The idea is to follow the CRISP-DM process. You need to do some reading on the ED operation and on the problems with the EDs. Then you *think* about it and - looking at the data - you formulate...

The idea is to follow the CRISP-DM process. You need to do some reading on the ED operation and on the problems with the EDs. Then you *think* about it and - looking at the data - you formulate questions which it is possible to answer by analysing the data. You also need to think about the data quality and data formats - the dataset is as it comes (this is "business understanding" and part of "data understanding"). Once you have your initial set of questions formulated, you propose which methods will be good to answer these questions (e.g. is age influencing the overall stay? For this, we need to define and construct the "overall stay" - something like "Overal stay is the time between arrival and departure" - or maybe better "time between triage and Ready-for-departure". Then we can look whether there is an association between age and overall stay. Now you need to do this for each triage category... NB: you need to justify your choices and decisions - how the partial question is going to contribute to the overall problem understanding/solution; why several factors are involved in the analysis etc.). Once you get an idea about methods you want to apply and data you need to construct, you include all required data transformations into your plan.