[Typ]Presentation[Sbj]LaboratoryDiagnosisDefined[Dte]20131028
- 1. THE LABORATORY’S ROLE IN
DIAGNOSIS AND DIAGNOSTIC DECISION MAKING
Mark Gusack, M.D., Staff Pathologist
VA Medical Center, Huntington, WV
Adjunct Clinical Professor
Marshall University Medical School
NORTHEAST LABORATORY CONFERENCE
15 OCT 2013
RISK
- 2. Mark Gusack, M.D., OCD [Overly Concerned Doctor]
Staff Pathologist,
Pathology and Laboratory
Medicine Service
VA Medical Center
Huntington, WV 25704
mark.gusack2@va.gov
SPEAKER
Dr. Gusack has almost 40 years experience in the Laboratory field starting as a Nuclear Medicine
Technologist in the early 1970’s, then working as a clinical engineer, and then becoming a
physician and pathologist. He is AP/CP boarded, has held positions in a variety of hospital and
reference based laboratories as a medical director and as staff pathologist. During this time he
has also been a consultant and practiced as a Licensed Health Care Risk Manager in Florida. Dr.
Gusack has been involved with all aspects of laboratory development and management including
startup, licensing, as well as designing integrated management systems for clinical laboratories.
The opinions expressed in this presentation are those of the author and do not necessarily
represent those of the Department of Veterans Affairs
- 3. OUR MISSION
HOW DO WE CARRY OUT THIS MISSION?
OUR MISION IS
TO PROVIDE
THE VERY BEST
CARE TO OUR
PATIENTS
- 4. SITUATION – PART I
HOW HAS THIS AFFECTED WHAT WE DO IN THE LABORATORY?
The meaning of diagnosis has changed radically over the past 60 – 70 years.
THEN:
In the past, there were few diagnostic tests and fewer effective treatments.
A diagnosis was a label placed on a set of signs and symptoms with a prognosis.
Since most therapy was ineffective error in diagnosis was rarely a critical issue.
NOW:
Since that time our fund of knowledge has grown prodigiously.
Our ability to apply this knowledge to health care has transformed medicine.
An error in diagnosis has serious consequences given more effective therapy.
THE CONSEQUENCES:
The patient can be harmed or die due to a false positive or negative diagnosis.
Increased morbidity and mortality in a population can cost a society greatly.
Increased numbers of failure in the face of unprecedented success.
- 5. SITUATION – PART II
SO WE HAVE INITIATED A CYCLE OF:
Increased subdivision of older diagnoses into numerous newer diseases.
Increase capacity to make an accurate diagnoses.
Increased ability to effectively treat more diseases.
LEADING TO:
A dramatic increase in the complexity of diagnostic medicine.
A dramatic increase in the perceived impact of a diagnosis.
CONSEQUENCES:
Increased expectations for a “perfect” outcome.
Increased apparent error in diagnosis.
Increased perception of harm by the patient and society.
Now standards of Acceptable Risk are much more stringent
- 6. SITUATION – PART III – THE RISE OF CHRONIC DISEASE
THIS HAS DRAMATICALLY CHANGED THE MEDICAL PROFESSION
Today most acute diseases have been eliminated through:
Civil Engineering: Waste, water, and food safety management
Improved Diet: Improving our capacity to survive an illness
Modern Medicine: Improved therapies to cure acute illness
As a result today we are dealing primarily with:
Chronic, progressive, debilitating diseases for which
There is often no definitive curative therapy which
Cost far more to treat than acute illnesses which
Escalate over time as the patient progressively declines in health
Present national budget problems are due,to a great degree to this progression
And what WE do will determine, to a great degree the final outcome.
- 7. PROBLEM
How can we reduce the impact of
Systematic Complexity of the Diagnostic Process
on how we generate test results in order to assure establishing:
Acceptable Risk For our Patients?
