When a control result falls outside its limits at the start of a run, patient results are not released until the control failure is evaluated. Laboratories assess quality across the entire testing path. During the analytic phase, quality control (QC) results, calibration records, and method-evaluation data show whether a measurement procedure is performing as intended. Proficiency testing compares the laboratory's results with an external target or peer group. A quality management system links this work to the laboratory's other processes.
The QC workstation: AST warning case applies the multirule explanation below to aspartate aminotransferase (AST) controls, from the first run to a patient-result reporting decision.
Statistical control of a measurement procedure
Repeated measurements of a stable control material produce a distribution of results around a central value. If that distribution is approximately Gaussian, about 68.3% of results fall within 1 standard deviation (SD) of the mean, 95.4% within 2 SD, and 99.7% within 3 SD. These proportions support common statistical control rules. Before applying them, confirm that your control data suit the selected rule system.1,2
- The mean is the arithmetic average.
- The standard deviation describes dispersion around the mean in the analyte's units.
- The variance is the SD squared.
- The coefficient of variation compares imprecision across concentrations or analytes:
CV (%) = SD ÷ mean × 100.
SD and CV can change with concentration. Many quantitative procedures show a larger SD at higher concentrations, and the CV may remain approximately constant over a limited interval. Transfer an SD or CV between control concentrations only when your laboratory's data support that assumption.2
A potassium control produces nine results: 4.1, 4.2, 4.2, 4.3, 4.3, 4.3, 4.4, 4.4, and 4.5 mmol/L.
Scroll sideways to see all columns
| Statistic | Calculation | Result |
|---|---|---|
| Mean | 38.7 ÷ 9 | 4.30 mmol/L |
| Median | Middle ordered value | 4.3 mmol/L |
| Mode | Most frequent value | 4.3 mmol/L |
| Variance | Σ(xi − mean)² ÷ (n − 1) | 0.0150 (mmol/L)² |
| SD | √0.0150 | 0.122 mmol/L |
| CV | 0.122 ÷ 4.30 × 100 | 2.8% |
The resulting control lines are:
Scroll sideways to see all columns
| Interval | Lower limit | Upper limit |
|---|---|---|
| ±1 SD | 4.18 mmol/L | 4.42 mmol/L |
| ±2 SD | 4.06 mmol/L | 4.54 mmol/L |
| ±3 SD | 3.93 mmol/L | 4.67 mmol/L |
These nine results demonstrate the calculations. Nine points are too few to establish that the underlying distribution is Gaussian. For a reported ±2 SD interval of 80 to 110 mg/dL, the midpoint is 95 mg/dL. The full interval spans 4 SD, so 1 SD is (110 − 80) ÷ 4 = 7.5 mg/dL.
Sigma metric and matching units
A sigma metric relates a chosen allowable total error to the procedure's estimated bias and imprecision at a stated concentration:
Sigma = (allowable total error % − absolute bias %) ÷ CV %
Subtract the magnitude of bias because either direction consumes the permitted error. Convert SD to CV before combining it with percentage inputs, or use allowable error, bias, and SD all in the same concentration units. The value depends on the selected goal, concentration, and performance estimates. It is one input to the assessment of assay suitability and does not establish suitability.3,4
For a control mean of 125 mg/dL and SD of 2.5 mg/dL, CV is 2.5 ÷ 125 × 100 = 2.0%. With a laboratory-adopted allowable error of 10% and bias of −2%, sigma is (10 − 2) ÷ 2 = 4.0. Subtracting the signed negative bias would incorrectly increase the allowance. The laboratory uses the metric with its risk assessment and applicable QC requirements.
Control materials
Control materials contain one or more analytes in a matrix that is the same as, or similar to, the patient specimen when an appropriate material is available. Laboratories commonly monitor at least two concentrations at clinically relevant points. Materials may be liquid, frozen, or lyophilized. Lyophilized material must be reconstituted with the manufacturer-specified volume and diluent because reconstitution error changes the observed concentration.
