== Processing the Patient Identifer Set during a Load. ==

This is work in progress.

=== Cardinality, with some examples. ===

==== __Example One__ ====
First of all, this is wrong. Every pid must have a patient_id.
{{{
<pid>
   <patient_map_id source="BRICCS">BPt00000040</patient_map_id>
   <patient_map_id source="UHLT">Snnnnnnn</patient_map_id>
</pid>
}}}

==== __Example Two__ ====
This is acceptable:
{{{
<pid>
   <patient_id source="BRICCS">BPt00000040</patient_id>
   <patient_map_id source="UHLT">Snnnnnnn</patient_map_id>
</pid>
}}}

==== __Example Three__ ====
The following example is also acceptable, but implies we know the participant is either:
 * already within the CRC (and that we know the internal identifier = 2) or:
 * this is a new participant and we are ourselves assigning a new i2b2 internal identifier (= 2) for them. 
Both situations are ones we can avoid by adopting the approach of Example Two above, and omitting the HIVE as a source.
{{{
<pid>
   <patient_id source="HIVE">2</patient_id>
   <patient_map_id source="BRICCS">BPt00000040</patient_map_id>
   <patient_map_id source="UHLT">Snnnnnnn</patient_map_id>
</pid>
}}} 

==== __Comment__ ====
As far as I can tell, a row in the temporary table covers a patient_id / patient_map_id combination. So:
 * Example Two would give rise to one row.
 * Example Three would give rise to two rows.

=== First Stage: Eliminate Duplicates. ===
Any "duplicates" are eliminated from the temporary table.
A duplicate is one where another row matches on:
 1. patient_id
 1. patient_id source
 1. patient_map_id
 1. patient_map_id source

=== Second Stage: Process HIVE as a Source. ===

=== Third Stage: Not using HIVE as a Source. ===

[[BackLinks]]