In terms of a standards based approach I think I would look at some generic metadata modelling / ontology tool for representing these complex relationships. Like RDF. Or even ISO topic maps, though even though they would be excellent for this type of domain they are a bit dated and never really caught on.
Using RDF I guess you can incorporate dublin core, HL7 registry stuff as well as newer kids on the block like IHE Care Services Discovery. But you are not going to find off the shelf a standard which encapsulates the link between orgunits, datasets, datelements and indicators like we have in dhis. So something more generic like RDF might be your friend.
As I mentioned earlier, I believe the challenge will be with categories.
Most grammars you might try to define are going to make use of codelists (and there are a few candidates for representing these eg sdmx, sts, cts2). You might want to look at these, perhaps particularly the latter two as they are used in IHE/HL7 so have some health credibility 
···
On 10 April 2014 15:10, Greg Rowles greg.rowles@gmail.com wrote:
Thank you Bob. This is really helpful!
I think I understand some of the complexities that are being worked on. I can see that HL7’s specification on organisationUnit only partially satisfy our data-warehousing interests - and creating our own aggregated-data specification will mean challenging (un)common terminology. I’m working on an architecture for a dictionary/register that incorporates orgunit, data element + indicator specifications - this information really helps kick start the process.
Thank you,
Greg
On Thu, Apr 10, 2014 at 12:30 PM, Bob Jolliffe bobjolliffe@gmail.com wrote:
Hi Greg
Not sure I have space in short email to give my orientation on this discussion, but I’ll try.
First of all you are right about the limited relevance of standardized clinical vocabularies to ‘our’ domain.
I say limited because if you are looking at calculating aggregate dataelements (whether these are indicators or not is another conceptual mismatch we come up against) from transactional data stored in an EHR, then there can be some relationship between the dataelement for “number of malaria cases” and the ICD10 B52.X codes used for coding diagnosis but its not necessarily a simple one. And when you stray into important routine reporting dataelements which don’t have an origin in an EHR calculation like ‘number of bednets distributed’, ‘number of nurses at facility’ etc then the clinical vocabularies are not useful at all. Sometimes its hard to explain this to someone who asks why are you not using LOINC, Snomed-CT or what have you.
So unlike clinical vocabularies, there is no international standard for routine reported aggregate health data. Which creates a problem because standards are being seen as the underlying fabric of building ehealth systems. Though note that when these same people talk of ehealth systems they are for the most part talking of EHR and HIE.
The IMR was I believe a useful starting point in trying to reach global harmonisation of indicators, particularly across donors and international organisations. Though it does not have any of the consensus building trappings and formalities of a standards organisation. Its an application. So it cannot of itself produce standard vocabularies.
Given the august position and standing of WHO, work that is coordinated by them does enjoy some sort of international legitimacy. So there is also work on things like the global health observatory and xmart which provides “standard” indicator definitions and codes and is some way linked to IMR.
Do not go down the SDMX-HD road. We have been down it (I was co-editor of the SDMX-HD normative document so I know). The underyling assumption of “cubular” data is a poor fit for the type of data we collect and exchange (I actually think it is more suited as an output standard for indicator publication, than as an input standard for dataelement value exchange).
DHIS2 does not use use an underlying SDMX model. We have borrowed some ideas in the implementation of our datavalueset which is used for exchange in dxf2.
Exchange of dataelement definitions and codes is relatively trivial compared to disaggregation. Our categorycombo model has proved hard to translate to things like sdmx codelists - for good and bad reasons.
There is a process (which is not moving fast) which was started at Helina workshop last year to address the standardisation problem and how it might impact on dxf2. Speak to Vincent.
There is also the openHIE process. Do not consider what you do to be different, or in competition to openHIE. We as HISP are active members within that community so what we do is also HIE. In fact openHIE have taken a decision back in February that we shall form an hmis sub-community within openHIE which might also concern itself with some of the issues above. That should really kick off very soon and the broader hisp community will be encouraged to participate in that.
Bob
–
Business Intelligence Planner
Health Information Systems Programme
**- - - - - - - ****- - - - - - - ****- - - - - - - ****- - - - - - - ****- - - - - **
Mobile : 073 246 2992
Landline: 021 554 3130
Fax: 086 733 8432
Skype: gregory_rowles
On 10 April 2014 08:57, Greg Rowles greg.rowles@gmail.com wrote:
Hi All
For Data Elements:
As far as I can understand DHIS2 makes use of SDMX cube model (http://www.w3.org/2011/gld/wiki/Data_Cube_Vocabulary)
For Indicators:
There isn’t much standardization although WHO developed a tool that references components of SDMX (http://apps.who.int/gho/indicatorregistry/App_Main/view_indicator.aspx?iid=3326)
For Data Sets:
I see [Dublin Core] are referred to for data set (meta data) specifications.
