Introduction: Improving Patient Safety Through Analytic Process Design

1
(No Transcript)
2
Introduction Improving Patient Safety Through
Analytic Process Design

• Joe Boone, PhDAssociate Director for
  ScienceNational Center for Health Marketing
• September 29, 2005

3
Opportunities for Improvement in Quality and
Safety in Healthcare

• Medical errors are 8th leading cause of death
• Cost 17 billion US per year
• To Err is Human, Institute of Medicine,
  September 1999
• 50/50 chance of proper health care
• Elizabeth A. McGlynn, Ph.D., et al.The Quality
  of Health Care Delivered to Adultsin the United
  States. NEJM, June 26, 2003
• Up to 15 of patients in five countries in past
  two years got incorrect test results or delays
  in being notified of abnormal test results
• Patients Report Barriers to Primary Care,
  Lab Test Errors or Delays, International
  Survey of Five Nations Health Affairs,
  October 28, 2004

4
Errors in Laboratory Medicine, Minireview
Bonini et al. Clinical Chemistry 2002 485,
691-698

• Focus mostly on analytical errors
• Inappropriate test ordering and interpretation is
  difficult to capture
• Errors are often difficult to detect
• 75 of test errors are for results within the
  reference interval
• 12.5 absurd and 12.5 effect on patient health
• New tests may not have a gold standard

5
Errors in Laboratory Medicine, Minireview
Common Elements of Studies

• No common nomenclature blunder, mistake,
  problem, defect, error
• Similar distribution of errors in Total Testing
  Process
• Method of data collection influences results
  Complaints lt Testing Process Review
• No definition of allowable error rate
• Bonini et al. Clinical Chemistry 2002 485,
  691-698

6
The Publics Health System
7
Institute for Quality in Laboratory
MedicinePromoting an Integrated System of
Services within and across boundaries
Across boundaries of public / private
organizations
Within institutions
Pre-analytic Analytic Post-Analytic
8
Laboratory Professionals
Clinicians
Patients
Improve Laboratory Testing and Services
Manufacturers
Government
Accrediting Bodies
Payers
9
Institute for Quality in Laboratory Medicine
Promote and Facilitate
Create an awards and grants program
Develop networks of laboratories and partners
Laboratory Networks
Awards
Mentoring and Implementing
Research
Quality Indicators
National Report
Develop indicators and monitor progress
Identify issues and best practices
Conferences and Pilot Studies
10
IQLM Progress since May 2005Abbott Web-cast

• Inaugural IQLM Conference
• Progress by Workgroups
• Incorporation of IQLM
• Call for Nominees for Board of Directors
• Newsletter
• Conference Proceedings Available
• iqlm.org
• Conference Highlights in MedGenMed
• medscape.com/viewarticle/506782

11
Improving Patient Safety Through Analytic
Process Design

• David L. Witte, M.D., Ph.D.

12
Test Cycle Must Focus on Outcome
13
Lab Medicines Many Rights

• Offering the RIGHT tests
• Sampling the RIGHT patient at the RIGHT time
• Using the RIGHT tube and labeling RIGHT
• Testing with RIGHT accuracy, precision,
  sensitivity, specificity
• Doing the test RIGHT with RIGHT controls
• Report results to RIGHT physician at the RIGHT
  time
• Providing the RIGHT reference information
• Evaluating with RIGHT outcome time and accounting
  frames

14

• GENIUS is hiding your sources (E/K)
• THEFT is stealing one idea (10C)
• SCHOLARSHIP is stealing many ideas (?)

15
Analytical Result is the Sum of

• Calibration errors (Bias)
• Random errors (imprecision)
• Interferences (specificity)
• All are determined by TEST CHOICE

16
Variation is Either

• COMMON CAUSE the usual variation in
• calibration or precision (SD)
• SPECIAL CAUSE a larger variation than usual,
• with a different explanation
• (BLUNDER RATE)

17
Method Imprecision

• Common Cause Variation
• measure by control s.d.
• likelihood of crossing cutoff (Acland)
• basis for analytical goals (C. Faser)
• impact on ability to control patient
• Special Cause Variation
• systematic change e.g. calibration
• random e.g. blunders

18
Common Cause ErrorsT. Badrick. Clin Bioch Rev
2481

• Systematic
• Method drift
• Instrument bias
• Reagent lot bias
• Calibration bias

• Random
• Matrix interference
• Mechanical variation
• Electrical interference
• Photometer/detector variation

19
Special Cause Variation

• Large errors of low frequency
• Increased by complexity
• Increased by weak points in process
• Reduced by design (mistake-proofing)
• Frequency not well known

20
Large Error in Labs
Steindel Arch Path 19961201094 1996 Clin Chem
Forum Witte Clin Chem 1997431352
21
Distributions and Blunders
Gaussian
Observe
Special cause variation exists
22
Others Have Studied

• 1 defect in 500 products below 250 ppm
• 211 ppm defects 90 pass the n500 tests
• WHAT HAVE VENDORS DESIGNED?
• ANSI/ASQC 21 4-1993, table X-N-1

23
Calibration Variability

• PSA
• Cholesterol
• INR
• Heparin (APTT)

24
Calibration Variability

• patients exceeding limit for action
• Population distribution
• Variance of calibration
• Random errors
• What can vendors deliver
• Can laboratorians monitor population
• Median result
• Fraction above cut-off

25
Calibration VariabilityG. Klee, Scan J. Clin Lab
Invest 199959509
26
PSA Population Stability (CLS)for values 0.1 to
10.0
27
Checking for Errors?

