A pool refers to a group of individuals subject to a
specific set of random selection parameters, such as the rate (i.e. 50%) and
periodicity (monthly) of selections.
The rate is the number of selections as a percentage of the
pool size. For example, if there are
100 people in the pool, and the annual rate is 50%, then 50 selections will
occur over a years time. Because the process is random, it is probable that a significant number of the 50
selections will “repeat”, meaning that some people get picked more than
once. So a random rate of 50% of a 100 person pool means that you’ll conduct 50 drug tests, not test 50 different
The program period refers to the period of time during which
the random rate will be calculated.The
easiest program period to use is one year, however, it’s possible and sometimes
advisable to have shorter program periods.
The program period is divided into a specific number of selection
periods, which is called the frequency.
Testing activity may fluctuate over the course of the program period,
but by the time the period closes, the number of completed tests should equal
the random rate.
Frequency is the number and spacing of selection periods
during the program period. Typical
frequencies are month, weekly, quarterly, or daily. Other frequencies are possible and sometimes helpful. A high frequency of selections, i.e. daily,
results in a very high level of deterrence.
However, it also tends to more difficult to administer. As a general rule for establishing
deterrence, you should use the highest frequency possible given your
administrative capabilities. For
example, if all of your pool members are located at a single facility with
on-site collection capabilities, then weekly, or even daily selections are
possible. But if the same number of
people are spread out over a large geographic area with diverse work schedules,
monthly or quarterly selections may be more appropriate.
The key factor that helps determine frequency is the ability
to locate, notify, and collect a sample from the individual selected for
testing. That ability is driven by your
communication abilities, management practices, geographical structure, and
collector arrangements. Of course, all
of these are related to the cost of the testing program.
The selection period is an interval within the program
period for which a given number of random selections are performed and their
corresponding tests completed. Typical
selection periods are one month, one week, one quarter, or one day. For example, if you have chosen a frequency
of “monthly” each month would be a new selection period. There are a couple of important things to
remember about selection periods:
When using simple random sampling with replacement, the prior
selection periods have absolutely no impact on the current selection
period! (Hint: Think of it as a new roll of the dice- the
dice have no memory of previous roll.)
The easiest and most objective way to administer testing is to
excuse all pending tests at the end of each selection period. “Carrying over” can introduce all sorts of
problems, most of which result from bias.
Remember, random testing is about performing a specific number of tests
on the subject population. If you’re
finding it difficult to accomplish the desired number of tests in each
selection period, you may have to adjust management practices, communication,
or the logistics of collection. You may
also simply need to increase the number of selections per period (over
sampling), or, at the start of the next program period, change the frequency.
Over sampling refers to the practice of selecting more
people for testing than the rate requires.
This is done in anticipation of some number of tests not being
completed. Over sampling is required in
almost every random testing program because it is simply not possible to
conduct a test on every person that is picked by the computer. People get sick, go on vacation or leave,
change responsibilities, or are otherwise unavailable for testing. An over sampling rate of 20% is quite
common. It can be much higher or lower,
depending on the situation. Again, the
emphasis of random testing is completing a given number of tests during the
program period in an unbiased fashion.
Some measure of over sampling is required to meet that goal.
Random selection is a mathematical process driven by several
parameters: pool membership, the program period, rate, and frequency.
A “pool group” is created. The pool group includes those personnel
subject to random testing. The operator
also specifies the rate of random testing, or a specific number of pool members
to be selected each period. For
example, the Department of Transportation requires a 50% annual testing
rate. This means that, over the course
of one year, at least 50 drug tests must be conducted for every 100 employees
in the pool.
It is important to understand that the 50 tests do not have to be
conducted on 50 different individuals.
