Video summary
The video discusses the importance of validating health data by comparing indicators that contain multiple denominators, such as the crude birth rate and infant mortality rate, against other reliable sources. Since these metrics are calculated using components like live births and total population, which are also used in separate equations, cross-referencing them with external datasets provides a crucial check on data accuracy. For instance, comparing domestic routine data from systems like DHIS2 with publicly available census figures or United Nations estimates allows analysts to assess the reliability of their own inputs. When these different sources yield values that are closely aligned, it significantly boosts confidence in the underlying data, whereas large discrepancies might suggest issues with specific components or warrant a closer investigation into potential errors.
Beyond static comparisons, the presenter emphasizes the value of analyzing trends over time by aligning local data with global estimates to ensure logical consistency. A prime example provided is Laos, where significant improvements in health services should logically result in a steady decline in crude birth rates over an extended period. If local data shows sudden increases or erratic fluctuations that contradict these broader global patterns and the known trajectory of service delivery, it may indicate data quality issues unless there are specific contextual factors like war or migration affecting the population. This approach helps distinguish between genuine demographic shifts caused by external factors and anomalies in the reporting system, ensuring that health systems accurately reflect the reality of service delivery.
The discussion also extends to calculating annual growth rates specifically from a health perspective, focusing on changes in live births rather than total population dynamics influenced by immigration or emigration. The video explains how to compute this rate by taking the difference between consecutive years' birth counts and dividing by the previous year's figure, noting that a negative result simply indicates fewer births in the current year compared to the last. It is important to understand that such negative growth rates do not necessarily imply an error in the data; rather, they reflect scenarios where the number of live births has decreased, which could be due to various demographic reasons. By examining these annual changes, health officials can identify whether their data shows stable trends or erratic spikes and drops that require further scrutiny.
In conclusion, the core message is that comparing implied rates within complex indicators against external benchmarks and global trends serves as a powerful tool for data validation. Whether checking if infant mortality rates are decreasing in line with system improvements or verifying that birth growth rates do not exhibit unexplained volatility, these comparisons help build a robust foundation for further analysis. The video illustrates that while minor differences between sources are expected and acceptable, significant deviations should prompt a review of the data collection methods or an understanding of specific local circumstances like conflict or migration. Ultimately, this rigorous cross-checking process ensures that health statistics remain accurate, reliable, and reflective of the true status of public health services in a given region.
Read the full video transcript
Okay, another example is comparing the
crude birth rate with other sources. So,
um in that same portal I showed you that
crude birth rate was possible to obtain.
You can do the same thing. Now the
reason these types of of comparisons are
useful is that it things like infant
mortality rate um and um crude birth
rate they combine multiple denominators
in the indicator itself. Okay. So as an
example here crude birth rate is
calculated by taking the number of live
births divided by our estimated
population our total population right
and these denominators
well well this indicator consists of
essentially two denominators that we
would use in other equations right in
other indicators a live burst is often
used in many other indicators we might
also use our total population in many
other indicators as well okay
so this can be an interesting comparison
comparison to take these indicators that
are made up of other denominators and
compare them with other sources as well.
And when you see that they are close
together um once again same thing it
gives you more confidence essentially um
with the data that you have um when
you're seeing large differences in this
case we we give a bit more leeway about
20%. Okay. Uh difference I just have an
example table here where we have some
some values from the LAO demo. Um this
is the the real population data is um
that's publicly available. Um and then
we have the UN data as well. Okay, that
we've taken for the same periods and we
can see they're very close. Okay. Um
there's very little difference between
them. There's some but they're very
minor, right? And this gives us a lot
more confidence when we think about our
live birth and when we think about our
estimated population um for our country,
right? And and we can give this
different weightings in our mind, right?
Depending on what we think of the source
we're comparing to um but uh you know
given that we have one method where
we've calculated data using our own
method, our live births are from our
routine data in this case, right? the
data that would be inside of DHIS2 and
our estimated population
um would be from our census, right? And
the UN estimates are completely
different, right? They're just estimates
of of things. Um they don't have any
necessarily or as much raw data to work
with necessarily depending on the
country, right? But we see the values
are very close and this gives us a lot
more confidence um in these denominators
to use in in further um analyses later
on.
