Harvey & Stockham - Stats, QA/QC - AKG Flashcards
When determining reference intervals and a Gaussian distribution is present, a minimum of ___ individuals should be assayed for statistical validity.
40
____ is the likelihood of a positive or abnormal test result occurring in animals with the disease being considered.
Sensitivity
*trust your negatives - negative result means you can rule the disease OUT
_____ is the likelihood of obtaining a negative or normal test result in a nondiseased animal.
Specificity
*trust your positives; if positive - you can rule the disease in
Calculate Sensitivity (%)
(TP x 100)/(TP + FN)
*calculates likelihood of a positive or abnormal test result occurring in animals with the disease (denomater = total diseased animals = TP + FN)
Calculate Specificity (%)
TN x 100/(TN + FP)
*likelihood of obtaining a negative or normal test in nondiseased animals (denominator = total healthy animals = TN + FP)
Please define and calculate PPV (aka PVPT).
Positive predictive value considers only animals in the population being studied that have a positive test result and determines what % of animals actually have the disease being considered.
PPV = TP x 100/ (TP + FP)
(denominator = all positive test results)
Please define and calculate NPV (aka PVNT).
Predicted value of a negative test (NPV) consideres only animals in the population being studied that have a negative or normal test results and determines what % of animals with negative test results do not have the disease being considered.
How likely is that an animal with a negative or normal test result will be free of the disease being considered?
PVNT (NPV) = TN x 100/ (TN + FN)
Define prevalance of a disease. Calculate using TP/FP/TN/FN
The percentage of animals in a given population that have a certain disease.
(TP + FN) x 100/ (TP + TN + FP + FN)
What affects prevalance:
A) Predictive values
B) Sensitivty and specificity
A) Predictive values
A low prevalance will keep the PVPT ___ and PVNT ____. The exception is tests with rare false positives, like PCR tests.
low and high
Tests with high positive predictive values are needed for a rule__ strategy when significant hazards are associated with treatment or euthanasia is being considered.
Tests with a high negative predictive value are theoretically important as a rule-___ startegy when the penalty for missing a dx is high (as with a disease for which terhapy is effective if begun quickly).
rule-in
rule-out
The best evidence for ruling out a disease is finding a negative test result for an assay that has a high _____ for the recognized disease.
sensitivity
____ is determined by how closely the result approaches the true value of the analyte being measured.
Low accuracy are said to have a negative or positive ____ based on if results are below or above the true value, respectively.
Bias
the ____ of a test reflects how reproducible the test results are when the assay is replicated.
precision
Precision is independent of accuracy. T/F
True
What is the coefficient of variation (CV)?
the amount of imprecision (aka random error) present in an assay
More specifically, it is the standard deviation of the repeated measurements expressed as a percent of the mean of the repeated measurements (SD/mean x 100)
It can be measured over time intervals to assess within-run, between-run, or between-day variation.
CV depends on accuracy. T/F
False. It measures impercision (aka random error), and therefore depends on precision.
What do Bland Altman plots show?
Describe the y and x axis.
What do the horizontal lines mean?
Bland-altman plot (aka difference plot) is a graphical method to compare two measurement techniques.
The differences between the two techniques (usually y axis) are plotted against the averages of the two techniques (usually x axis). Alternatively, the differences can be plotted against one of the two methods, if this method is a reference or “gold standard” method.
Horizontal lines are drawn at the mean difference, and at the limits of agreement (SD). The line of equality is at 0 on the y axis (where measurements of method 1 = method 2).
The graph is an example of ______ error/bias.
proportional
Interpret this bland-altman plot
Variation of at least one method depends strongly on the magnitude of measurements.
The graph is an example of ______ error/bias.
constant
This test is a robust, nonparametric method for fitting a straight line to two-dimensional data where both variables, X and Y, are measured with error (y = bx + a). It is useful when you have two devices that should give the same measurements and you want to compare them.
Passing Bablok regression for method comparision
How do you interpret Passing Bablok regression for method comparision ?
If 0 is within the confidence interval (CI) of the y-intercept, then there is no evidence of constant bias.
If 1 is within the CI for the slope, then you infer that there is no proportional bias.
If both are met, then the two methods are comparible within the investigated concentration range.
EDTA chelates which ions?
Calcium, Mg, Cu, and Pb
EDTA serves as an anticoagulant becuase it chelates calcium, which is a cofactor for many parts of the coagulation cascade. What is the main mechanism of action for heparin as an anticoagulant?
Heparin chelates calcium as well, but it’s major action as an anticoagulanat is by activating antithrombin III –> this will inhibit coagulation factors including thrombin.
What are the three major disadvantages of using heparin for blood smear evaluation?
- Alters morphologic features and staining of leukocytes
- Allows clotting as effects are slowly overriden by the coagulation system
- Allows platelet clumps to form
Plasma consists ~ 95% water and ~ 5% solids. List some of the components that contribute to total solids.
