Tools, Technologies and Training for Healthcare Laboratories

Guidance for Improving Statistical QC Frequency Strategy for Multitest Analyzers

The body of literature on QC Frequency continues to grow. Change is coming, get ready. It's time to think about QC Frequency.

Guidance for Improving Statistical QC Strategies - including QC Frequency - for Multitest Analyzers

James O. Westgard, PhD
November 2021

We are in an interesting situation concerning the adequacy of SQC procedures in medical laboratories, particularly here in the US. On one hand, there is evidence based on a survey of large academic medical centers conducted by Rosenbaum et al [1] that many laboratories embrace a practice of using 2 SD control limits. This is surprising given the known false rejection problems when using 2 SD limits, e.g., approximately 9% when 2 levels of controls are analyzed per run, as required by CLIA. This practice is defended as being a Repeat:2s sampling strategy [2], wherein controls are repeated, effectively employing a 2:2s rejection criterion that reduces the false rejections. This Repeat:2s strategy also increases the total number of controls analyzed, which is logistically difficult particularly for the high-volume continuous production analyzers that are the workhorses in large medical laboratories.

Meanwhile, recommendations for improving QC in the recent scientific literature seem to focus on Patient Based Real Time Quality Control procedures (PTRTQC) [2]. For example, July’s issue of Clinical Chemistry provides an article on “Average of patient deltas: Patient based quality control utilizing the mean with-patient analyte variation” [3]. This Average of Deltas approach (AoD) extends the use of delta checks by determining the average of individual patient deltas that are collected over a 20-28 hour period. Such intra-day repeats are limited to tests for frequent monitoring; those studied that provide good performance include bicarbonate, urea nitrogen, chloride, creatinine, glucose, magnesium, potassium, and sodium. For the best performing electrolytes, 10 to 20 intra-day repeats were sufficient.

Another example is an article on “Regression-adjusted real-time quality control” [RARTQC?] that has just been published online by Clinical Chemistry [4]. RARTQC, in contrast to PBRTQC, provides an additional regression step prior to the use of a common control algorithm, such as moving averages, etc. The approach is claimed to reduce the number of patient samples needed by 50% compared to other PBRTQC techniques.

For laboratories looking to improve their QC processes, these articles provide good examples of the emerging patient based QC approaches. They also provide a good source of references for other recent work on PBRTQC.

We also believe they demonstrate that these patient data approaches are complicated and not ready for prime time until manufacturers and middleware providers decide to provide the necessary support for optimization, validation, and implementation. Patient based QC is an important option for making improvements for those tests/methods with low Sigma quality, particularly Sigmas ≤ 3.5. But for tests/methods with Sigmas ˃ 3.5, traditional SQC procedures currently offer the best approach for providing practical risk-based SQC strategies.

To that end, we provide some guidance and support for those laboratories wishing to improve their SQC strategies for high-volume testing systems. The guidance comes in the form of a recommended process for developing an SQC strategy [6]; the support comes in the form of online QC Frequency calculators that perform the necessary calculations [7].

Process for Developing an SQC Strategy for a Multitest Analyzer

  1. Define desired run size in terms of number of patient samples between QC events or the desired reporting interval for a continuous production analyzer.
  2. Define quality requirement in form of allowable Total Error (TEa) at a medically important decision concentration.
  3. Determine intermediate imprecision. Also determine method bias if the laboratory has more than one method for performing the same tests and both/all methods are expected to achieve the same TEa requirement.
  4. Calculate Sigma-Metric as Sigma as TEa/SD or %TEa/%CV when laboratory has only a single method for performing the test. For tests with multiple methods, calculate Sigma as (TEa - |Bias|)/SD or as (%TEa - |%Bias|)/%CV.
  5. Identify candidate SQC procedures that would be practical for implementation and operation in your laboratory, (e.g., 1:3s, N=2; 1:3s/2:2s/R:4s multirule with N=2; 1:3s/2:2s/R:4s/4:1s multirule with N=4).
  6. Determine run sizes for the observed Sigma quality of your tests/methods.
  7. Sort the tests based on Sigma values high to low to identify those methods that satisfy your desired run size or reporting interval.
  8. If run sizes are not acceptable for a practical SQC strategy, then review TEa, consider whether to change quality requirement (minimal, desirable, optimum); if so, recalculate Sigma and sort again high to low, and assess acceptability of available run sizes. If acceptable, then stop.
  9. If run sizes are not acceptable, then review risk factors and consider whether those tests with too short run sizes can tolerate higher patient risk (increase to 2.0 to 5.0). If run sizes are acceptable, then stop.
  10. If run sizes are not acceptable, then consider whether performance of those methods can be improved; this could involve reducing bias between methods or improving precision by use of control material having more optimal concentration (medical decision concentration) or possibly by performing duplicate analyses. Recalculate Sigmas and repeat steps above.
  11. If run sizes are not acceptable, then consider whether QC procedures with higher numbers of control measurements (N) could be implemented, such as multirules with Ns of 3, 4, or 6. If so, determine run sizes for these higher N QC procedures and assess whether a practical SQC strategy is available.
  12. If run sizes are not acceptable, consider adding patient data QC algorithms. There are many recommendations for implementing Patient Based Real Time Quality Control (PBRTQC) procedures. Depending on the capabilities of your middleware and laboratory computer, it may be possible to implement such procedures. Prioritize your low Sigma methods and begin by assessing the population distribution and determining the ratio of the SD of the population to the SD of the analytical method, SDpop/SDmeth. See reference 5 for further guidance. Those low Sigma tests with low population/method ratios are the best candidates for PBRTQC. One design strategy is to match your desired run size, or reporting interval, to the number of patient samples required for the PBRTQC procedure.
  13. If the above approaches do not lead to an acceptable SQC strategy, it may be necessary to move those low Sigma testing processes to a different analytical system.
  14. If that is not possible, it will be necessary to recognize that those low Sigma tests may produce some erroneous results and it will be essential to provide more aggressive medical review of those test results to assess whether they fit with the clinical conditions of those patients.

Example applications

Please see references 6 and 7 for detailed examples showing how this planning process can be applied to multitest chemistry and enzyme analyzers. The Clinica Chimica Acta Journal has provided online access to these papers for a limited time.

Online Calculators

The QC Frequency calculators can be found at the following addresses:

Discussion/Directions about Online Calculators


  1. Rosenbaum MW, Flood JG, Melanson SEF, et al. Quality control practices for chemistry and immunochemistry in a cohort of 21 large academic medical centers. Am J Clin Pathol 2019;150:96-104.
  2. Parvin CA, Kuchipudi L, Yundt-Pacheco JC. Should I repeat my 1:2s QC rejection? Clin Chem 2012;58:925-29.
  3. Cembrowski GS, Xu Q, Cervinski MA. Average of patient deltas: Patient-based quality control utilizing the mean within-patient analyte variation. Clin Chem 2021;67:1019-1029.
  4. Duan X, Wang B, Zhu J, Zhang C, Jiang W, et al. Regression-adjusted real time quality control.
  5. Westgard JO, Smith FA, Mountain PJ, Boss S. Design and assessment of average of normal (AoN) patient data algorithms to maximize run lengths for automated process control. Clin Chem 1996;42:1683-8.
  6. Westgard SA, Bayat H, Westgard JO. A multi-test planning model for risk based statistical quality control strategies. Clin Chim Acta 2021; 523:216-223.
  7. Westgard JO, Bayat H, Westgard SA. Planning SQC strategies and adapting QC frequency for patient risk. Clin Chim Acta 2021; 523:1-5.