Week 1 Flashcards
What is quality management composed of:
Error detection - quality control
error prevention - quality assurance - audit
what is quality control
when you run QC you run a test in the same way you would a patient sample but you use known values
-helps detect issues with testing processes before results are released
-includes things like reagents, standards, water quality, temperature and pipettes, gram stain control slides, antisera to known cells, commercial hemoglobin control
-QC issues are important in identifying potential errors
what is accuracy and how is it measured
-close to true value
-correct values are obtained from standardized test materials, procedures and conditions
measure using comparison/test method vs reference method (what we got vs what we should have gotten)
-shown as a % error and inaccuracies are due to systemic error
% error = what is got - what I should get /what I should get x(100)
-error can be negative or positive which tells you the direction of the difference (bias) and is reported to 1 decimal place
what is bias
methods that regularly give results that are higher/lower than the reference
positively 3.0% or negatively -2.5%
what is precision
reproducibility - when multiple aliquots of same sample having a constant concentration repetitively. However, there will be a tolerance because you wont always get the same result
expressed as standard deviation (SD or s) or coefficient of variation (CV)
if the mean of two methods are the same use s
if the mean of two methods are different use CV
imprecision is random error
most important
what is descriptive statistics
foundation for monitoring performance (QC)
-basic data measures like shape, center - most common and spread - most impactful
for example when you measure glucose 100 times in a row there can be lots of results. But the results form patterns so you can assess predictability
how do you measure center
Mean - average, measure of central tendency, needs units and has the same decimal places as the data provided
Median- middle value after you arrange numbers in ascending order
Mode - most common number - rarely used as datas center but used to describe date that is bimodal
how to measure spread
-Standard deviation (s, SD) - precision
-describe how the data is distributed
-SD is married to mean so if if the mean changes so does the SD
-reported to one more decimal place than data provided
-units must be included always and use the same units as the data
-used for isolated set of data or to compare two data sets with equal means
Coefficient of variation -CV - for precision and scatter
-higher the CV- lower the precision (more scatter)
-lower the CV - better precision (less scatter)
-report to ONE decimal place
-used more often
-does not matter the values of mean- so can be used anytime
-Range - largest minus the smallest
how to measure shape
Gaussian distribution - symmetric distribution
68% of the data fall between ±1SD;
95% of the data fall between ±2SD;
99% of the data fall between ±3SD
what is random error
unpredictable
-no pattern
-change in s but not change in mean (precision is affected not accuracy)
what is scatter
-excessive variation with few coming close to the mean
-use s or CV
-points can be within +/- 2SD
-indicates precision issue and random error
-can be caused by instrument, operator, reagent or environmental issues
what is systemic error
- constant (difference between test and comparative value regardless of analyte concentration)
-proportional (when differences between test and comparative values are proportional to analyte concentration
-measurable
-affects results in the same direction
-give a change in mean but not s (accuracy is affected not precision)
what is trend
-gradual change over time that looks like a progressive problem - more than 6 values that are increasing or decreasing
-control chart indicate trends or warn of errors
-can be caused by faulty instrument -light source, accumulation of debris on electrode
-can be caused by change in control sample (gradual decrease in bilirubin when exposed to light)
-can be caused by change in calibrating standard (loss of constituents)
Indicates systemic error
what is shift
-change in ONE direct ; 4-6 values lying on one side of the mean
-loss of accuracy ; a sudden and consistent change
-can be caused by :
inaccurate sample volume
inaccurate standard
change in reagent or incorrectly prepared reagent
incorrect wavelenth setting on spec
change in light source
systemic error
what is an outlier
random error- does not affect mean or s if it is a large data set
-a value that is randomly too high or too low
-if it happens in the first few points there can be a precision issue
-usually human error- transcription error, improper wavelength setting
difference in what was expected vs what was delivered
