Support for multiple hierarchies is one of the main features of DHIS 1.x (the MS Office based version) that the core team decided NOT to include with DHIS2, presumably because non of the early adopters of DHIS2 required that. SA used up to three different hierarchies for many years during transition periods, although gradually the interest in the “older” hierarchies was reduced and we changed over to using “backups” to cater for the remaining long term trend analysis needs (ref 1 below).
There are some dhis2 strategies you can use to achieve similar results, though:
For historical comparisons across significant boundary change events, you can use “backup” copies of the same data existing on your servers under different names. Whether such comparisons will enable to you determine if a positive or negative health indicator change is primarily due to the restructuring or something else - well, that will depend a lot on how radical the re-structuring is and of course what kind of re-structuring and increased resources are used on the ground. Example:
in a very corrupt country, where the bulk of health funding are stolen by politicians and bureaucrats, a radical re-structuring eliminating the most corrupt groups might yield significant and verifiable results.
for your typical admin boundary changes for political reasons, like when a president creates more and more districts as part of a jobs for votes programme, I think you will find it a lot more difficult to prove that it results in poorer service delivery.
the same will be the case for boundary changes that relates to longer-term changes in populations and habitats. E.g. it is unlikely that suburb A and suburb B, having grown 50% over 10 years and is now split into suburb C and D and E, will provide you will enough data that you can attribute changes in health status of the split to the split itself.
For multiple hierarchies currently in use, you can run concurrent instances with different hierarchies but where the lower level data is kept in synch using automated tools (e.g. HISP-SA and HISP-India are developing Apps for meta-data synch and data synch).
Depending on the level of difference, you can also use OrgUnit attributes combined with some sql views and custom reporting to deal with partially different hierarchies.
Again depending on the type of data and hierarchies involved, you can possibly “merge” the hierarchies into a new hierarchy with more detailed, composite orgunits at the lowest level. This method is actually a generic solution to many seemingly incompatible classification systems - you break the different classification systems down into more uniform base components for which you can attach the relevant data. Moving from routinely collected aggregated data to Tracker data, or to an Electronic Medical Record system, is a variant of this principle. The further your data collection moves towards the real world, the easier it becomes to aggregate and analyse the data from multiple perspectives (including organisational perspectives).
My 2c worth
On 3 June 2016 at 00:19, Kristof Bostoen email@example.com wrote:
We just posted our desire to add multiple hierarchies to DHIS2 in Launchpad.
We think we need this to be able to analyse the data according to significantly different hierarchical systems, in our case a geo-administrative one and one based water basins which do always follow administrative lines.
With multiple hierarchies we also want to be able to keep historical (past) hierarchies when administrative boundaries change so we can see if changes to results is due to restructuring our based on real changes.
If you identified the same problem, have already found a solution to this problem or have implementation advice we would very much appreciate.
Mailing list: https://launchpad.net/~dhis2-devs
Post to : firstname.lastname@example.org
Unsubscribe : https://launchpad.net/~dhis2-devs
More help : https://help.launchpad.net/ListHelp
46D Alma Road, 7700 Rosebank, SOUTH AFRICA
Tel/fax (home): +27-21-685-6472
Iridium SatPhone: +8816-315-19119