Patient safety is an important topic but non-trivial to measure, as it comprises very different phenomena. One important example of patient safety issues is delirium. Many approaches for the detection and prediction of delirium are described in the literature. However, additional effort is often needed for a comprehensive data collection and for the avoidance of potential biases. To systematize a process for the detection and prediction of delirium reusing available clinical routine data, we aim to develop a minimum data set (MDS) for delirium. By combining a top-down and bottom-up approach, we compiled a comprehensive delirium map containing potential delirium elements based on evidence. The alignment with clinical routine data led to a specific version of an MDS for delirium and revealed that most of the delirium elements could be identified within available nursing routine data.