The way in
which data are gathered can affect whether you are wrongfully accused of crimes
and can affect your success in challenging the evidence. This chapter considers
the way in which drug tests should be and are required to be conducted.
The technology
to identify criminal activity is finding increasing use in both the courtroom
and the workplace. Whether the method be a breathalyzer test, a test for illicit
drug use, DNA fingerprinting, or other, the average citizen (as juror, witness,
or defendant) can easily be overwhelmed by the scientific, technological aura
of these methods. Yet, technology does not guarantee accuracy, and the public
should not place blind trust in the application of such technologies,
especially when someone's liberty or life is at stake. As a society, we want
the benefits of these new technologies, but at the same time, we want to make
certain that they are applied in a fair manner.
Short of
learning the technologies yourself, there are a few simple steps you can
undertake to determine whether such technologies are being applied fairly. The
law does not explicitly guarantee an individual's right to sound scientific
practices when gathering evidence to convict them, but one's constitutional
right to a fair trial can certainly be argued to include these rights. Any test
used against you should have the five relevant features of ideal data.
Unfortunately, although a good defense attorney might be able to insist on the
criteria of ideal data in court, corporate use of scientific data to fire
employees is not subject to such scrutiny: a drug test may be used as the basis
of job termination without the employee's knowledge of the basis for dismissal.
On the positive side, however, the Department of Transportation has instituted
a lengthy and detailed protocol for lab-based drug-testing and alcohol-testing,
which incorporates several features of ideal data. In what follows, we list the
elements of what we regard as a fair drug test and note whether the features
are included in the Dept. of Transportation rules (DOT rules). By and large,
the DOT rules go a long way toward ensuring that ideal data criteria are met. The last part of the chapter deals with
DWI data that are not part of lab testing under the DOT.
The following
paragraphs consider each of our ideal data features, first describing our model
of a fair drug test and then followed by the relevant features prescribed by
the DOT. A summary of whether the DOT rules match the features of ideal data is
offered below.
GOAL: assess whether the
concentration of a drug exceeds a threshold |
||
MODEL being tested: the
drug level is below the threshold |
||
DATA feature |
Relevant? |
Status |
EXPLICIT
PROTOCOL |
yes |
present |
REPLICATION |
yes |
present |
STANDARDS |
yes |
present |
RANDOMIZATION
|
yes* |
present |
BLIND |
yes |
present |
* Randomization
is relevant for certain civil rights issues but is not a major factor in data
accuracy in this case
Ideally. The protocol for gathering,
analyzing, recording and reporting the findings should be published before any
drug tests are conducted. It should explicitly cover: 1) the procedures used to
collect the specimens, 2) the criteria or methods used to decide which
employees will be tested, 3) how the specimens will be handled between
collection and being received by the testing laboratory, and 4) the
experimental protocol employed by the testing laboratory, and 5) what
information and specimens will be stored as a permanent record.
DOT rules. The Department of
Transportation has developed a comprehensive protocol for its drug testing
program. The details are presented in the Federal Register 49 CFR PART 40 and
described in the 100+ page pamphlet "Guidelines for Implementing the FHWA
anti-drug program" published by the DOT in 1992. These rules specify which
records are to be maintained, how long samples must be kept, procedures for
handling and transferring specimens, what kinds of tests are to be performed on
which classes of employees, and many other features needed for a program of
this nature. More important for our purposes, these rules specify a number of
features that pertain to our five additional design features, as explained
next.
Ideally. Each batch of samples
sent to the testing laboratory should contain several standards, some known to
be non-positive and some known to be positive. The testing laboratory should
not be told which samples are standards. The company should publicize the
results of the standards.
DOT rules. Standards are included
at two levels. The testing laboratory itself is required to include samples of
known concentrations in each run. These standards enable them to calibrate the
test procedure and thus to determine drug concentrations in the unknowns. As an
additional standard, each manager whose employees are tested is required to
include 3 known blank samples (lacking drugs) for every 100 unknowns; managers
with more than 2000 drivers must also include some standards known to contain
drugs. The rules do not specify publication of the results but they require
that the DOT be notified of all "false positives," so the information
should be accessible. Whether the data are, in fact, treated this way would be
difficult to determine.
Ideally. Each sample should be
divided into at least two tubes. One should be retained as a voucher for
retesting, if necessary. The other should be sent to the testing laboratory. An
even better design, however, would be to split the sample three ways, sending
two for independent testing at different laboratories. After the tests are
completed, the company should publicize what percent of the replicated samples
came back with inconsistent results.
