Introduction
Previous articles1,2,3,4 in this series on quality management for in-clinic laboratories have introduced the need for a total quality management system for in-clinic laboratory testing, quality planning, and a quality plan; discussed some aspects of facilities, instrumentation, health and safety, training, and improvement opportunities; discussed standard operating procedures; and considered various aspects of equipment or instrument maintenance and analytic performance assessment. The purpose of this final article is to provide additional information and examples regarding statistical QC for in-clinic laboratory testing and to expand on the quality-assurance concepts introduced in the fourth article4 of the series.
Laboratory error can occur in any phase of in-clinic laboratory testing, whether the preanalytic (from test ordering through sample transit and preparation), analytic (sample analysis), or postanalytic (from results reporting to interpretation) phase.1 Statistical QC refers to the use of statistical methods in the monitoring and maintenance of the quality of products and services.5 Within the context of veterinary clinical laboratory testing, statistical QC applies to the analytic phase. Traditionally, such QC has been used in laboratories to provide peace of mind that a laboratory system (ie, the instrument, reagents, and operator) is performing stably prior to testing patient samples. It can be applied to many of the testing systems within the laboratory and is commonly used for hematologic and biochemical analysis and dipstick or automated urinalysis.
Statistical QC is a long-standing practice in commercial diagnostic, university, and research laboratories because it provides evidence-based, numeric documentation of stable laboratory system performance and complements qualitative, nonstatistical approaches to QC that may include the use of skilled analysts, additional training of laboratory staff, concordance of measured data with clinical signs and other laboratory findings, and analysis of cohorts of data (eg, from individual patients over time or from all testing performed over specified periods).
This article will focus on the application of statistical QC by use of data collected from the analysis of commercial QCMs. Other approaches, such as repeated patient testing or the use of pooled patient samples, have been described6,7 but are beyond the scope of this article.
Reasons to Perform Statistical QC
It is difficult to look at any single test result and know whether that number or result is accurate and reliable. Statistical QC helps ensure that laboratory results can be used for diagnosis, identification of patients that may require further diagnostic investigation, and monitoring disease progression or response to treatment. For a laboratory result to be clinically useful, the measurement error needs to be smaller than the decision threshold used to determine whether the result is within reference limits or abnormal; that is, the result should be able to discriminate between health and disease.8 Once a result has been shown to be clinically useful, ongoing statistical QC helps to ensure that results are the product of a stable laboratory system that can be relied on for continued use. Statistical QC is but a single tool in a total quality management system.1 Adherence to statistical QC can ensure that the TEobs remains less than the TEa, which are concepts explained in the fourth article4 of this series. Statistical QC is a more advanced but also a more reliable way of monitoring instrument performance, compared with methods involving manufacturer-supplied target means and ranges.
Statistical QC is based on the principle that repeated measurements of a QCM by a stable testing system should yield a set of results with a Gaussian (normal or bell-shaped) distribution9 (Figure 1). Approximately 68% of results will fall within 1 SD of the mean, approximately 95% within 2 SDs, and almost all (99.7%) within 3 SDs.
Standard Gaussian (normal or bell-shaped) data distribution showing the percentage of data points that fall within 1, 2, or 3 SDs of the mean.
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
Standard Gaussian (normal or bell-shaped) data distribution showing the percentage of data points that fall within 1, 2, or 3 SDs of the mean.
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
Standard Gaussian (normal or bell-shaped) data distribution showing the percentage of data points that fall within 1, 2, or 3 SDs of the mean.
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
Basics of Statistical QC
Statistical QC is performed by measuring 1 or more QCMs with the automated laboratory instrument just as would be done for a patient sample (QCMs are typically designed to be run as a panel, such as hematology or biochemical panels). The results are then compared with the acceptable ranges previously established by the manufacturer of that particular instrument or with acceptable ranges established by in-house QCM measurements (the latter is preferred and explained subsequently).
