[Gnumed-devel] Re: Lab import fundamentals

[Jim Busser]

(replying back onto the list, on account of this thread nearing maturity)

On 2010-08-28, at 6:36 PM, Karsten Hilbert wrote:

>> It
>> only becomes a problem if multiple labs would each misspell many people's
>> names *multiple* different ways.
> 
> Oh, they will.

I agree that patients can easily provide each praxis, and each lab, with a 
slightly different name.

The GNUmed importer that is being developed will accept name variants, and 
these will boil down potential matches on a combination of

        date of birth
        surname
        some combination of "first names portions".

Where the above matches resolve down to a single person (who may have more than 
one matching name) already in the praxis, GNUmed will accept that match as the 
person against whom to link the results.

Each praxis can define its own "equipoise" with respect to lab-generated match 
difficulties... very likely different labs can spawn the same variants on 
patient names. As soon as one of these variants is accepted into GNUmed, the 
name-alias will permit results from any lab that would use this variant to 
match.

Problems will become evident as soon as any one praxis observes excess 
(frequent) patient auto-creations that must be merged. As soon as a praxis 
identifies patients who have

        the same date of birth

        AND
        the same surname
        the same first or second name (maybe even just initials)
        the same sex 

on any alias (as Liz suggests will occur in very large praxes), then the 
importer will smartly refuse to resolve what it cannot resolve uniquely, and 
will instead auto-create a patient and let the praxis decide which patient to 
merge the results under. The interesting thing here is the possibility of two 
persons in a single praxis, where both persons have

        date of birth = July 1, 1960
        name = Smith, Mary Jane
        sex = F

in which case the merging of the results into the correct person need not 
bother to create another alias "Smith, Mary Jane", because that would be 
redundant.

Where the above proves at a praxis level to be a frequent event, it would 
constitute grounds to pressure the labs to begin using some other presumably 
unique person identifier for the lab results, such as an "identity 
issuer-unique number" combo typified by public and private health insurance 
issuers. This we do already in BC (CA) using the "Personal Health Number".

My only other suggestion is that variability in a date of birth will denote 
either a mis-entered version of a single truth, or competing versions of 
"truth". Both are probably worth to "settle" at the praxis-lab administrative 
level. 

-- Jim