ISO Clinical Data Warehouse draft Technical Specification
Whitaker, Patrick
Patrick
Whitaker
Technical Officer
Health Care Informatics
Unit
Health
Statistics and Informatics Department
World Health
Organization
Tel. direct: +41 22 791
1372
E-mail:
whit...@who.int
From: Andrew Grant [mailto:Andrew...@usherbrooke.ca]
Sent: 16 October 2008 11:12
To: Whitaker, Patrick
Cc: Bailey, Christopher; Boucher, Phillipe; jeremy...@nhs.net; Spohr, Mark; Mark Fuller
Subject: RE: Clinical Data Warehouse
(Voir fichier joint : DW - Process and outcomes.ppt)
Thanks
Patrick,
On behalf also of Mark Fuller (Canada) and Jeremy Thorp (UK) who
are leading their respective national data warehouse programs, and myself as a
medical medical informatics practitioner / researcher in this area responsible
for coordinating this technical specification content, we would be very grateful
for comments to all parts of the proposal. Following our discussions in Istanbul
I am particularly interested in including certain points of view that could be
particularly relevant to developing countries.
It is a technical
specification hence it is the generalisable component which is of most interest
which can be supported by examples of particular scenarios.
As regards
part 2, concerning data abstraction and modelling which will continue to evolve
in its specificity before publication, I am interested in your remarks on how
already abstracted data might become the input into a data repository/warehouse
rather than primary data, as well as your points of view on indicators. As
regards indicators, data warehouses are a future basic source of performance
measures of different sorts. From the point of view of data quality the relation
to primary data is a very important component and normally it is recommended
that the primary data be in the data warehouse. However we should also be able
to describe contexts where this may not be possible. As regards formulation of
indicators we also should be descriptive rather than prescriptive. I would also
suggest you take into account as a reference point the ISO TS 21667 on the
Health Indicator Conceptual Framework which is being promoted to become an
International Standard.
I am including an additional slide from my
presentation as it illustrates how a data warehouse can act across different
levels of care, as well as take into account not only cross-sectional but also
longitudinal (pathway / process) data views and I would like to suggest this
integrative perspective be taken into account in your comments and from your
perspective.
This is a great and important contribution from your part
and I am delighted that we can address these issues.
With very best
wishes
Andrew Grant MD PhD
Professeur titulaire, université de
Sherbrooke, Québec
Ngeno, Titus
Hi Patrick,
Not sure how help this would be but in about a year ago I was privilege to work with University of Utah Hospitals & Clinics. The objective for this project was set up a testing environment to determine the amount of backend hardware resources required to execute data-warehouse ( my case PowerInsight by Cerner) extracts against a client sized database and the length of time associated with each type of extract.
1 Test Configuration:
The test is comprised of a 3 node configuration. Extracts will be pulled from the clinical environment node (Clinical Database). And you must have two other nodes one for loading and another for informatica database. The two nodes used for running the load will have a Health management systems footprint running and one node will have the informatica database executing on it. The load nodes will also connect across to pull the extract after it has completed on the clinical environment and populate the data warehouse.
II Hardware:
Back-End Clinical Environment (used for extracts) I would think Aix P570 with 4-CPU’s 26 GB of memory . The loads environment should also be AIX P570 with 2-CPU’s 6.5 GB of memory and the second node (informatica database) should be AIX P570 and 2 CPU’s and 6.5 GB of memory
III Software:
Back-End Software: Aix 5.3 with MQseries (for message transfers) or Linux and Health management software (my case Cerner Millennium) of your choice and the load Environment should have HMS software , informatica version 7.12 and ( Cerner Powerinsight) data-warehouse EDW alpa code
IV Workflows:
Workflow 1
Workflow 1: Test Data Extracts from Clinical Database source
Identify a date range in which the number of required encounters was updated. Configure the extract scripts to pull data from the ***table_name*** for that date range, Run the extract scripts. The data extracted from the tables are written to flat files.
Workflow 2
Workflow 2: Test data loads from flat files to EDW tables
Move the generated data files to the Informatica server and load the data into the data warehouse Gather performance statistics around time measurements and other data useful in evaluating performance improvement (CPU, Memory, etc).
V) Business Objects
You would Need business Objects for the end users
I strongly believe with a good Database this could be emulated in a third world country!
I hope this helps
Titus Ngeno