1 / 45

Control Charts

Control Charts. Michael Koch Michael Gluschke. Assuring the Quality of Test and Calibration Results - ISO/IEC 17025 – 5.9. The laboratory shall have quality control procedures for monitoring the validity of tests and calibrations undertaken.

Download Presentation

Control Charts

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Control Charts Michael Koch Michael Gluschke

  2. Assuring the Quality of Test and Calibration Results - ISO/IEC 17025 – 5.9 • The laboratory shall have quality control procedures for monitoring the validity of tests and calibrations undertaken. • The resulting data shall be recorded in such a way that trends are detectable and, where practicable, statistical techniques shall be applied to the reviewing of the results.

  3. Assuring the Quality of Test and Calibration Results - ISO/IEC 17025 – 5.9 • This monitoring shall be planned and reviewed and may include, but not be limited to, the following: • regular use of certified reference materials and/or internal quality control using secondary reference materials; • participation in interlaboratory comparison or proficiency-testing programmes; • replicate tests or calibrations using the same or different methods; • retesting or recalibration of retained items; • correlation of results for different characteristics of an item.

  4. Control Charts • Powerful, easy-to-use technique for the control of routine analyses • ISO/IEC 17025 demands use wherever practicable • It is hard to imagine quality management systems in laboratories without control chart

  5. History • Introduced by Shewhart in 1931 • Originally for industrial manufacturing processes • For suddenly occurring changes and for slow but constant worsening of the quality • Immediate interventions reduce the risk of production of rejects and complaints from the clients

  6. Principle • Take control samples during the process • Measure a quality indicator • Mark the measurement in a chart with warning and action limits

  7. Control Charts in Analytical Science • Assign a target value • Certified value of a RM/CRM (if available) • Mean of often repeated measurements of the control sample (in most cases)

  8. Control Charts in Analytical Science • Warning / action limits • If data are normally distributed • 95.5% of the data are in µ ± 2σ • 99.7% are in µ ± 3σ • ± 2s is taken as warning limits • ± 3s is taken as action limit

  9. Action Limits • There is a probability of only (100-99.7) 0.3 % that a (correct) measurement is outside the action limits (3 out of 1000 measurements) • Therefore the process should be stopped immediately and searched for errors

  10. Warning Limits • (100-95.5) 4.5% of the (correct) values are outside the warning limits. • This is not very unlikely. • Therefore this is only for warning, no immediate action required.

  11. Calculation of Standard Deviation • Measurements marked in the control chart are between-batch • Standard deviation should also be between-batch • Estimation from a pre-period of about 20 working days • Repeatability STD  too narrow limits • Interlaboratory STD  too wide limits

  12. Limits  Fitness for Purpose • Action and warning limits have to be compatible with the fitness-for-purpose demands • No blind use • Limits should be adjusted to fit-for purpose requirements

  13. concentration upper action limit upper warning limit target value lower warning limit lower action limit date 22.06.2006 27.06.2006 12.06.2006 13.06.2006 14.06.2006 15.06.2006 16.06.2006 19.06.2006 20.06.2006 26.06.2006 23.06.2006 21.06.2006 Out-of-control Situation 1 • Suddenly deviating value, outside the action limits

  14. concentration upper action limit upper warning limit target value lower warning limit lower action limit date 22.06.2006 27.06.2006 12.06.2006 13.06.2006 14.06.2006 15.06.2006 16.06.2006 19.06.2006 20.06.2006 26.06.2006 23.06.2006 21.06.2006 Out-of-control Situation 2 • 2 of 3 successive values outside the warning limits

  15. concentration upper action limit upper warning limit target value lower warning limit lower action limit date 22.06.2006 27.06.2006 12.06.2006 13.06.2006 14.06.2006 15.06.2006 16.06.2006 19.06.2006 20.06.2006 26.06.2006 23.06.2006 21.06.2006 Out-of-control Situation 3 • 7 successive values on one side of the central line Not so critical as 1 and 2

  16. concentration upper action limit upper warning limit target value lower warning limit lower action limit date 22.06.2006 27.06.2006 12.06.2006 13.06.2006 14.06.2006 15.06.2006 16.06.2006 19.06.2006 20.06.2006 26.06.2006 23.06.2006 21.06.2006 Out-of-control Situation 4 • 7 successive increasing or decreasing values Not so critical as 1 and 2

  17. Advantages of Graphical Display instead of in a table • Much faster • More illustrative • Clearer

  18. Different Control ChartsX-chart • Synonyms are X-control chart, mean control chart or average control chart • Original Shewhart-chart with single values • Mainly for precision check • For trueness control synthetic samples with known content or RM/CRM samples may be analysed • It is also possible to use calibration parameters (slope, intercept) to check the plausibility (constancy) of the calibration

  19. Different Control ChartsBlank Value Chart • Analysis of a sample, which can be assumed to not contain the analyte (blank) • Special form of the X-chart • Information about • The contamination of reagents • The state of the analytical system • Contamination from environment (molecular biology laboratories) • Enter direct measurements of signals, not calculated values

  20. Different Control ChartsRecovery Rate Chart - I • Reflects influence of the sample matrix • Principle: • Analyse actual sample (unspiked) • Spike this sample with a known amount of analyte (ΔX) • Analyse again • Recovery rate:

  21. Different Control ChartsRecovery Rate Chart - II • Detects only proportional systematic errors • Constant systematic errors remain undetected • Spiked analyte might be bound differently to the sample matrix  better recovery rate for the spike • Target value: around 100%

  22. Different Control ChartsRange Chart • Synonyms are R-chart or Precision chart. • Absolute difference between the highest and lowest value of multiple analyses • Repeatability Precision check • Control chart has only upper limits

