Laboratory Statistics and Quality Control Flashcards
What is sensitivity and how is it calculated?
Probability of correctly classifying a diseased person as diseased
TRUE POSITIVE RATE
% Sensitivity = # of Diseased Persons with positive test result / Total # Diseased Persons X 100
OR
of True Positives / (# True Positives + # False Negatives) X 100
What is Specificity and how is it calculated?
Probability of correctly classifying a nondiseased person as nondiseased
TRUE NEGATIVE RATE
%Specificity = #Nondiseased Persons with negative test result / Total # Nondiseased Persons X 100
OR
True Negatives / (True Negatives + False Positives)
How is positive predictive value calculated?
True Positives / (# True Positives + # False Positives) X 100
How is negative predictive value calculated?
True Negatives / (# True Negatives + # False Negatives) X 100
What is the False Positive Rate? How is it calculated?
False Positives / Total # Nondiseased Persons X 100
Probability of incorrectly classifying a nondiseased person as diseased
OR
1 - Specificity (True Negative Rate)
What is the False Negative Rate and how is it calculated?
The probability of incorrectly classifying a diseased person as nondiseased
False Negatives / Total # Diseased Persons X 100
OR
1 - Sensitivity (True Positive Rate)
What is the difference between accuracy and precision?
Accuracy = closeness of measured value to its true value (ability to get close to bullseye)
Precision = closeness of repeated measurements of the same quantity (ability to hit the same spot on the target repeatedly)
What is a Normal Distribution?
AKA Gaussian Distribution = Normal Bell Curve
-Single peak representing mean of the population of observations
- 68.26% of observations (values) that make up distribution will fall between +/- 1 SD of the mean
- 95.46% between +/- 2 SD
-99.73% between +/- 3 SD
What are the standard deviation distributions/frequencies in a normal bell curve?
1 SD = 68.26%
2 SD = 95.46%
3 SD = 99.73%
Where is the mean in a unimodal vs skewed data set?
In unimodal/symmetrical data set, mean will be in the center of the curve
In skewed data set mean will be shifted left (negative) or right (positive) of peak of the curve
What is the Median of data set?
Divides all data points exactly in half with one half higher and one half lower than the median number
- Median is also 50th percentile
-Non parametric method (Not calculated from formula)
- Mean is parametric (from formula)
-If odd number of data points, median will be value in the middle
-with even number of data points, median is average of two middle values
What is the Mode in a dataset?
Most commonly occurring value in dataset
-not very useful measure for describing or comparing data sets
-shows when data set consists of two or more different populations that result in more than one mode
- BIMODAL = two separate populations are present in data set
- Mode always represented by highest point on a bell-shaped curve whether the data are normal or skewed!
Where are the Mean, Median and Mode found in a Gaussian curve verses a Skewed Curve?
In gaussian curve data is unimodal - Mean, Mode and Median overlap
In a skewed data set - Mean, Mode and Median will not lay on top of one another
What is the Standard Deviation and how is it calculated?
Defined as root mean square deviation of the values from their mean or as square root of the variance
Numerical value used to indicate how widely individuals or points in a group vary from the mean of the group.
How does the formula used for Standard Deviation vary depending on the population?
If the data set is a POPULATION of its own, divide by the number of data points, N
If data set is a SAMPLE from larger population, divide by one fewer than the number of data points in the sample, n-1
Sample deviation is more common calculation in statistical analysis
What is the Coefficient of Variation? How is it calculated?
CV is the ratio of the standard deviation to its arithmetic mean
CV = (SD/MEAN) X 100
What is a Confidence Interval?
Estimated range of values which is calculated from the mean and standard deviation
Estimates statistical probability of values falling above or below the mean value on a normal bell=shaped Gaussian curve
1 SD : 68%
2 SD : 95%
3 SD : 99%
What is the Student’s T-Test?
T-Test is often used to determine how significant the differences between two groups are.
-Could the differences have occurred by chance?
-used statistically to compare group means
What is the t-score and its significance?
Ratio of difference between two groups and difference within the groups
-Larger t-score, the greater difference between the groups
-smaller t-score, the more similarity between the groups
- a t-score of 3 means groups are THREE TIMES as DIFFERENT FROM EACH OTHER as they are within each other
- When you run t-test, the bigger the t-value, there likely it is that the results are repeatable
- Large t-score tells you the groups are different, while small t-score suggests the groups are similar
What does a large t-score vs small t-score tell you?
Large t-score suggest that the groups being compared are different from each other
Small t-score shows that the groups are more similar
What are the 3 main types of t-test?
1) Independent samples t-test: compares the means for two groups. most common
2) Paired sample t-test: compares means from the same group at different times. AKA correlated pairs t-test, paired samples t-test or dependent samples t-test) Dependent samples are connected (tests on same person or thing)
3) One sample t-test : tests the mean of a single group against a known mean
What is a p-value?
probability that the results from sample data occurred by chance
-low p-values are good: indicate data did not occur by chance
-E.G. p-value of 0.01 means there is 1% probability that the results from an experiment happened by chance
-p-value of 0.05 is generally accepted to mean data is valid
What is the relationship between the p-value and the null hypothesis?
p-value is probability of finding the observed or more extreme results when the null hypothesis of study question is TRUE
what is the null hypothesis?
usually hypothesis of no difference between two groups
What is the significance level (alpha)?
pre-chosen probability versus P value, which is used to indicate probability that is calculated after a given study
What is the alternative hypothesis?
opposite of the null hypothesis
usually the hypothesis under investigation
What do you conclude if P value is less than chosen significance level versus higher?
