Organizing standardized electronic healthcare records data for mining

Digitized and standard health data is highly desirable in healthcare domain. Current research started with the aim of mining standardized electronic health records (EHRs) data. It is observed that the available column-modeled database lacks in volatility and ability to deal with sparseness. Due to highly sparse nature of EHRs, there is need of organized standard based EHRs data to ascertain the benefit of mining and also to improve storage capacity. This paper proposes an architecture for preprocessing of data that deals with the issues of volatility and sparseness, and then transforming the data in a form which is desirable for data mining. The architecture proposed in this paper can be used by the data administrator to organize standardized EHRs for the application of any type of mining which can reveal knowledge to benefit various healthcare users, such as, doctors, patient and insurance company users. Simulation of proposed architecture is done on a sample set of 6700 instances of EHRs. Results demonstrate benefits of our approach in terms of storage required and time consumed (reduced by a factor of 2.5 approximately) to retrieve the desired piece of information. Consequently, data mining techniques capable of automatically extracting useful knowledge to provide clinical decision support may be applied. The decision support thus developed can be used by stakeholders, policy makers and other persons involved directly or indirectly with EHRs to take accurate and calculative decisions.