In the project report on Defining and Evaluating Possible Database Models to Implement the Sentinel Initiative, the contractor suggests utilizing a common data model for the initial implementation of the Sentinel System. The report describes a common data model as "a set of specific definitions for the structure of databases and of data elements (i.e., specific variables)" and states that "most implementations of common data models require each contributing partner (in this case, each data owner) to transform its data into the common model, either virtually or physically."
Some advantages of a common data model cited in the report include that it "substantially reduces the need to centrally create and maintain complex metadata, ontology, and data mapping activities." It also addresses issues related to "syntactic and semantic interoperability." It creates an environment in which much of the data validation is conducted during the "initial transformations instead of requiring complex data checking effort after execution of each query." Finally, the common data model allows you "the ability to centralize analytics as opposed to requiring each data owner to interpret and implement a set of written instructions." However, The report also notes that a "common data model requires initial effort and expense to transform data...and to store it... In addition, to the extent that the common model does not capture the full granularity of the source data, the common model approach could result in the loss of information."
As the FDA works toward implementing the Sentinel Initiative, please share your thoughts of the advantages and disadvantages of the common data model.