In fact, this is highly improbable, if not impossible. At a 50% selection rate, the actual
probability is that 37 or 38 different individuals will be selected for the 50
tests. This means that 12 or 13 of the
100 individuals in the pool will be selected at least twice or more.
beginning the selection process, Drug Test America figures out how many tests
need to be conducted for the “selection period.” The selection period is usually a week or month. The Department of Transportation requires
that each member of the pool have an equal chance at being selected for a test
every selection period. When figuring
out how many tests are needed for the period, Drug Test America takes into
account absenteeism, incomplete tests, etc. to make sure that the minimum
number of required tests is accomplished.
Drug Test America
uses a random algorithm, or mathematical equation, to assign an “index number”
to every one in the pool. The
employee’s index number is usually different every selection period, however,
it is possible for the computer to assign the same index number two or more
periods in a row. The number of index
numbers is always equal to the number of people in the pool for the selection
period. The index number becomes the
“identity” of each member of the pool group for the selection period.
For example, if the pool group has 100 members, then each member in the
pool will receive a randomly assigned index number between 1 and 100.
Using a random algorithm,
Drug Test America generates a series of random numbers equal to the number of
tests required for the period.
Drug Test America then looks at the index numbers that are randomly
assigned to the pool group members and matches up the numbers.
For example, if Drug Test America determined that 5 tests were needed
for the period in a pool group of 100 members, it would pick at random 5
numbers between 1 and 100. For illustrative purposes, let’s assume that
the numbers 34, 45, 67, 35 and 10 were picked by Drug Test America. Drug Test America would then search through
the 100 index numbers and find out which pool group members were assigned the
index numbers of 34, 45, 67, 35 and 10.
Those five individuals would be selected for a test.
PROCESS CANNOT BE UNDONE. ONCE
Drug Test America HAS ASSIGNED INDEX NUMBERS AND MADE SELECTIONS, A PERMANENT
RECORD FOR EACH SELECTION IS CREATED.
It is also important to know that the random algorithm used
by Drug Test America has been thoroughly tested and documented. Drug Test America’s random number generator
verification is available upon request.
Statistical analysis has also determined that computer algorithms are
the “best” random generators because they are free from physical biases and can
thoroughly document the random selection process.
If the explanation above seems a little confusing, the following
example will help illustrate how Drug Test America selects individuals for
Let’s assume that there are 52 people in a room that are subject to
random testing. Let’s also assume that
5 people need to be picked for random tests.
We can accomplish this goal fairly with two decks of playing cards. First, we would shuffle both decks of
cards. We then take the cards from one
deck and pass out one card, face down, to each person in the room. Next, we would draw five cards from the
second shuffled deck and place them face up on a table. Everyone in the room would then turn their
playing card face up. The five cards on
the table from the second deck will match up to five individuals in the room
holding cards from the first deck.
These five individuals are now “picked” for a test.
We could repeat this exercise time and time again, shuffling both decks
each time and passing out the cards.
The odds are that some individuals will never get “picked”, and, in like
manner, some individuals will be picked several times.
Why are some people picked for testing more than once and
others are not picked at all?
The card analogy explains it in detail, but there are many
other familiar random processes that help you understand how this happens. When you flip a coin thousands of times,
it’s probable that you’ll get “heads” as many times as “tails”. Each incidence of flipping the coin has
nothing to do with the previous flip.
However, it is unlikely that each flip will have the opposite result of
the preceding flip. You may get a whole
series of “heads” before another “tails” flip.
So, if everybody in your pool participated long enough, it’s likely that
selections would end up being evenly distributed. However, in the real world, people move in and out of the pool at
a rate that makes that impossible. This is especially true given the relatively
low rates of selection used in drug testing (i.e. 10%-50%).
Some words of caution: It’s possible for someone to have
increased chances of being picked if they are entered in the pool more than
once. The computer won’t let someone
with the same unique ID be entered in the pool twice, but if a mistake is made
whereby an individual exists in the pool under two or more unique ID’s, the
odds of that person being picked go up (like having two raffle tickets).
Another common problem is not removing “unavailable” people
from the pool on a regular basis. If
the pool contains ID’s of individuals that cannot be tested (no longer working,
extended leave, etc.), those that are available will be subjected to a higher
incidence of testing events.