INTRODUCTION-WHAT IS A DATA MART?
A data mart is a collection of data in a customized format in a data warehouse focused on a specific report or functional area such as hospital’s census, hospital’s charge activity, labor and delivery outcomes, or Nursing quality measures. (Oracle, 2012) Data marts are usually build by department of information services or an EHR vendor. Data marts usually draw data from more than one source. These data sources can be internal systems used by operational units or external sources. The internal source doesn’t always have to be an internal operational system; it can be another internal data warehouse. The data contained in a data mart is single-subject focused, can vary in sizes and complexity, although it would be limited in scope. (Oracle, 2012)
Data marts are classified in two categories. Dependent data marts draw data from a center data warehouse, sometimes known as enterprise data warehouse, where as independent data marts draw data from operational systems or external sources. For the purpose of this presentation, we are promoting the use of dependent data marts because of the benefits of the Extraction-Transformation-and Loading (ETL) process is much similar in that the data is already clean, summarized, and loaded into the data warehouse. The ETL process consists of mainly identification of the subset of relevant data and moving it to the data mart. With independent data marts on the other hand, the ELT process has to be managed in its entirety, meaning that the data has to be staged every time to be normalized, integrated, and in dimensional formats. (Ariyachandra & Watson, 2012) For these reasons, I am advocating building dependent data marts to maximize performance and to increase data availability for long-term information need.
There are five steps in data mart implementation that needs to be followed (Oracle, 2012):
1. Designing the schema
2. Constructing the physical data storage
3. Populating the data
4. Accessing the data to make decisions
5. Managing the data
The design process involves the following tasks:
• Gathering the requirements from users of the information
• Identifying relevant data sources based on pre-defined requirements
• Identifying the physical design of the source data
• Identifying the business logic of the source data
• Transform the source data into a logical and physical design of the data mart
The design step in the data mart process doesn’t need to include all of the source data; usually just a subset of data is needed for the data mart as specified by the business requirements. (Oracle, 2012).
The process of constructing the logical structures of the physical database for data access as related to the data mart involves the following tasks:
• Creating database storage structures (e.g. table spaces)
• Creating schema objects (e.g. tables, primary keys, table linkages, indexes)
• Setting up table structures for data access (e.g. cross tab tables)