Skip to content
SearchProgress
Display

Display

Theme
Density
Text size

Sign in

Add your earlier progress to your account?

Study progress is waiting to be saved

How well a method performs

14 min

  • Calculate a sigma metric with compatible error, bias, and imprecision units
  • Tell method agreement from a high correlation coefficient
  • Test proficiency samples through the routine patient workflow

Read the full reference

Try first

Try first

At one control concentration a method has a CV of 2.5% and a bias of −1.5%. The laboratory's allowable total error is 10%. What is the sigma metric?

The next section explains it.

The next section explains it.

Right. The next section explains why.

The next section explains it.

Get the idea

Daily QC shows whether a method is behaving today. Three other tools show whether the method is good enough for its purpose and whether it agrees with other methods and laboratories.

The sigma metric

The sigma metric counts how many CVs of the method's imprecision fit into the allowance left after bias:

sigma = (allowable total error % − |bias %|) ÷ CV %

  • Bias is taken as a size, whatever its sign, because a method reading 3% low wastes as much of the allowance as one reading 3% high.1
  • All three numbers must be in the same form. Convert an SD to a CV with CV = SD ÷ mean × 100, and a bias in concentration units to a percent of the same mean, or keep all three in concentration units.
  • The metric belongs to one concentration and one choice of allowable error, and it is one input to choosing QC rules.

Read the metric this way:1

Sigma metricWhat it means for QC
Near 6Room to spare
Near 3Tighter QC rules, and more of them

Method agreement

Method agreement comes up when a laboratory compares a new method with the one it replaces. Paired patient specimens are run on both, and the correlation coefficient, r, is often the first number printed.

  • A high r says only that the pairs fall close to a straight line.
  • Two methods can sit on a near-perfect line with one reading consistently higher, so r mostly tells you whether the specimens covered a wide enough range to fit the line well.2
  • Agreement is judged from the bias at the concentrations where clinical decisions are made, compared with the bias the laboratory decided in advance it could accept.2,3

Proficiency testing

Proficiency testing (PT) sends blinded samples from an outside program, and the laboratory's answers are graded against a target or a peer group. The point is to see the laboratory's routine error, so the samples go through the routine workflow. Under CLIA they are tested:4

  • with the regular patient workload
  • by the people who usually run the test
  • with the usual method
  • the same number of times as a patient specimen

Laboratories must not:4

  • discuss PT results with each other before the reporting deadline
  • send PT samples to another laboratory for testing they are certified to do themselves

Special handling makes the grade describe that handling and hides the error in patients' results.

References
  1. Westgard QC. Best practices for Westgard rules. Accessed September 23, 2026.
  2. Westgard JO. The comparison of methods experiment. Westgard QC. Accessed September 23, 2026.
  3. Clinical and Laboratory Standards Institute. Measurement Procedure Comparison and Bias Estimation Using Patient Samples. 3rd ed. CLSI guideline EP09c. Clinical and Laboratory Standards Institute; 2018. Accessed September 23, 2026.
  4. Electronic Code of Federal Regulations. 42 CFR §493.801: Condition: Enrollment and testing of samples. Accessed September 23, 2026.

Watch one

A cholesterol control has a mean of 200 mg/dL and an SD of 4.0 mg/dL.

  • A peer comparison shows the method reads 6 mg/dL low at that level.
  • The laboratory's allowable total error is 10%.

What is the sigma metric?

  1. Convert the SD to a CV: 4.0 ÷ 200 × 100 = 2.0%.

    Sigma needs imprecision as a percent when the allowable error is a percent.

  2. Express the bias as a percent of the mean: 6 ÷ 200 × 100 = 3.0%.

    Bias must be in the same form as the other two numbers.

  3. Take the size of the bias, 3.0%, and ignore that the method reads low.

    A low bias uses up the allowance just as a high one does.

  4. Calculate (10 − 3.0) ÷ 2.0 = 7.0 ÷ 2.0 = 3.5.

    What is left of the allowance, divided by the CV, is the number of CVs that fit.

  5. Check in mg/dL: the allowance is 20 mg/dL, and (20 − 6) ÷ 4.0 = 3.5.

    Working in concentration units must give the same answer if the units match.

The sigma metric is 3.5 at 200 mg/dL.

Your turn

Problem 1 of 3

A procedure has CV 2%, bias −2%, and a laboratory-adopted allowable total error of 10%. What is its sigma metric at this concentration?

Correct. Use the magnitude of bias: (10 − 2) ÷ 2 = 4.

Incorrect. (10 − (−2)) ÷ 2 incorrectly treats negative bias as extra allowable error. Either bias direction consumes the allowance.

Incorrect. 10 − 2 gives the remaining allowance in percentage points. Divide it by CV 2% to obtain sigma 4.

Hint
  1. Bias in either direction uses up part of the allowable error.
  2. Use the size of the bias, 2%, and subtract it from 10%.
  3. Then divide what is left by the CV.

Review Sigma metric and matching units

Problem 2 of 3

A laboratory compares a new creatinine method with its current one on 40 patient specimens. The correlation coefficient is 0.998. Across the range, the new method reads about 0.3 mg/dL higher. Before the study the laboratory set an allowable bias of 0.1 mg/dL at 1.0 mg/dL. Can the new method replace the old one without further action?

An r of 0.998 says the pairs lie close to a line. A line that runs 0.3 mg/dL above the old results still has them lying close to it.

Called methods interchangeable because correlation was high

A correlation coefficient measures how closely paired results follow a line. In one calcium comparison, r = 0.9997 sits beside a bias of +0.15 mg/dL at 10.5 mg/dL. Calling the methods interchangeable on r alone lets that bias reach patient results when the laboratory switches methods.

Agreement is judged from the bias at decision concentrations against the limit set in advance, and this bias is well beyond it.

Forty specimens with a consistent offset already answer the question. More pairs would not remove a bias three times the allowed limit.

Hint
  1. Ask what r measures: closeness to a line, or closeness of the results.
  2. Compare the bias you see with the bias the laboratory decided it could accept.

Review Comparing a candidate method

Problem 3 of 3

A chemistry PT event arrives on a busy Tuesday. Which way of handling the samples fits the rules?

That is the routine workflow, so the grade reflects the error that patients' results carry.

Extra repeats and a hand-picked analyst describe special effort. PT samples are tested the same number of times as patient specimens by the people who usually do the work.

Gave proficiency material a special workflow

Running PT samples in replicate, assigning them to the most experienced person, or checking with another laboratory makes the result describe special effort. CLIA requires the routine personnel, method, and number of repeats and forbids referral or interlaboratory communication before the deadline, so special handling risks CLIA sanctions and hides the errors that routine testing makes.

Talking with another laboratory about PT results before the reporting deadline is not allowed, even when the testing itself was routine.

Gave proficiency material a special workflow

Running PT samples in replicate, assigning them to the most experienced person, or checking with another laboratory makes the result describe special effort. CLIA requires the routine personnel, method, and number of repeats and forbids referral or interlaboratory communication before the deadline, so special handling risks CLIA sanctions and hides the errors that routine testing makes.

A PT sample must not go to another laboratory for testing the laboratory is certified to perform itself.

Review Proficiency testing and external quality assessment

Keep

Sources checked