Blood work: how do I normalize values over time between different labs?

November 30, 2013 9:29 AM

How do I normalize similar values from multiple medical tests with different reference ranges?

I have around 20 lab reports for my blood from the past 10 years. I want to centralize the more important values, like testosterone and good/bad cholesterol, in a spreadsheet so I can chart my progress over time. However, the ranges for each test differ between the labs (sometimes w/in the same lab). The beginning and ending values of each range are often different, as well as the total range.

My idea was to:
1. subtract starting value from recorded value
2. subtract starting value from end value
3. divide 1 by 2, giving me a percentage X

Ex: Total Cholesterol: 160, from 125-200:
1. 160 - 125 = 35
2. 200 - 125 = 75
3. 35/75 = 47%

And then (finally) I can compare total cholesterol across all my labs.

My questions (for any medical-minded people out there):
1. Does this make sense? Is this a fool's errand and I should just go by the most recent discussion with my GP?
2. If the tests are measuring the same total value (mg/dL), do I just disregard the ranges?
3. can I do this for all my values, or is the distribution not linear for some tests? (ala, for some tests, it only matters if you fall on the straight part of the hockey stick, and not the curved end)
4. When I only have one reference number, like (>= 40), and that changes, am I out of luck?

Thanks for your help-
cgs

posted by cgs to Health & Fitness (7 answers total)

As long as the results are reported in the same units, you can compare the values directly without any kind of conversion. The reference range reported is typically generated by the lab based on a large sample of all of the tests they've run -- it has nothing to do with you specifically (and sometimes it doesn't even reflect a good "normal range", depending on the population utilizing the lab).
posted by telegraph at 9:42 AM on November 30, 2013 [3 favorites]

I second that there's no reason for conversion factors. Diagnostic blood testing is not all cut and dry for so many analytes - hence the great variation in 'normal' reference ranges and the pretty large range that is considered 'normal' on most tests.

Before making any grand conclusions from the data over time, be sure to check with your doctor on them. A big example on this that I can think of is Prostate Specific Antigen. Granted, it's not the best predictive test, but another factor is that different 'brands' of tests will give slightly different results on the same person. The brand used for testing is not always provided to the patient. Trying to conclude anything from a small change (especially in the 'normal' or 'borderline' result area) would be impossible and inappropriate.

There are other factors besides 'brand' that can also cause variation. Assume a lot of the variation that you see in most analytes is not due to your own overall levels varying, but to normal variation in the test procedure itself.
posted by Tandem Affinity at 9:54 AM on November 30, 2013

More info and a fun graphic

see especially the "Variability" section
posted by Tandem Affinity at 9:57 AM on November 30, 2013

I would use a bar & whisker chart for this sort of situation, but second the idea that you get a physician to help you properly contextualize what you come up with.
posted by Emperor SnooKloze at 10:22 AM on November 30, 2013

As long as the results are reported in the same units,

using the same test run on the same type of machine calibrated in exactly the same way...

Absolute units may be directly comparable between labs but they also may not. Also, for many health parameters being within the normal range is what counts, being near the top or bottom etc doesn't necessarily mean much as long as you're within the range (either due to poor correlation between test output and health or due to variation within the test or etc etc). Or the variation does mean something about your health but it's specific to you, so e.g. while you're near the bottom of the 'normal' range for general population that specific thing may still be too high for you. Or, for some tests, there is a direct health-correlation to the result such that tracking changes over time is meaningful. Which situation fits is going to depend on the test, how and where it was done, your genetic and lifestyle background, and any other health things you have going on. One standard calculation for tracking stuff probably isn't going to cut it.

In other words, yeah discuss this with your doctor and get their guidance on what each test really means for you specifically, then decide what kind of tracking is relevant and appropriate with their input.
posted by shelleycat at 10:39 AM on November 30, 2013

Way before spreadsheets (electronic ones, anyway), I was keeping track of our family's medical tests like this:

Draw a scale line for one test, low figure for normal on left, high figure for normal range on right

Find center point and mark it, then do the same with each half, so you have quarter values

Now mark the point at which your current test result falls. (say, just below dead center)

A few years later, same test, different lab, different figures for range:

Do exactly the same as above, using the new range, quartered, and post your new number (say, just barely within normal range on the low side)

Compare the two scales.

Your xxxx study has dropped significantly in the last few years. Do you need to explore this? Find out more about the test - does the value vary dramatically depending on what time of day the blood is drawn or depending on when you ate last or depending on some other variable factor, or should the two values have stayed in the same general vicinity on the scale, thus indicating a possible need to follow up on this?

From my own experience, expect some variation in all the tests, say a quarter range from one time to another. It's good information to have, especially as the years go by, and when some unwelcome change shows up one day you'll notice it. Or, alternatively, running consistently at the very top or very bottom of "normal" range could be significant after all.

I've no idea at all how one would go about doing this on a computer, however.
posted by aryma at 11:13 AM on November 30, 2013

I'm wondering if the 'norm' values have anything to do with time -- for instance, originally it was considered 'normal'/desirable to have total serum cholesterol under 220 (maybe even 240 can't remember) -- but today's value is < 200. Those things could be tracked on the same curve -- and I'm betting that's the case for most of the values.
posted by MeiraV at 4:44 PM on November 30, 2013