dhis2-users team mailing list archive

[Lars Helge Øverland <larshelge@xxxxxxxxx>]

On Fri, Apr 27, 2012 at 12:22 AM, Olav Poppe <olav.poppe@xxxxxxxxx> wrote:

> Hi,
> came over and issue today regarding the calculation of indicators based on
> organisation unit groups. Not sure if there are any easy ways to solve it,
> but though it was worth discussing.
>
> The problem is ownership reports with indicators using population.
> Population data is often entered for districts (of similar). These have
> "government" ownership, thus when comparing performance based on
> ownership/population, figures will be wrong.
>
> One way to solve this would be to have catchement population for all
> facilities, giving a population figure that acutally have an ownership. But
> in cases with several aggregation levels for population (facility +
> district), the same problem would probably occur in a district or regional
> reports?
>
> Another option would be to avoid the ownership of population data, either
> by (somehow) not having an ownership for non-facilities, or by computing
> such indicators "cumulatively" for org unit groups. This is probably best
> explained through an example:
> A district has a population of children < 1 years of 100. 60 children are
> given BCG at government facilities, 30 at private facilities. An indicator
> showing BCG coverage for that district would currently (from how I
> understand it) show a coverage of 60/100 = 60% for government and 30/0
> = 0% for private. Maybe it would make more sense to show this as 60/100 =
> 60% for government and 30/100 = 30 % for private, giving the the total
> (correct) coverage of 90%? Without knowing the actual catchment population
> for the private and government facilities, it's impossible to calculate the
> real coverages, but as a whole I believe this method could be better on
> average?
>


Hi Olav,

your observation is correct... org unit group aggregation for indicators
does not work well when facilities do not have denominator data. Currently
in those situations we use it mostly for data element analysis for
comparison of utilization through stacked column charts or pie charts (e.g.
how many was immunized at public facilities out of total immunized).

One could come up with a method for improving this like you suggest. To me
it becomes a bit too complex to create a general rule for it - one would
assume that members of org unit groups are on the same level, then say
something like "look at the level above for this data element for
denominator data"... And the aggregation rules will become different from
when denominator data exists for facilities.

This also raises some interesting questions, like do private facilities
really have a catchment area? Should one use geographical bordes or adjust
for attendance and migration? Anyway I think the only way to solve this is
to apply a reasonable catchment population to facilities. Sorry for not
being more helpful.

Lars