8. QUALITY CONTROL

INTRODUCTION TO QUALITY CONTROL

  • Definition: The process of monitoring the characteristics of an analytical system to detect error and prevent the release of inaccurate results.

  • The Goal:
      - Ensure Precision (reproducibility) and Accuracy (closeness to the true value).

  • QC Materials:
      - Utilize control samples with known concentrations (Low/Normal/High) that mimic patient samples.

STATISTICAL FOUNDATIONS

  • Mean (x): The average of the data points.

  • Standard Deviation (SD): Measures the “spread” or imprecision of data.

  • Coefficient of Variation (CV): Expresses SD as a percentage of the mean.
      - Formula: CV=racSDximes100CV = rac{SD}{x} imes 100
      - A lower CV indicates better precision.

QUALITY CONTROL STATISTICS

  • Reference range for a particular measurement is generally related to a normal bell-shaped curve.

MEAN, MEDIAN, AND MODE

  • Mean (Average):
      - Commonly used in laboratory measurements.
      - Calculation: Summing values and dividing by the number of data points.
      - Essential for estimating precision and control levels.

  • Data Collection:
      - Obtain a minimum of 20 data points from separate testing runs.
      - If 20 runs are not feasible, use at least seven runs (three replicates per run).

  • Provisional Ranges: Set using mean and standard deviation (SD).

  • Updating Mean and Limits:
      - Replace mean and limits derived from abbreviated data when 20 separate runs’ data become available.
      - Regular reassessment is crucial for accuracy and reliability.

MEDIAN

  • Definition: Represents the middle value in a dataset.

  • Calculation: Arrange data in order of magnitude and locate the value halfway between the highest and lowest variables.

  • Example: For the list 2, 2, 3, 4, 5, 6, 7, the median is 4.

MODE

  • Definition: The most frequently occurring value in a dataset.

  • Example: In the list 2, 2, 3, 4, 5, 6, 7, the mode is 2.

STANDARD DEVIATION

  • Definition: Measures the spread or variability in a dataset.

  • Calculation:
      - Calculated as the square root of the variance of values.

  • QC Tasks:
      - Establish QC targets and ranges (SD) for assay controls.
      - Determine mean and SD for new QC lots.
      - Revised standards allow calculating target mean using 10 control specimens over 10 days.

  • Statistical Interpretation:
      - In a normal population:
        - 68% of values fall within ±1 SD from the average.
        - 95% fall within ±2 SDs.
        - 99.7% fall within ±3 SDs.

  • Reference Values:
      - Reference interval includes 95% of test results for a healthy population.
      - Replaces terms like “normal values” or “normal range.”

CONFIDENCE INTERVALS

  • Definition: Confidence Intervals (CIs) represent 2 SDs around the mean and encompass 95% of values.

  • Considerations: Acknowledge day-to-day shifts in analytical procedures.

  • Purpose of 95% CI: Accounts for sampling variability and method imprecision.

  • Laboratory Reference Ranges:
      - Manufacturer’s ranges provide initial guidance.
      - Individual labs adjust mean and QC limits based on patient population.

  • Control Material Analysis:
      - New control lots should be analyzed alongside existing material to ensure accurate mean and limits for quality control.

COEFFICIENT OF VARIATION

  • Definition: CV, expressed as a percentage, is the ratio of SD to the mean; it quantifies variability relative to the mean in a data set.

  • Purpose: Normalizes variability, allowing comparison of SDs across different means.
      - Useful for assessing precision differences in assays and methods.

  • Comparison:
      - When comparing SDs, consider the mean to avoid misleading impressions.

  • Control Limits:
      - Setting control limits at ±2 SDs is common; however, it may lead to high false rejection rates.

EXAMPLE OF COEFFICIENT OF VARIATION

  • Calculation:
      - SD=0.36SD = 0.36,
      - Mean=3.14Mean = 3.14
      - % CV:
       ext{% CV} = rac{SD imes 100}{Mean} = rac{0.36 imes 100}{3.14} = 11.5 ext{%}

DETERMINATION OF CONTROL RANGE

  • Control Solution Assessment:
      - Laboratories must determine an acceptable control range for a specific analysis after purchasing an unassayed control solution.

  • Method for Establishing Range:
      - Assay an aliquot of the control serum alongside regular batches of assays over 15 to 25 days, treating the control sample like an unknown specimen.

  • Calculating Acceptable Limits:
      - Repeated determinations yield a normal bell-shaped curve. Calculate mean (x) and standard deviation (SD).
      - Most labs use ±2 SDs from the mean as the allowable range, while others use it as a warning limit.