For in-clinic instruments, QCMs can be commercially available solutions or patient samples. Commercially available solutions are often stabilized so that they can be used over a specified period. The shelf life of a vial of QCM after opening will vary, depending on the manufacturer, components of the control material, and instrument on which it will be used. More information is available elsewhere.4
Frequency of Statistical QC
The Quality Assurance and Standards Committee of the American Society for Veterinary Clinical Pathology recommends testing of 1 or more levels of QCM each day that a given test is performed.10 A minimum of 2 QCM levels (each level is a different bottled product)—1 representing healthy patient values and 1 representing abnormal patient values—is recommended for initial start-up validation of the instrument, as described in the fourth article of this series.4 At least 1 level is recommended for routine statistical QC monitoring. The QCM should be run through the largest testing panel that the instrument offers. Various manufacturers of in-clinic laboratory instruments suggest that QC monitoring on a weekly or monthly basis may be sufficient to demonstrate stable performance of an in-clinic instrument. However, this brings into question what should be done with results generated between the time an acceptable QC result is obtained and the time an unacceptable QC result is obtained, given that the problem causing the unacceptable result may have occurred at any time during the interval between QC runs (7 to 30 days) and sample stability may not be sufficient to retest the patient sample once a QC failure is noted. Furthermore, responsible and ethical laboratory testing entails frequent checks of laboratory system performance to ensure that reliable results are generated and sustained.
Desirable Features of Statistical QC
Ideally, a QC procedure would consistently detect deviations of stable performance within a laboratory system with a high probability of error detection and a low probability of false rejection of valid test results. This concept can be compared with that of a fire alarm in that it should not be so sensitive that it alerts when a single match is lit (false alarm or false rejection of a valid test result), but it should alert when the fire is still small and able to be eliminated before damage ensues (ie, high probability of error detection, before erroneous patient results are reported or acted on). Another analogy would be to compare the probability of error detection with diagnostic sensitivity: a highly sensitive test will have few false-negative results and detect almost all patients with a given disease. Similarly, a statistical QC strategy with a high probability of error detection will detect almost all clinically important analytic errors, missing few.
The ability to provide a high probability of error detection and a low probability of false rejection will depend on the actual performance parameters of the instrument (imprecision and bias), as obtained from instrument performance assessments covered in the fourth article4 of this series. This ability will also depend on the calculated control limits for QCM test results, which are the upper and lower boundaries of the range for acceptable QCM results. High and low control limits are established by determining a certain number of SDs above and below the mean of a set of QCM measurements (an example of these calculations is provided subsequently). Narrow control limits, such as the range within 2 SDs of the mean, will provide a high probability of error detection but also a high probability of false rejection of a valid result. Wider control limits, such as the range within 3 SDs of the mean, will provide a slightly lower probability of error detection with an almost negligible probability of false rejection. The number of SDs used to calculate control limits is referred to as a control rule, whereby limits representing the range within 3 SDs of the mean are termed the 1-3s control rule and those representing the range within 2 SDs are called a 1-2s control rule (the 1 refers to the fact that 1 QCM measurement is being used for evaluation).11 It should be noted that more complicated control rules exist involving more than just the SD range11; however, those rules are beyond the scope of this article.
Design and Execution of Statistical QC
Designing an in-clinic statistical QC strategy involves assessment of instrument performance, deciding whether the 1-3s control rule is suitable for each measurand, calculating control limits, and applying the new control limits to future (daily or routine) QCM measurements (Figure 2).
Diagram showing steps involved in the design and execution of statistical QC.
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
Diagram showing steps involved in the design and execution of statistical QC.
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
Diagram showing steps involved in the design and execution of statistical QC.
