Measurement Uncertainty 101

Understanding and calculating uncertainty in the clinical laboratory

Whether you are measuring an object with a ruler, weighing items on a scale, or running complex assays, you will never be able to achieve exactly the same measurement every time. Each time something is measured, the result will vary slightly, and it is therefore impossible to attain a single true measurement value.

Learn more about data management

Why do we calculate measurement uncertainty?

In clinical laboratories, measurements need to be as accurate as possible to facilitate high quality patient care. When test results directly affect patient treatment, it is important to understand exactly what those results mean. Because it is impossible to conclusively achieve one specific result for any given measurement, clinicians need to understand how much potential error exists. This way, they can accurately interpret results and determine if a patient’s condition has changed significantly or is outside of clinical decision limits. This is why we calculate the doubt, or uncertainty, of a result.

Uncertainty: A calculable range that provides context for the accuracy of a result. The true value of the result is expected to lie within that range.

For example: A vial weighed on a scale measures 10.2 ml, but depending on relevant variables like scale sensitivity and precision, the result could actually be 10.2 ± 0.1 ml. This is the calculated uncertainty range for this measurement.

For quantitative diagnostic tests, it is important to calculate this amount of uncertainty. Quantifying the doubt that exists for a measured value provides an estimate of that result’s accuracy. Unity Real Time provides three built-in uncertainty formulas to help you easily calculate measurement uncertainty and manage a comprehensive result database.

“The laboratory shall determine measurement uncertainty for each measurement procedure in the examination phase used to report measured quantity values on patients’ samples.”

–ISO 15189:2012 5.5.1.4 Measurement uncertainty of measured quantity values

For the purposes of this document, we will be using the ISO 15189:2012 standard to discuss measurement uncertainty. ISO 15189 specifies that laboratories should calculate uncertainty values for quantitative tests. It is important to understand the basic principles of measurement uncertainty in order to ensure the laboratory is in compliance with the relevant regulations or policies.

The basics of measurement uncertainty

So what exactly is measurement uncertainty, and where does it come from?

Uncertainty exists because, no matter how carefully assays are controlled or instruments are maintained, there will always be variation in a measurement process. When running tests in a laboratory, there are many different variables that can influence instrument performance. Factors like sample storage and handling, environmental conditions, operator changes, and calibrator conditions can all affect assay results. These factors are all sources of uncertainty.

img1

There are many sources of uncertainty, and while attempts should be made to control them when possible, assay results will inevitably still vary. This is unavoidable, as variation and error are inherently involved in any measuring process. Defining and calculating uncertainty ranges provides a useful context for understanding how much variation we are working with.

Uncertainty sources exist in the pre- and post- analytical phases, but those sources are often difficult to identify and quantify. According to the ISO 15189 standards, uncertainty calculations only need to take into account uncertainty sources during the analytical phase, where the measurement actually occurs.

“The relevant uncertainty components are those associated with the actual measurement process, commencing with the presentation of the sample to the measurement procedure and ending with the output of the measured value.”

–ISO 15189 5.5.1.4

When we calculate uncertainty, we are attempting to evaluate error.

Error is any deviation in a measured value from the true value. Because this “true value” is impossible to precisely and conclusively determine, we can instead calculate an uncertainty range to estimate an interval of values within which the true value probably lies.

There are two types of error involved in the measuring process: random and systematic error.

Error Deviation of a measured value from the true value. Since it is impossible to determine the true value, error is an abstract concept that cannot be quantified

Even if every possible aspect of a measurement process was controlled, measurements would still vary slightly due to random error. However, random error can be addressed through repeated measurements. If a specific measurement is repeated enough times, averaging the results will minimize the influence of random error and bring the measured value closer to the true value.

Systematic error reflects an underlying issue with a consistent effect on the measurement process. It can result from irregularities in instruments, assay materials, environmental characteristics, or other factors. Because systematic error has a constant effect on the overall process, it cannot be minimized with repeated measurements. Instead, corrections are applied as needed once the influence of systematic error is identified. These corrections can be derived from information like reference values or established calibrator uncertainties.

Why is the true value impossible to determine?True values are impossible to identify because results will always vary upon repeated measurements, making it impossible to conclusively measure one single value. This is because there are many different factors that influence the measurement process (ex: tiny fluctuations in instrument performance, shifting environmental conditions, operator changes,etc.). It is impossible to completely control every factor that affects a measurement. Even if a method is extremely precise, the results will always involve a certain amount of error. Therefore, the true value of a measurement is an abstract, unmeasurable value.

