AN EVALUATION OF POCKET-MODEL, NUMERICAL READOUT
BREATH ALCOHOL TESTING INSTRUMENTS
A Dissertation
by
WILLIAM EDWARD VAN TASSEL
Submitted to the Office of Graduate Studies of
Texas A&M University
in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY
August 2003
Major Subject: Health Education
AN EVALUATION OF POCKET-MODEL, NUMERICAL READOUT
BREATH ALCOHOL TESTING INSTRUMENTS
A Dissertation
by
WILLIAM EDWARD VAN TASSEL
Submitted to Texas A&M University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Approved as to style and content by:
___________________________ ___________________________
Maurice E. Dennis Walter F. Stenning
(Chair of Committee) (Member)
___________________________ ___________________________
Quinn R. Brackett, Jr. Steve M. Dorman
(Member) (Head of Department)
___________________________
Buster E. Pruitt
(Member)
August 2003
Major Subject: Health Education
iii
ABSTRACT
An Evaluation of Pocket-Model, Numerical Readout
Breath Alcohol Testing Instruments. (August 2003)
William Edward Van Tassel, B.B.A., Texas A&M University;
M.B.A., Texas A&M University; M.A. University of Central Oklahoma
Chair of Advisory Committee: Dr. Maurice E. Dennis
Eight small-scale breath alcohol measurement devices were tested for accuracy,
precision and the ability to not yield false positive and false negative readings. These
pocket-sized breath testers (PMBTs), which provided numerical readout of BrAC to the
100th of a percent, were smaller than evidential and preliminary breath test instruments
(EBTs and PBTs). The smallest devices were approximately the same size as a cigarette
lighter. Designed to provide drinkers feedback about their individual alcohol levels, the
PMBTs ranged in price from $40-100 USD.
The devices were first tested under laboratory conditions with alcohol solution
simulators providing the alcoholic samples. They were then tested with human drinkers,
under controlled field conditions. Each device was tested at multiple alcohol levels.
Two of the eight PMBTs failed to complete all levels of testing and were excluded
from the study. All PMBTs demonstrated the ability to not yield false positive and false
negative readings. No device met NHTSA performance criteria for accuracy (systematic
error) in testing EBTs at every alcohol level tested. An interaction between PMBTs and
the alcohol test levels was found. Thus, accuracy was found to be dependent upon the
iv
alcohol level at which the devices were tested. No device met NHTSA performance
criteria for precision in testing EBTs at every alcohol level tested. Further, precision
varied depending on the testing condition, as there was less precision under controlled
field conditions than under laboratory conditions. Five of the six PMBTs that completed
the testing overestimated BrAC; only one device read below actual BrAC.
Ramifications of the findings are discussed, regarding the overestimation and
underestimation of BrAC and the possibility of manufacturers intentionally calibrating
the devices to overestimate BrAC. Potential PMBT users are discussed and areas for
future research are addressed.
v
ACKNOWLEDGMENTS
The completion of this study was possible thanks to the support offered by many
individuals. I am very grateful to Dr. Maurice E. Dennis, Chair of the Advisory
Committee. His support throughout the dissertation process, as well as through my
entire doctoral program, has been absolute and of great magnitude. He has been tireless
in his encouragement and never-ending in his commitment to me as a student. I could
not have asked for a better mentor under which to study. His approach will serve as my
model as I continue in the academic environment.
One’s family certainly is a substantial factor in the ability to pursue a doctorate. To
this end, I express special thanks to my mom, who has been wholly supportive in my
academic efforts. Her dedication in capturing and relaying to me virtually all traffic and
alcohol-related events in the media has been especially helpful. It’s great having one’s
own personal clipping service! To my father goes special appreciation. His input
regarding the study and draft documents was of great use and he consistently confirmed
that this research was of real value. Somehow he always knew I would make it!