DOT rules. As of 1994,
replication is required in the form of a "split sample" collection
(at least for urine tests). Split sample collection simply involves the
partitioning of the original sample into two vials. One vial is sent for
testing, whereas the other is retained for retesting in the event of a positive
result on the first vial. The rules specify that the retest is to be conducted
by a different laboratory than did the first test. Retesting is not automatic,
however: the person being tested must request the retest. In addition, the lab
uses the original sample for multiple-stage testing - the specified procedure
is that all samples are tested with an initial and rapid screen, and only the
positive ones are analyzed more comprehensively. The rules specify that written
records must be kept, but we are not aware that the results of replications
must be published or made generally available.
Ideally. The decision about
which employees are tested should be random. There may be special circumstances
which invariably warrant testing so that no element of choice is involved
(e.g., a driver involved in some kind of accident), but if a choice is to be
made from a group of otherwise equivalent individuals, that choice should be
random. Random choice here is not so much to reduce error but to avoid abuses.
DOT rules. One form of required
testing is random. The rules specify that the choice in "random"
testing must be made in a strictly random manner, with suggested methods being
the use of random number tables or names on slips of paper drawn from a bin. The
rules further specify that sampling with replacement must be used: an
individual tested in the previous round is eligible to be chosen in the next
round.
Ideally. All samples should be
labeled with a number, not a person's name, when sent to the testing
laboratory. Standards should be indistinguishable from the unknowns so that the
lab is not explicitly careful just when processing the standards. Additionally,
the testing laboratory should receive no correspondence concerning which
specimens the company suspects might be positive.
DOT rules. Blind testing is used
at two levels. Most importantly, names are not included with the samples; the
name of the person being tested is kept only on a single form (one of 3 copies)
with the manager requesting the test. In addition, the standards are to be
labeled in ways that make them appear as ordinary samples.
In science, the
penalty for being caught falsifying evidence is complete and total
ostracization from the scientific community. However, perhaps the major
deterrent against fraud in academic is the high probability of being detected:
any important discovery will be checked and verified by other scientists. Drug
tests are different because there are no repeated cycles of evaluation and
revision. That is, if your blood or urine tests positive by a fraudulent
procedure, the authorities will not come to you later and take another sample
to confirm the result. It is thus imperative that procedures be implemented to
guard against falsification of test results. Of course, the use of replicates,
standards, and blind procedures help in this respect. But the rules also
further safeguard against falsifying results by requiring tamper-proof labels
and requiring that all tests results be sent to an independent 3rd party (the
medical review officer) before being returned to the office originating the
test. In general, the DOT rules are far more specific about matters of personal
liberties and rights than is the typical procedure in science, largely because
most matters of pure science do not impinge on people's rights and freedoms,
and they can be verified at a later date.
DWI field testing
Providing a blood sample for DWI testing should subject the analysis to the usual DOT rules because the sample is processed in a lab. However, providing breath samples or taking the SFST are done on site and don't necessarily have all those safeguards. At least in Austin, a breathalyzer test involves two types of standards, with known concentrations of alcohol and alcohol-free samples. Furthermore, the subject provides two breath samples (replication). So sampling error, human/technical error, and bias should be minimized; there is little opportunity for bias, so blind and randomization are not necessary.
A more problematic measurement is the score on the SFST. To ensure data accuracy, the SFST involves:
When there is no video record of the test, the score will be almost impossible to challenge because there is no basis to evaluate the officer's scoring. This is a serious drawback of the test, although video cameras are now standard features of most patrol cars in this state. Perhaps the main drawbacks of the SFST when a video record is obtained are (i) the HGN test cannot be reliably interpreted on the video, and (ii) there are no baseline data for the SFST, either for the population at large or for the individual being tested. In the few rigorous studies of SFST performance, there are no data for our population that establish failure rates for sober people, which would be considerable, especially among older people. Such a database might work against using the SFST in some cases, because it is undoubtedly the case that thresholds for arrest are stringent. (I would speculate that a majority of people over 65 would fail it when sober because coordination declines with age.) Although baseline data for an individual could be provided at a later time, they never are. Nor would the performance of the arrested individual when sober be a blind measure of their ability, as there would be a strong incentive to perform poorly when establishing a baseline. However, as the HGN is involuntary, baseline data could be provided after the fact.
Copyright
1996-2000 Craig M. Pease & James J. Bull