- Mostly protein
- GLucose
- Urea
- Electrolytes
When evaluating serum, when should you centrifuge the blood sample?
After clot retraction - usually takes ~30 minutes naturally, but clot activator will speed up the clotting time
Which is higher, serum [K+] or plasma [K+]?
Serum [K+] - cells (especially platelets) will release intracellular contents during clotting
RIs take this into consideration, and this is only a problem if patient has thrombocytosis
The major protein (on a weight/volume basis) that is absent in serum but present in plasma is _____.
Fibrinogen
How many significant figures do the numbers below have:
- 124
- 124.0
- 0.12
- 0.020
- 120
- 124 - 3 SF
- 124.0 - 4 SF
- 0.12 - 2 SF
- 0.020 - 2 SF
- 120 - 2 or 3 SF
When performing calculations, the answer should have the _____ (fewest or most) significant figures present in the values used in the calculation.
fewest
e.g. 1.23 x 2.4 = 2.952 –> report as 3.0 becuase 2.4 has the fewest significant figures at 2.
If values have a Gaussian distribution, the mean, median, and mode values will be _____.
Equal
If the reference distribution is positive skewed, the tail of the graph pulls out to the _____ and the peak of hte graph is shifted to the ____. List in order, highest from lowest value, the mean, median and mode.
right
Left
Mode > median > mean
If the reference distribution is negatively skewed, the tail of the graph pulls out to the left and the peak of the graph is shifted to the right. List in order, lowest from highest, the mean, median and mode.
mean < median < mode
Sample collection and sample handling are examples of _____ errors.
Preanalytical
List examples of analytical error
- Method appropriate for species
- Quality of instruments and equipment
- Quality of reagents
- Quality of laboratory technique
- Quality control program
Transcriptional erros and incorrect interpretation are examples of _______ errors.
Postanalytical
When evaluating and comparing laboratory assays, five properties can be assessed. What are those five properties?
- Analytical precision
- Analytical accuracy
- Analytical specificity
- Analytical sensitivity
- Detection limit
Analytical ____ is the ability of an assay to get the same result if a sample is analyzed several times. Aka random analytical error or reproducibility.
precision
*analytical impercision is expressed by CV
Calculate CV %
Standard deviation/mean x 100
A ____-assay CV represents the random error that is expected when one sample is analyzed multiple times in one run of an assay.
A _____-assay CV represents the random error within one run of the assay plus the error from additional runs of the assay by using the same sample.
Which of the assays will be smaller?
within-assay
between-assay
within-assay will be the samller value
An assay’s CV will typically vary with the analyte’s concentration. Higher CV values may be found at the lower and upper limits of the assay’s analytical range. WIthin the analytical range, CV values are typically _____ at the lower analytic concentrations becuase CV values are expressed at percentages.
higher
E.g., assay has a CV of 10% at all concentrations –>
- then assay’s SD would be 0.1 mg/dL at an analyte concentration of 1 mg/dL
- and 10 mg/dL at an analyte concentration of 100 mg/dL
E.g., assay has a CV of 10 mg/dL at all concentrations
- CV would be 50% at a mean concentration of 20 mg/dL
- CV would be 10% at a mean of 100 mg/dL
_____ is an average of SD values from 3-6 consecutive months of quality assurance values.
USD
If change in patient’s data is < 2 x USD –> the changes can be explained by _____ variation alone
If the change in patient’s data is > 3 x USD –> the changes are likely due to _____ variation.
analytical variation
biological variation
How are control solutions used to assess precision?
Repeated analysis of control solutions
Analytical _____ is defined by the closeness of the agreement between the measured value of an analyte and its ‘true’ value.
accuracy
How do you test for analytical accuracy?
Typically assessed by comparison of the assay’s results to results of an accepted reference method, a method that has been accepted by a standardization group as providing a true value.
You can measure the analyte using a reference standard solution whose concentration was determined by a reference method
Analytical _____ defines the ability of an assay to detect only the substance of interest (analyte) or freedom from interfering substances.
Specificity
Analytical specificity is related to analytical ____.
accuracy
_______ is defined by the smallest concentration or quantitiy of an analyte that can be detected with reasonable certainty for a given analytical range.
Detection limit
Analytical ______ is defined by the slope of the calibration curve and the ability of an analytical procedure to produce a change in the signal for a defined change of the quanitity
sensitivity
in other words - how much change of the analyte is needed for the assay to detect the change
In Westgard rules, what does 12s rule mean?
A run is rejected when a single control measurement exceeds the mean plus 2 SDs or the mean minus 2SD of previous control sample values, or it serves as a warning system to initiate additinol inspection of control data.
In Westgard rules, what does 22s rule mean?
A run is rejected when two consecutive control measurements exceed the same mean plus 2SD or the same mean minus 2SD of the previous control sample values.
In Westgard rules, what does 13S rule mean?