- 8. SOLUTION PART I
THIS IS NOT GOING TO BE EASY
In order to establish Acceptable Risk we need to:
Define what the term diagnosis means
Understand the difference between the concept of a diagnosis
and the process of making a diagnosis
Delineate what a diagnosis implies clinically and societally
Define the logic by which diagnoses are made
Determine what affects how reliable our diagnoses can be
Define the role of the laboratory test in the diagnostic process
- 9. SOLUTION PART II
Application of Evidence Based Medicine [EBM] to the
implementation and management of laboratory methods and
instrumentation
Is the best approach to achieving an:
Acceptable Balance
Between benefits and risks when applying complex diagnostic
modalities to our patient’s best interests. That is:
Acceptable Risk
- 10. IMPLEMENTATION
TO IMPLIMENT THIS WE WILL HAVE TO BE THE LEADERS
Successful implementation of EBM requires both an educational and
organizational effort to bring together the following activities:
CRITICAL REVIEW OF THE SCIENTIFIC LITERATURE
OBJECTIVE VALIDATION THROUGH LOCAL IMPLEMENTED ACTIVITIES
REALISTIC MEASUREMENT OF USEFULNESS OF THESE ACTIVITIES
MONITORING AND REPORTING OF TEST RELIABILITY AND ITS IMPACT
Into a single integrated structure that assures close cooperation between those
engaged in improving health care today:
Basic Researchers
Vendors
Clinicians
Laboratory
Government Regulatory Agencies
- 11. THE AXIS OF EVIDENCE BASED MEDICINE
HOW DO WE BEGIN THE PROCESS?
AXIS GOAL MEASURABLE OBJECTIVE
SCIENCE ESTABLISH SCIENTIFIC FACT REPRODUCIBLE PROTOCOLS
ANECDOTAL EVIDENCE VALIDATE AT THE PATIENT LEVEL MONITOR PATIENT OUTCOMES
SOCIETAL REFERENCE FRAME FIT TO SOCIETAL EXPECTATIONS ACCEPTANCE BY PATIENTS
VALUES
GENERAL
PATIENT
The Definition of Evidence Based Medicine varies greatly. This is the most applicable
- 12. DEFINITIONS – DIAGNOSIS THEN
A label applied to a cluster of seemingly related signs and symptoms that lead to
the same or similar clinical outcomes. The components of a diagnosis were:
Presenting signs and symptoms
Natural progression of illness
Final outcome of recovery, chronic morbidity, and/or mortality
Therefore, there was:
Little or no need for laboratory testing and
Little in the way of therapy and so
Little need to monitor patient response or to guide patient management.
- 13. DEFINITIONS – DIAGNOSIS NOW
TODAY THE LABORATORY IS CENTRAL TO THE DIAGNOSTIC PROCESS
A cluster of one or more of related illness complexes:
Signs
Symptom
Vital Measurements
Radiologic Studies
Pathologic diagnoses, and
Laboratory findings
That leads to [to name a few]:
Clinical categorization
Prognosis
Therapy
Therapeutic monitoring for outcomes
There is now great need for laboratory testing at every step of the process.
- 14. PURPOSE OF A TEST IS NOW MULTIFACETED
FROM THIS WE CAN SEE THAT PRESENT METHODS OF VALIDATION ARE INADEQUATE
There are eight main purposes of a laboratory test :
PURPOSE DESCRIPTION
SCREENING To identify a high risk subpopulation of people for definitive diagnostic workup
RISK Stratification of subpopulation with increased probability of progression to disease
DIAGNOSIS Establishing or ruling out a specific actionable cause for an illness complex
CLASSIFICATION Determining the type of pathophysiologic mechanisms causing the illness
PROGNOSIS Determining probable outcome if disease is left untreated
STAGING Stratification for extent and aggressiveness of disease at time of diagnosis
TREATMENT Determining most effective therapeutic modality
MONITORING Observing outcomes to confirm diagnosis, document progress, and validate
effectiveness of therapy
Each type of activity will require a different level of minimum resolving power and
reliability both acutely and over long periods of time between diagnosis and final
disposition of the patient.