For most quantitative nonwaived procedures, the CLIA default is at least two control levels once each day patient specimens are assayed. Qualitative procedures use a negative and a positive control. Graded or titered, extraction, and molecular procedures have additional method-appropriate control requirements. More stringent manufacturer instructions apply, and an approved Individualized Quality Control Plan can provide an eligible test system with a different risk-based plan. CLIA also requires control procedures after a complete reagent change and after major preventive maintenance or replacement of a critical part that can affect performance. The laboratory's written procedure may call for controls after additional events, such as troubleshooting or calibration.5
CLSI C24 uses separate schedules to estimate the initial SD and the target mean of a stable control material for a quantitative measurement procedure. When no suitable historical SD exists, collect at least 20 data points on separate days while the procedure is stable and in control. The study should follow routine operating conditions, including the usual opened-vial use pattern. Inspect a Levey-Jennings chart for pronounced drift, shift, and outliers before using the 20-point SD as the initial estimate. One result per day supports the ordinary sample-SD calculation. With multiple results per day, one-way analysis of variance can separate within-day and between-day components and avoid underestimating long-term imprecision.2
For a new control lot at a similar target concentration, a medically suitable long-term SD from the stable procedure can provide the initial imprecision estimate. If the new mean differs and the CV is reasonably constant at the two concentrations, estimate the new SD as new mean × established CV ÷ 100. Establish the new lot's target mean separately by testing it in parallel with the current lot. In most cases, 10 measurements on separate days spanning at least 10 days provide an initial target. Include several calibration events and reproduce the intended opened-vial use period. A target assigned over fewer days with multiple measurements per day is temporary.2
A long-term SD from the current procedure may also initialize control limits for a replacement procedure when that SD remains medically suitable and the replacement procedure has similar or better imprecision. A less precise replacement procedure requires an estimate that represents its own performance.2
Update the target mean and SD from stable data accumulated over the first several months. This longer interval can represent calibration cycles, reagent bottles and lots, maintenance, component replacement, environmental conditions, and material stability. A control-specific matrix shift after a reagent-lot change can inflate a cumulative SD. In that setting, data from one reagent lot or pooled SDs from separate stable intervals may be appropriate. Manufacturer values are guides. Choose the C24 pathway that fits the measurement procedure, the control material, the historical data available, and the laboratory's adopted statistical-control procedure.2
The official C24 correction notice clarifies that the updated SD should reflect long-term variability and directs the reagent-lot discussion to CLSI EP26. The schedules above remain unchanged.2
Commutability describes whether a material and authentic patient specimens show equivalent relationships among different measurement procedures. A noncommutable control can still monitor within-method stability when the laboratory establishes suitable local limits. It is unsuitable for demonstrating calibration traceability across methods. Because commercial processing, stabilizers, and added analyte can affect commutability, evaluate a material by its intended use and supporting evidence, regardless of the supplier.2
Levey-Jennings charts
A Levey-Jennings chart plots successive control results on the vertical axis against time or run number on the horizontal axis. Horizontal lines mark the mean and the ±1, ±2, and ±3 SD limits. The following hemoglobin values form an arithmetic example. Each laboratory derives control targets from its own procedure:
Scroll sideways to see all columns
| Line | Value |
|---|---|
| +3 SD | 19.5 g/dL |
| +2 SD | 18.0 g/dL |
| +1 SD | 16.5 g/dL |
| Mean | 15.0 g/dL |
| −1 SD | 13.5 g/dL |
| −2 SD | 12.0 g/dL |
| −3 SD | 10.5 g/dL |
Plot results from one defined control lot against that lot's mean and SD. A new lot needs its own evaluated target and limits. Overlap testing helps separate a true method change from a difference between control lots.2
Random error creates unpredictable scatter. Examples include an air bubble, an inconsistent pipette delivery, or transient electrical noise. Systematic error moves results in one direction. Reagent deterioration, a calibration problem, or a temperature shift can cause systematic error. An abrupt sustained displacement appears as a shift. Gradual movement in one direction appears as a trend. The number of points used to define a shift or trend belongs in the laboratory's written control policy because conventions differ among procedures.2
Multirule interpretation
The printable common QC rule violations guide places six control-chart examples beside their rule definitions for practice with the QC workstation.
