Delta$Days
$C_{\tau - \Delta}$The cumulative [observations per 100K] value on date $\tau - \Delta$


Now, consider the graph of curves below. Each point on the 14 days curve, for example, is the percentage change relative to 14 days ago; the same logic applies to other "days ago", each having its own curve.


stick
Running percentage changes


It is the accumulative behaviour of these curves that is of interest due to the following observations/hypotheses

Positives
  • A divergence point, i.e., an x-axis point from whence curves start diverging, is probably a warning of an impending outbreak. And the probability of the outbreak occurring increases if there weren't any mitigating measures [coincidentally] in place around the same time.
  • Converging curves that are approaching zero suggests an improving state of affairs.
Hospitalizations
  • A divergence point is probably a warning that there might be a substantial increase in required hospitalizations. Such a point should set off a contingency planning alert.


Graphs

More will be written about percentage change curves in future. Meanwhile, the links below lead to a few graphs. Note, the most consistently available data sets are the tests, positives, and deaths data sets.