Page Comparison

OpenMRS has few tools in place allowing for easier analysis of concept, patient, location, encounter or visit data in an aggregated, dimensional manner. OLAP (Online Analytical Processing)  is one technology encompassed under the umbrella of business intelligence that facilitates rapid answers to multi-dimensional querying of data.

Click on the image at right for a simple sample of what dimensional modeling looks like at a high level. 

This functionality extends beyond traditional reporting in several ways:

• OLAP tends to be more adhoc in nature, although static queries can be issued to the underlying engine;
• OLAP deals specifically with querying at an aggregate level (whether that means simply counting patients meeting requested dimensional criteria or performing complex calculations on some other measure), and drilling into interesting intersections (or tuples) for more detailed data;
• OLAP requires a significantly different data model than the underlying OpenMRS data model in order to craft the dimensional models that the OLAP engine understands. 

The community edition of the Pentaho Business Intelligence suite includes Pentaho Analysis, an OLAP engine (specifically ROLAP) project named Mondrian. This suite also includes the powerful and easy to use Pentaho Data Integration, an ETL engine project named KETTLE. The Pentaho BI platform is architected as a pluggable and componentized set of pillars (or engines) that can operate in the Pentaho BI Server as an integrated set of tools, or can be utilized as embedded engines in other applications individually.

The purpose of this project is to research the feasibility of integrating the Pentaho toolset with OpenMRS to provide advanced ETL and analytic modeling within an OpenMRS module.The design should take into consideration the fact that data semantics can vary significantly between OpenMRS implementations, which makes standard ETL and modeling challenging. The module will require rules definition and logic that guide the ETL processes in the movement of dimensional members and aggregation of fact data. Design considerations for this logic should include utilizing existing forms, rules engines, fuzzy matching, semantic modeling and other techniques to bridge the gap between concrete facts and data-relationship semantics.

The project will include ongoing development of a set of prototype ETL transformations and models in order to flesh out detailed requirements and validate design decisions. This requires a decent sized implementation of OpenMRS to partner with, and Andy Kanter from the Millenium Villages (Institute, Columbia University) project has agreed to participate.