So let's look at an example of COVID-19
epidemic curve in the early stages.
And this is taken from a Canadian
province just as an example.
It shows that it's climbing to a
peak and starts to diminish again,
as you would expect from a
It's going to start again
for a second wave after that.
We can look to this curve to
learn about the ways in which
changes in public health
are manifesting in the data
and how the effects of interventions
might be noticeable in the Epi curve.
For example, sometime in early
March, the lab testing criteria changed
we're identifying and defining what is a case.
And so in this case it lowered
the threshold for identifying a case.
As a result, the numbers increased a little bit.
If you didn't know that happened you'd
think oh what happened on that one day?
Why was that one day an outlier
that so many people got sick?
And it wasn't a case of a lot more
people getting sick, it was a case of
defining what a case is a bit differently.
The next day, we changed
the policies in the community.
Gatherings were limited to
50 or fewer, and that caused
a diminishing of transmission.
and so while the cases were still
increasing, they're increasing a bit more slowly
several days later.
So again we can use epidemic curves to
see how various interventions in a community
might be leading to
changes in transmission rates.
So epidemic curves are lagging indicators.
They're a lagging indicator
of an epidemic because
you only know about an illness
when you test for an illness.
You typically only test for an
illness when there's a reason to test,
if someone has symptoms and
when they present to be tested.
So it's possible for someone to become
infected, not show symptoms for several days
and then will delay even longer until the
symptoms are bad enough to go seek out care
and then get tested and then we wait
even longer for the test results to come back.
And maybe a bit longer for those results to be
coded by an epidemiologist into the Epi curve.
So it's always late.
Curve is always a lagging
indicator because of all these delays.
Keep in mind too that for some diseases,
there's always going to be a background
number of cases that are always there.
They're always percolating in the community
but never enough to really be a crisis of any kind.
An outbreak is when we have more
cases starting to mount in a short time period.
So if background cases are always there,
it's difficult to say exactly
which case is the index case,
which case is the one that started
Of course, that's not true for all
kinds of diseases but for some.
For some cases, the date
of illness onset is not known
because that takes a while for public
health investigators to actually figure that out.
As a result, some of these
data points will be wrong.
And it can be difficult to determine
when a curve starts to go down
because of the reporting delay.
We don't know when the
epidemic has really plateaued.
More cases start mounting,
test results come back
So it's not uncommon to go back and revise
the numbers we have from a few days ago
This is very common in COVID-19 for example,
so it's dangerous look at the daily case counts
and assume that's the
way it's always going to be.
It's safer to go back a couple
days in the past and see,
"ah that's what the real number is
because we have more test results back now".
As a result we never really know
when the curve has plateaued
until after the plateau has been observed.
You never know the full shape of
the curve until the outbreak is over.
Epi curves are powerful tools
but they are not complete tools
and it's really important to remember
that they are incomplete data
and only are suggestive of what's going on.
They are not complete.
Here's an example, erratical example
of a COVID-19 epidemic curve showing
infections and deaths.
Now a typical Epi curve looks just
but nothing prevents you
from plotting deaths if you want.
Deaths are a lagging indicator to infection
because it takes time for someone to become ill
then gets sick enough to be
hospitalized and then unfortunately to die.
So in the case of COVID-19 is about 21
and 28 days between infection and dying and
maybe some days between that
death being no test encoded in the data.
So deaths are a lagging indicator.
So it's not unusual then to
see the Epi curve coming down
as it is in this case, the curve is
coming down fairly early in this epidemic
but the deaths are still mounting.
Mounting deaths in this case should not be
taken as a case of the outbreak is out of control,
In fact, in this case the outbreak is under
control but because deaths are lagging,
it gives a false impression of what's going on.
So epidemics are difficult to keep track of
for many reasons and one of them is this,
nature of time delays and lagging indicators.