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
Step 1: assessment of instrument performance
The first step in the statistical QC process is to assess the performance of the instrument, specifically the bias and CV associated with each measurand. This assessment should be performed on acquisition of the instrument and then annually thereafter or more frequently to coincide with any software updates. The purpose of this first step is to not only evaluate the performance of the instrument but also to gather the data that will be used to generate control limits. This assessment requires at least 5 repeated analyses of 1 or more assayed commercially available QCMs, over 5 days,12,13 to yield sufficient data for calculation of mean, SD, and CV values, as described in the previous article.4 This is the simplest and most common approach to statistical QC. Although the target mean provided by the manufacturer of the QCM is used to determine the bias of the instrument for each assessed level of QCM, the instrument mean, SD, and CV are derived solely from in-clinic measurements.
Step 2: deciding whether 1-3s control limits are suitable
As suggested elsewhere,4,10 at a minimum, results measured with purchased QCMs should be within the acceptable range reported by the QCM manufacturer. However, these ranges for commercial QCMs cannot be expected to adequately serve for error detection regarding individual instruments because such ranges are often quite wide and do not allow a high probability of error detection, as shown in a recent study.14 This situation is clearly undesirable and does not represent efficient and effective use of QCMs within in-clinic laboratories.
As noted previously, when performing statistical QC, control rules are chosen that will provide a high (at least 85%) probability of error detection and low (< 5%) probability of false rejection of a valid test result. Use of the 1-3s control rule has been recommended for point-of-care instruments and will be used in the procedures described here as an assumed starting point for the statistical QC of in-clinic instruments.10,13 The 1-3s control rule, based on the use of control limits that represent 3 SDs from the calculated mean of a single QCM, provides a probability of error detection > 85% and a probability of false rejection of 0%.
After the mean, SD, bias, and CV have been determined from the QCM test results, the calculated CV, bias, and TEa can be used to assess the suitability of applying the 1-3s rule for each measurand15,16 (Table 1). For this step, bias should be < 5%. At least 2 levels of QCM should initially be used to determine their individual biases and CVs; the CV and bias data corresponding to the more clinically relevant QCM level should be used to assess the suitability of applying the 1-3s rule. For example, a high BUN concentration is more clinically relevant than a low or normal concentration, so the BUN data from a high QCM sample should be used for this evaluation, not data from a low or normal QCM level.
Instrument performance specifications suitable for the use of a 1-3s control rule.
| TEa* (%) | Absolute bias (%) | CV (%) |
|---|---|---|
| 50 | < 5.0 | < 7.5 |
| 33 | < 5.0 | < 4.6 |
| 25 | < 5.0 | < 3.3 |
| 20 | < 2.5 | < 2.8 |
| 20 | 2.5–5.0 | < 2.5 |
| 17 | < 2.0 | < 2.6 |
| 17 | 2.0–4.0 | < 2.2 |
| 17 | 4.0–6.0 | < 1.8 |
| 16 | < 2.5 | < 2.2 |
| 16 | 2.5–5.0 | < 1.8 |
| 14 | < 2.0 | < 2.0 |
| 14 | 2.0–4.0 | < 1.7 |
| 14 | 4.0–6.0 | < 1.4 |
| 13 | < 2.0 | < 1.8 |
| 13 | 2.0–4.0 | < 1.5 |
| 13 | 4.0–6.0 | < 1.1 |
| 11 | < 2.0 | < 1.5 |
| 11 | 2.0–4.0 | < 1.2 |
| 11 | 4.0–6.0 | < 0.8 |
| 10 | < 2.0 | < 1.3 |
| 10 | 2.0–4.0 | < 1.0 |
| 10 | 4.0–6.0 | < 0.6 |
| 5 | < 1.0 | < 0.6 |
| 5 | 1.0–2.0 | < 0.5 |
| 5 | 2.0–3.0 | < 0.3 |
| 5 | 3.0–4.0 | < 0.1 |
See Table 1 in the previous article4 in this series for recommended values of TEa for various hematologic and serum or plasma biochemical measurands.