• LIMIT beyond values expected
• DELTA big change from previous
• INTER not consistent with other results
• eg anion gap, globulins, Ifts, BUN CRT
• GROUP not the averages expected
• CAN WE DESIGN LOCAL CRITERIA?

28
Response to Errors

• Over-reacting to common cause variation usually
  increases the variation
• Responding to special cause errors will help
  reduce complexity and weak process parts

29
Reduce Analytical Variation

• Purchaser plus vendor goals
• Reduces misclassification of patients
• Acland Lipton J. Clin Path 196720780
• C. Fraser goal vs. biological variation
• Ability to keep drug or INR in range
• B. Steele on drugs
• Lassen on INR
• Hemoglobin A1c?
• Always strive to lower SDs (common cause)

30
Challenges to Lab Medicine

• Have we understood the common cause variation in
  our calibration?
• Do we have a measure of our blunder rate?
• Do we have adequate means to identify important
  shifts in the process?

31
What you Measure (test selection)

• D-dimer-quantitative/qualitative
• Chlamydia-antigen/gene
• HCV or HIV-antibody/gene
• Calculated Tests-GFR and AVG-GLU
• Examples could be endless

32
All Tests Have 4 Results
Condition
Present
Absent
TP
FP
Pos
Test
FN
TN
Neg
33
D-dimer for EXCLUSION

• High sensitivity to minimize false negatives
• Specificities vary more by test (40-70)
• Lower pre-test probability essential

34
D-dimer One Example
DVT Present?
NO
YES
19 5 TP
75 FP
Test abnl
Negative
3 FN
302 TN
Found in 6 mos. Follow-up Ann Int Med
2005142490
35
D-dimer Second ExampleTest Wells Score low
intermediate
DVT Present?
NO
YES
106 TP
417 FP
Test abnl
Negative
8 FN
297 TN
Ann Int Med 2005143100
36
Two Studies of D-dimer and DVT
37
Two Studies of D-dimer and DVT
38
D-dimer Test Choice

• Influences who can be tested
• Influences frequencies of results
• FP increase with poorer specificity
• FP predominate at low pretest probability
• FN goal is below 2 of all Negatives (PE)
• Chest 20031241116

39
D-dimer Systematic ReviewBMJ 30 July 05, 331

• PTP allowing FN rate of 2 varies
• Sensitive test usable to 40 PTP
• Specific test usable to 10 PTP
• No test both sensitive and specific
• PTP is pretest probability
• estimates from graphs of Figure 5.

40
D-dimer a Local Decision

• Reduce Radiology? Maximize TN
• No Misses? Minimize FN
• Choice of Analytical System Matters

41
Patient/Control Discordance

• I have too many neutropenias, I know you have a
  problem even if controls are okay
• 3 of 4 of my patients have T3 over 250 even
  though your controls are okay

42
Total Triiodothyronine Saga (TT3)

• 8/22/03
• 8/29/03
• 8/20/04

• Four ordered TT3 232-281
• TT3 controls SDI .52?.81
• New reagent formulation began
• 43 wellness samples
• Range 101 ? 279 13 over 200
• Previous reference range 87 ? 178
• Tubes are a problem
• Resolution to follow

43
Derm Patient Neutropenia
PMN qc SDI is 0.9 (0.05 PMNs) remains
unexplained
44
Be All We Can Be?

• HbA1c predicts average glucose?
• MDRD equation predicts GFR?
• Globulins over 4 g/dL are rare?

45
Globulin T.P. minus Albumin

• 99.8 of wellness samples below 4.0 g/dL
• 96.4 of ordered panels below 4.0 g/dL
• 98.7 of ordered panels below 4.1 g/dL
• CLS data

46
MDRD for GFR?
Rule Ann Int Med 2004141929
47
DCCT HbA1c and MEAN Glucose

• Glu 1.98 (HbA1c) 4.29 r0.82
• highest point for 8 A1c 290 mg/dL
• mean point for 8 A1c 208 mg/dL
• lowest point for 8 A1c 135 mg/dL
• most points for 8 A1c 165-260
• What standardization and precision?

avg glucose in mmol/l, multiply by 18 Rohlfing
Diab Care 200225275
48
Avoid Risk Homeostasis

• Reduction in risk A induces taking risk B
• Munich CABS with ABS temporary reduction
• Sweden to right lane temporary reduction
• Gladwell New Yorker, Jan 22, 1996, p. 32 Blow-Up

49
Avoid False Reliance on Technology

• METHOD
• manual tight sequence
• random with barcodes
• barcode plus sequence

• MISDIRECTED SAMPLES
• 1/3000 or 330 ppm
• 30/3000 or 9900 ppm
• 1/3000

Gambino Lab Report 19891189
50
What to Wish for?

• reduce common cause variation
• reduce complexity
• eliminate weak points
• invent checking processes
• limit, delta, interresult, group
• invent repeat test criteria
• increase customer feedback

51
NEW IDEAS?

• If SD analysis is 1/20th current, what qc?
• If special cause most important, qc?
• Increase redundancies and feed back loops?
• Limit check
• Delta check
• Inter check
• Group check
• Would preventive system design predominate?

52
Design in Analytical Improvement

• Right Analyte
• Right Precision
• Right Cal Variability
• Reduce Blunder Rate
• Increase Analytical Specificity
• Connect Test to Right Outcomes

53
QA