Re: Help finding natural keys
Date: Mon, 20 Jan 2003 12:01:25 +0000
Message-ID: <3K3QGa2VU+K+Ew3G_at_diamond9.demon.co.uk>
In message <OniW9.10233$I55.385455_at_news2.east.cox.net>, Alan Gutierrez <ajglist_at_izzy.net> writes
>--CELKO-- wrote:
>
>AG>>> In the real world, patients are identified by name by someone who
>> is responsible for a dozen or so patients. <<
>>
>> Maybe the name and the name of the responsible person form a key?
>>
>>>> These health care organizations use questionnaires designed by the nurses,
>>>> usually in Word ... questionnaire is a model of a blank questionnaire, it
>>>> is meta data, a completed questionnaire is an instance of the
>>>> questionnaire ... a questionnaire can have sub-questionnaires ... <<
>
>> Arrrgh!!
>
>That's what I needed to hear...
>
>> This is looking like more of a problem than we can do in a newsgroup
>> thread ...
>
>...I wanted to make sure that it wasn't as simple as it sounded in some of
>these debates.
Each newsgroup posting deals with small points in isolation from all of the others.
>
>In the archives I'd read that the lack of well defined natural keys indicates
>a weak model. This was the cause of much hand-wringing on my part. It seems
>that this is not a hard and fast rule. It is reassuring that people are asking
>me the same questions I asked myself to determine natural keys, and that the
>answers are not forthcoming. SSN, for example, has even more caveats than I'd
>imagined. Illegal immigration? My goodness. Better to learn that here than in
>production.
You are facing the same problem that many different organisations have faced, how to identify a person. The SSN is the government's attempt to fix that. They assign an SSN to everyone they have to deal with. If they discover a duplication they can change someone's SSN. If you discover a duplicate SSN you have to live with it.
The alternative is for you to issue your own codes, call them Case
Numbers. You have to make sure that your case number gets assigned to
one person for the duration of their stay in hospital, and that there is
no way that the Case Number can be confused with any other person, even
if there is another patient with the same name (or even one with the
same name and date of birth).
>
>That the application is to be shared by dozens of small organizations makes
>the choice of natural keys that much more difficult. Some organizations will
>naturally identify a patient by first name, last name, and to break a tie,
>room number. Home health care organizations won't have rooms however. In
>emergency situations, the name might not be known at all.
>
>I better understand the costs of a surrogate key (maintianence, an extra
>index, etc.), and I am prepared to pay them. I will reconsider how these
>surrogate keys are generated.
-- Bernard Peek bap_at_shrdlu.com www.diversebooks.com: SF & Computing book reviews and more..... In search of cognoscentiReceived on Mon Jan 20 2003 - 13:01:25 CET