The term Quality Assurance refers to the total program that the laboratory uses to establish and maintain
high standards of performance. It includes the licensed professional
laboratory analysts employed by the lab, the continuing training they receive, the quality and upkeep of the instruments
and glassware used in the lab, specifications and Certificates of Analysis for the reagents and standards purchased,
records of reagent and standard preparations, instrument log and maintenance books and more.
This program is generally formalized in written form as a “Quality Assurance Manual.”
Quality Control is the single procedure used by the laboratory on a regular basis to obtain quantitative data about some aspect of the validity of the specific test procedure. EPD requires documentation of the accuracy of the calibration curve, performance of reagent blanks, standards and spiked samples if they are specified in the approved method and participation in DMR QA studies for all major dischargers (greater than 1 MGD) and some minor dischargers. There are five quality controls that are strongly recommended by regulatory agencies, such as EPA and the Georgia EPD, to be performed on each parameter test and records kept of the performance:
Reagent Blanks,
Accuracy assessment,
Precision assessment,
Control Charts, and
Performance Evaluation Samples.
Calibration is not on this list. Each and every test must be performed with equipment that has been calibrated, or the results may be meaningless. Reporting data from such a procedure is illegal. Intentional reporting of data gleaned from uncalibrated equipment is noncompliance with the approved method, a major violation of the NPDES permit and a criminal offense. Going through the motions to come up with bad numbers which others use for deciding changes in process control will change careers.
Reagent Blanks confirm the amount of laboratory contamination. They should not be used as part of the calibration. They consist of a sample of analyte-free water (reagent water) which is taken through the entire analytical procedure and analyzed as a sample. The results for the blank should fall below the lowest concentration point on the calibration curve. This indicates that the glassware, reagents and solutions used in the procedure are free from contamination by the target analyte. Reagent blanks should be performed every day that testing is performed.
Feedwater, tap water, drinking water, bottled water, or distilled water is not to be used indiscriminately. Distilled
water is actually toxic to humans, in large amounts. Make reagent water from feedwater. There are grades of water
used in the lab. Mostly classed into three grades. Resistivity, conductivity, and SiO2
amounts determine which grade a 'water' will fall into. Use of these are:
|
Type of parameter |
Low |
Medium |
High |
|
Resistivity, megohm-cm @ 25 °C |
.01 |
>1 |
>10 |
|
Conductivity,µmho/cm @ 25 °C |
10 |
<1 |
<.01 |
|
SiO2, mg/L |
<1 |
<0.1 |
<0.05 |
Accuracy in analytical chemistry is finding out how close your answer is to the actual amount of stuff in the sample. Accuracy assessment is normally calculated as Percent Recovery (%R) and is done by analyzing Matrix Spikes. Matrix spikes are the adding of a known concentration of the target analyte to a real sample and then analyzing the matrix spiked sample at the same time the unspiked sample is analyzed. The solution used for preparing a matrix spike is normally the calibration standard stock solution. For example, the calibration stock solution for preparing fluoride standards is normally 100 mg/L. Addition of 1.00 mL of this solution to a 100 mL volumetric flask and then diluting to volume with sample gives a matrix spike of 1.00 mg/L of fluoride. The calculation of %R is done in one of two ways, depending on conditions. The first condition is when the sample contains no detectable levels of target analyte. The calculation is simply:
, where the observed value is the analytical result for the matrix spiked
sample and the true value is the known concentration of the matrix spike.
When the sample does contain target analyte, its concentration must be taken into account to calculate accuracy. The following equation
is used: 
, where (again) the observed value is the analytical result for the
matrix spiked sample, the background value is the analytical result for the unspiked sample and the true value
is the known concentration of the matrix spike. To be absolutely
correct, the background value should be adjusted for the volume added by the matrix spike solution.
In the fluoride example above, the matrix spike was prepared by combination of 1.00 mL of matrix spike solution
with 99.00 mL of sample in the 100 mL volumetric flask. The background
is most correctly equal to 0.99 times the sample result, however, many times in practice this correction is ignored. A perfect result for %R is 100.
There are some cases where this is not true, if there is final volume adjustment, say, following a digestion
or distillation. For instance, for total phosphate, you take a 50
mL sample, digest it, neutralize it, then dilute to 100 mL for colorimetric analysis.
If you add a spike before digestion, the volume of the spike solution causes no problems.
There is little justification in reporting %R to any more than integer values because of the imprecision in the performance of the analytical method. 95.34 calculated as %R with a calculator should be rounded to 95%.
There are several procedures in the laboratory which are not amenable to the matrix spike procedure. Examples are pH, conductivity, residual chlorine and dissolved oxygen. For these analytes there are commercially available prepared solutions and samples which can be substituted for the matrix spiked sample to obtain %R values.