Westgard multirules combine control observations to improve detection of random and systematic error. Laboratories commonly include them in their control policies. CLIA requires effective control procedures and does not mandate this rule set.6
Scroll sideways to see all columns
| Rule | Trigger | Usual interpretation |
|---|---|---|
| 12s | One result exceeds ±2 SD | Warning that prompts review of the other rules |
| 13s | One result exceeds ±3 SD | Reject; large random error or systematic error is possible |
| 22s | Two consecutive results exceed the same 2 SD limit in the same direction | Reject; systematic error is likely |
| R4s | Two results in one run differ by more than 4 SD, with one above +2 SD and one below −2 SD | Reject; random error is likely |
| 41s | Four consecutive results exceed the same 1 SD limit in the same direction | Reject; a small systematic change is likely |
| 10x, 9x, or 8x | The specified consecutive results fall on one side of the mean | Reject under the selected policy; systematic change is likely |
| (2 of 3)2s | Two of three control levels exceed the same 2 SD limit in the same direction | Reject; systematic error is likely |
Current simultaneous multirule use evaluates every selected rule with each new control result and the applicable earlier results. A value within ±2 SD does not bypass 41s, 10x, or another sequence rule. Under the laboratory's policy, the 12s rule remains a warning. Check every other selected rule whether or not a 12s warning occurs.
The control patterns below suggest whether an error is random or systematic, giving the laboratory a starting point for investigating the cause:
- A glucose control has a mean of 120 mg/dL and SD of 4 mg/dL. Ten results are 123, 125, 124, 126, 122, 127, 124, 125, 126, and 123 mg/dL. All lie within ±2 SD, but all are above the mean. The sequence satisfies the 10x rule and indicates a shift.
- Creatine kinase (CK) controls have targets of 60 U/L with SD 3 and 180 U/L with SD 9. Results of 66.3 and 199.8 U/L in one run are +2.1 SD and +2.2 SD, satisfying 22s. Results of 53.7 and 199.8 U/L in one run are −2.1 SD and +2.2 SD, a 4.3 SD spread that satisfies R4s.
- A sodium control has a mean of 145 mmol/L and SD of 1.5 mmol/L. Four results of 147.0, 147.5, 147.0, and 147.5 mmol/L are all above +1 SD and satisfy 41s.
The AST warning case applies the distinction between a warning and a rejection rule to two control levels and the preceding results.
Resolving rejected QC
When a rejection rule is met, the laboratory stops reporting affected patient results, investigates the control failure, corrects the cause, and documents acceptable control performance before resuming testing. It evaluates all patient results from the unacceptable run and all results reported since the last acceptable run to determine whether any were adversely affected. Corrections follow the laboratory's procedure and the clinical significance of the error.5
The calcium control-material case follows a rejected run through investigation, corrective action, rerun assessment, and patient-result review.
Individualized quality control plans and calibration
An Individualized Quality Control Plan (IQCP) is a CLIA quality-control option for eligible nonwaived test systems. Its three parts work together:7
Scroll sideways to see all columns
| Part | Required work |
|---|---|
| Risk assessment | Evaluate hazards associated with the specimen, test system, reagent, environment, and testing personnel |
| Quality control plan | State the control practices that reduce the identified risks |
| Quality assessment | Monitor the plan, investigate failures, and revise the assessment when relevant conditions change |
Before implementation, the laboratory director approves, signs, and dates the plan. The plan must meet manufacturer requirements and applicable regulatory requirements. Some testing is excluded from IQCP, so check eligibility before developing a plan.
Calibration establishes the relationship between an instrument signal and analyte concentration. Calibrators must be specified or validated for the measurement system. Calibration mainly addresses systematic bias. Routine control results are still needed to detect later changes and imprecision. When a higher-order reference system exists, traceability links the patient result through an unbroken calibration chain to that reference.5
Agreement with the material used to fit a calibration curve shows only that the curve fits its own calibrators. Controls assess the resulting system separately and must meet the applicable acceptance criteria before patient reporting. Under 42 CFR §493.1256(d)(9), calibration material used as a control must come from a different lot from the material used to calibrate the system or establish its cutoff.5
For each applicable nonwaived test system, unless another provision of the same subpart specifies otherwise, CLIA requires calibration verification according to the manufacturer's instructions and at least once every six months. The materials must include at least a minimal or zero value, a midpoint value, and a maximum value near the upper limit of the reportable range. Calibration verification is also required after a complete reagent change unless the laboratory demonstrates that reagent-lot changes leave the reportable range and control values unaffected, after major maintenance or replacement of a critical part, and when an unusual control shift, trend, or unacceptable value remains unresolved after other assessment and correction. The laboratory's written schedule may require more frequent verification.5