- 15. WHAT TYPE OF TEST IS NEEDED?
IT IS CLEAR THAT TESTS MUST MATCH THEIR INTENDED PURPOSE(S)
ACTIVITY WHAT HOW DETERMINED
SCREENING Identify high risk population Highly sensitive tests
DIAGNOSIS Filter out those without disease Highly specific tests
PROGNOSIS Rank by risk for morbidity and mortality Highly discriminatory tests
STAGING Rank by degree of appropriate therapy Tests that allow stratification
MONITORING Determine therapeutic effect/diseases recurrence Tests that detect small changes
Below is a brief example of what type of test would be needed for some of the
purposes it might be applied to:
- 16. TEST VALIDITY
There are four main types of laboratory test validation to consider
TYPE DESCRIPTION
ANALYTICAL Defines the ability of a test to reliably measure the analyte of interest
CLINICAL Defines the ability of the test to establish if a disease state is present or not present
UTILITY Defines the ability of a test to improve clinical outcomes
IMPLICATIONS Societal implications and perceptions of the test’s value
In essence, the validity of any particular laboratory test is determined by a number
of interacting, synergistic, and conflicting issues that we face in today’s healthcare
environment.
So, the laboratory needs to understand all of these and to choose an optimal
course of action when validating and monitoring test methodologies.
- 17. VALUE OF A TEST
Once validity has been established the actual value of a test needs to
be addressed.
There are three main aspects to determining the value of a test:
Degree of association of a test result with a disease state
The effectiveness of a test in decision making activities
Utility in establishing/monitoring a the pathophysiologic state
- 18. There is very important clinical risk information implied by each test result that is acted
upon by our clinicians often without fully understanding the implications…and circularity:
BY DEFINITION, LAB TESTS THEMSELVES MEASURE CLINICAL RISK!
SO; TESTS ARE A SOURCE OF RISK AND MEASURE RISK…
TYPE RISK DESCRIPTION
Reference ranges [intervals] LOW Population based screening for presence or risk of
one or more diseases or physiologic states often in
the asymptomatic patient.
Medical Decision Points [MDP’s] MODERATE Statistical cut off for presence of a single disease
or clinical state requiring timely treatment or
modification of treatment - INSULIN
Alert [Critical/Panic] Values HIGH Physiologically dangerous levels of an analyte
requiring immediate evaluation and possibly
intervention – OVER INSULINIZATION.
How a test is applied for each level of risk is based on the following:
Clinical trials in the literature and/or personal anecdotal experience
Practice standards boards and deemed institutions
Cost benefit analysis by insurers and risk managers
This determines the degree of reliability a test must have so that clinicians have
confidence it meets their standard of Acceptable Risk
- 19. CATEGORIZATION OF RISK IN LABORATORY TESTING IS CRITICAL
Risk specifically related to a laboratory testing is determined at three levels:
IN THIS PRESENTATION WE WILL SEE HOW THIS CAN AFFECT US
THEORETICAL OPERATIONAL TECHNICAL
Analyte Significance Sensitivity Accuracy
Relative Value Specificity Precision
Basic Research Clinical Trials Laboratory Use
AVOID THE
WRONG TEST
PREVENT
INAPPROPRIATE USE
MITIGATE EFFECT OF
ANALYTICAL ERROR
NOTE: ALTHOUGH THERE IS OVERLAP BETWEEN EACH LEVEL
The type of risk that should be managed:
AVOIDANCE Is this test even worth commercializing?
PREVENTION If so what do we need to do to minimize adverse events?*
MITIGATION Where prevention is not enough can we monitor and correct?
IS THIS WHAT ACTUALLY HAPPENS?