To use this table, identify the TEa closest to the recommended value for the applicable measurand.15,16 Then compare the bias and CV, as calculated from QCM results for the most clinically relevant level of QCM for that measurand, with the (absolute value) bias and CV percentages provided in this table. If both the calculated CV and bias are within the ranges given in any 1 row for that TEa, then the 1-3s rule can be used and control limits can be calculated. If the CV and bias are not within the ranges of any of the rows provided for that TEa (both must conform in a single row), then an alternative analytical method must be used, the TEa must be relaxed, or the measurand should be controlled through use of nonstatistical QC.14
Several possible combinations of bias and CV (as determined from QCM test results) are provided for each TEa value (Table 1).13 In any single row of the table for a given TEa, both the bias and the CV cutoffs should be satisfied to use the 1-3s rule. If either the CV or the bias indicates that a 1-3s rule is unsuitable to use for a particular level of TEa, several possibilities exist.12 One possibility is that the instrument may not perform well enough to provide reliable results for clinical interpretation. A different instrument may need to be chosen and evaluated to provide results that can be interpreted with confidence.
Another possibility is that the recommended TEa could be relaxed (ie, increased) if the TEobs is only slightly greater than the desired TEa. The revised TEa can then be used with the determined CV and bias to select a 1-3s rule. This increase in TEa should be taken into account when interpreting patient results.4 For example, the recommended veterinary TEa for serum glucose concentration is 20%.16 If this TEa value cannot be achieved with current instrument performance (TEobs) and the TEa is increased to 30% to match the TEobs, a reading of 170 mg/dL (9.4 mmol/L) could be anywhere within 51 mg/dL (2.8 mmol/L) of the true patient value, meaning the true value could be as low as 119 mg/dL (6.6 mmol/L), which might influence clinical decision-making (eg, adjustment of insulin dosing) more than it would if the recommended 20% TEa was used.
Yet another possibility would be to place higher emphasis on nonstatistical QC rather than statistical QC. This may require extensive assessment of correlations (such as laboratory hematology data with results for microscopic review of blood smears or endocrinology results with clinical signs and biochemistry results), additional training of personnel performing the analyses, or repeated analyses of individual patient samples, averaging the obtained results to achieve a more accurate result.
Step 3: calculating control limits for each measurand
If the 1-3s rule is found to be suitable as per Table 1, then the next step in statistical QC is to calculate the control limit range for each measurand by use of the previously calculated mean and SD as follows:
To provide an example, sample data for serum albumin concentration will be used (Supplementary Appendix S1, available at: avmajournals.avma.org/doi/suppl/10.2460/javma.258.7.733). The mean value of the level 1 (normal) QCM measurements as shown in this appendix is 30.7 g/L, and the mean is 37.2 g/L for level 2 (abnormal) QCM measurements. A serum albumin concentration of 30.7 g/L is more common in veterinary medicine than a value of 37.2 g/L, so the data for the level 1 QCM will be used. The recommended TEa for this measurand is 15%.16 No TEa of 15% is listed in Table 1, so the nearest value (14% or 16%) can be used. Data in Supplementary Appendix S1 indicate that the bias of the instrument for serum albumin concentration measurements is 1.7% and the CV is 1.6%. These results fit the criteria for a TEa of 14% (Table 1; TEa = 14%, absolute bias < 2.0%, and CV < 2.0%); therefore, a 1-3s rule would be suitable for this measurand and will be applied to both QCM levels. For the level 1 QCM data, the mean is 30.7 g/L and the SD is 0.5 g/L, so the control limits would be 29.2 to 32.2 g/L (30.7 ± [3 × 0.5] g/L). For the level 2 QCM data, the mean is 37.2 g/L and the SD is 0.6 g/L, so the control limits would be 35.4 to 39.0 g/L (37.2 ± [3 × 0.6] g/L).
In another example, serum urea concentration has a recommended TEa of 12%.16 According to Table 1, the CVs for both QCM levels (QCM 1 CV = 4.1% and QCM 2 CV = 2.6%; Supplementary Appendix S1) are too high to allow use of the 1-3s rule to detect error within this TEa of 12%. For the more clinically relevant level 2 QCM, the CV of 2.6% and absolute bias of 0.4% would allow for a TEa of 17% to be detected. This more lax TEa value could be adopted as a new TEa for urea, but such a change must be considered when interpreting instrument results, particularly those close to the upper limit of the reference interval. Control limits can then be calculated as described in the previous example for serum albumin concentration.