Method verification and performance establishment
Before reporting patient results, the laboratory documents that a procedure performs adequately in its own setting. Its responsibility depends on the status of the procedure.5
Scroll sideways to see all columns
| Procedure | Laboratory responsibility before patient testing |
|---|---|
| Unmodified FDA-cleared or FDA-approved system | Verify accuracy, precision, and reportable range comparable with the manufacturer's claims; verify that the reference interval is appropriate for the laboratory's patient population |
| Modified FDA-cleared or FDA-approved system, laboratory-developed procedure, or procedure without FDA clearance or approval | Establish accuracy, precision, analytical sensitivity, analytical specificity including interfering substances, reportable range, reference intervals, and other performance characteristics needed for test performance |
When the laboratory performs the same test with different methods, instruments, or testing sites, it evaluates the relationship between results at least twice each year. Define acceptance criteria before reviewing the data so the results cannot influence the criteria.5
Each performance characteristic has a distinct definition:
- Accuracy: broad closeness of agreement between a measured result and an accepted reference value; it reflects both systematic and random effects, which makes it broader than trueness
- Trueness: closeness of the mean of replicate results to a reference value; it addresses the systematic component of measurement error
- Bias: an estimate of the systematic difference between the mean result and a reference or comparator value
- Precision: closeness of agreement among repeated results; imprecision is expressed with statistics such as SD and CV
- Analytical sensitivity: the method's detection capability, including the limit of blank, limit of detection, or limit of quantitation when applicable
- Analytical specificity: the CLIA performance characteristic that includes the effect of interfering substances on the intended analyte result
- Selectivity: the extent to which a measurement system obtains values for the intended measurand independently of other quantities present in the specimen
- Linearity: the ability to produce results directly proportional to measurand concentration across a stated interval, within allowable deviation
- Reportable range: the interval the laboratory can report after applying its validated direct-measurement and specimen-processing procedures
- Reference interval: the range that describes the designated reference population under defined conditions
Total analytical error (TAE) describes the combined analytical effect of bias, imprecision, nonlinearity, interferences, specimen-matrix differences, and other error sources on individual quantitative results. CLSI EP21 estimates TAE across the analytical measuring interval or clinically relevant subintervals. Total error across the whole testing process also includes preanalytic and postanalytic error.8
Set acceptance criteria before examining study results. CLSI EP46 describes allowable total error (ATE) goals and limits based on clinical outcomes, biological variation, or current analytical performance, according to the evidence available for the measurand and intended use. The laboratory sets these goals for its own testing and derives QC limits and method-evaluation criteria separately from the CLIA criteria used to grade PT events.9
The short calculations below are teaching examples. Full studies use the current CLSI design suited to the task, with appropriate specimens or materials, concentrations, days, runs, replicates, statistical models, and predefined acceptance criteria.10-12
Agreement and sampling uncertainty
Complete agreement in a finite sample leaves uncertainty about the population. If all 60 comparator-positive specimens agree, observed positive agreement is 100%, and the one-sided 95% exact binomial lower bound is 0.05^(1/60) × 100 = 95.13%. The bound expresses sampling uncertainty only. It gives neither a second point estimate nor proof of perfect agreement in the population. The calculation assumes independent observations and cannot correct biased sampling.13
A routine comparator may not establish true condition status. Report the result as agreement with that comparator and assess negative agreement separately. Comparison with a non-reference method explains the denominators and terminology.
Comparing a candidate method
A paired t-test evaluates whether paired results have a statistically detectable mean difference. For each specimen, calculate d = candidate − comparison, then use:
t = d̄ ÷ (SDd ÷ √n)
Eight paired white blood cell (WBC) specimens produce differences of +0.2, −0.1, +0.3, −0.2, +0.2, −0.1, +0.3, and −0.1 × 103/µL. Their mean difference is 0.0625 × 103/µL and their SD is 0.207 × 103/µL. Therefore t = 0.0625 ÷ (0.207 ÷ √8) = 0.86, with 7 degrees of freedom. The two-tailed critical value at P = .05 is 2.365, so this set does not show a statistically significant mean difference. A nonsignificant result does not demonstrate equivalence.