*While maximizing utility value in diagnosis and management
- 20. ASSESSMENT OF CLINICAL VALUE FOR SYMMETRICAL BINARY MEDICAL DECISION POINT
4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5
Frequency Distribution of Dichotomous Patient Population
A1c
No Diabetes Type II Diabetes
TRUE
NEGATIVES
NOTE: In a perfect world the two populations are symmetrical and equal in prevalence with very small
overlap leading to high sensitivity and specificity as well as positive and negative predictive value
TRUE
POSITIVES
FALSE
POSITIVES
FALSE
NEGATIVES
IN ACTUAL FACT, THE TWO DISTRIBUTIONS ARE NOTHING LIKE THIS
- 21. ASSESSMENT OF CLINICAL VALUE FOR ASYMETRICAL BINARY MEDICAL DECISION POINT
4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5
Frequency Distribution of Dichotomous Patient Population
A1c
No Diabetes
Type II Diabetes
NOTE: Although small on this graph, the absolute number of false positives will be unacceptably high
when the MDP is set at the intersection of the two distribution curves.
TRUE
NEGATIVES
TRUE
POSITIVES
FALSE
POSITIVES
FALSE
NEGATIVES
SO WHERE SHOULD THE MDP BE PLACED?
- 22. THE EFFECT OF TEST BIAS AND IMPRECISION
THEREFORE IT IS CRITICAL THAT THE LABORTORY ESTIMATE RELIABILITY AND REPORT THIS
Assuming that we have an optimally placed MDP:
Systematic upward bias of A1c will:
Increase false positive diagnosis and therefore:
Systematic increase in diagnosis of healthy persons leading to
Increase glycemic therapy leading to
Increased hypoglycemic events and short term morbidity and mortality
Systematic downward bias of A1c will:
Increase false negative diagnoses and therefore:
Systematic decrease in diagnosis of diabetic persons leading to
Decrease glycemic therapy leading to
Increased hyperglycemic events and long term morbidity and mortality
- 23. THE RESPONSIBILITY OF OUR CLINICIANS
THEREFORE IT IS CRITICAL THAT THE LABORTORY EDUCATE OUR CLINICIANS OF THIS NEED
We can estimate and report out reliability data until the cows come home.
However, until our clinicians:
Recognize the significance of this data and
Begin to critically evaluate the establishment of Medical Decision Points
We will not have the critical information regarding the actual risk for adverse
outcomes that will occur for any combination of:
Placement of the Medical Decision Point
Resulting Sensitivity and Specificity
Prevalence of the Disease being Diagnosed
Types of Adverse Outcomes at Risk
Frequency of Adverse Outcomes at each MDP/Reliability pair
Severity of each Adverse Outcome
So we will not be able to reliably establish Acceptable Risk
- 24. AND ASSIGNMENT OF THE MDP ESTABLISHES CLINIAL RISK(S) TAKEN
WE NEED TO KNOW WHAT THE CLINICAL RISK IS BEYOND THE MDP
Clinical trials determine that, an analyte rises the number of adverse outcomes rises.
However, the trials do not determine specifically which patient is at risk.
At any level there will be a certain incidence of events.
As Upper MDP is moved intervention will be increased outcome1/ Outcome2
As Upper MDP is moved intervention will be decreased outcome1/ Outcome2
Risk for a Outcome 1
Risk for Outcome 2
POPULATION
FREQUENCY
ANALYTE
Risk for a Outcome1
Risk for Outcome2
Medical Target Point
Medical Decision Point to
Increase Intervention
Zone of Balanced Risk
Medical Decision Point to
Decrease Intervention
An idealized presentation with a Gaussian distribution of patient test results for any particular intervention
assuming all risks are of equal severity.
- 25. THE MDP DOES NOT DEFINE AN ALL OR NOTHING CLINICAL RISK
WHAT IF THE DISTRIBUTION OF RISK ISN’T GAUSSIAN?
Not all patients will have an event the moment their analyte rises above the upper MDP.
Instead, there will be a distribution of significant events occurring over a range above the
MDP with a maximum incidence of events beyond which there are less and less patients
who have not yet had a bleed.
Distribution of Risk for a
Outcome1
OUTCOME 1
FREQUENCY
ANALYTE
Medical Target Point
Medical Decision Point to
Decrease Intervention so as
to Maximize Benefit while
Minimizing Risk of Outcome1
An idealized presentation with the affect of comorbid states not considered.