  • Control Implementation:
      - Include control specimens in each batch once the acceptable range is established.
      - If a control value falls outside the limits, repeat the procedure before reporting patient results.
      - Accreditation bodies mandate written procedures for monitoring and resolving out-of-control situations.

SOURCES OF VARIANCE OR ERROR

  • Achieving identical results for specific specimens is generally impossible due to inherent variability arising from:
      - Sampling Factors:
        - Collection time, patient position, physical activity, fasting duration, and storage conditions may all contribute to variance.
      - Procedural Factors:
        - Aging of chemicals, personal bias, variations in standards, reagents, environment, methods or apparatus can also influence variance.
        - Experimental errors may arise from method changes, instrument variations, or personnel shifts.

LEVEY-JENNINGS CHARTS

  • Definition: The primary visual tool for monitoring QC over time.

  • Components:
      - X-axis: Time/Date
      - Y-axis: Concentration, marked with the Mean, ±1SD, ±2SD, and ±3SD.
      - Value: Enables technicians to see shifts and trends immediately before they lead to out-of-control situations.

DAILY CONTROL SPECIMEN VALUES

  • Laboratories are required to plot these values on a Quality Control (QC) chart.

  • Many modern instruments automatically generate QC charts daily, flagging out-of-control results.

  • Purpose of Control Charting:
      - Ensures reliable and stable laboratory processes and identifies unacceptable runs and deviations effectively.

DRIFT DETECTION

  • If a drift is detected, checks should include reagent age, calibration status, and potential adjustments to the mean.

GLUCOSE LEVEL EXAMPLE

  • Levey-Jennings Graph for Glucose Level: Marks concentration against time, demonstrating the process for visualizing QC data.

CONTROL SAMPLES

  • QC involves using at least two different control samples for a specific analyte, including both normal and abnormal control specimens.

  • Control Limits:
      - The mean value and acceptable error limits are typically set at ±2 SDs or ±3 SDs from the mean.
      - 2-SD serves as a warning limit; 3-SD acts as an action limit.

  • Daily Monitoring: Allow for easy identification of values that fall “out of control,” helping detect trends or drift over time.

VALIDATION OF NEW PROCEDURES

  • Laboratories validate new procedures before routine use, determining reproducibility and confidence limits while establishing acceptable variation limits for control specimens.

  • Quality Control (QC) Program:
      - Calculates mean and SD for each procedure and generates control charts for performance monitoring.
      - Regular assessments help detect issues promptly and correct them timely.

SHIFTS, TRENDS, AND DISPERSION IN LEVEY-JENNINGS CHARTS

  • Shifts:
      - Definition: A sudden and sustained change in one direction in control sample values.
      - Indication: May be caused by sudden instrument malfunction.

  • Trends or Systematic Drift:
      - Defined as gradual change in control sample results over time.
      - Indication: If direction consistently shifts from the mean for at least 3 days, signaling potential issues like reagent deterioration.

  • Dispersion:
      - Refers to increased random error or lack of precision, which can undermine data reliability.

WESTGARD RULES (MULTI-RULES)

  • Multi-rules are employed before reporting patient data and designed to detect random and systematic errors, maximizing detection ability while minimizing false rejection rates.

WESTGARD RULES (“THE DECISION TREE”)

  • 1₂S: One control exceeds 2SD, often used as a warning.

  • 1₃S: One control exceeds 3SD, indicating random error; the run must be rejected.

  • 2₂S: Two consecutive controls exceed the same 2SD limit, pointing to systematic error.

  • R₄S: The range between two controls in the same run exceeds 4SD (e.g., one QC is +2SD, the other is -2SD), indicating random error.

  • 4₁S: Four consecutive results exceed the same 1SD limit, which signifies systematic error.

  • 10x: Ten consecutive results fall on the same side of the mean, indicating a shift or systematic error.

TABLE OF WESTGARD QUALITY CONTROL RULES AND INTERPRETATION

  • 22s Rule:
      - One control observation exceeding the mean ±2s may be used as a warning rule to initiate additional testing.

  • 13s Rule:
      - One control observation exceeding the mean ±3s recommends rejecting patient results, sensitive to random error.

  • 2s Rule:
      - Two consecutive control observations exceeding the same mean plus or minus 2s recommends rejecting patient results, sensitive to systematic error.

  • R4s Rule:
      - One observation exceeding the mean plus 2s and another exceeding the mean minus 2s recommends rejecting patient results, sensitive to random error.

  • 4s Rule:
      - Four consecutive observations exceeding mean plus or minus 1s recommend rejecting patient results, sensitive to systematic error.

  • 10x Rule:
      - Ten consecutive control observations falling on one side of the mean recommend rejecting patient results, sensitive to systematic error.