A complete method comparison evaluates bias across the measuring interval, examines differences between paired results, and uses regression suited to the data and error structure. High correlation can coexist with unacceptable bias.10
Ten calcium specimens measured by the comparative method (x) and the candidate method (y), in mg/dL, illustrate the statistics:
Scroll sideways to see all columns
| Specimen | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
|---|---|---|---|---|---|---|---|---|---|---|
| Comparative, x | 7.6 | 8.0 | 8.4 | 8.8 | 9.1 | 9.5 | 9.9 | 10.3 | 10.8 | 11.2 |
| Candidate, y | 7.7 | 8.1 | 8.5 | 8.9 | 9.2 | 9.6 | 10.0 | 10.4 | 11.0 | 11.4 |
These pairs give a least-squares line of y = 1.025x − 0.118, r = 0.9997, and a standard error of the estimate of about 0.031 mg/dL. The fitted line predicts a bias of about +0.10 mg/dL at 8.5 mg/dL and +0.15 mg/dL at 10.5 mg/dL. The slope and intercept describe the fitted relationship. Judge the method by the estimated bias at clinical concentrations, because neither coefficient alone proves a particular error source. Correlation measures association between the methods. The standard error of the estimate describes residual scatter around the fitted line. Acceptance depends on the predefined allowable bias across the interval.
Evaluating precision
The F statistic compares two variances. A one-sided test specifies its direction before the results are examined. A two-sided question requires both tails of the F distribution.
F = variance B ÷ variance A
Suppose the one-sided question is whether coagulation analyzer B has greater variance than analyzer A. Each analyzer measures one prothrombin-time control 18 times, with SDs of 1.2 seconds for A and 1.6 seconds for B. Their variances are 1.44 and 2.56 seconds². The calculated F is 2.56 ÷ 1.44 = 1.78, with degrees of freedom (17, 17). The one-sided upper-tail critical value at α = .05 is about 2.27, so these data do not show greater variance for B. This single-level calculation illustrates the F statistic. CLSI EP05 addresses establishment of precision. EP15 addresses user verification of a manufacturer's precision claim.11,12,14
Recovery, interference, linearity, and reference intervals
The dilution-factor calculation explains how the prepared volumes affect concentration.
A recovery experiment adds a known amount of analyte and measures the observed increase:
Recovery (%) = [(spiked result − unspiked result) ÷ amount added to the final aliquot] × 100
For five sodium specimens spiked to add 12 mmol/L in the final aliquot, observed increases of 12.2, 11.8, 12.1, 12.3, and 11.9 mmol/L give recoveries of 101.7%, 98.3%, 100.8%, 102.5%, and 99.2%. Mean recovery is 100.5%. The study design must account for dilution and matrix effects. This arithmetic result applies only to the tested spike conditions and is judged against a predefined recovery or bias criterion.15
An interference study compares an interferent-spiked aliquot with a paired aliquot containing an equal volume of diluent:
Bias = mean(interferent-spiked aliquot) − mean(diluent-control aliquot)
Five potassium pairs show biases of +0.4, +0.5, +0.6, +0.4, and +0.5 mmol/L after adding hemolysate. The mean bias of +0.48 mmol/L describes the effect at this hemolysate level with this procedure. Establishing the effect at other interferent levels requires additional evidence.16
Linearity and reportable range are related but distinct. A linearity study evaluates the response pattern and clinically allowable deviation across selected concentrations. The analytical measuring interval covers direct measurement without dilution, concentration, or other specimen pretreatment beyond the procedure's routine preparation. A validated dilution or spiking procedure may extend the interval that the laboratory can report. Study design follows the manufacturer's instructions, the laboratory's protocol, and current guidance. CLIA specifies no universal number of concentrations or replicates. Results beyond the validated reportable range follow the laboratory's approved above-range or below-range reporting convention.5,15,17