Without the complete support and assistance of Margaret Parker with the Texas
Department of Public Safety, this research simply would not have been possible. Since
the project’s inception, she has given freely of her time and effort to help ensure the
success of the research. It is always great to work with someone who shares an interest
in answering unanswered questions!
vi
The members of my graduate advisory committee also deserve my appreciation.
Special thanks goes to Dr. Quinn Brackett, Dr. Buster Pruitt and Dr. Walter Stenning.
Their individual and collective input substantially strengthened the dissertation.
I would also like to thank Dr. Charles Shea, Dr. Lindsay Griffin, III and Dr. Homer
Tolson for all their help with the statistical design and analysis of the data. My special
thanks goes to Sergeant Rod Gullberg of the Washington State Patrol, who has been
tireless in his support and input regarding the design and analysis phases. Through these
professionals I have learned much that will be of value in my future research.
Without a strong network of supportive friends and colleagues this effort would not
have been possible. Great thanks goes to Robert “Dr. Bob” Reinhardt, who provided
great assistance in all phases of this project and who yielded special insights into the
consequences of subjecting one’s self to this endeavor. Dr. Michael Manser also served
as a substantial resource throughout the dissertation. His, as well as Dr. Bob’s periodic
kicks in the butt kept me motivated to make it all the way through. Thanks to the
Bondurant Crew!
I extend my great appreciation to Dr. Stacey Stevens, Dr. John Green, Dr. Danny
Ballard, Dr. Jim Robinson and Dr. Ben Cranor for their friendship and support
throughout every phase of this effort. Their view that it was always a matter of when,
rather than if, I would complete this phase was a source of constant hope. Special thanks
go to Dr. Robert Armstrong, Dr. Stephen Crouse and Dr. Steve Dorman for their support
and for facilitating use of departmental resources in this research.
vii
I would also like to recognize the supportive input and assistance from Becky Davies
and Dave Willis, of the Center for Transportation Safety at the Texas Transportation
Institute. They were both wholly and actively supportive of my use of TTI facilities
during this research. My thanks also goes to Dr. James Frank of NHTSA. His support
was total, active at the earliest stages of the effort. Appreciation is also due to Dr. A.W.
Jones, who provided very useful input regarding study design and comments regarding
the devices to be tested.
Doing research with human participants can greatly complicate the conduct of
research. However, several people helped reduce the burden of acquiring qualified
participants for this research. My thanks go to Dottidee Agnor, Gayden Darnell, Beth
Tessandori and Dr. Jerry Elledge. My appreciation is also extended to those who helped
me corral and monitor the participants, including Donna Adler, Romona Clark, Jodi
Luecke and Christie Dickson.
Most of all, I thank the good Lord, who has guided me to and through this research.
viii
TABLE OF CONTENTS
Page
ABSTRACT…………………………………………………………………… iii
ACKNOWLEDGMENTS…………………………………………………….. v
TABLE OF CONTENTS……………………………………………………… viii
LIST OF FIGURES…………………………………………………………… x
LIST OF TABLES…………………………………………………………….. xi
CHAPTER
I INTRODUCTION…………………………………………………... 1
1.1 Statement of the Problem…………………………………. 9
1.2 Purpose of Study………………………………………….. 9
1.3 Value of Study…………………………………………….. 10
1.4 Research Hypotheses……………………………………… 11
1.5 Independent Variables………………………….….……… 13
1.6 Dependent Variable…………………………….……….… 13
1.7 Operational Definitions…………………………………… 13
II REVIEW OF LITERATURE………………………………………. 15
2.1 Modern Breath Alcohol Testing…………….……………. 15
