Height measurement errors
I've noticed some issues with height measurements for a brief period of time. This is improtant as BMI is calculated incorrectly and it affects calculated risks based on BMI.
I would hypothesise that the measurement device was incorrectly positioned on the wall at a certain centre and the recorded height was about 2 inches too small.
Comments
6 comments
I assume you are looking at field 50. The Notes say “Standing height was measured using a Seca 202 device”. See https://www.seca.com/fileadmin/documents/manual/seca_man_201_int.pdf .
The data does look a bit worrying, especially as the same pattern is in both male and female, but I think it is unlikely that anyone would have set it up 2 inches wrong (or even 2 cm), and if they had done so then I think when it was noticed there would have been corrections made, or at least a warning added.
I notice that the month of concern has a lot fewer data points than the following months, only about a third. I am hoping that what we see is random fluctuation. Notice that the last month on the graph for Males is also much smaller than the overall average, and that probably also has fewer data points. You could maybe do some stats to check how unlikely this would have been.
If you are still concerned, it might be helpful to look at the participants measured in April 2007, find any that have returned for repeat visits (i1, i2, i3) and compare these i0 heights with the repeat values on an individual basis. (Participants will naturally shrink as they age, but if we find a set that appear to be getting taller, then we know we have a problem.)
Field 39 Height measure device id might also be useful. I expect that more than one Seca 202 device was being used. If so, and if just one were set up incorrectly, that should show in the data.
Please let us know if you find more evidence that it is a real issue.
It is definitely a real issue. It shows up in my risk models for diabetes wrt to BMI. The issue seems to be in Mancheter There are fewer data points because there is only one testing centre and the first after the trial, so having it set up wrong would be an easy mistake to make (I'm not sure how you would notice it). I don't think there is anyway it could be random fluctuation given that there is no observed difference in the weight. Note:The last month on the first plot has very few data points but the males here also appear to weigh less as well.
I agree that there does seem to be something wrong at the start of the Manchester visits. I still think it is unlikely that this would be due to badly-set-up Seca instruments: looking only at the Manchester data, the problem seems to be concentrated in the first few weeks (though there may also be an ongoing smaller error). It doesn't seem to vary with instrument, nor with staff member. I'm currently taking my own advice and looking at the participants with i2 visits, and the differences in measured heights, but it is complicated by the fact that all members of the cohort are shrinking, so I will need to take account of the time between visits.
I think it's pretty conclusive there was a 2cm under measurement. Hence no one should use the given BMI for these people.
Hi,
I'm just wondering if there has been any investigation of this height issue and you can share any light on how it actually happened?
Also if it has been corrected could you let me know or my own code will double correct it.
Thanks in advance
Hi Philip,
I don't have any useful update, sorry. There has been some further investigation, but nothing conclusive.
It has definitely not been corrected yet, and if any correction is applied it will not be ready in time for the Q1 2025 update, ie UKB-RAP v19.
I will post again here when there is any news.
Please sign in to leave a comment.