- 26. THE DISTRIBUTION OF ADVERSE EVENTS DEFINES CLINICAL RISK
COMPARE THIS TO SKEWING IN THE OPPOSITE DIRECTION
If the distribution of adverse events is this way then
Acceptable Total Error must be very small.
Otherwise even a small upward bias will greatly increase the risk for Outcome 1.
Distribution of Risk for
Outcome 1
OUTCOME 1
FREQUENCY
ANALYTE
Medical Target Point
Medical Decision Point to
Decrease Intervention so as
to Maximize Benefit while
Minimizing Risk of Outcome 1
An idealized presentation with the affect of comorbid states not considered.
.
- 27. THE DISTRIBUTION OF ADVERSE EVENTS DEFINES CLINICAL RISK
AND WE NEED TO KNOW WHAT THE CUMMULATIVE RISK IS!
If the distribution of adverse events is skewed this way then the
Acceptable Total Error can be much larger because
Even a moderate upward bias will not greatly increase the risk for a bleed.
Distribution of Risk for
Outcome 1
OUTCOME 1
FREQUENCY
ANALYTE
Medical Target Point
Medical Decision Point to
Decrease Intervention so as
to Maximize Benefit while
Minimizing Risk of Outcome 1
An idealized presentation with the affect of comorbid states not considered.
- 28. CUMMULATIVE RISK PROVIDES A MEANS OF CLINICAL DECISION MAKING
FINALLY, A USEFUL VISUAL CONCEPTUALIZATION OF RISK
Note, depending on the width and height of the “Outcome” curve, there will be a point of inflection
and asymptotic rise. While attempting to maximize benefits of therapy, the MDP should to be placed
before this rise to prevent the majority of adverse outcomes.
Finally, we have a means of determining more exactly a maximal total error we can tolerate
OUTCOME 1
FREQUENCY
ANALYTE
1.0 2.5 3.5
Medical Decision Point to
Decrease Intervention
2.0
An idealized presentation with a Gaussian distribution of bleeding risk with distribution of ANALYTE
Medical Target Point
Cumulative Clinically
Observed Frequency for
Outcome 1
In Addition, the steepness
of the slope of the
Cumulative Distribution will
Greatly Affect Degree of
Risk at the MDP.
Distribution of
Laboratory Results
Around the MDP can
be compared to risk
frequency to
determine
acceptability.
- 29. A CLOSE UP OF THE EFFECT OF IMPRECISION TAKEN ALONE
THIS SHOWS TWO RISK COMPONENTS – CLINICAL & LABORATORY
Below is the case where there is no bias but significant imprecision causing many results to lie far
from the actual patient value that, in this case, lies just above the MDP. If we know the actual
frequency at any point then we can determine overall risk.
Cumulative Clinically
Observed Frequency for
Outcome 1
OUTCOME 1
FREQUENCY
ANALYTE
An idealized presentation with a Gaussian distribution of test results.
Distribution of Laboratory
Results Around the Patient
Value
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
Actual Patient ANALYTE
Medical Decision Point to
Decrease Intervention
Cumulative Frequency
of Patients with Test
Results that are
Erroneously Below the
MDP Placing them at
Risk for Over Treatment
and Outcome 1.
Cumulative Frequency
of Patients with Test
Results that are
Erroneously High and May
Place them at Risk for
Under Treatment and
Outcome 2.
Frequency for Outcome 1 at
1 and 2 SD’s Away from the
MDP can be Calculated .
- 30. A CLOSE UP OF THE EFFECT OF BIAS TAKEN ALONE
THIS SHOWS TWO RISK COMPONENTS – CLINICAL & LABORATORY
Below is the case where there is significant bias but no significant imprecision causing many results to
lie to one side of actual patient value that, in this case, above the MDP. This type of error leads to
systematic over treatment and, therefore risk for a Outcome 1.
