Statistics And Quality Control

Front 1

Descriptive statistics

Back 1

- summarizes features of a collection of data
- frequency histogram
- Gaussian distribution

Front 2

IntRAindividual variation

Back 2

- observed with repeated determinatinos on the same specimen
- due to reagen, instrument, environment, and operator technique
- IntERindividual variation will be greater

Front 3

Gaussian distribution

Back 3

- symmetric, bell shaped distribution
- allows for use of standard parametric statistical analysis
- or skewed: requires distribution-free (non-parametric) statistics
- Most common form of skewness: increase in one of the tails

Front 4

Cumulative-frequency histogram

Back 4

- A plot of the number of observations less than or
equal to a certain observation (y-axis) in % vs. the value of
that observation (x-axis)

Front 5

Central tendency measures

Back 5

- Mean: arithmetic average with a gaussian distribution, the mean, mode, and median are approximately equal
- Mode: most frequent observation
- Median: the P50, or 50th percentile, half of the results are above this value, half are below

Front 6

Standard deviation

Back 6

- descriptor of dispersion of a single group of observations
- The SD, or S is added and subtracted from the mean to obtain “confidence limits”
- The usual reference range for an analytic method includes 95% of the healthy population, or +/- 2SD

Front 7

Coefficient of variation

Back 7

- expresses SD as percentage
- allows comparison of ranges of different magnitude
- CV (%)= SD/mean x 100

Front 8

Method evaluation

Back 8

- using PAIRED groups of observations
- the perfect relationship: y=x, m=1, and y-int=0
- old, comparative method on x-axis
- new, test method on y-axis

Front 9

Altman-Bland difference plot

Back 9

- the differences between the comparative method and the test method values are plotted against the old, comparative method
- simple comparison of the differences to establish max limits
- any concentration-dependent differences can easily be seen

Front 10

Linear regression analysis

Back 10

- to analyze paired-method comparison data to calculate the best fit line

Front 11

Standard error

Back 11

- standard deviation of the regression line
- Sy/x

Front 12

Correlation coefficient

Back 12

- "r"
- a measure of the strength of the relationship between the 2 sets of data
- result ranges from -1 to +1
- for most clinical chemistry comparisons r should be > 0.98 (or < -0.98)

Front 13

Random error

Back 13

- estimated by Sy/x
- a characteristic of all measurements
- occurs in + and – directions
- due to instrument, operator, reagent, and environmental variations
- the error inherent to an analytic system

Front 14

Systemic error

Back 14

- consistent, in one direction
- defined by slope and y-intercept
- should not be present in a method

Front 15

Constant systematic error

Back 15

- y-int other than zero

Front 16

Proportional systemic error

Back 16

- slope deviates from 1.0

Front 17

Total Analytic Error

Back 17

- random and systemic error

Front 18

Normal Values

Back 18

- correspond with the health associated population
- Reference population studies: large study or follow guidelines using smaller population and demonstrate fit with suggested reference range
- Statistical determination: central 95% distribution: 2.5%-97.5% and if gaussian distribution: mean +/- 1.96 S
- lower limit may be lowered if few or no pathologic consequences
- some reference intervals are not based on normal populations

Front 19

Accuracy

Back 19

- closeness to true value
- laboratory determination: recovery studies, sample comparison, or interference determination

Front 20

Precision

Back 20

- closeness of repeated results (SD, CV)
- perform analysis on aliquots of same samples, at more than one concentration, over a 20-day period

Front 21

Proficiency testing

Back 21

- Participation mandated by CLIA 88
- Laboratories perform specific tests on samples
- Mean and SD is calculated for all participating laboratories using similar methods for a given analyte
- Target values and acceptable ranges are determined

Front 22

Quality Control Materials

Back 22

- lyophyllized or frozen solutions
- should resemble the specimen physically and chemically
- test in the same manner as patient specimens
- matrix effects: interference due to differences between samples and control or standardization materials

Front 23

Statistical quality control

Back 23

Levy Jennings Chart

Front 24

1 2s

Back 24

- Westgard multirule control procedure
- use as rejection or warning when 1 control exceeds 2SDs
- overused
- should only be used with manual assays with low number of analytes or control materials

Front 25

1 3s

Back 25

- reject a run when 1 control exceeds 3SDs
- detects random error and large systematic error

Front 26

2 2s

Back 26

- reject a run when 2 consecutive controls are on the same side of the mean and exceed 2SDs
- detects systematic error

Front 27

4 1s

Back 27

- reject a run when 4 consecutive controls are on the same side of the mean and exceed 1SD
- detects small systematic error

Front 28

10 x

Back 28

- reject a run when 10 consecutive controls are on the same side of the mean
- detects very small errors
- do not use

Front 29

R 4s

Back 29

- reject a run if the range between max and min control of the last 4-6 controls exceeds 4SDs
- detect random errors
- use within run

Front 30

A gaussian distribution is usually
A.) rectangular
B.) bell-shaped
C.) uniform
D.) skewed

Back 30

B.) bell-shaped

Front 31

On doing a correlation study, the test method and reference method generate equal data points. When plotted, the slope equals____and the y-int equals____.
A.) 0, 1
B.) 1, 1
C.) 1, 0
D.) 0, 0

Back 31

C.) 1, 0

Front 32

Interference studyies typically use_____as an interferent.
A.) Hemolyzed red blood cells
B.) Intralipid
C.) Highly icteric specimens
D.) all of the above

Back 32

D.) all of the above

Front 33

Which Westgard rule detects random error?
A.) 1 3s
B.) 4 1s
C.) 2 2s
D.) 10 0

Back 33

A.) 1 3s

Front 34

Which of the following rules probably detect small systemic error and should hardly be used?
A.) R 4s
B.) 10 x
C.) 2 2s, 4 1s
D.) 1 3s

Back 34

B.) 10 x

Front 35

Which rules can be used to evaluate sets of five proficiency testing results?
A.) Mean > 1.0 SDI
B.) 1 result > 2 SDI
C.) 5/5 results > +/- 1.0 SDI and mean > 1.5 SDI
D.) 1 result > 3 SDI

Back 35

D.) 1 result > 3 SDI

Front 36

The primary reason for implementing POCT is
A.) reduced testing cost
B.) enhanced outcomes of patient care
C.) lower central laboratory workload
D.) use of nonlaboratorians as analysts

Back 36

B.) enhanced outcomes of patient care

Front 37

A trend occurs when QC results fall on one side of the mean or the other over a period of 6-7 consecutive days:
True or False

Back 37

False

Front 38

Diagnostic sensitivity refers to the probability that only people who do not have the disease will test negative for the disease.
True or False

Back 38

False

Front 39

Random error relates to method precision and systematic errors relates to method accuracy.

Back 39

True