Results from data mining in a radiology department: the relevance of data quality.

This work is part of an ongoing effort to examine and improve clinical workflows in radiology. Classical workflow analysis is time consuming and expensive. Here we present a purely data-driven approach using data mining techniques to detect causes for poor data quality and areas with poor workflow performance. Data has been taken from a operational RIS system. We defined a set of four key indicators for both data quality and workflow performance. Using several mining techniques such as cluster analysis and correlation tests we were able to detect interesting effects regarding data quality and an abnormality in the workflow for some organizational units of the examined radiology departments. We conclude that data-driven data mining approaches may act as a valuable tool to support workflow analysis and can narrow down the problem space for a manual on-site workflow analysis. This can save time and effort and leads to less strain for clinicians and workflow analysts during interviews.