Okay. Um, another thing we can do is
compare the overall trend of your rates
with just global rates that you see or
globally calculated rates. So, our demo
is Lao, okay? And the Lao Ministry has
graciously kind of given us their their
geography to use in our demos. Okay. Um,
but you can see here in Lao there's been
a a very steep decrease in crude birth
rates over time, right? the the services
are improving significantly. So, we
wouldn't want to see in our own crew
birth rates uh some type of increase all
of a sudden, right? That's not reflected
um in in the global estimate at all. And
that's not maybe a reflection of the
services um that are being delivered,
right? We know if we zoom out, maybe
there are small hiccups here or there,
but if we zoom out over an extended
period of time, we know that the
services are significantly improving. Um
so uh we wouldn't want to see a a a
different trend necessarily in our own
data. Now there are of course places
where this is not the case unfortunately
um you know places where there has uh
you know things like war etc have have
affected these services um and then you
would maybe want to see the data closely
more closely aligned um to those
findings. Okay. But the whole idea is
that when you're observing these global
trends and you're thinking about your
own services um just kind of logically
you know ideally some of these implied
rates should match um with these uh kind
of global um global service uh global
estimates to some extent. Okay.
[clears throat] And and knowledge of
course about your own services and how
they're being delivered. Um infant
mortality uh rate is another u one you
could compare. um uh infant mortality
rate is taken by uh taking the live
births minus your surviving infants um
that are less than one and dividing them
by your live bursts. Now surviving
infants I mentioned
was a possible uh denominator for some
of your indicators right it's not always
used the way it should be um but uh it
can be a very useful indicator to use um
in many cases okay and these are the
number of infants that are surviving um
over a certain one-year period that are
less than one right um and this could be
different than your live birth as as
some of those children may unfortunately
um die right so so the idea here is once
again to make these comparisons and it's
the same kind of thought process as your
crew birth rate,
right? And same kind of thing. If you
see that your infant mortality rate is
decreasing over time, generally
speaking, okay, um then this should
align with with what you see in your
system.
Um another thing we can look at is the
annual growth rate. Okay? And what we
can do from a health perspective is look
at the annual growth rate as a result of
births. Um there's other factors that
affect the growth rate, right? And we're
not asking you to necessarily take those
into account. So you could have
immigration, people coming in,
immigration, people leaving. Um you
know, there are other factors here as
well. We're we're purely focused on the
growth rate from a kind of health
perspective, health systems perspective,
which is therefore looking at the bursts
in this case because we would have
access to live bursts for every year,
right? And what we would want to see is
based on this growth rate, um you know,
is it erratic? Is it changing too much?
Right? Are the differences in bursts
from year to year causing uh large
spikes or increases or large decreases?
Okay? and what's the overall trend here?
Um, and then I'll go through some
examples of how we calculate this
essentially. Okay, but the whole idea is
once again you're looking for erratic
kind of differences, large differences,
whether they're increases or decreases.
Um, couple notes on growth rate. You can
have a negative growth rate, right? And
what this means is you have less less
births in the next year compared to the
previous year. This means that just as a
result of bursts at least um your
population would be decreasing. But
there could be other factors um where
the that that mean your population is
not decreasing necessarily. Um of course
people coming into the country through
other means um as well could increase
could increase that rate. Okay. But uh a
negative growth rate is not incorrect
necessarily. Okay. It just means that uh
your your um population is decreasing
based on Barcelona.
Okay, here's an example and we'll walk
through uh an example um of how to
calculate this. So if I wanted to
calculate in this example the growth
rate for 2024,
okay, I would take the number of live
births in 2024.
I would minus the number of live births
in 2023. I would divide this by the live
births in 2023, okay, and times it by
100%. In this case, I have a negative
growth rate. You can see here my number
of live births is less in 2024 when
compared to 2023. And that's why I have
a negative rate. Okay? Doesn't mean it's
wrong. It just is the scenario um in the
country. Okay?