  23. Different Control ChartsDifference Chart - I • Uses difference with its sign • Analyse actual sample at the beginning of a series • Analyse same sample at the end of the series • Calculate difference (2nd value – 1st value) • Mark in control chart with the sign

  24. Different Control ChartsDifference Chart - II • Target value: around 0 • Otherwise: drift in the analyses during the series • Appropriate for repeatability precision and drift check

  25. Different Control ChartsCusum Chart - I • Highly sophisticated control chart • Cusum = cumulative sum = sum of all differences from one target value • Target value is subtracted from every control analyses and difference added to the sum of all previous differences

  26. T = 80 s = 2.5 Nr. x x-T Cusum 1 82 +2 +2 2 79 -1 +1 3 80 0 +1 4 78 -2 -1 5 82 +2 +1 6 79 -1 0 7 80 0 0 8 79 -1 -1 30 9 78 -2 -3 10 80 0 -3 20 11 76 -4 -7 10 12 77 -3 -10 13 76 -4 -14 0 0 2 4 6 8 10 12 14 16 14 76 -4 -18 -10 15 75 -5 -23 -20 -30 Different Control Charts - Cusum Chart - II 90 85 80 75 70 0 2 4 6 8 10 12 14 16

  27. 30 30 in control out of control 20 20 10 10 0 0 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 -10 -10 -20 -20 -30 -30  d Different Control Charts - Cusum Chart - III • V-mask as indicator for out-of-control situation • Choose d and  so that • Very few false alarms occur when the process is under control but • An important change in the process mean is quickly detected

  28. Different Control ChartsCusum Chart - IV • Advantages • It indicates at what point the process went out of control • The average run length is shorter • Number of points that have to be plotted before a change in the process mean is detected • The size of a change in the process mean can be estimated from the average slope

  29. Different Control ChartsTarget Control Charts - I • In the contrary to classical control charts of the Shewhart-type the target control charts operates with fixed quality criterions and without statistically evaluated values • The limits for this type of control charts are given by external prescribed and independent quality criterions (fitness for purpose)

  30. Different Control ChartsTarget Control Charts - II • All types of classical control chart (X-chart, blank value, recovery, R-, R%-chart etc.) can be used as a target control chart • A target control chart is appropriate if: • There is no normal distribution of the values from the control sample due to persisting out of control situations (e.g. blank values) • There are not enough data available for the statistical calculation of the limits (rarely analysed parameters) • There are external prescribed limits which have to be applied to ensure the quality of analytical values

  31. Different Control ChartsTarget Control Charts - III • The control samples for the target control charts are the same as for the classical control charts • The limits might be given by • Requirements from legislation • Standards of analytical methods and requirements for internal quality control • The (minimum) laboratory-specific precision and trueness of the analytical value, which have to be ensured • The evaluation of laboratory-internal known data of the same sample type

  32. Different Control ChartsTarget Control Charts - IV • Constructed with an upper and lower limit • Pre-period is not necessary • Out-of-control only, if the analytical value is higher or lower than the respective limit • Nevertheless trends in the analytical quality should be identified and steps should be taken against them

  33. Different Control ChartsTarget Control Charts - V (example) only two limits and one out-of-control situation

  34. EXCEL-Tool for Control ChartsExcelKontrol 2.1 • X-/mean-charts • Blank value chart • Range chart with absolute ranges • Range chart with relative ranges • Recovery rate chart • Differences chart

  35. Control Samples • No control chart without control samples • Requirements: • Must be suitable for monitoring over a longer time period • Should be representative for matrix and analyte conc. • Concentration should be in the region of analytically important values (limits!), if possible • Amount must be sufficient for a longer time period • Must be stable for several months • No losses due to the container • No changes due to taking subsamples

  36. Control SamplesStandard Solutions • To verify the calibration • Control sample must be completely independent from calibration solutions • Influence of sample matrix cannot be detected • Limited control for precision (no matrix effect) • Very limited control for trueness

  37. Control SamplesBlank Samples • Samples which probably do not contain the analyte • To detect errors due to • Changes in reagents • New batches of reagents • Carryover errors • Drift of apparatus parameters • Blank value at the start and at the end allow identification of some systematic trends

  38. Control SamplesReal Samples • Multiple analyses for range and differences charts • If necessary separate charts for different matrices • Rapid precision control • No trueness check

  39. Control SamplesReal Samples Spiked with Analyte • For recovery rate control chart • Detection of matrix influence • If necessary separate charts for different matrices • Substance for spiking must be representative for the analyte in the sample (binding form!) • Limited check for trueness

  40. Control SamplesSynthetic Samples • Synthetically mixed samples • In very rare cases representative for real samples • If this is possible  precision and trueness check

  41. Control SamplesReference Materials • CRM are ideal control samples, but • Often too expensive or • Not available • In-house reference materials are a good alternative • Can be checked regularly against a CRM • If the value is well known  good possibility for trueness check • Retained sample material from interlaboratory tests

  42. Which One? • There are a lot of possibilities • Which one is appropriate? • How many are necessary? • The laboratory manager has to decide! • But there can be assistance

  43. Choice of Control Charts - I • The more frequent a specific analysis is done the more sense a control chart makes • If the analyses are always done with the same sample matrix, the sample preparation should be included. If the sample matrix varies, the control chart can be limited to the measurement only

  44. Choice of Control Charts - II • Some standards or decrees (authority decisions) include obligatory measurement of control samples or multiple measurements. Then it is only a minimal additional effort to document these measurements in control charts • In some cases the daily calibration gives values (slope and/or intercept) that can be integrated into a control chart with little effort

  45. Benefits of Using Control Charts • A very powerful tool for internal quality control • Changes in the quality of analyses can be detected very rapidly • Good possibility to demonstrate ones quality and proficiency to clients and auditors

More Related