-If p value is less than chosen significance level, then you REJECT the null hypothesis = you ACCEPT that the sample gives reasonable evidence to support the alternative hypothesis
-Choice of significance level to reject H0 is arbitrary
-5%, 1% and 0.1% are commonly used
When is ANOVA used instead of T-Test?
-When only two groups are compared t-test is used
-When the means of more than two groups are compared analysis of variance (ANOVA) is preferred
What are 4 types of error?
1) Random error
2) Systematic error
3) Active error
4) Latent error
What are the differences between the 4 types of error?
-Systematic: influences values consistently in ONE DIRECTION (Examples 22S
-Random: exists in all measurements and is due to CHANCE (Example 13S)
-Active: occurs at the interface between a laboratories and patient samples
-Latent: related to the organization or design of a laboratory
-
What is variance?
Describes factors that affect the measurement of an analyte
What are sources of variation?
Sample error
Random analytical error
Method/instrument bias (analytical systematic error)
operator error
sample variation within patient day to day
patient variation
What is a Reference Range? How is it derived?
-range of values derived from a group of normal persons free from disease
-comprised of 95% of the population of normal persons
-mean +/- 2 SD
AKA normal values, normal range, reference range, reference interval
What is the Chain of Custody?
chronological documentation showing procurement, custody, control, transfer, analysis and disposition of specimen
Required for specimen used as legal evidence in court
also ensures proper identification and handling of samples for all testing
What are some major elements of an effective QC program?
-educated, trained certified personnel
-Operation and calibration of instruments
- Proper and consistent Techniques and methods
- Proper procedure, policy, and safety manuals
-Documentation and record-keeping
-chain of custody
-system for appraisal of test performance
-correction of deficiencies and implementation of advances and improvements
What are some CLIA required components of QC program?
-Follow manufacturer instructions
-Procedure manual
-Calibration procedures every 6 months
-Perform QC procedures using 2 levels of controls each day of testing
-Document remedial action
-maintain QC records for 2 years, immunohematology for 5 years
What are ways labs keep track of chances in test performance over time?
Levy-Jennings and custom plots
Westgard Rules
External QC PT
What is a Levy-Jennings Plot?
plot of control results on the Y-axis versus time on the X-asis
Compare dispersion vs trend vs shift
Dispersion = Increased frequency of both high and low numbers
Trend = Progressive drift of reported values from prior mean
Shift = An abrupt change from established mean
What is a trend and some potential causes?
Gradual loss of reliability and usually subtle
-Deterioration of light source
-GRADUAL:
accumulation of debris
aging of reagents
deterioration of controls
deterioration of incubation temperature
deterioration of light filter
deterioration of calibration
What is a SHIFT and potential causes?
Abrupt changes in control mean. Sudden and dramatic positive or negative change in test system performance
-Sudden failure or change in light source
-change in reagent formulation
-change of reagent lot
-major intstrument maintenance
-suddden change in temperature
- change in room temperature or humidity
-failure in sampling or reagent dispensing system
-innacurate calibration or recalibration
-improper mixing of controls
-vial to vial variation
-controls left at room temperature too long
What is dispersion and examples of causes?
Dispersion is observed when random errors or lack of precision increases.
indicates inconsistency in technique or a stability problem.
-fluctuating eletrical voltage
-poor mixing of sample or control
-Different or poorly trained personnel running same analysis
-clots/bubbles in sample
What is a Cusum Plot? What does it show?
Expected value is subtracted from actual value each day of measurement and the difference is summed together over time
When control data are randomly scattered about their expected mean value, custom will wander above and below zero to yield horizontal line
When SYSTEMATIC ERROR is present, custom value will steadily increase above or below zero
What are Westgard Rules?
Multirule QC uses combination of decision criteria or control rules to decide whether an analytical run is in control or out of control
Westgard multirule QC procedure uses 5 different control rules to judge acceptability of an analytical run
Westgard decision tree is helpful in process
Single-rule QC uses single criterion or set of control limits
How many control measurements are used for Westgard rules?
generally used with 2 or 4 control measurements per run
I.E. When two different control materials are measured 1-2 times per material
Alternative control rules are more suitable when 3 control materials are analyzed such as in hematology, coagulation, and immunoassays
What is a 1-2-s control rule violation?
typically used as warning
What is 2-2-s rule ?
Reject when 2 consecutive control measurements exceed the same mean + 2s or the same mean minus 2s control limit
What is R-4-s rule?
Reject when 1 control measurement in a group exceeds mean plus 2s and another exceeds the mean minus 2s
What is 4-1-s rule?
reject when 4 consecutive control measurements exceed the same mean plus 1s or the same mean minus 1s control limit
What is the 10x rule?
Reject when 10 consecutive control measurements fall on one side of the mean
What is the 1-3-S rule?
control rule commonly used with levy-jennings chart when control limits are set as mean plus 3s or mean minus 3s
run rejected when single control measurement exceeds mean plus 3s or mean minus 3s control limit
What is six sigma and its purpose?
quality improvement process to identify and remove causes of defects
Goal: 3.4 defects or errors per million opportunities (tests)
Normal distribution of 6 standard deviations from the mean
99.9997%
Follows DMAIC method of identifying and eliminating errors
What is the DMAIC method?
Identifies and eliminates errors. Followed by six-sigma process. Five phases.
Define: identify high level processes for improvements
Measure: baseline data on current processes, pinpoint problems and areas of improvement
Analyze: root causes, validate root causes against captured data
Improve: implement improvements that address root causes
Control: perform before and after analysis, monitor systems, document results and determine next step recommendations