Matrix spikes should be done every day that testing is performed on each different type of sample. If the lab tests effluent from three different treatment plants, then three different matrix spikes should be done each testing day.
Standard additions is a common technique for checking test results. Other names are “spiking” and “known additions” and “matrix spikes.” The technique can test for interferences, bad reagents, faulty instruments, and incorrect procedures.
Perform Standard Additions by adding a small amount of a standard solution to your sample and repeating the test. Use the same reagents, equipment, and technique. You should get about 100% recovery. If not, you have a problem (which should be identifiable and the attempt must be made to find and solve the problem).
If you get about 100% recovery for each addition, everything is working right, your technique is good, there are no interferences and you can believe your results represent the sample faithfully.
However, when the results show other than 100% (e.g.. 80% or 120%) some kind of problem exists. Now use deionized water [sometimes it could be a problem] as the sample, add the Standard Addition, analyze. If 100% recovery for each addition is the result with the deionized water, something is interfering with the reagent in the raw sample. On the other hand, if 100% recovery with the deionized water is not the analysis result, then the procedure and the equipment must be reviewed and inspected. The problem may be bad reagent or bad Standard Addition. After finding something suspect and making adjustment, another run another test to prove what the problem was.
Suppose a single standard addition to the sample did not give the correct concentration increase (100% recovery). A possible cause could be interferences. If interfering substances are present, it may be possible to arrive at an approximate result if the increases are uniform.
Carefully check the instructions for the test. Make sure to use the correct reagents in the correct order. Be sure the colorimeter is adjusted to the correct wavelength and the glassware in use is what is required. Be sure time for color development and the sample temperature are as specified. If the procedure technique was incorrect, make copious notes then rerun the test again.
Check the reagent performance. This may be done by obtaining a fresh lot of reagent or by using a known standard solution to run the test. Make sure the color development time given in the procedure is equal to the time required for the reagent in question.
Apparent interferences may also be caused by a defect in the instrument or standards. Before assuming the interference is chemical, start with equipment checks or other basic elements of the procedure. After determining the procedure, reagents, instrument and /or apparatus are correct and working properly, you may conclude the only possible cause for standard additions not functioning correctly in deionized water is the standard used for performing standard additions. Obtain a new standard and repeat test.
Run a proof of accuracy check on a standard solution. Take a known concentration of a standard solution through the same steps as for the original sample.
where:
Cu
= measured concentration of the unknown sample
Vu
= volume of the unknown sample
Cs
= concentration of the standard
Vs
= volume of the standard.
where:
Xs
= measured value of the spiked sample
Xu
= measured value for the unspiked sample
adjusted for the dilution of the spike volume
K = known value of the spike in the sample.
In some situations, it may not be possible to dilute out an interference without diluting out the parameter (target analyte). In this case, you may need to use a different chemistry or an ion selective electrode.
-------- USEPA pub. SW-846 ---------------------
pH interference
Many of the procedures only work within a certain pH range. Measure the pH of your analyzed sample. Prepare a blank with all the reagents called for in the procedure. Measure its pH. If there is little difference in the values, then pH interference is not a problem.
Two terms often used to describe uncertainty in measurements are precision and accuracy. Although these words are frequently used interchangeably in everyday life, they have very different meanings in the scientific context. Accuracy refers to the agreement of a particular value with the true value. Precision refers to the degree of agreement among several measurements of the same quantity. Precision reflects the reproducibility of a given type of measurement. In quantitative work, precision is often used as an indication of accuracy. We assume that the average of a series of precise measurements (which should “average out” the random errors because of their equal probability of being high or low) is accurate, or close to the “true” value. However this assumption is only valid if systematic errors are absent.
Precision is the consistent ability to obtain the same answer when
testing a sample. Precision assessment is reported as Relative Percent
Difference (RPD). Data for calculation of RPD is obtained from performing
the analysis on two aliquots of the same sample, known as Duplicates. The
calculation for RPD is:
,
where A and B are the analytical results from testing the sample and the duplicate.
Perfect precision gives a RPD of zero. RPD is simply the absolute
difference between the two results for a duplicate pair, divided by the average result, and multiplied by 100 to
express as a percentage. Precision can be determined on a real sample
on a daily basis if the target analyte is consistently found in the sample, such as BOD and TSS.
In cases where the analyte is infrequently found in the sample, a better test for precision is:
duplicate the matrix spike sample and calculate the RPD on both of the results of the matrix spike and the
matrix spike duplicate. Good precision indicates that the analyst
is controlling laboratory variation in the test procedure by using careful techniques in a reproducible, defensible
manner.