Cumulative Clinically
Observed Frequency for
Outcome 1
OUTCOME 1
FREQUENCY
ANALYTE
An idealized presentation with a Gaussian distribution of test results.
Distribution of Laboratory
Results Around the Patient
Value
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
Actual Patient ANALTYE
Medical Decision Point to
Decrease Intervention
Cumulative Frequency
of Patients with Test
Results that are
Systematically Below the
MDP Placing them at Risk
for Over Treatment and
Outcome 1.
Frequency for a Outcome 1
at 1 and 2 SD’s Away from
the MDP can be Calculated .
- 31. ASSESSMENT OF RISKS – CUMULATIVE FREQUENCY OF EVENTS/OUTCOMES: A1C
4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5
.20
.30
.40
.50
.60
.70
.80
.90
Cumulative Frequency of Event/Outcome
A1c
1.00
.10
Hypoglycemic
Events
Hyperglycemic
Events
Overlap Zone
NOTE: Access to this quantitative data would allow for us to determine an overlap zone for placement
of our Medical Decision Point to minimize one or the other Event/Outcome as needed.
CUMULATIVE FREQUENCY IS CRITICAL FOR RATIONAL DECISION MAKING
- 32. ASSESSMENT OF RISKS – CUMULATIVE DEGREE OF SEVERITY: A1C
4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5
.20
.30
.40
.50
.60
.70
.80
.90
Cumulative Degree of Severity [Severity x Frequency]
A1c
1.00
.10
Severity of
Hypoglycemic Event
Severity of Hyperglycemic
Outcome(s)
Overlap Zone
NOTE: Developing a useful ranking system would allow for us to determine an overlap zone for
placement of our Medical Decision Point to minimize severity.
Mortality
Morbidity
CUMULATIVE SEVERITY IS CRITICAL FOR PERSONAL AND SOCIETAL DECISION MAKING
- 33. In order to fully integrate the clinical and laboratory rolls in choosing and utilizing test
methodologies in the diagnosis and treatment of disease we need to revisit that part of
FMEA that causes us the greatest problem:
Calculating a Risk Priority Number [RPN] to assign levels of severity upon which decision
making could be more realistically be based in defining Acceptable Risk.
Although there is no one best way to approach this matter, we can create models upon
which to base further clinical and laboratory research to validate their applicability.
Below is an approach I recommend. I don’t claim that it is necessarily the best.
However, if nothing else, this approach clearly delineates all of the problems we face in any
particular medical activity whether Point-of-Care Testing or in utilizing other studies in
classifying a patient’s clinical status upon which to base further diagnostic and/or therapeutic
efforts.
By doing this we define the limits of our knowledge and our capabilities
CLINICAL BASIS FOR DETERMINING SEVERITY OF A RISK
CONSIDER A COMBINED QUALITATIVE/QUANTITATIVE MODEL
- 34. We really need a way to easily categorize risk by clearly applicable criteria that can
incorporate clinician perception. I find that the following two descriptive sets, when
combined as a matrix, provide a more operationally useful approach:
TYPE OF RISK: [Including but not necessarily limited to]
Pain, Discomfort, Malaise, GI Upset, Etc.
Physical Damage – Structural and Cosmetic
Functional Limitation – Orthopedic, Muscular, Neurologic, Cognitive, Immunologic, etc.
Physiologic – Endurance, Handling of Catabolic Waste, Organ Functionality
Psychologic – The Impact of all of the Above Factors
DEGREE OF ADVERSITY FOR EACH TYPE:
Morbidity – Temporary – Relatively Short Lived with Full Resolution
Morbidity – Persistent – Intermediate Term Course with Full or Partial Resolution
Morbidity – Permanent – Long Term Course with Partial or No Resolution
Mortality – Distant, Delayed, or Immediate – Due to Disease or Complications
COMPONENTS OF A SEVERITY OF A RISK ASSESSMENT
THESE RISK AXIS’ PROVIDE A MEANS OF OBTAINING CLINICIAN RANKING
- 35. Below is my proposed stratification of severity creating a 10 point scale that is open ended
allowing the medical community and society to define the descriptive terms:
PROPOSED MODEL FOR DETERMINING SEVERITY OF A RISK
RANKING SYSTEMS OFTEN IMPLY A LINEAR PROGRESSION. IS THIS APPROPRIATE?
