The linked definition doesn't preclude redundant entries resulting from
multiple encounters taking place on the same day (and these are not
rare, at least at the two GPC sites I've worked with so far). This is
deeper than terminology. The pragmatic question is: what should be in
each row of the analysis-ready result-set: patient, patient-encounter,
or patient-day?
Perhaps a study that delves deeply into the timing and ordering of
individual specialist visits would treat each encounter as a separate
collection of observations, even if those encounters were on the same
day. Presumably someone with enough domain knowledge to do a study like
that will also know how to resolve redundancies and discrepancies. I don't.
If the research question is more general, like "is vital sign X
different between patients with diagnosis A and diagnosis B?" or "does
the prevalence of diagnosis Y increase in patients from demographic
group C compared to patients from demographic group D?", I would start
to worry about whether I sufficiently understand if it's possible and
what it means for A, B, C, D, X, and Y to show up on /multiple
encounters that are really administrative abstractions representing the
same physical visit to the clinic or hospital/.
I too want to hear from clinicians about any universal rules governing
uniqueness of observations from different encounters on the same day.
Until then, when preparing data for analysis, I play it safe by
collapsing visits into one row per patient-day. Codes of a given scheme
(after being filtered for any applicable modifiers) get concatenated
into comma- or semicolon-delimited strings (resulting in one column each
for distinct diagnoses, procedure orders, medications, etc.) /except/
specific codes that the researcher has singled out for individual
attention. Each of those gets a separate column, however many columns
that takes. Numeric values (e.g. each type of lab result) also get
individual columns.
I'm not advocating that the definition of encounter should or shouldn't
be changed. Only that encounters as our systems define them aren't
always the right thing to count for power analysis or feasibility studies.
On 09/29/2014 09:04 AM, Dan Connolly wrote:
I asked for clarification on this issue and haven't gotten any. Would somebody
please give a specific technical reason why, for example, this Wikipedia
definition of encounter doesn't suffice?
Within the medical record, individual medical encounters are marked by
discrete summations of a patient's medical history by a physician, nurse
practitioner, or physician assistant and can take several forms.
-- http://en.wikipedia.org/wiki/Medical_record#Medical_encounters
_______________________________________________
Gpc-dev mailing list
Gpc-dev@listserv.kumc.edu
http://listserv.kumc.edu/mailman/listinfo/gpc-dev