Reference-interval verification determines whether a proposed interval fits the procedure and the laboratory's reference population. The current CLSI implementation guide describes a small-sample design using 20 qualified reference individuals. The interval is verified when no more than two results fall outside it. If three or more fall outside, that first sample has not verified the interval and the CLSI follow-up process applies. Establishing a new nonparametric reference interval generally uses at least 120 qualified individuals for each partition.18
Selected two-tailed critical t values
Scroll sideways to see all columns
| df | P = .10 | P = .05 | P = .01 |
|---|---|---|---|
| 5 | 2.015 | 2.571 | 4.032 |
| 9 | 1.833 | 2.262 | 3.250 |
| 10 | 1.812 | 2.228 | 3.169 |
| 15 | 1.753 | 2.131 | 2.947 |
| 20 | 1.725 | 2.086 | 2.845 |
| 25 | 1.708 | 2.060 | 2.787 |
| 29 | 1.699 | 2.045 | 2.756 |
| 30 | 1.697 | 2.042 | 2.750 |
| 40 | 1.684 | 2.021 | 2.704 |
| 60 | 1.671 | 2.000 | 2.660 |
| 120 | 1.658 | 1.980 | 2.617 |
| ∞ | 1.645 | 1.960 | 2.576 |
As degrees of freedom increase, the t critical values approach the corresponding standard-normal values. The calculation uses the standard error because it evaluates uncertainty in the estimated mean difference. The SD describes variation among individual observations.19
Proficiency testing and external quality assessment
External quality assessment includes several kinds of external comparison. Proficiency testing is a regulated form in which an outside program sends blinded samples for testing. The laboratory compares its results with an assigned target or with a peer group using similar instruments, methodologies, and reagent systems. Material commutability determines how confidently results can be compared across measurement procedures. Accuracy-based programs use commutable material and reference-method targets when available.20
For analytes covered by CLIA proficiency-testing requirements, the laboratory enrolls in an HHS-approved program. For other testing, it verifies accuracy at least twice yearly when CLIA requires an alternative assessment. During a PT event, the personnel who ordinarily perform the test process samples through the routine patient-testing workflow. The laboratory uses its routine method and number of repeats and keeps PT records for at least two years. Participating laboratories must not communicate with one another before the submission deadline.5,21
If the patient-testing procedure normally calls for reflex, distributive, or confirmatory testing at another laboratory, PT testing stops at the point where the patient specimen would be referred. A PT specimen must not be sent to another laboratory for analysis the sending laboratory is certified to perform. A first referral limited to otherwise routine reflex, distributive, or confirmatory testing is still improper. CLIA handles it with alternative sanctions and classes it separately from intentional referral. A laboratory that receives a PT specimen must notify CMS.21
An unacceptable result requires a documented investigation. The laboratory reviews transcription, sample handling, calculations, reagents, calibration, control performance, instrument function, and personnel technique. Corrective action addresses the identified cause, evaluates possible effects on patient testing, and includes evidence that the action worked. The laboratory director ensures that appropriate staff review PT performance and take corrective action. The investigation records its conclusion, including cases in which the evidence supports random error as the likely explanation. An initial PT failure can lead to corrective or regulatory action according to the analyte, event pattern, investigation, and applicable CLIA provisions.5
Standard deviation index
The standard deviation index (SDI) locates one laboratory result relative to a peer-group mean:
SDI = (laboratory result − peer-group mean) ÷ peer-group SD
A positive SDI indicates a result above the peer-group mean, and a negative SDI a result below it. One result does not establish persistent bias. Use the peer group and grading rule for that event. An SDI limit used by an external assessment scheme does not replace the applicable CLIA analyte-specific PT criterion.22
For an illustrative scheme accepting absolute SDI of 2 or less, a result of 108 U/L against a peer mean of 100 U/L and SD of 3 U/L gives 8 ÷ 3 = +2.67, outside the stated limit. If the next event has mean 100 U/L and SD 4 U/L, its upper acceptable result is 100 + 2 × 4 = 108 U/L. A different peer SD changes the concentration limit even when the index limit stays the same.