Two types of errors are defined: Random and Systematic. A random error means that a measurement has an equal probability of being high or low. This type of error occurs in estimating the value of the last digit of a measurement. A random error is also known as an indeterminate error. The second type of error, systematic (or determinate) occurs in the same direction. The preceding is an old definition of accuracy and precision. Accuracy is affected by two components, bias and precision. Bias is the result of SYSTEMATIC error. For example, if you are analyzing a pH standard that you know to be pH 6.00SU, and the average of 20 analyses is 5.50SU, your measurement system (pH meter, probe, buffers) is BIASED and reading too low. Imprecision is the result of RANDOM error. For example, in analyzing that pH standard, if you got a 5.51, 5.49, 5.50, 5.53, etc., your PRECISION was good (i.e., values were close to each other), but still you were not doing ACCURATE work because the system was BIASED. You can verify that this argument 'that accuracy is affected by bias and imprecision' is gaining acceptance by comparing the last paragraph of several methods in Standard Methods in the 17th Edition, and then looking at the same paragraph in the 18th or 19th. The older methods titled the subparagraph "Accuracy and Precision" while the new methods say "Bias and Precision." This is not just semantics ... you cannot measure accuracy directly (even though your manual says you can by analyzing matrix spikes ... again, an old wives tale) ... you have to measure bias and precision, the two components. You estimate bias by analyzing a standard several times (at least 20), and comparing the average value to the true value. You estimate precision by computing the standard deviation of those 20 results and comparing it to what the method says is acceptable. For example, in the BOD test, if the average of 20 G/GA (glucose-glutamic acid) tests is close to 198 mg/L, the lab is doing unbiased work. If the standard deviation of those results is less than 30.5 mg/L, the lab's precision is better than the bunch of labs that did the study reported in Standard Methods. If you read the method carefully in the 19th Edition, you will see that they recommend the standard deviation for a SINGLE lab should be close to 10 mg/L.
The purpose of doing matrix spikes is to reveal interference because of matrix effects. If the results of analysis of a standard in a given batch are within the limits on a control chart, analytical bias and precision are in control. If the matrix spike in that same batch comes out way below what you would like to see (say 50%), it means only one thing ... the matrix has interfered with the analysis. That is a source of systematic error that the analyst has no control over, and it should not be thought of as an analytical problem.
Of the two assessments, precision is the more important value as an indicator of laboratory performance while accuracy is a reflection of the sample character. Most test procedures have known interferences from substances other than the target analyte. These interferences can give falsely high or low results. Chloride, when testing for COD; nitrite when testing for cyanide and background color in samples being examined by colorimetric analysis are examples of interfering with the accuracy of the result. The matrix spike procedure serves as an evaluation of known and unknown interferences present in the sample.
Control Charts are graphical representations of the results obtained from a quality control measure performed on a regular basis. The most common control charts which are required to be kept are those for precision and accuracy. An example of a %R control chart
Performance Evaluation (PE) samples are samples of known concentration which are analyzed by the laboratory as unknown samples. The EPA each year sends to participating wastewater laboratories with NPDES permits, performance evaluation samples called DMR QA Studies. The PE samples are analyzed by the laboratory and the results sent to the DMR coordinator at EPA. After evaluation, the results are graded as ‘acceptable’, ‘check for error’ or ‘unacceptable’. The ‘acceptable’ results are within 2 standard deviations of the mean result submitted by all the participating laboratories. ‘Check for error’ results are between 2 and 3 standard deviations of the mean. ‘Unacceptable’ results are outside 3 standard deviations of the mean and the laboratory is required to respond to the State EPD NPDES coordinator with an explanation as to why the result was found unacceptable.
Besides the DMR PE samples, the EPA prepares and evaluates two other studies each year called the water pollution (WP) studies. Each WP parameter is provided at a single concentration for evaluation. A laboratory can apply to join the WP studies by contacting the head of the EPD laboratories (Kerry Wilkes, 1-404-853-7978)(may be 770 now). Acceptance into the WP check sample program is at the discretion of EPA. Commercial supply houses, such as APG and ERA, also prepare and evaluate PE samples on a fee basis.
Questions for Chapter 5
1. A sample was performed in duplicate and the results obtained were 34 mg/L and 38 mg/L. Calculate the
precision.
2. What is the most common cause for low matrix spike recoveries?
3. A sample which contains 148 mg/L sulfate was matrix spiked with 81 mg/L sulfate. An accuracy of 87% was
calculated from the analysis of the matrix spiked solution. What was the result obtained from analysis of the
matrix spiked solution?
4. Within how many standard deviations of the mean should 95% of your results lie?
5. What is the quality control for monitoring laboratory contamination? How is it performed and with what
frequency?
6. Describe the differences between accuracy and precision.
| Back to top | On to Chapter 6 |
| My question sets for self exam | errors? |