2.2 Breath Alcohol Detection Technology…………………… 19
2.3 Breath Alcohol Sample Simulators………………………. 23
2.4 Breath Alcohol Measurement Devices…………………… 26
III METHODOLOGY…………………………………………………. 42
3.1 Test Devices……………………………………………… 42
3.2 Pilot Testing………………………………………………. 43
3.3 Experiment One…………………………………………... 48
3.4 Experiment Two………………………………………….. 48
3.5 Limitations of the Study………………………………….. 61
3.6 Delimitations of the Study………………………………... 63
ix
CHAPTER Page
IV RESULTS…………………………………………………………… 64
4.1 General……………………………………………………. 64
4.2 Pilot Testing………………………………………………. 64
4.3 Experiment One…………………………………………... 66
4.4 Experiment Two………………………………………….. 71
V DISCUSSION………………………………………………………. 87
5.1 Hypothesis Testing……………………………………..… 87
5.2 Precision………………………………………………….. 90
5.3 Accuracy…………………………………………………. 91
5.4 Relationship to Intoxilyzer……………………………….. 93
5.5 Summary…………………………………………………. 94
VI RECOMMENDATIONS AND CONCLUSION………………….. 98
6.1 Recommendations for Future Research………………….. 98
6.2 Recommendations for PMBT Manufacturers……………. 102
6.3 Recommendations for Addressing Research
Methodological Issues………………………………..…... 105
6.4 Conclusion…………………………………………….….. 107
REFERENCES………………………………………………………………... 109
APPENDIX A: SAMPLE IMAGES OF QUANTITATIVE PMBT
MODELS…………………………………………………….. 120
APPENDIX B: CLOSED SYSTEM WET BATH SIMULATOR…………… 129
APPENDIX C: THE ALCOHOL USE DISORDER INVENTORY TEST….. 131
APPENDIX D: THE NUMERICAL DRINKING PROFILE………………… 138
APPENDIX E: INFORMED CONSENT FORM…………………………….. 145
APPENDIX F: LATIN SQUARE COUNTERBALANCING SYSTEM…….. 149
APPENDIX G: PARTICIPANT DATA SHEET……………………………... 151
VITA…………………………………………………………………………… 153
x
LIST OF FIGURES
FIGURE Page
1 Mean BrAC Results of Each Device Tested at .032……………….. … 65
2 Mean BrAC Results of Devices at Each Simulator Concentration.…... 67
3 Mean Results for Each Device at Each Simulator Concentration..…… 68
4 Standard Deviations of Each Device at Each Simulator
Concentration…………………………………………………………. 70
5 Mean BrAC Results for Devices at Each Alcohol Concentration……. 76
6 Mean Intoxilyzer and Device Results for Each Device at
Each Concentration …………………………………………………… 77
7 Standard Deviations of Each Device at Each Concentration………….. 78
8 Correlations Between the Intoxilyzer and Each Device…….…………. 86
xi
LIST OF TABLES
TABLE Page
1 Participant Variables…………………………………………………... 56
2 Acceptable Range for Individual Testing………………..…………….. 59
3 Mean Test Results for Each Device at Each Concentration…………… 66
4 Standard Deviations for Each Device at Each Concentration…………. 69
5 Means and Standard Deviations for Each Device at Each
Concentration………………………………………………………….. 72
6 Amount Added to Each Device, Based on Testing Order…………….. 74
7 Corrected Means and Standard Deviations for Each Device at Each
Concentration………………………………………………………….. 75
8 Results of Mauchly’s Test…………………………………………….. 79
9 Results of Repeated Measures Analysis of Variance………………….. 80
10 Results of Repeated Measures Analysis of Variance, Lower Bound
Values………………….……………………………………………….. 81
11 Simple Main Effects…………………………….……………………… 82
12 Differences Between Intoxilyzer and Each Device at
.08 Concentration………………………………………………………. 85
13 Correlations Between Intoxilyzer and Each Device…………………… 85
1
CHAPTER I
INTRODUCTION
While automobiles have only recently reached the century mark of existence, they
have provided a degree of personal mobility never before experienced. Automobiles
provide rapid transportation, permit infinite recreation options and afford their users
wide choices regarding where to live and work. This individual mode of transport is
widely available, relatively affordable and facilitated by solid roadway infrastructures.
In the United States (US) and many other nations, automobiles have become the
mechanized equivalent of freedom.