I’ll try get hold of Patrick Whitaker - he seems to be a domain specialist. Everything points to SDMX and ISO/TS 17369:2005.
Regards,
Greg
On Wed, Apr 9, 2014 at 10:41 PM, greg.rowles@gmail.com greg.rowles@gmail.com wrote:
Hi Everyone
We’re in the process of developing a design spec for a (master) data-dictionary that caters for the many variations of datasets we have here in South Africa. I’ve been challenged on my claim that there are no ‘data standards’ around aggregated data elements or indicator definitions in our space. Jason suggested I put this forward to the DHIS2 community…
I cannot find anything useful. Everything is focused on standardization & interoperability at patient-level. HL7, LOINC, SNOWMED, ICD10 are all down a level on aggregated data. Does anyone have any articles or reference material on the issue?
For over a decade now HISP-SA has been the proud custodian of SA’s (unofficial) MFL + (official) NIDS - in the hands of Calle. Now that we are moving to DHIS2, with an expanded National Indicator Data Set, we also need to start managing a Master[Facility]List at OU5, a M[Ward]L at OU6, a M[Environmental-Health-Services]L at OU5, a M[School-Health-Services]L at OU5, a M[Reporting-Unit]L at OU6, a M[Ward-Based-Outreach-Team]List at OU7.
There are standards and guidelines for implementing a MFL but we have ‘extended circumstances’ here in SA. We are looking at building a sophisticated solution using DHIS2 for each of the unique Hierarchy arrangements. We are also looking towards the Australian DD for hints on building an ‘information services’ type-of platform out of our hybrid MFL+DD. Unfortunately there are feelings that we will be replicating what is already available in the openHIE software suite or we may risk losing buy in.
Anyone have info on standards for aggregated data? I’m fully comfortable driving ahead with these plans but derailment seems a possibility without a ‘standards oriented’ defense. There are also strong feelings that openHIE is a superior offering to our conceptual design (even though DHIS2 would be the underlying platform and data would continue to flow through transactional instances of DHIS2).
Any ideas, comments or feedback?
Greg
----- Forwarded message -----
From: “Jason Pickering” jason.p.pickering@gmail.com
To: “Greg Rowles” greg.rowles@gmail.com
Subject: Standards for Aggregated Data Definitions (Element and Indicators) in Health
Date: Wed, Apr 9, 2014 20:05
Hi Greg,
I think it is a very valid question and not something to be ashamed of at all. In terms of what WHO offers, you can check this out
http://www.who.int/gho/indicator_registry/en/
In addition, there are some internal HISP reccomendations (which I have heard of but never actually seen) in terms of how to create names of data elements and indicators. For instance, not using “Number of confirmed malaria cases” but rather “Confirmed malaria cases”.
I think in general though, there is a gap, and something which needs a lot more attention. HISP has been very focused on the technology, and less so on the semantics and M&E processes. I think Bob Joliffe and Ola though would have something to contribute here, so would encourage you to send this mail to the user/dev list for more feedback from the community.
Cheers,
Jason
–
Business Intelligence Planner
Health Information Systems Programme
**- - - - - - - ****- - - - - - - ****- - - - - - - ****- - - - - - - ****- - - - - **
Mobile : 073 246 2992
Landline: 021 554 3130
Fax: 086 733 8432
Skype: gregory_rowles
On Wed, Apr 9, 2014 at 7:01 PM, Greg Rowles greg.rowles@gmail.com wrote:
Hi Jason
I’m a little embarrassed to copy in any other DHIS2 heavy-weights because I might embarrass myself with this question:
Do you know of any existing standards or guideliness on the development of aggregated data-element and indicator definitions? Everything I find points to patient-centric data definitions (ICD10, LOINC, SNOWMED). HL7 appears to be more about interoperability standards and not about aggregation terminology.
Aggregated data is required for reporting to WHO so I assume there are standardization guidelines or existing codified definitions somewhere? If not - there is a gap…?
Regards,
Greg
Business Intelligence Planner
Health Information Systems Programme
**- - - - - - - ****- - - - - - - ****- - - - - - - ****- - - - - - - ****- - - - - **
Mobile : 073 246 2992
Landline: 021 554 3130
Fax: 086 733 8432
Skype: gregory_rowles
–