Data compilation has a long history in the field of medicine, and domains for which data are pooled include, among others, epidemiological studies and quality assessment. Unfortunately, data may be of low quality with poor validity and reliability, and consequently, decisions based on statistics may be unreliable. The range of applications of information use and reuse is expected to extend from aggregation of information to retrieval of specific cases, which further emphasizes the importance of high quality data. The present thesis deals with aspects of information quality in Swedish primary health care registries. Such information is commonly encoded according to standardized coding systems.
In order to improve coding quality, it is necessary to study the content, structure, and semantics of the coding systems, as well as the functionality of the tools used to access them. In particular, the thesis discusses the effect of the coding system structure - as an instrument for code retrieval as well as data aggregation - on the reliability of coding. In summary, the functionality of a tool that supports coding of medical problems must meet numerous requirements. Flexible structures for browsing the coding system and different types of lexical tools are needed, as are rules for guiding the correct choice of code with respect to a particular medical problem. As an instrument for data compilation, the structure of the coding system may support aggregation in ways that reflect dimension with low variability.
The main contribution of this thesis is to increase understanding of the complexity of disease concepts and the coding systems used for representing them. In addition, it also encompasses the measurements of reliability in coded Swedish primary health care data that were carried out in both a smaller prospective and a larger retrospective study. Further, the thesis proposes a metric for retrospective reliability studies. Finally, the presented coding tool - the Classification Browser - is in itself a platform for further discussion and development in the domain of diagnostic coding.
Linköping: Linköpings universitet , 2003. , 65 p.
2003-04-11, Patologsalen, Universitetssjukhuset, Linköping, 09:15 (Swedish)