A run is rejected when a single control measurement exceeds the mean plus 3s or the mean minus 3s of previous control sample values.
In Westgard rules, what does R4s rule mean?
A run is rejected when one control measurement in a group exceeds the mean plus 2s and another exceeds the mean minus 2s
In Westgard rules, what does 41s rule mean?
A run is rejected when four consecutive control measurmements exceed the same mean plus 1s or the same mean minus 1s control limit.
In Westgard rules, what does 10x rule mean?
A run is rejected when ten consecutive control measurements fall on one side of the mean
Which westgard rules are sensitive to systematic error vs random error?
- 13s
- 22s
- R4s
- 41s
- 10x
- 13s - random error
- 22s - systematic error
- R4s - random error
- 41s - systematic error
- 10x - systematic error
A common method of monitoring the results of the control samples is to plot them in a ________. These are used to monitor values, but do not help us decide what is acceptable or unacceptable variation.
Levey-Jennings control chart
Sigma stands for the greek letter σ - it stands for _____ and is a statistical index used to express random deviation from the mean.
standard deviation
A sigma classifcation for a test refers to its _____ relative to the amount of acceptable error. The higher the sigma value, the ____ (more or less) precise and ____ (more or less) random error there is.
A sigma classifcation for a test refers to its precision relative to the amount of acceptable error. The higher the sigma value, the more precise and less random error there is.
For a theoretical assay B that has a TEa of 10 mg/dL, analysis of control solution with an analyte concentration near the decision threshold reveals a SD of 2 mg/dL. Assuming there is no systematic error (bias), this assay woudl be called a ______-sigma assay. In this system, the ideal assay is is a ____-sigma assay because unacceptable error would occur only 0.002 times per million assays.
five
six
What does TEa stand for?
allowable total error - calculated based on biological variatoins and the decision threshold
What is the difference between NCCLS Bias plot and the bland-altman bias plot?
The bland-altman plot plots the mean of methods 1 and 2 on the x axis, while the NCCLS Bias plot plots values for one method on the x axis. The NCCLS bias plot is preferend when the second method is being compared to a reference method or an established method.
The NCCLS bias plot or the bland-altman bias plot displays a positive bias when mean difference is > 0 and a negative bias when mean difference is < 0. If there is a difference between the results that is constant over all values, then this is considered ____ bias. If there is a difference between the results that changes proportionately with manigtude of the result, this is considered _____ bias.
constant
proportional
Describe the graphs in a Deming method comparison and passing bablock method comaparison
How do you interpret the graphs?
Values obtained form new method - y axis
Values obtained from old method or comparison method - x axis
Equation is formed for the best fit line (y = mx + b) and the 95% confidence intervals for the slope and y-intercepts are calculated.
Proportional bias = detected if 95% confidence interval for the slope does not include 1
Constant bias = detected if 95% of the confidence interval for the y-intercept does not include 0
What are the main differences between the deming method and passing bablock methods for method comparison?
Both Deming method and Passing Bablock method allow for imprecision, however the deming method requires impercision to be normally distributed (aka parametric analysis) while the passing bablock method does not require that the variances int he data sets have normal distributions (aka it is a nonparametric method).
____ is a method comparison that can be used when laboratory results are in a categorical scale, such as 1+, 2+, or trace, mild, moderate, marked. A 2+ value may not be twice that of a 1+ and a 4+ may not be twice the value of a 2+.
Kappa agreement
These are distribution curves of healthy and diseased animals.
- Which graph has the best decision threshold?
- Which one would have a diagnostic sensitivity of 100%?
- Which one woul dhave a diagnostic specificity of 100%?
- C
- A (b/c there are no false negatives)
- B (b/c there are no false positives)
ROC curves display the relationship between a TP rate and a FP rate. What are TP rate and FP rate equal to?
- TP rate = diagnostic sensitivity expressed as decimal
- when diagnostic sensitivity is 90%, TP rate is 0.9 - results are psotivie in nine of ten diseased animals
- FP rate = 1 - diagnostic specificity expressed as a decimal
- WHen diagnostic specificity is 70%, the FP rate is 1-.07 = 0.3 - 3/10 nondiseaed animals woudl have a FP result
How do you evaluate a ROC curve?
The more the curve is to the top left, the better becuase nearly all results are TP - and it means the test is more accurate. You can also look at it as area under the curve, and the more AUC, the better.
At what point along the ROC curve would you use use for a decision threshold if you want the best balance between diagnostic sensitivity and specificity?
At what point woudl you use for a decision threshold if you want the test to screen for a diseae?
At what point woudl you use for a decision threshold if you want the test to be a confirmatory test?
- 40
- 20
- 50
What bias is being illlustrated by this Bland-Altman Difference Plot?
Constant Bias (all differences below zero)
AND
Proportional bias (trend where differences get more negative as Hgb concentration increases)
*side note: mean difference only tells you about constant bias, it does not take proportiona bias into consideration