Step 4: use the new control limits for future QCM measurements
Once control limits have been determined for a given measurand, they should be used to replace the acceptable ranges supplied by the manufacturer. This is because the manufacturer's acceptable ranges are often derived from data for multiple instruments, whereas control limits derived from an individual instrument are more specific to that instrument. Subsequent daily QCM measurements should fall within these limits. In some situations, in-clinic laboratory personnel may need to consult with a veterinary clinical pathologist or laboratory expert with experience in statistical QC if there are questions about the appropriateness of the chosen control limits. These experts may include members of the Quality Assurance and Laboratory Standards Committee of the American Society for Veterinary Clinical Pathology, members of the Laboratory Standards Committee of the European College of Veterinary Clinical Pathology, or professional laboratory personnel at various commercial, university, or government laboratories.
If possible, sufficient QCM from the same lot should be ordered such that it would last for approximately 6 months to a year to eliminate the need to reestablish control limits frequently. If the QCM lot number changes or the manufacturer provides a software update, instrument performance should again be evaluated and new control limits established.
Recording Routine Ongoing Statistical QC Measurements
Once control limits for laboratory measurands have been established, routine recording of QCM data should be performed each day that patient samples are run for that test. Some instruments have a QC mode, into which QCM target values are uploaded by the user for each new QCM lot, either manually or by barcode transfer. The QCM measurements are recorded within this mode and presented as a chart (eg, Levey-Jennings chart) that usually displays the QCM target mean, manufacturer's acceptable range, and individual data points at each level of QCM tested (Figure 3). The measured QCM data mean, SD, and CV are often also reported by the instrument. It is important to note that for many point-of-care devices, and even for many reference laboratory instruments, the manufacturer's acceptable ranges programmed into the instrument cannot be replaced, necessitating the creation of novel charts or spreadsheets for statistical QC of every measurand by use of the more accurate calculated control limits. Online tools,17 spreadsheets (eg, Supplementary Appendix S1), or manual forms can be used to generate Levey-Jennings charts. A single chart typically records 1 week or 1 month of control data with a graphic representation of the measured results, compared with the determined control rule. A Levey-Jennings chart of the sample level 1 QCM data for serum albumin concentration, as generated by use of an online tool,17 is provided (Supplementary Figure S1, available at: avmajournals.avma.org/doi/suppl/10.2460/javma.258.7.733).
Example of a Levey-Jennings chart showing 40 measurements of control data for 1 hypothetical QCM measurand, typically performed at a rate of 1 measurement/d. The mean and SD were calculated from repeated QCM measurements as described in the fourth article4 of this series and not from manufacturer-supplied values. The dotted orange lines represent the control limits (3 SDs from the mean) for the 1-3s control rule, and the magenta dotted lines represent 2 SDs from the mean. The green line represents the manufacturer's target mean for the measurand. One QC failure (red square) occurred at measurement 18. At that time point, troubleshooting (as shown in Figure 4) would have taken place and patient samples would not have been run before the QCM measurement was within the control limits again (measurement 19).
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
Example of a Levey-Jennings chart showing 40 measurements of control data for 1 hypothetical QCM measurand, typically performed at a rate of 1 measurement/d. The mean and SD were calculated from repeated QCM measurements as described in the fourth article4 of this series and not from manufacturer-supplied values. The dotted orange lines represent the control limits (3 SDs from the mean) for the 1-3s control rule, and the magenta dotted lines represent 2 SDs from the mean. The green line represents the manufacturer's target mean for the measurand. One QC failure (red square) occurred at measurement 18. At that time point, troubleshooting (as shown in Figure 4) would have taken place and patient samples would not have been run before the QCM measurement was within the control limits again (measurement 19).
