This “methodological review” highlights how nursing knowledge bases are to be constructed in the future. Instead of “feeding” a decision-making system with an expert nurse’s knowledge, as is done in the creation of classical expert systems, data mining should be used to extract knowledge from nursing data, and to embed this knowledge into the daily functions of clinical information systems.

The first part of the paper draws attention to nursing knowledge, rather than nursing procedures, which has to do with the purported fact that many expert nurses explain their tasks in a different way from how they perform them. This supports the need for constructing and managing nursing knowledge bases in health information systems, but many open questions remain, such as what the exact role of nurses in the construction of this knowledge should be, what additional capabilities or training they will require, and whether they should devote part of their care-taking and monitoring time to this endeavor, or if there should be a specialized expert nurse role.

The second and longer part starts with a more classical view of the opportunities of knowledge discovery in clinical databases, a very short overview of what data mining is, and a case study on pre-term birth prediction. Apart from the great work of synthesis done to make sense of all this in a few pages, the most important issues of data mining for nursing knowledge are emphasized well in the example: data privacy, data quality, data standardization, sequential and repeated measures, missing values and variables, overfitting and metrics, and variable selection and dimensionality reduction.