ABSTRACT
Accessing electronic health record (EHR) data is challenging. Sources of EHR data are varied, each containing implicit strengths and weaknesses about the appropriateness of the data for research purposes. EHR data may be obtained from a single institution or multiple institutions, and the data may be directly accessible to the researcher or require a collaborator or third party to access the data. Research data may be further complicated by the inclusion of non-EHR data, such as claims data or census data. In this chapter, we consider each of these data sources in turn and cover the methods needed for building a research database from the EHR. This discussion includes approaches for obtaining data, observation and variable selection, and underlying data architecture, with particular attention given to understanding the various methods of data abstraction, from manual chart review and reporting tools to directly interfacing with the EHR's database.
