The design and structure of clinical research information systems. Implications for data retrieval and statistical analyses

Abstract

Data management software designed to support clinical data bases typically provides the user with the ability to "enter" and "retrieve" information according to simple user-specified criteria. In the medical research environment, such data base management systems can be self-limiting unless the user has carefully structured the data base schema to be consistent with subsequent statistical procedures used for analysis. For statistical purposes, the data base schema must be configured such that the dependent and independent variables are structurally situated to facilitate the use of statistical application programs. Furthermore, the analysis of time-oriented, prospective studies often requires the data base to be "relational." This may be inconsistent with data collection procedures that result in "hierarchical" schemata. Methodology for ensuring compatibility between the data base schema and subsequent statistical analyses is presented using examples derived from a multicenter clinical trial of diabetes and an observational data bank approach to disease surveillance in rheumatology.