Doing a Quality Job In the Lab
by Joel Gordon
On a recent day, I measured a TSS of 4.6 mg/L for my plant final effluent. How sure am I that it was not really 5.6 or 3.6? What would give a state inspector confidence that my result was correct?In industry, manufacturers keep the number of defective products to a minimum by means of a quality control (QC) program. When we do laboratory work, our products are our test results, and we also need a QC program to make sure that they are trustworthy.
Some of the basic quality control elements in the lab are: proper training for analysts; clear, written standard operating procedures; proper maintenance and calibration of laboratory equipment (pH and D.O.meters, balances, etc.); attention to temperatures of ovens and incubators; properly cleaned glassware; use of laboratory pure water and fresh, analytical grade reagents; checking of all calculations, etc. But even then, just running procedures and accepting the results is not enough. We need to perform continual “tests on our tests”, so that we can spot problems when they develop— not later, when some one questions the data we have reported .
Problems with analyses are usually classified as causing inaccuracy (bias) or poor precision (not reproducible). The classic picture is of a dartboard: If your results are accurate and precise, the darts will all be in a tight pattern around the bull’s eye. If they are accurate, but not precise, they are still centered, but scattered in a looser pattern. They can also be precise, but not accurate—a tight pattern of darts, but off-center. And in the worst situation, they will be widely scattered and off-center, as well—poor precision and accuracy.
What can cause these problems? Poor accuracy can come from contaminated, outdated, or improperly prepared reagents or standards; from not mixing samples well before pouring; from poorly calibrated, malfunctioning, or misread instruments; from not following procedures correctly; from incorrect calculations or transcription errors. Poor precision (and accuracy) can come from sloppy measurements (of volumes, for instance); poor mixing of samples; dirty or scratched optical cells; not noticing holes in filters or large, unrepresentative “chunkies” (TSS); contaminating samples or glassware during handling, etc.
So, how do we check for these problems? Accuracy is checked by running blanks, standards, and spikes, while precision is usually checked by running duplicates. We can track the values we get for these QC measurements over time and construct charts which tell us whether our procedures are in control, or heading for trouble.
A blank is run by carrying out the test without a sample, usually using laboratory pure water, instead. This can spot contamination in your reagents, or introduced during handling. A standard is a reference material of known composition. These can be purchased from a laboratory supplier, or carefully made up in your lab. Examples might be a 1.00 mg/L copper solution, or glucose-glutamic acid for BOD’s. A spike is a standard which has been added to one of your samples. You compare the amount detected to the amount added to see if there is any interference caused by the sample. For instance, if you added 0.50mg/L of phosphorus (P) to a sample of your effluent which you have measured as containing 0.80 mg/L of P, then you would expect to measure 1.30 mg/L in the spiked sample. If you measure only 1.20 mg/L, then you are recovering only 0.40 mg/L, or 80% of the spike, and there may be a problem.
To find out how precise your measurements are, you clearly have to run the same sample (or standard) more than once. Consider two laboratories, each of which makes 5 replicate measurements of a 1.00 mg/L ammonia standard. Lab A gets 0.83, 0.82, 0.85, 0.82, and 0.83, while lab B gets 0.80, 1.21, 0.88, 1.12, and 0.99. What calculations would we do to see how they compare? Well, we can check for accuracy (lack of bias) by figuring the average for each lab. The average, also called the mean, is calculated by adding up the values and dividing by the number of results. For lab A, the average is 0.83; for lab B, 1.00. So lab B seems to be more accurate. But lab B obviously has more scatter in its results.
The usual measure for scatter is the standard deviation. You can calculate it easily with a $10 statistical pocket calculator or a spreadsheet program such as Microsoft Excel. (If you want to calculate it manually, it is equal to the square root of the sum of the squares of the difference of each result from the mean, divided by one less than the number of replicates--- Let the calculator do it!) The standard deviation for lab A works out to be only 0.01 mg/L; for lab B, 0.17 mg/L. So lab A is precise, but inaccurate, while lab B has less bias, but poorer precision. Neither one is ideal. With lab B, you have a chance they will get the right answer—but only a chance. With lab A, you won’t get the right answer, but you can see that some aspect their technique is very well done—maybe they just need to check their calibration standard.
Next time: How to construct and use a quality control chart for your lab work…