RANK RISK STRATIFICATION DESCRIPTION
0 No significant affect
Morbidity - Temporary Pain, Physical, Functional, Physiologic, Psychologic
1 Mild
2 Moderate
3 Marked
Morbidity – Persistent Pain, Physical, Functional, Physiologic, Psychologic
4 Mild
5 Moderate
6 Marked
Morbidity – Permanent Pain, Physical, Functional, Physiologic, Psychologic
7 Mild
8 Moderate
9 Marked
Mortality By Disease or By Complications of Therapy
10 Distant
10 Delayed
10 Immediate
- 36. PROPOSED TRANSLATION OF THE SEVERITY “CURVE”
IT SHOULD LOOK LIKE FALLING OFF A CLIFF OR OVER A WATER FALL
The severity value assigned to each succeeding severity rank should increase at an
accelerating rate until the value reaches its maximum at death. If we also translate
frequency from linear to a power function then multiplying them with detection rate to
obtain a Risk Priority Number would result in a more realistic valuation to define
Acceptable Risk.
By applying a reversed power
function such as a parabolic
or other similar function we
more accurately assign a
severity value to each rank.
From this an FMEA Risk
Priority Number can be
calculated to use in
positioning the Medical
Decision Point and, along
with the cumulative
distribution of adverse
outcomes, develop a Maximal
Total Error Target.
SEVERITY
VALUE
SEVERITY
RANK
A Linear valuation
would not match the
perceived loss on the
part of patients, family,
and society.
It over values mild levels
of severity and under
values marked levels of
severity.
An idealized presentation with the affect of comorbid states not considered.
0 1 2 3 4 5 6 7 8 9 10
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
- 37. Below is a proposed set of steps our clinicians should take in establishing Acceptable Risk before we
can attempt to determine what level of Acceptable Analytical Error we need to meet in choosing and
implementing a test methodology:
CLINICAL BASIS FOR DETERMINING ACCEPTABLE RISK
ONCE THIS IS ESTABLISHED THEN CRITERIA FOR TEST METHODOLOGY IS IN PLACE
# TASK
CLINICAL RESPONSIBILITIES
1 Define disease to be investigated clinically.
2 Define diagnostic/therapeutic criteria.
3 Determine optimal laboratory analyte to test criteria clinically.
4 Define therapeutic modalities that will be initiated or withheld based on test results.
5 Determine benefits of therapy versus withholding therapy.
6 Determine risks of therapy versus withholding therapy.
7 Assign Severity Rank and determine Severity Value.
8 Measure frequency of adverse outcomes and determine the Risk Priority Number.
9 Determine affects of comorbid states on frequency and on severity as needed.
10 Assign Medical Decision Point(s) based on Cost/Benefit analysis of above information.
11 Determine maximum variance around the MDP that avoids Unacceptable Risk.
12 Establish maximum total Acceptable Analytical Error for the laboratory.
LABORATORY RESPONSIBILITIES
13 Evaluate Available Testing Systems for Acceptable Analytical Methodology.
14 Carry out appropriate validation studies and reject or implement test methodology.
- 39. VALIDATION OF INSTRUMENTS SHOULD INCLUDE:
CLINICAL:
Determine specifically what your facility is going to use the test for.
Determine the environment in which the instrument will be used.
Establish clinical standards for Acceptable Risk for that facility if possible.
LABORATORY:
Validate by linear regression/bias plot to establish reliability over a range of values
Validate by modified Cohen’s statistic at MDP’s to estimate our part of risk
COOPERATIVE:
Determine if the instrument meets clinician perception for Acceptable Risk
If so, implement with full orientation, training, and procedural documentation
Initiate period of full monitoring to assure operational reliability
Monitor estimated technical error to identify any significant shifts
Validate clinical protocol to assure it works in THAT facility.
