CLINICAL CHEMISTRY LABORATORY QUALITY CONTROL: STATISTICS & QC CHARTS OUTLINE • Statistics o Descriptive Statistics o Inferential Statistics • QC Charts o Gaussian Curve o Youden/Twin Plot o Shewhart Levey-Jennings Chart o Westgard Chart o Interpretation of QC Result • Terminologies STATISTICS • Quality control data are analyzed by statistic and presented through QC Charts DESCRIPTIVE STATISTICS • Used for monitoring test performance and quality control o Commonly used in the laboratory • Assessment of data dispersion allows laboratorians to assess predictability in a laboratory test or measurement • Involves: o Measures of center • Mean - average • Median – middle data after the data have been ranked order • Mode – the most frequently occurring data o Measure of dispersion ▪ Standard deviation ▪ Coefficient of Variation ▪ Variance • Mean ( x̄) o aka Average o Measure of central tendency o Summation of X (σ x) o The only measure of accuracy o Formula o Sample problem: What is the mean of the control values of glucose obtained after 5 consecutive days? • Standard Deviation (SD) o Measure of dispersion of values from the mean. o Measure of distribution range. o It helps describe the normal curve. o Best measure of random error (indeterminate error) o Ideal : 2SD or lower ▪ Reject when above the ideal o Measures precision ▪ Inversely related to SD ▪ ↑SD = dispersed values = ↓ precision ▪ ↓SD = values close to each other = ↑ precision o Formula n-1 = degree offreedom o Sample problem: What is the standard deviation of the control values of glucose obtained after 5 consecutive days? • Coefficient of Variation (CV) o Percentile expression of the mean o Index or 2nd measure of Precision o Ideal: 2 to 4% o Formula o Sample problem: What is the coefficient of variation of the control values of glucose obtained after 5 consecutive days if: • Variance (V) o aka standard deviation squared o Square of dispersion o Measure of variability o Represents difference between each value and mean o Formula: V = (SD) 2 o Sample problem: What is the variance of the control values of glucose obtained after 5 consecutive days if: SD = 2.65 V = (SD) 2 V = (2.65) 2 V = 7.02
INFERENTIAL STATISTICS • Used to compare means or SD of two groups of data • Used to determine significant difference between two groups of data** • Commonly used in experimental and quantitative research • T-test o used to determine whether there is a statistically significant difference between the means of two groups of data • F-test o used to determine whether there is a statistically significant difference between the SD of two groups of data QC CHARTS • Used to observe control values over time to determine reliability of the analytical method. • Utilized to observe and detect analytic error such as inaccuracy and imprecision. GAUSSIAN CURVE • Aka Bell-Shaped Curve, Normal Frequency/Distribution Curve • Obtained by plotting the values from multiple analyses of a sample. • Bell-shaped curve – occurs when data elements are centered around the mean with most elements close to the mean • 1.0/100% - total area under the curve • ** • Values from -2SD to -3SD & 2SD to 3SD are invalid CUMULATIVE SUM GRAPH (CUSUM) • Calculate the difference between QC results and the target means • Common Method: V-mask method • Identifies the earliest indication of systematic problem • Very sensitive to small, persistent errors. • >45 ̊ slope (>2.7 SD) – out-of-control! YOUDEN/TWIN PLOT • Used to compare results obtained on a high and low control serum from different laboratories. • Displays results: o Y-axis – mean values of one specimen o X-axis – mean values of another specimen • Detects systematic error: o Proportional Error – points fall on the center but on the 45 ̊ line. o Constant Error – points fall on the center but NOT on the 45 ̊ line SHEWHART LEVEY-JENNINGS CHART • aka Dot Chart • A graphical representation of the acceptable limits of variations in the results on an analytical method. • Only control values are plotted in LJ Chart o Most widely used QC chart in the clinical laboratory. • Best Internal QC chart o Detects both random and systematic errors o Allows laboratorians to apply multiple rules without computer aid
