1. IQC
Edward Kearney

3. Why is IQC/EQA important?
To ensure the test result is fit for purpose

4. EQA
Historic
Accuracy
Repeated samples may give precision

5. Who determines if a patient gets a statin?
We do !

6. Who determines if a patient gets a Dx of DM?
We do !

7. Who determines if a patient gets a prostrate biopsy?
We do !

8. Who determines if a patient gets dialysis?
We do !

9. How do we know we are right?
We do !

10. Westgard .com “1575”

11. Fasting Glucose 6.8 mmol/L = IFG Fasting Glucose 7.2 mmol/L = DM
Diabetes
Fasting Plasma Glucose > 7.0 mmol/L
On two or more occasions can make the diagnosis of Diabetes Mellitus
Fasting Glucose 6.8 mmol/L = IFG
Fasting Glucose 7.2 mmol/L = DM

12. Acute Coronary Syndrome
Troponin T
> 0.05 ug/L = Biochemical evidence of cardiac damage
< 0.05 ug/L = No Biochemical evidence of cardiac damage

13. How do we know we are right?
Standardisation
Calibrators
Reagents
Test conditions
Quality Control
External Quality Assurance (EQA)
Internal Quality Control (IQC)

14. IQC Real time Stable process Bias Precision Statistical QC
Electronic QC

15. IQC Materials Mimic Patients sample
Cover normal and pathological range
Ideal – decision points

16. Controlling the quality - IQC
Compare process performance with what is expected under stable operation
Stable performance determined by assaying control material over time : Calculate mean and SD
Measurements are made continually and compared with the original distribution
Unexpected results are identified to alert the analyst to possible changes in process performance

17. Multi-rule QC
Westgard rules

18. Use and Interpretation of Common Statistical Tests in Method-Comparison Studies - all 2 versions »
JO Westgard, MR Hunt - Clinical Chemistry, 1973
Performance characteristics of rules for internal quality control: probabilities for false rejection
JO Westgard, T Groth, T Aronsson, H Falk, CH de … - Clinical Chemistry, 1977
A multi-rule Shewhart chart for quality control in clinical chemistry –
JO Westgard, PL Barry, MR Hunt, T Groth - Clinical Chemistry, 1981
Assuring analytical quality through process planning and quality control.
JO Westgard - Arch Pathol Lab Med, 1992
European specifications for imprecision and inaccuracy compared with operating specifications
JO Westgard, JJ Seehafer, PL Barry - Clinical Chemistry, 1994
Internal quality control: planning and implementation strategies
JO Westgard - Annals of Clinical Biochemistry, 2003
The Quality of Laboratory Testing Today, An Assessment of σ Metrics for Analytic Quality Using Performance Data From Proficiency Testing Surveys and the CLIA Criteria for Acceptable Performance
James O. Westgard, and Sten A. Westgard. Am J Clin Pathol 2006;125:

19. X Conceptual basis of the control chart
Very Unexpected
Somewhat Unexpected X
250
260
245
240
255
235
265
Run Number (or Time, Date)
Determine the expected distribution of control values
Calculate mean and SD from control data to establish control limits for control chart
Expect control values to fall with certain control limits
95% within 2 SD
99.7% within 3 SD
Plot control values versus time to provide control chart
Identify unexpected values
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

20. Take Corrective Action
Flowchart and logic for the multirule IQC procedure commonly known as “Westgard Rules”
QC Data
12s
13s
22s
R4s
41s
10x
Report Results
Take Corrective Action
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

21. IQC 2 SD rule for 1 QC sample
Statistically 1 in 20 will be outside this range
This is not unexpected
5% of runs (false) rejected

22. IQC 2 SD rule for 3 QC samples 18% of runs (false) rejected
This is not unexpected

23. What actually happens?
EM Kearney. The South Thames (East) Quality Assurance Liasion Group (1998)
James O Westgard. Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611
David Housley. North Thames audit and QA group: Audit of Internal Quality Control practice and processes. (2005)

24. 5. NUMBER OF SAMPLES USED TO ASSIGN QC VALUES
1 = provisionally 3, then re-assessed at 15
2 = 10 – 15
1 = 12
11 = 20 (one 20 days ?); one with minimum of 1 month run in (major variations discussed with manufacturer); one re-assess after 1000; some variation in number of batches / days etc
1 =
1 = > 30
3 = at least 50
1 = adjusted every month 9 ( no number stated)
David Housley. North Thames audit and QA group: Audit of Internal Quality Control practice and processes. (2005)

25. EP5-A Clinical and Laboratory Standards Institute (NCCLS) EP5-A
Evaluation of precision performance of Clinical Chemistry Devices: Approved guidelines.
Precision Evaluation Experiment
Recommendations
Minimum of 20 working days
Batch analysers: two runs per day with two test samples at each of at least two levels.
Random access: four samples at each level should be analysed through out the day.
At the end of each 5 days the control limits are recalculated and all data checked for acceptability.
The cause of outliers should be determined.
Data may not be rejected without valid justification.
Learning curve data may be rejected and replaced by a equal number of points at the end.
Other methods:
Not recommended:
Single run of 20 samples for within run
Single observation each day for 10 – 20 days for day to day
Because:
Single runs do not reflect usual operating parameters, thus adversely affecting the estimate.
Single observations will be highly dependent on the number of days used.

