Not long after digital computers started becoming widespread, computer scientists began studying the problem of how best to store the collections of data these machines would inevitably be tasked to work with. It wasn’t long before they were able to lay down formalisms which, when put into practice, would help to ensure that the databases embodying them would remain consistent and reliable. In fact, these relatively early researches into data management of this sort served the world of computing well for decades.
In time, however, applications in particular industries came to show some of the shortcomings of approaches incorporating these insights. For example, the modern move toward so-called “document-oriented” databases in some sectors is inspired by the greater ease with which these seem to handle some sorts of data sets when compared to the relatively rigid ones that were formerly more widespread.
In the health-care industry, a similar transformation has been taking effect. Being fundamentally reflective of not-so-easily reducible human subjects as they are, medical records and similar data points seem to resist easy incorporation into rigid, early-binding database schemes. Even when they are forced to fit in traditional databases, in fact, they must often be stripped down in important ways in the process. This is a problem for those interested in analyzing such data because it means that their efforts will inevitably have to do without the full load of information such documents originally contained.
A solution known as the late-binding data warehouse has been making waves in the industry of late. At a first glance, it bears some resemblance to those document-oriented databases that have been gaining in popularity elsewhere. In this case, however, the capabilities of the underlying database engine are tuned specifically for the needs of those in the health-care sector.
This development allows health-care analytic systems which are built on top of such technology to offer far more flexible analytic tools than they could otherwise. At the same time, because the database technology they work with has not completely done away with structure and normalization, it can still offer the levels of reliability and consistency that are rightfully demanded when such important sorts of data are involved.