ERRORS OBSERVED IN LJ CHART • Trend o Gradual loss of reliability in the test system o Seen visually in LJ Chart ▪ at least 6 consecutive increase or decrease in control values o Cause: ▪ Deterioration of reagent ▪ Deterioration of light source/control materials. • Shift o Shift in reference range is due to transient instrument differences o at least 6 consecutive control values situated on either side of the mean. ▪ Neither decreasing or increasing ▪ Do not cross the mean line o Cause: ▪ Improper Calibration of machine ▪ Change in reagent formulation • Outlier o Control value is far from the main set of values o Highly deviating value (≥3 +/- SD) o Caused by random or systematic error in analytical method WESTGARD CHART • aka Multiple Rule Chart • States that the use of simple upper and lower control limits are not enough to identify analytical problems • Error detection can increase without increasing the false rejection rate. • Control rule – indicates if the analytical process is out-of- control. WESTGARD CONTROL RULES • 1 2s o One control value exceeds the mean +/- 2SD. o Warning Rule (for screening purposes) ▪ Only rule that do not require repeat • 1 3s o One control value exceeds the mean +/- 3SD. o Determines random errors o If occurs, repeat quality control • 2 2s o Two consecutive control values exceed the mean +/- 2SD. o Determines systematic errors • 4 1s o Four consecutive control values exceed mean +/- 1SD. o Determines systematic errors
• R 4s o The range or difference between two control values exceed 4SD. o Determines random errors o Similar to LJ Chart Outlier • 6 x o 6 consecutive control values fall on one side of mean. o Do not cross the mean line o Similar to LJ Chart Shift • 7 T o Seven control values trend in the same direction (progressively increasing or progressively decreasing). o Similar to LJ Chart Trend • 8 x o 8 consecutive control values fall on one side of mean. o Do not cross the mean line • 9 x o 9 consecutive control values fall on one side of mean. o Do not cross the mean line • 10 x o 10 consecutive control values fall on one side of mean. o Do not cross the mean line • 12 x o 12 consecutive control values fall on one side of mean. o Do not cross the mean line • 3 1s o 3 consecutive control values exceed mean +/- 1SD. • 2 of 3 2s o If 2 out of 3 consecutive control values exceed mean +/- 2SD • Reject control run if it meets any of westgard rule criteria (except 1 2s ) • Multirule Shewart (LJ Chart) – control chart + control rules INTERPRETATION OF QC RESULT • +/- 2SD – acceptable reference limit • If control value exceed +/- 2SD – run new set of control and repeat specimen testing** • If control value is between +/- 2SD and +/-3SD – sign of potential problem • If control value is outside +/- 3SD – perform corrective action o If the problem is already identified – retest new set of control material
• Continuous QC failures requires more troubleshooting: o Preparation of new reagents o Recalibration o Instrument maintenance and repair TERMINOLOGIES • Delta Check o Patient-based quality control technique o The difference between two consecutive measurements of the same analytes on the same individual o It requires computerization of test data so that current results can be compared with past results. • Analytical Run o A set of control and patient specimens assayed, evaluated, and reported together. • Reference Value o aka Reference Limit, Reference Interval o previously termed as “Normal value” o Range of values into which 95% of non-diseased individuals will fall o The usual values of a healthy population that represents 95% central tendency. HOW TO CREATE A SHEWART LEVEY-JENNINGS CHART • Create a Shewhart Levey-Jennings Chart for the following control values of creatinine performed for 15 consecutive days. • Step 1: Tabulate the Control Values o Tabulation of data must be in order of analyses (day 1 to day x) • Step 2: Calculate for the Mean and SD. o You may compute using scientific calculator or using EXCEL. • Step 3: Calculate the upper and lower control limits (±1SD to ±3SD) • Step 4: Form your chart o X-axis – day 1 to day x ▪ number of control values o Y-axis ▪ Upper Control Limits • +1SD, +2SD, +3SD ▪ Mean ▪ Lower Control Limits • -1SD, -2SD, - 3SD • Step 5: Plot your values inside the chart. o Make sure that the values are within the range. • Step 6: Analyze your chart using Westgard Multirule System. o To detect random and systematic errors. o To detect accuracy and precision.