Carinci F, Rose W,
Monash Centre for Health Systems Research, 2003
Objectives
To develop software specifically designed to
improve the ability to perform biostatistical analyses of data routinely
collected in a variety of settings of care. To encourage distributed
processing by deploying a prototype that could be widely applicable
and amenable to further extensions.
Materials and Methods
We have designed a prototype defined as H+
MetaBase, which has the following features: a statistically oriented
metadata representation; functionality on a distributed environment;
ability to include clinical concepts; distribution of the processing
load across sites; integration of the results through a specialised
statistical engine; secure communication between sites; a common
interface; based on open source.
The system has been implemented through the
development of five fundamental components: a query builder, a query
spooler, a partial analyser, a meta-analyser and a results viewer.
Standards, programming languages and protocols
used include XML, Python, OpenSSH and the Apache Web Server for functions
related to the interface and related queries. Statistical routines have
been developed using SAS.
The server side application was based on a
Linux platform, while three Windows NT/XP workstations were used to
simulate a distributed system.
Results
The final prototype has the ability to execute
distributed queries and produce results in terms of summary statistics,
univariate analysis and multivariate graphs.
The system is also extensible and allows
users to customise its use to their particular case.
An example using a large database of hospital
admissions has been used to test its functionality. This application
has shown that statistical analysis using H+ is feasible and reliable,
and that the system can be used on a routine basis in real life
situations.
Software and demo are freely available
as open source at a web repository that also includes other
libraries: http://hplus.med.monash.edu.au.
Conclusions
Application of statistical analysis on
distributed health databases is possible with minimal hardware/software
requirements and limited assistance from skilled technicians.
The H+ MetaBase prototype, although limited in
the available options, provides a foundation platform whose capabilities
may be further developed and advanced.
Open source solutions may be essential to
deliver easy to use, inexpensive and effective solutions.
However, knowledge exchange and
multidisciplinary collaboration for health information systems need to
be enhanced.
Development of a H+ web repository may
constitute a strategic step in this direction.