Star schema design question

Question

Hi!
I`m trying to build a dw with patient data. I`m mostly interested in the patients` visits table, which will be used for a facts table and the patient table.
The patient table contains among others the following data:
patient (id, sex, marital_status, nationality, profession, zip).
My question is what should I do with the "patient" table. There seem to be three options:
1. Use it to build a one level dimension and just use the above fields as simple dimension attributes.
2. Create separate hierarchies for each attribute in the patient dimension
(e.g. patient -> sex, patient -> nationality, patient -> zip -> city e.t.c.)
3. Create separate dimensions for each attribute (e.g. a professions dimensions which will be joined with the patient visits facts table) and no patient dimension.
What would you do?
Thank you...

Answer 1

I guess my answer would be based on how you intend to physically implement. If you are deploying this relationally, then I would definitely have a single patient dimension table. Perhaps build a hierarchy for patient -> zip -> city, but I'd keep gender and nationality as simple "attributes" of the dimension.
If you're going to deploy this through an OLAP cube, its a bit more complex, because the standard Oracle OLAP functionality makes it impossible to do sums over attributes unless they are part of the hierarchy. In that case, I guess I'd deploy multiple dimensions (geneder, nationality, one a generic patient one that includes the hierarchy patient -> zip -> city).
If you attempt to do your option #2, you'll be in trouble if you need to show queries that cross the hierarchies, i.e. "Male American patients who live in zip code 12345".
Hope this helps,
Scott
p.s. you could also deploy this relationally, and then put an AW on top of the relational model.