How can we calculate uncertainty?

When calculating uncertainty values, both random and systematic error should be taken into account. Random error plays a significant role in affecting measurements, but when systematic error is identifiable, it should also be factored in.

We can measure random error with imprecision, represented by standard deviation. We can measure systematic error with bias.

Imprecision: How much each result differs from the other results.

Bias How much the result mean differs from a reference mean.

Uncertainty is often calculated by evaluating the standard deviation of measurement data over time, and other values (like bias estimates) can be included in the calculation when applicable.

It is important to use data collected over an extended period of time in order to account for as many uncertainty sources as possible. If uncertainty is calculated using data from a single week, for example, factors like instrument maintenance, shifting environmental properties, or operator changes might not be properly taken into account.

img 2
“Measurement uncertainties may be calculated using quantity values obtained by the measurement of quality control materials under intermediate precision conditions that include as many routine changes as reasonably possible in the standard operation of a measurement procedure (e.g. changes of reagent and calibrator batches, different operators, scheduled instrument maintenance, etc.).”

–ISO 15189 5.5.1.4

Straightforward uncertainty calculations with Unity Real Time

Measurement uncertainty can be a confusing concept, but Unity Real Time’s simple preset calculations help your laboratory easily calculate uncertainty values. For ease of use and convenience, Unity Real Time provides three calculation methods consistent with various requirements and recommendations.

img3

Measurement uncertainty values evolve constantly, and laboratories should routinely review measurement uncertainty values and establish an appropriate frequency for calculating uncertainty for each assay.

Example of a UnityReal Time screen showing expanded uncertainty with calibration uncertainty

The measurement uncertainty value is provided both as an absolute value and as a percentage

img4 Unity screen

When Unity calculates uncertainty, all available data for a given assay and lot is automatically taken into account, but users can select a specific period of time to evaluate if desired. The uncertainty interval is plus or minus the calculated uncertainty value, which is displayed in both absolute value and percentage.

What is expanded uncertainty?

Expanded uncertainty refers to uncertainty calculations that have a higher confidence level. Since most uncertainty formulas are based on standard deviation, a basic uncertainty value would represent the potential range of error plus or minus one standard deviation. According to the rules of normal distribution, approximately 68% of results lie within one standard deviation of the mean. This means that, for a basic uncertainty calculation, there is a 68% chance that the true value lies within that uncertainty range.

In order to increase the confidence interval, expanded uncertainty values are calculated so that their range comprises plus or minus two standard deviations. This means that there is a 95% chance the true value lies within that range. To achieve this higher confidence level, expanded uncertainty calculations are simply multiplied by 2.

img5 screen


“The laboratory shall define the performance requirements for the measurement uncertainty of each measurement procedure and regularly review estimates of measurement uncertainty.”

–ISO 15189 5.5.1.4

Applying measurement uncertainty

Clinical laboratories typically calculate and keep records of their measurement uncertainty, providing uncertainty values to physicians and auditors upon request.

Physicians can use uncertainty ranges to make more informed treatment decisions. For example, when a result is near a clinical decision value, physicians can use uncertainty intervals to determine whether or not a result is definitively lower or higher than that decision value. Similarly, if a result is close to a previous patient result, evaluating those results in the context of their uncertainty intervals can help determine if they are significantly different enough to warrant a change in treatment or diagnosis. Taking all of this information into account can help a physician confidently make the appropriate patient care conclusions.

A calculable range of uncertainty exists for any measurement, and once we understand how to calculate that range, we can accurately represent the laboratory’s confidence level for any given result. Calculating uncertainty is only a few clicks away with Unity Real Time, which provides three built-in uncertainty formulas for quick and convenient calculations. In addition to helping laboratories comply with regulations, logging uncertainty values also demonstrates a commitment towards a higher standard of laboratory quality with careful and thorough result records.

“Examples of the practical utility of measurement uncertainty estimates might include confirmation that patients’ values meet quality goals set by the laboratory and meaningful comparison of a patient value with a previous value of the same type or with a clinical decision value.”

–ISO 15189 5.5.1.4

Reference

Further reading

 

Find Products & Information

Log In / Register

Log In / Register
PLEASE LOG IN NOT REGISTERED?
Login ID
Password