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
Example of a Levey-Jennings chart showing 40 measurements of control data for 1 hypothetical QCM measurand, typically performed at a rate of 1 measurement/d. The mean and SD were calculated from repeated QCM measurements as described in the fourth article4 of this series and not from manufacturer-supplied values. The dotted orange lines represent the control limits (3 SDs from the mean) for the 1-3s control rule, and the magenta dotted lines represent 2 SDs from the mean. The green line represents the manufacturer's target mean for the measurand. One QC failure (red square) occurred at measurement 18. At that time point, troubleshooting (as shown in Figure 4) would have taken place and patient samples would not have been run before the QCM measurement was within the control limits again (measurement 19).
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
When Statistical QC Results Are Outside Specified Limits
If QC fails (ie, if routine QCM measurements fall outside the specified limits after those limits have been established), a systematic approach should be used to determine the underlying root cause for the QC failure, establish the appropriate course of correction, and then demonstrate subsequent stable performance prior to measuring patient samples. A stepwise approach to determining possible underlying root causes for QC failures may be of benefit and should include items related to the instrument, reagents, and operator (Figure 4). A backup plan should be in place for sending patient samples to a reference laboratory if an in-clinic instrument fails QC testing without a cause readily identified. In this situation, consultation with technical support from the instrument manufacturer, servicing of the instrument, updating the instrument software, or replacement of the instrument altogether may be necessary.
Algorithm for troubleshooting when 1 or more QC measurements fall outside of chosen control limits. Results must be restored to within control limits before patient samples can be run.
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
Algorithm for troubleshooting when 1 or more QC measurements fall outside of chosen control limits. Results must be restored to within control limits before patient samples can be run.
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
Algorithm for troubleshooting when 1 or more QC measurements fall outside of chosen control limits. Results must be restored to within control limits before patient samples can be run.
Citation: Journal of the American Veterinary Medical Association 258, 7; 10.2460/javma.258.7.733
Clinical Bottom Line
Statistical QC is a valuable tool in a total quality management system for any medical laboratory. It provides an evidence-based method to show that laboratory system performance is stable and trustworthy over time. Many aspects of statistical QC that have been described in this article can be implemented in the in-clinic laboratory and are further elaborated in the cited references. Use of a checklist may be helpful in ensuring that various aspects of statistical QC have been addressed (Appendix; Supplementary Appendix S2, available at: avmajournals.avma.org/doi/suppl/10.2460/javma.258.7.733).
Veterinary clinical pathologists and professional laboratorians are available to help in-clinic laboratory personnel become comfortable with statistical QC. A commitment to, and practice of, statistical QC can help provide the best possible patient care and ensure that laboratory systems are functioning well and continue to do so over time.
Acknowledgments
No third-party funding or support was received in connection with this study or the writing or publication of the manuscript. The authors declare that there were no conflicts of interest.
Abbreviations
| CV | Coefficient of variation |
| QC | Quality control |
| QCM | Quality control material |
| TEa | Total allowable analytic error |
| TEobs | Total observed analytic error |
References
- 1. ↑
Cook JR, Hooijberg EH, Freeman KP. Quality Management for In-Clinic Laboratories: the total quality management system and quality plan. J Am Vet Med Assoc 2021;258:55–61.
- 2. ↑
Hooijberg EH, Freeman KP, Cook JR. Quality Management for In-Clinic Laboratories: facilities, instrumentation, health and safety, training, and improvement opportunities. J Am Vet Med Assoc 2021;258:273–278.
- 3. ↑
Freeman KP, Cook JR, Hooijberg EH. Quality Management for In-Clinic Laboratories: standard operating procedures. J Am Vet Med Assoc 2021;258:477–481.
- 4. ↑
Hooijberg EH, Freeman KP, Cook JR. Quality Management for In-Clinic Laboratories: equipment maintenance and instrument performance. J Am Vet Med Assoc 2021;258:725–731.
- 5. ↑
Westgard QC. Glossary of QC terms. Available at: www.westgard.com/glossary.htm#r-s. Accessed Aug 25, 2020.