VIOLATIONS OF QUALITY CONTROL RULES

  • Figures visualize various violations of established control rules.

  • Guidelines regarding violations help maintain QC integrity and reliability in lab practices.

MULTIRULES FOR THREE CONTROLS

  • Purpose: Monitor lower and upper analytical range critical for clinical values.
      - Level 1: Low (Abnormal) - monitors sub-normal concentrations.
      - Level 2: Normal - validates routine sample accuracy.
      - Level 3: High (Abnormal) - tests linearity and protects against toxic levels.

RESOLUTION FOR OUT-OF-RANGE CONTROLS

  • Laboratory policies dictate procedures for managing out-of-range controls, detailing steps to identify and correct issues before reporting patient results.

RATIONALE BEHIND THE RULES

  • Multirules integrate individual rules with lower false rejection rates, raising error detection rates.

  • 12s Rule Discussion: 95% of results fall within 2 SDs; 5% exceed it, raising concern regarding true errors.

  • 13s Rule Discussion: 99.7% of results fall within 3 SDs; a violation indicates a probable error, balancing false rejection and detection effectively.

  • Multirules are recommended for increased complex error detection capability.

EXTERNAL QUALITY ASSESSMENT (eQA)

  • Definition: A process where an external agency sends “blind” samples to the laboratory.

  • The Process: Labs treat eQA samples like patient samples and submit results for comparison against peer group means.

  • Purpose: Ensures long-term accuracy and inter-laboratory comparability.

INTERNAL QC vs EXTERNAL QA

  • Features:
      - Internal QC: Conducted daily; aims for immediate actions regarding report decisions.
      - External QA: Done periodically; focuses on long-term accuracy and bias detection through peer comparisons.

ANALYZING eQA RESULTS

  • SDI (Standard Deviation Index): A common metric for evaluating eQA performance.
      - Formula: SDI=racLabresult−PeerGroupMeanPeerGroupSDSDI = rac{Lab result - Peer Group Mean}{Peer Group SD}
      - Interpretation:
        - 0.0: Perfect agreement
        - ±1.0 to 1.5: Acceptable performance
        - > ±2.0: Marginal; requires investigation
        - > ±3.0: Unacceptable; requires immediate corrective action

eQA PROVIDERS

  • IQMH: Common in Ontario for proficiency testing and accreditation services.

  • QMP-LS: Associated with provincial quality mandates.

  • One World Accuracy: Provides oversight for tests not available from IQMH.

  • CAP: An international “gold standard” for specialized testing.

CORRECTIVE ACTION FOR eQA FAILURES

  • A systematic investigation is necessary when a “red flag” (unacceptable result) is reported.

  • Steps involve:
      - Clerical Check: Confirm accuracy in transcription and units.
      - Specimen Integrity: Ensure the proper handling of eQA samples.
      - Instrument/Reagent Review: Assess documented shifts or QC failures on testing day.
      - Peer Group Analysis: Verify if others in the peer group experienced similar results.
      - Technical Competency Check: Confirm adherence to SOPs by lab staff.

DOCUMENTATION AND RESOLUTION

  • The Paper Trail:
      - Every eQA failure requires a Corrective Action Report (CAR).
      - Root cause analysis must identify the reason for variance and outline preventative actions.
      - Final sign-off is done by the lab manager or medical director prior to review.

CASE STUDIES

Case Study #1: The Chemistry Shift

  • Scenario: Glucose control level 2 above mean for 9 days, today at +2SD.

  • Question: Which Westgard rule is violated and what is the likely cause?
    Answer: Possibly a 10x warning, indicating systematic error, likely due to reagent lot change.

Case Study #2: The Random “Hiccup”

  • Scenario: WBC Level 1 at -2SD, Level 2 at +3SD.

  • Question: What is the rule violation and immediate action?
    Answer: This is a 1₃S violation indicating random error. QC should stop, check for bubbles or clots, and repeat the QC.

Case Study #3: The eQA Surprise

  • Scenario: Lab result of Sodium at 138 mmol/L, peer group mean 142 mmol/L, peer group SD 1.0.

  • Question: Calculate the SDI.

  • Answer: SDI=rac138−1421.0=−4.0SDI = rac{138 - 142}{1.0} = -4.0 indicating unacceptable performance, requiring a thorough review.

SUMMARY

  • Levey-Jennings Chart: Visualizes control data, aids in detecting shifts and trends.

  • Westgard Rules: Key for identifying anomalies in control values.

  • Problem-Solving Approach: Utilize tools to troubleshoot and maintain analytical accuracy.

  • Patient Care: Quality control is ultimately about ensuring patient care and safety.