Information Technology Reference
In-Depth Information
Finally, there are some significant technical issues including:
How to express a clear clinical question in a way that disparate computer systems can all
properly interpret it
Developing a standard for expressing clinical terms and concepts
Overcoming the lack of consistency among providers in data coding
In summary, the challenge is how to collect, aggregate and meaningfully analyze
clinical data from such a complex set of systems and technologies in an environ-
ment with no real standards for data representation.
This question is important enough that it's been given a name: distributed popu-
lation queries. It was the subject of a series of web meetings sponsored by ONC in
the summer of 2011 and is an initiative of the Standards and Interoperability col-
laborative community of volunteers from the public and private sectors. [
1
] This
group has created a set of proposed Query Health standards. The details are quite
technical but basically they sit on top of the already agreed to standards for data
representation (e.g. coding systems like SNOMED) and packaging into documents
(e.g. CDA).
Query Health involves two new standards. The first is referred to as the “enve-
lope” and it contains the query and any associated policy requirements. This must
include information needed for systems at the receiving location to make sure their
internal security and privacy guidelines are not violated by the query.
The second new standard is the Query itself. It is expressed in an XML format
similar to that already used in CDA documents thus assuring it can be read by virtu-
ally any EHR while avoiding the need for an entirely new query standard.
There are at least three existing open source implementations of distributed query
PopMedNet [
2
] , i2b2 [
3
] and hQuery [
4
] .
PopMedNet
™
was developed for Harvard Pilgrim Healthcare Institute and the
Harvard Medical School, Department of Population Medicine. It consists of two
components, the Portal and the DataMart Client. The system assumes a group of
entities have created a collaborative network for shared research. There is a single
Portal for each such network where requests are made and where all system com-
munications, security, and governance policies are managed. Each collaborator on
the network has their own local DataMart Client. A collaborator could be a hospital,
an individual clinic or any other entity that has stored clinical data. A key goal of
this design is that PHI never leaves each collaborator's system. Queries get pro-
cessed by the DataMart Client and the local system and only the results go back to
the Portal.
PopMedNet is being used in the FDA Mini-Sentinel project [
5
] [
6
] to assess the
safety of already approved drugs, biologics, and devices in actual clinical practice.
As of December, 2011 the population being assessed consisted of 126 million
patients. An example of its use was the rapid evaluation of cardiac outcomes of
patients using drugs for smoking cessation. This was a proof of concept that was not
designed to produce useful results but it did work without the need to transfer any
PHI over the network.