26. EP5-A
Random access:
Four samples at each level should be analysed through out the day.
At the end of each 5 days the control limits are recalculated and all data checked for acceptability.
The cause of outliers should be determined.
Data may not be rejected without valid justification.
Learning curve data may be rejected and replaced by a equal number of points at the end.

27. Criteria use to determine acceptability
Written procedure 1
Professional judgement 6
Combination of above 12
EM Kearney the South Thames (East) Quality Assurance Liasion Group (1998)

28. Is the procedure always adhered to?
Yes 2
Mostly 1
No 5
Blank 11
EM Kearney the South Thames (East) Quality Assurance Liasion Group (1998)

29. QC rules used for calcium?
2SD
3SD
12s + Multi-rules 1
Westgard 3
Limits professional judgement, Vitros F2/F3
EM Kearney the South Thames (East) Quality Assurance Liasion Group (1998)

30. Calcium QC rule violated describe the process for dealing with this
Check QC 3
Rerun 12
Recalibrate 4
Inspect QC history 1
EM Kearney the South Thames (East) Quality Assurance Liasion Group (1998)

31. Westgard Rules 1. Check control 2 Make up new 1. Repeat 2 Make up new
1. Repeat 2 Recalibrate
EM Kearney the South Thames (East) Quality Assurance Liasion Group (1998)

33. 9. ARE WESTGARD RULES USED AS PRIMARY FORM OF DATA EVALUATION?
YES = 18 (62%)
NO = 11 (38%)
1 = Mean and target range (< 2SD)
5 = single 2SD rule
1 = single 1.5 SD rule
1 = not formally but simplified Westgard used
1 = under consideration
1 = Instrument set flag points, but no rule stated
1 = 2SD in first instance, then Westgard for detail bias profile
David Housley. North Thames audit and QA group: Audit of Internal Quality Control practice and processes. (2005)

34. REASONS FOR ACCEPTING FAILED QC
Only if not clinically significant – 7 sites
Lack of reagent prevents change
Lack of staff, pressure to reduce TAT, demotivated staff
Checked with EQA
Various. eg. very close to limit, one QC in/ one QC out, QC deterioration before new QC is ready e.g. lot no. change until new range established.
Only if all attempts to correct it have failed
Only after consultation with manufacturer
Only if calibrant returns result to normal
Only if no reason can be found and 1 QC out
eg. If LDH QC is out, the reagent is changed the next day
David Housley. North Thames audit and QA group: Audit of Internal Quality Control practice and processes. (2005)

35. Planning strategies (doing the right IQC)
Define the quality requirement for the test
Determine the method precision and bias
Identify candidate IQC procedures
Predict IQC performance
Select goals for IQC performance
Select an appropriate IQC procedure

36. Operating Specifications (smeas, biasmeas, control rules, N)
A system of quality requirements and operating specifications
Clinical Outcome
Criteria (DInt)
Analytical Outcome
Criteria (TEa)
Operating Specifications
(smeas, biasmeas, control rules, N)
Proficiency
Testing
Criteria
Medically
Important
Changes
Total
Biologic
Goals
Individual
Performance Criteria
SDMax, BiasMax
Diagnostic
Classification
State
of the
Art
Arbitrary
Control
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

37. Power curves for commonly used IQC procedures having two control measurements per run. Pfr, or the probability for false rejection, is given by the y-intercept. Ped, the probability for error detection, depends on the size error occurring, which is illustrated by the vertical line, and estimated by reading the y-value at the point of intersection with the power curve.
Systematic Error (SE, multiples of s)
Probability for Rejection (P)
Power Function Graph (SE)
Pfr given by
y-intercept
Ped depends on
size of
systematic error
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

38. Probability for Rejection (P) Power Function Graph (SE)
Example use of a power function graph to assess IQC performance for a cholesterol test where the medically important systematic error is equivalent to 2.85 times the standard deviation of the method.
Probability for Rejection (P)
Power Function Graph (SE)
Systematic Error (SE, multiples of s)
Probability
of false
rejection, Pfr
given by
y-intercepts
Probability of
error detection,
Ped , depends
on size of
systematic error
Pfr is
0.07 to
0.01
Ped is 1.00 to 0.40
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