RECOMMENDED STEPS IN ESTABLISHING POC TESTING
AN INTEGRATED APPROACH BETWEEN CLINICAL AND LAB STAFF IS BEST
- 40. Ideally, the laboratory should travel down the following path to the point where it can
evaluate clinical risk with great confidence.
RECOMMENDED STEPS IN VALIDATING TESTING METHODOLOGIES
THIS REQUIRES AN INTEGRATED SOFTWARE APPLICATION TO ASSIST US IN VALIDATION
STEP QUESTION TO ANSWER TYPE STATISTICAL TOOL RECOMMENDED*
1 Is the data unimodal? MDP Data Distribution Plot
2 Is the data normally distributed? MDP χ2-Test
3 Is estimated precision low? MDP SD CV / Range
4 Is estimated accuracy high? MDP Reference – Mean / Reference - Median
5 Are the SD’s the same? COM F-Test
6 Are the Means the same? COM T-Test
7 Is there a linear relationship? COM Regression
8 What is the range of linearity? COM Bias Plot evaluation
9 What is the overall reliability? COM Correlation Coefficient
10 What is the concordance at MDP’s? MDP Simplified Cohen’s Kappa
* NOTE: There are many others in the literature.
- 41. SUGGESTED FLOW CHART TO ASSURE CLINICAL RELIABILITY
ACCEPTABLE RISK SHOULD DRIVE CLINICAL AND LAB TEST MONITORING
Selection of Analyte
Selection of Test Methodology
Clinical Application Defined
Risks for Each State Defined
Clinical States Defined
Acceptable Risk Defined
Technical Error Limits Defined
Validation of Methodology
Acceptable Performance?
YES
NO
Develop Utilization Policies
Institute Adjustments if Needed
Write Technical Procedures
Publish Estimated Reliability
Orient and Train Users
Monitor Analytic Reliability
Monitor Clinical Results
Carry out Ongoing Upgrades
Retire at End of Life Cycle
Risk Distribution(s) Defined
- 42. PUTTING IT ALL TOGETHER
AN INTEGRATED APPROACH IS BEST
In his 1989 article on Controlling Error in Laboratory Testing Donald Forman, Ph.D puts it
all together in an integrated overview of problems which is applicable to all Clinical
Laboratory Testing:
“…correct interpretation of test results requires knowledge of analytical and biological as
well as pathophysiological sources of variation, their expected magnitude, and the time
course over which changes can occur in health and disease.”
In just one sentence Dr. Forman includes all the critical factors we as laboratorians and our
clinicains must consider when implementing, measuring, reporting, and applying any test
analyte result.
The operational and technical problems we face in doing so are very significant. Add in
cultural resistance on the part of laboratorians and physicians and you have a serious
obstacle to improvement in the laboratory.
Forman, D.T., MLO Supplement 1989.
- 43. THERE IS MUCH STILL TO DO:
It is clear that, today despite miraculous advances in the knowledge and application of that
knowledge to the clinical laboratory we are still faced with significant challenges when it
comes to the appropriate validation of the tests we place in the hands of our clinical staff.
KNOWLEDGE AND UNDERSTANDING IS LACKING:
Presently our clinicians are not educating themselves adequately in the fundamental science
underlying these tools in order to understand in detail their strengths and weaknesses and
this has lead to health care by cook book and not medicine by clinical acumen placing the
laboratory in a difficult situation for expected accuracy and precision.
WHAT IS TO BE DONE?
If we are going to maintain our position as a highly regarded profession that is dedicated to
service to our clinicians and patients then we must begin to rethink how we apply all this
technology to our efforts at clinical evaluation and management. Otherwise, soon, we will
be submerged under a collective system of government regulations and accreditations
standards that may do more harm than good.
CONCLUSIONS AND WARNINGS
FINIS! QUESTIONS?