#210: Query by BMI percentile among children.
-------------------------------------------------+-------------------------
Reporter: bokov | Owner: gkowalski
Type: enhancement | Status: assigned
Priority: major | Milestone: obesity-
Component: data-stds | survey-def
Keywords: BMI pediatric-cohort obesity-cohort | Resolution:
Blocking: 252 | Blocked By:
-------------------------------------------------+-------------------------
Comment (by huhickman):
I've been mulling over these calculations - and think we have a problem
here. The code posted here computes BMI percentiles based upon values in
the blueherondata observation_fact table. These dates are '''after'''
adjustment. If you consider a child of 36 months (real) age and subtract
the date adjustment from their birthdate - then all of sudden they are
'younger' on the percentile curve and hence will tend to land in a bucket
that could be higher than it should be.
To prove my point, I ran our calculations based upon the real dates and
then on the post-adjusted dates.
The numbers I came up with for UNMC are as follows:
||= PERCENTILE =||= Adjusted Dates =||= Real Dates =||= Diff =||
|| 3|| 4067|| 4298|| -231||
|| 5|| 1490|| 1615|| -125||
|| 10|| 3550|| 3781|| -231||
|| 25|| 11325|| 11880|| -555||
|| 50|| 23932|| 25027|| -1095||
|| 75|| 34681|| 35004|| -323||
|| 85|| 19480|| 19318|| 162||
|| 90|| 12283|| 12043|| 240||
|| 95|| 16345|| 15771|| 574||
|| 97|| 9722|| 9217|| 505||
|| 100|| 26816|| 24466|| 2350||
The skew is a noticeable shift up the BMI percentile scale - fewer
readings in the lower percentile ranges and more in the upper percentiles.
I haven't done it but if I broke down my counts based upon patient age, I
suspect the skew among the younger patients would be more pronounced than
those who tend toward the higher age range.
Hence, I think our BMI Percentiles lookups should be computed off the real
dates, and not adjusted dates.
--
Ticket URL:
<http://informatics.gpcnetwork.org/trac/Project/ticket/210#comment:24>
gpc-informatics <http://informatics.gpcnetwork.org/>
Greater Plains Network - Informatics
_______________________________________________
Gpc-dev mailing list
[email protected]
http://listserv.kumc.edu/mailman/listinfo/gpc-dev
