This is one of my pet peeves. It started for me back in the 1970’s when the Intensive Care Units where I was working were evaluating thermal dilution cardiac output meters. This was at a time when digital displays were just starting to become common. One of the meters showed cardiac output with two digits after the decimal point (i.e. 0.12) and the other one had three digits after the decimal point (i.e. 0.123).

Thermal dilution cardiac output works by threading a catheter (a Swan-Ganz is what was used at the time) with a thermistor in its tip through the right side of a patient’s heart into their pulmonary artery. A small amount of iced saline is then injected through the catheter and the system times how long it takes for this cold pulse to go around the patient’s body and return. There are a number of uncertainties involved so it’s not a terribly accurate technique and the very best you could ever expect would be a precision of about 1/10th of a LPM and that’s being very generous.

We had no ability to actually determine if either meter was accurate and the best we could see was that both meters gave similar results on the same patient. Neither meter was particularly harder or easier to use than the other. Nevertheless the cardiac output meter with 3 digits after the decimal point won the evaluation hands down because everybody said it had to be more accurate. This may say something about human nature but it’s also just nonsense. Simply because somebody places extra digits after the decimal point doesn’t make the measurement more accurate.

I’ve seen many times where a test result is reported with more digits after the decimal point than you could reasonably expect to get from the equipment or the measurement. When new devices (software or smartphones for example) add new features that are of little value (other than probably as ad copy) it is called feature creep. When this happens with digits I think this should be called decimal creep.

When I review our PFT reports I see a number of results that are reported with too many digits. The one that bothers me the most is DLCO. I’ve written about a number of the problems we’ve encountered with DLCO measurements and in particular analyzer offset and gain. The manufacturer of our equipment claims the analyzer has an accuracy of 1%. Assuming that this means 1% of full scale (and not 1% at any reading) and plugging more or less normal values into the DLCO calculation, a single 1% error in either CO or the trace exhaled gas concentrations causes an error of approximately 0.8% in DLCO. A 1% error in both the CO and trace exhaled gas concentrations causes up to a 2.5% error in DLCO. A 1% error in both inhaled and exhaled gas concentrations can cause an error of up to 13% error in DLCO.

What this means is that even a single 1% error in any of the DLCO gas concentrations (let alone the inspired volume or the breath-holding time) will cause an error of approximately 0.3 ml/min/mmHg and therefore its accuracy can’t be any better than that. So why is DLCO reported out to two digits after the decimal point?

Similarly, our test systems report MIP and MEP with two digits after the decimal point. Our equipment manufacturer doesn’t provide any specifications for the pressure transducer but that doesn’t really matter. We calibrate our mouth pressures using a U-tube manometer that uses colored oil rather than water. The surface tension of the oil is less than water so in some ways its height is easier to read than water would be (due to a flatter meniscus) but even so, the scale lines are 0.2 cm H2O water pressure apart and you have to eyeball it carefully to get it to the pressure used to calibrate the transducer (10 cm H2O). I’ve calibrated transducers using U-tube manometers for years and don’t think that I can get it any closer than +/- 0.1 cm H2O. Unless you’re with the National Bureau of Standards I don’t see how it’s possible to calibrate a transducer with an accuracy of 0.01 cm H2O so again, why are results reported with two digits after the decimal point?

I will mention in passing that our lab’s database stores results as single-precision numbers which means that volumes and gas concentrations are stored out to at least five digits after the decimal point (i.e., 0.00001 L). This may be a reflection of the output from the test system’s A/D converters, but really, five digits?

There is at least one result however, that I think is reported with too few digits after the decimal point and that is the FEV1/FVC ratio. Our software reports both the observed and the predicted values as integers, i.e. with no digits after the decimal point. I realize that it is traditional to report the FEV1/FVC ratio as a percent, but to me it is a ratio and for this reason I’d like to see at least one digit after the decimal point. My concern is that I don’t know when the FEV1/FVC ratio is being rounded up or rounded down. Realistically this doesn’t matter for more than a small handful of patients but when you have somebody that is literally on the borderline between normal and abnormal it would help to make an informed decision.

One final peeve is that too often I see staff members both within and without the PFT Lab report decimal fractions without the leading zero (i.e., .78 rather than 0.78). Not only does this make it harder to read but it doesn’t meet the Joint Commission standards for clarity in reporting results. I can correct the lab’s technicians when I see them do this, but when it’s somebody outside the lab all I can do is let it pass.

This problem could be solved (at least to my satisfaction) if our reporting software allowed us to adjust the number of digits after the decimal point that appear on reports, but it doesn’t.  Decimal points are preset and cannot be altered.  We’re stuck with it at the moment and the best I can do is to add this to my list of gripes about our reporting software.

We need to get real about our decimal points and acknowledge that there are limits to the accuracy of our test measurements. Worrying about this may be a bit OCD-ish on my part but I think the number of digits after the decimal point clearly implies the level of accuracy of the result and too many digits after a decimal point gives a sense of accuracy and precision that just isn’t there. Let’s stop these creepy decimals!

PFT Blog by Richard Johnston is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.