| Consideration of the usefulness of data in the design of systems to reuse clinical information is primarily concerned with computational obstacles such as missing values and the mathematical format of data. This computational approach is based on the tacit assumption that, once properly formatted, the data will be suitable for its proposed secondary use or uses. In the complex and heterogeneous clinical information environment this is often not the case. The medical record is a socially constructed artifact, shaped by clinical, economic, legal, and technical influences. These contextual influence determine the form and contents, and therefore quality, of extant clinical data. In this dissertation, techniques for the secondary use of clinical data are explored that incorporate both computational and contextual perspectives on data quality.;Four related and cumulative studies are conducted using the data from two hospitals to identify process-based correlates to outcomes for the prostatectomy: surgical removal of the prostate. In the first study, a system is designed to automatically extract the prerequisite "before and after" information from pathology reports. In the second study, interviews are conducted with urologists to discover the contextual influences that shape the format and contents of narrative free text surgical operative reports. Based on the results, the third study introduces a method to discover correlates to outcomes in surgical operative reports. In the final study, the results of the previous studies are used to compare the outcomes and procedural variations of urologists.;The outcomes of this dissertation include a system to automatically extract known quality measures from clinical records, an understanding of the influences shaping dictated operative reports, a technique to automatically discover correlates to outcomes in operative reports, and the identification of correlates to outcomes as well as significant differences in surgical approaches used by urologists. |