- 6. ↑
Flatland B, Freeman KF. Repeat patient testing shows promise as a quality control method for veterinary hematology testing. Vet Clin Pathol 2018;47:252–266.
- 7. ↑
Rishniw M, Pion PD. Evaluation of performance of veterinary in-clinic hematology analyzers. Vet Clin Pathol 2016;45:604–614.
- 9. ↑
Westgard JO. What's the idea behind statistical quality control? In: Westgard JO, ed. Basic QC practices. 3rd ed. Madison, Wis: Westgard QC Inc, 2010;15–26.
- 10. ↑
Flatland B, Freeman KP, Vap M, et al. ASVCP guidelines: quality assurance for point-of-care testing in veterinary medicine. Vet Clin Pathol 2013;42:405–423.
- 11. ↑
Westgard QC. “Westgard Rules” and multirules. Available at: www.westgard.com/mltirule.htm. Accessed Aug 29, 2020.
- 12. ↑
Lester S, Harr KE, Rishniw M, et al. Current quality assurance concepts and considerations for quality control of in-clinic biochemistry testing. J Am Vet Med Assoc 2013;242:182–192.
- 13. ↑
Rishniw M, Pion PD, Maher T. The quality of veterinary in-clinic and reference laboratory biochemical testing. Vet Clin Pathol 2012;41:92–109.
- 14. ↑
Manzocchi S, Furman E, Freeman K. Comparison of 3 options for choosing control limits in biochemistry testing. Vet Clin Pathol 2017;46:120–125.
- 15. ↑
Nabity MB, Harr KE, Camus MS, et al. ASVCP guidelines: allowable total error hematology. Vet Clin Pathol 2018;47:9–21.
- 16. ↑
Harr KE, Flatland B, Nabity M, et al. ASVCP guidelines: allowable total error guidelines for biochemistry. Vet Clin Pathol 2013;42:424–436.
- 17. ↑
Westgard JO, Schilling P. Westgard QC online QC calculator. Available at: tools.westgard.com/qccalculator.html. Accessed Aug 25, 2020.
Appendix
Checklist for statistical QC for in-clinic laboratories.
| Recommendation | Compliant? | Additional comments by auditor |
|---|---|---|
| QCMs obtained for all automated instruments. | □ Yes □ No | |
| Lot numbers and expiration dates of QCMs recorded. | □ Yes □ No | |
| QCMs stored in accordance with manufacturer's instructions and are not used past their expiration date. | □ Yes □ No | |
| QCMs run through the largest panel that the instrument offers. | □ Yes □ No | |
| At least 2 levels of QCM (1 representing healthy and at least 1 other representing abnormal) used for initial in-clinic instrument evaluation. This evaluation is repeated annually or more frequently if software updates occur. (These measurements help determine the correct control range [bound by control limits] that will produce a high probability of error detection and a low probability of false rejection for patient sample measurements.) | □ Yes □ No | |
| Control limits for measurands chosen on the basis of bias, CV, and TEa. The 1–3s rule (control range bound by 3 SDs of the mean) is used if feasible (as described in the main text). | □ Yes □ No | |
| After control limits established, at least 1 level of QCM used for routine statistical QC monitoring on each day a given test is performed. | □ Yes □ No | |
| Statistical QC measurements serially recorded in a spreadsheet or on a Levey-Jennings chart that shows the mean and calculated control limits (number of allowable SDs from the mean). | □ Yes □ No | |
| QC measurements falling outside of chosen control limits addressed by troubleshooting before any patient samples are run. | □ Yes □ No | |
| If QC measurements continue to fall outside control limits, if TEobs > TEa, and if these trends persist after assistance from the manufacturer, purchase of a new instrument considered or nonstatistical QC added to test evaluation. | □ Yes □ No | |
| Control means, SDs, TEobs, and control limits reevaluated with changes in QCM lot numbers or software updates. | □ Yes □ No |