39. Probability for Rejection (P) Power Function Graph (SE)
Sizes of systematic errors that can be detected with 90% assurance by different IQC procedures
Probability for Rejection (P)
Power Function Graph (SE)
Systematic Error (SE, multiples of s)
Lines show SE
detectable with
90% chance
by different
QC procedures
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

40. Mathematical basis for a chart of operating specifications that shows the bias and imprecision that are allowable for different IQC procedures, whose control rules and Ns are given in the key at the right.
Allowable Inaccuracy (bias, %)
Allowable Imprecision (s, %)
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

41. Plot your observed CV and bias
Application of the OPSpecs tool by plotting the observed method bias as the y-coordinate and the observed method precision as the x-coordinant to describe the “operating point” of an analytical method.
Allowable Inaccuracy (bias, %)
Allowable Imprecision (s, %)
OPSpecs Chart TEa 10% with 90%AQA(SE)
Plot your observed CV and bias
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

42. A practical procedure for planning and selecting IQC procedures.
1. Define quality required
for diagnostic test, %TEa
2. Assess method
%bias, %CV
3. Calculate normalized
operating point
4. Plot on normalized
OPSpecs charts or EZ rules3
5. Inspect normalized
OPSpecs charts
6. Select control rules,
number of measurements
7. Adopt Total QC strategy
8. Reassess for changes
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

43. Total quality control strategies (TQC) that illustrate the proper combination of IQC with instrument and function checks (Other QC) and need for quality improvement (QI).
HI-Ped
Strategy
IQC
Other QC QI
MOD-Ped
LO-Ped
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

44. A detailed flowchart to guide the development of TQC strategies.
Minimize cost of
statistical QC
non-statistical QC
Maximize
error detection
Improve method
performance
Document TQC
system
Optimize QC for
process stability
Deploy skilled
analysts
Add patient
data QC
Apply QP process
with OPSpecs
charts
AQA or
Ped
HI-Ped (90%)
MOD-Ped (50%)
LO-Ped (<50%)
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

45. Results of an example application of IQC design for a multitest chemistry analyzer.
Probability for Rejection (P)
Critical-Error Graph (SE)
Systematic Error (SE, multiples of s)
Albumin
CO2
Chloride
OTHER TESTS
Crit SE > 4.0
Creatinine, T Bil, UA
Potassium, GGT,
ALP, Phosphorous
BUN, Cholesterol
AST, LD, Glucose
Total Protein
Calcium
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

46. Implementation of a multistage IQC strategy.
Select
QC design
from
drop-down
list
material
Enter
control
result
Program
calculates
z-value
Out-of-control
identified by
red bar
STARTUP
QC design uses
multirule
MONITOR
2.5s limits
displays
on scale
-4 to +4
Multi-chart display
shows multiple materials
and multiple designs with
multi-rule or single-rule
QC procedures
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

47. Validation Reports, Peer Reports
A Total Quality Control Support System that integrate IQC and EQA
Interfaces or
Manual data entry
Laboratory
testing
processes
Internal QC
Accept/Reject
EQA data transfer
EQA
Data Analysis
IQC Planning
And Design
Inservice training
Validation tools
Documentation
Validation Reports, Peer Reports
Toubleshooting
Inventory, ordering
Production, shipping
Supplies & materials
Internet
Support
Services
James O Westgard Internal quality control: planning and implementation strategies Ann Clin Biochem 2003; 40: 593–611

48. What to do? Do define the quality required for the test
Do select QC procedures that minimise false rejections
Do select QC procedures that detect medically important errors
Do adopt modern QC planning tools and techniques
Do standardise QC operations
Do calculate control limits from your own laboratory data
Do provide computer support to analyse and interpret QC data
Do reject out-of-control runs, identify the problem, and eliminate the cause
Do adopt a Total QC strategy to maximise the cost-effectiveness of QC
Do calculate daily/weekly/monthly patient means

49. What not to do? Don’t use 2SD control limits
Don’t just repeat the controls
Don’t use the same control rules for all tests
Don’t use bottle values to calculate control limits
Don’t use medical decision limits as control limits
Don’t rely on electronic QC alone
Don’t eliminate SQC in POC applications

50. Additional methods of IQC
Average of Normals (AoN, Daily patient means)
Standard signals
Experience
Expertise

54. IQM Monitors system performance real time
Automatically detects potential failures that affect the analytical performance
Automatically performs corrective action
Automatically documents failures and actions taken
Provides quality control reports

55. IL GEM Premier 4000