dhis2-devs team mailing list archive

[Bob Jolliffe <bobjolliffe@xxxxxxxxx>]

2010/2/1 Lars Helge Øverland <larshelge@xxxxxxxxx>

>
>
> 2010/2/1 Bob Jolliffe <bobjolliffe@xxxxxxxxx>
>
> you are taking about me?  What silly names do we have in mind?
>>
>
> Yes. Just kidding, no silly names this time.
>

Oh well.  I guess we can always try silly indentifiers instead :-)

We are going to implement system generated patient identifiers for the
> patient module.
>
> VN team has suggested:
>
> - Set of characters dependent on the organisation unit
> - Set of digits, include date and number of patients in the day.
>
> Viet has suggested an algorithm dependent on patient information and the
> time of creation.
>
> Jason has suggested an UUID.
>
> I think identifier for a person and identifier for a person's file are
subtley different.  Patients may already have a number of personal
state-issued identifiers (hence the flexible identifier type).  Many of
these might well provide the quality of uniqueness but not necessarily the
anonymity you would look for when for example tracking lab samples.  But it
seems what you are looking at is the generation of a file number for the
patient.  Like would be written on top of the cardboard folder in "real
life".  In which case I would be tempted to follow the simplicity of the VN
team suggestion.  Of course patients can migrate between org units which
needs to be taken into account.  If privacy is more of a concern then this
information can instead be hashed, probably along similar lines to Viet's
algorithm.

A few things to consider:
1.  there should be a check digit built in to the identifier
2.  regular expression validation can be useful
3.  if the identifier is meant to be human readable (and writable) as well
as machine readable then you don't want to go much over 8 characters
4.  before designing an identifier we should be very clear what the
identifier should be used for.  There are lots of examples of "scope creep"
where identifiers primarily designed for social security, tax, national
identification etc are "repurposed" as health identifiers.  Often they do
not have the required characteristics for this.  So maybe that's the
starting point - what exactly can we say this identifier is to be used for
and not used for.

The openmrs guys have some experience of issuing patient identifiers (and
printed barcodes) in Kenya.  We should look at what they have done.  I have
no idea if what they do represent s best practice or not but they do have
concrete experience to share.

A couple of broader references:
A discussion on the use of SSN which also contains some useful criteria to
frame thinking about identifier design:
http://epic.org/privacy/medical/hhs-id-798.html
A report commisioned in Ireland which specifically addresses the problem of
re-purposing:
http://www.hiqa.ie/media/pdfs/Unique_Health_Identifier_Report.pdf

That's all the thoughts I can muster for now.

Cheers
Bob