Selected CLIA performance limits
These limits reproduce selected criteria in 42 CFR §§493.931 (routine chemistry, including blood gas), 493.933 (endocrinology), 493.937 (toxicology), 493.941 (hematology, including coagulation), and 493.959 (immunohematology), as amended by 87 FR 41194 (July 11, 2022; effective July 11, 2024). They are a study aid. The regulation and the proficiency-testing program's grading instructions govern each event. These criteria apply to PT grading for the listed analytes, and each laboratory sets its QC limits and method-evaluation acceptance criteria separately.23
Scroll sideways to see all columns
| Discipline | Analyte | Acceptable limit |
|---|---|---|
| Chemistry | Glucose | ±8% or ±6 mg/dL, whichever is greater |
| Chemistry | Sodium | ±4 mmol/L |
| Chemistry | Potassium | ±0.3 mmol/L |
| Chemistry | Total calcium | ±1.0 mg/dL |
| Chemistry | Urea nitrogen | ±9% or ±2 mg/dL, whichever is greater |
| Chemistry | Total bilirubin | ±20% or ±0.4 mg/dL, whichever is greater |
| Chemistry | Total cholesterol | ±10% |
| Chemistry | ALT or AST | ±15% or ±6 U/L, whichever is greater |
| Chemistry | Alkaline phosphatase or amylase | ±20% |
| Blood gas | pH | ±0.04 |
| Blood gas | pCO2 | ±8% or ±5 mm Hg, whichever is greater |
| Endocrinology | TSH | ±20% or ±0.2 mIU/L, whichever is greater |
| Endocrinology | Free T4 | ±15% or ±0.3 ng/dL, whichever is greater |
| Endocrinology | hCG, excluding waived visual urine pregnancy tests | ±18% or ±3 mIU/mL, whichever is greater, or positive or negative |
| Toxicology | Digoxin | ±15% or ±0.2 ng/mL, whichever is greater |
| Toxicology | Blood lead | ±10% or ±2 µg/dL, whichever is greater |
| Hematology | Hemoglobin or nonspun hematocrit | ±4% |
| Hematology | Platelet count | ±25% |
| Hematology | Prothrombin time, reported in seconds or INR, or partial thromboplastin time | ±15% |
| Immunohematology | ABO grouping, D typing, unexpected antibody detection, and compatibility testing | 100% accuracy |
| Immunohematology | Antibody identification | At least 80% accuracy |
Quality management system
A quality management system coordinates the processes that support reliable laboratory service.24
Scroll sideways to see all columns
| Term | Scope |
|---|---|
| Quality control | Evidence that a particular measurement procedure or item of equipment performs as expected |
| Quality assurance | Evaluation of a complete process, such as specimen transport or critical-result notification |
| Quality management | Organization-wide planning, control, assessment, and improvement of laboratory quality |
CLSI organizes the laboratory quality management system into twelve quality system essentials:24
Scroll sideways to see all columns
| Quality system essential | Example activity |
|---|---|
| Organization | Quality planning and management review |
| Customer focus | Complaint investigation and service feedback |
| Facilities and safety management | Hazard assessment and emergency preparation |
| Personnel management | Training and competency assessment |
| Supplier and inventory management | Incoming inspection and lot tracking |
| Equipment management | Installation, calibration, and preventive maintenance |
| Process management | Procedure design and method evaluation |
| Documents and records management | Version control and record retention |
| Information management | Access control, data integrity, and downtime procedures |
| Nonconforming event management | Detection, documentation, and investigation of failures |
| Assessments | Internal audits and proficiency testing |
| Continual improvement | Corrective action and measured process improvement |
A quality indicator is a defined measurement of a laboratory process. Examples include specimen-rejection rate by cause, critical-result notification time, turnaround time for a defined test group, amended-report rate, and PT performance. For each indicator, the laboratory states the numerator, denominator, population, data source, review frequency, threshold, and responsible owner. Review frequency depends on the risk and the laboratory's quality plan. No universal monthly committee rule applies to every indicator.24
After a process failure, the laboratory documents the event and contains the immediate risk. It investigates contributing causes, evaluates affected patient results, implements corrective action, and checks effectiveness. Root cause analysis is one tool for serious or recurring events. CLIA requires investigation and correction of identified problems. Whether a formal root cause analysis is mandatory depends on the organization's accreditation and safety rules.5
The cost-of-quality model groups resources and losses into four categories:25
Scroll sideways to see all columns
| Category | Examples |
|---|---|
| Prevention | Training, method evaluation, preventive maintenance |
| Appraisal | Control materials, calibration verification, PT, internal audit |
| Internal failure | Recollection, repeat testing, discarded reagents, downtime |
| External failure | Corrected reports, patient follow-up, complaints, recalls, legal costs |
These categories let a laboratory compare the cost of preventing or detecting a problem with the cost created after failure. Their relative size depends on the process and the event.
Practice
References
- National Institute of Standards and Technology. The normal distribution. NIST/SEMATECH e-Handbook of Statistical Methods. Accessed August 31, 2026. Back to text
- Clinical and Laboratory Standards Institute. Statistical Quality Control for Quantitative Measurement Procedures: Principles and Definitions. 4th ed. CLSI guideline C24Ed4E. Clinical and Laboratory Standards Institute; 2016. Published September 29, 2016. Clarifications and corrections. Published February 8, 2017. Back to text
- Westgard QC. Best practices for Westgard rules. Accessed September 7, 2026. Back to text
- Westgard JO. The comparison of methods experiment. Westgard QC. Accessed September 7, 2026. Back to text
- Electronic Code of Federal Regulations. 42 CFR part 493: Laboratory Requirements. Updated through August 27, 2026. Accessed August 31, 2026. Back to text
- Westgard JO, Barry PL, Hunt MR, Groth T. A multi-rule Shewhart chart for quality control in clinical chemistry. Clin Chem. 1981;27(3):493-501. Back to text
- Centers for Medicare & Medicaid Services. Individualized Quality Control Plans. Accessed August 31, 2026. Back to text
- Clinical and Laboratory Standards Institute. Evaluation of Total Analytical Error for Quantitative Medical Laboratory Measurement Procedures. 3rd ed. CLSI guideline EP21. Published June 16, 2025. Back to text
- Clinical and Laboratory Standards Institute. Determining Allowable Total Error Goals and Limits for Quantitative Medical Laboratory Measurement Procedures. 1st ed. CLSI guideline EP46. Published June 16, 2025. Back to text
- Clinical and Laboratory Standards Institute. Measurement Procedure Comparison and Bias Estimation Using Patient Samples. 3rd ed. CLSI guideline EP09c. Published June 20, 2018. Back to text
- Clinical and Laboratory Standards Institute. Evaluation of Precision of Quantitative Measurement Procedures. 4th ed. CLSI guideline EP05. Published December 11, 2025. Back to text
- Clinical and Laboratory Standards Institute. User Verification of Precision and Estimation of Bias. 3rd ed. CLSI guideline EP15. Published September 11, 2014. Reaffirmed September 2019. Back to text
- National Institute of Standards and Technology. Exact binomial. Dataplot Reference Manual. Accessed September 7, 2026. Back to text
- National Institute of Standards and Technology. Upper critical values of the F distribution. NIST/SEMATECH e-Handbook of Statistical Methods. Accessed August 31, 2026. Back to text
- Clinical and Laboratory Standards Institute. Establishing and Verifying an Extended Measuring Interval Through Specimen Dilution and Spiking. 1st ed. CLSI guideline EP34. Published August 13, 2018. Reaffirmed March 2023. Back to text
- Clinical and Laboratory Standards Institute. Interference Testing in Clinical Chemistry. 3rd ed. CLSI guideline EP07. Published April 30, 2018. Reaffirmed October 2022. EP07 Guide published March 17, 2026. Back to text
- Clinical and Laboratory Standards Institute. Evaluation of Linearity of Quantitative Measurement Procedures. 2nd ed. CLSI guideline EP06. Published November 24, 2020. Back to text
- Clinical and Laboratory Standards Institute. Defining, Establishing, and Verifying Reference Intervals in the Clinical Laboratory. 3rd ed. CLSI guideline EP28-A3c. Published October 19, 2010. Reaffirmed April 2020. Verification of Reference Intervals in the Medical Laboratory Implementation Guide. Published 2022. Back to text
- National Institute of Standards and Technology. Critical values of the Student's t distribution. NIST/SEMATECH e-Handbook of Statistical Methods. Accessed August 31, 2026. Back to text
- College of American Pathologists. Proficiency Testing/External Quality Assessment Frequently Asked Questions. Accessed August 31, 2026. Back to text
- Electronic Code of Federal Regulations. 42 CFR §493.801: Condition: Enrollment and testing of samples. Updated through August 27, 2026. Accessed August 31, 2026. Back to text
- Kim S, Lee K, Park HD, Lee YW, Chun S, Min WK. Schemes and performance evaluation criteria of Korean Association of External Quality Assessment (KEQAS) for improving laboratory testing. Ann Lab Med. 2021;41(2):230-239. doi:10.3343/alm.2021.41.2.230 Back to text
- Electronic Code of Federal Regulations. 42 CFR part 493, subpart I: Proficiency Testing Programs for Nonwaived Testing. Up to date as of October 6, 2026. Accessed October 7, 2026. Criteria as amended by Centers for Medicare & Medicaid Services. Clinical Laboratory Improvement Amendments of 1988 (CLIA) proficiency testing regulations related to analytes and acceptable performance. Fed Regist. 2022;87:41194. Published July 11, 2022; amendments to §§493.801 through 493.959 effective July 11, 2024. Back to text
- Clinical and Laboratory Standards Institute. A Quality Management System Model for Laboratory Services. 5th ed. CLSI guideline QMS01. Published June 19, 2019. Back to text
- American Society for Quality. Cost of quality. Accessed August 31, 2026. Back to text