Unfortunately, each time a driver operates a motor vehicle there is a risk of serious or
fatal injury. With over 220 million registered vehicles in the US, crashes are inevitable
(National Safety Council, 2002). Causal factors include vehicle malfunctions, poor
environmental conditions and human error (Fell, Hendricks & Freedman, 2000; Shinar,
1978). These negative events can cause sudden and violent impacts, resulting in
property damage and injury or death of vehicle occupants.
Each year in the US, over 40,000 people die as a result of motor vehicle collisions.
This degree of loss equates to approximately three commercial jet aircraft crashing each
week. While losing so many lives in “chunks” in such an aviation scenario would likely
_______________
This dissertation follows the style and format of Accident Analysis and Prevention.
2
cause widespread panic, the yearly number of automotive-related deaths appears to be
far less a social concern. One factor may be the pattern of automobile crashes: the
pattern is diffused, with injuries and deaths occurring over many small, discrete crashes
throughout the year, spread over the entire US (”Low Priority,” 2002).
In fact, automobile crashes are the leading cause of death for Americans age 1-34
(Insurance Institute for Highway Safety, 2002). In addition, over two million people
annually suffer from disabling injuries as a result of car collisions (National Safety
Council, 2001). Further, the economic costs of motor vehicle collisions are staggering.
Such costs include medical expenses, productivity losses, employer costs and property
damage (National Safety Council, 2001). The U.S. Department of Transportation
(2002a) estimates that the annual cost of such crashes to society exceeds $150 billion.
This equates to a yearly cost average of approximately $790 per licensed driver in the
US. Regardless of how automobile crash results are measured, motor vehicle collisions
represent a major threat to public health and an enormous drain on the U.S. economy.
Only partially consoling is the fact that the motor vehicle crash rate was even worse in
years past. From the US peak death rate per 100,000 population of 30.8 in 1937, the
national rate has decreased to a rate of 15.4 in 2001, a decline of 50% (National Safety
Council, 2002). Many factors are credited with reducing injury and death rates,
including:
3
(a) Vehicle advances such as padded interiors, airbags, anti-lock braking systems,
improved tires and traction control systems (Davis, 2002; National Highway
Traffic Safety Administration, 2003a);
(b) Roadway improvements such as improved guardrails and lighting, and rumble
strips (National Highway Traffic Safety Administration, 2003b);
(c) Legislative initiatives aimed at raising the cost of unsafe driving behavior
(National Highway Traffic Safety Administration, 2003b);
(d) Advances in enforcement operations such as electronic speed measurement,
improved communications and accident reduction efforts (National Highway
Traffic Safety Administration, 2002).
In general, most crashes have been found to be attributed to human error/impairment
as opposed to vehicular or environmental factors (Fell, Hendricks & Freedman, 2000;
Moskowitz, 2002). Contributing to this finding is the myriad of ways that drivers can be
impaired, including:
(a) Distraction (AAA Foundation for Traffic Safety, 2001);
(b) Fatigue/drowsiness (National Highway Traffic Safety Administration, 1998);
(c) Road rage/aggressive driving (AAA Foundation for Traffic Safety, 1997);
(d) Drugs, including alcohol (Dennis, 1995; Moskowitz & Robinson, 1988).
Of all the causes of motor vehicle crashes, alcohol-related crashes remain the single
largest factor. Alcohol, a legal depressant drug, is widely available and widely abused in
the US (Hanson, Venturelli & Fleckenstein, 2002). Almost 40% of all motor vehicle
fatalities result from alcohol’s deleterious effects on driving ability (National Safety
4
Council, 2001). The National Highway Traffic Safety Administration (NHTSA) defines
an alcohol-related crash as one in which “either a driver or a nonoccupant (e.g.,
pedestrian) had a blood alcohol concentration (BAC) of 0.01 grams per deciliter (g/dl) or
greater in a police reported traffic crash” (Mothers Against Drunk Driving, 2003, p.2).
Compared to total crashes, state fatal alcohol-involved crash rates range from 23.9%
(Utah) to 50.4% (Texas) (National Safety Council, 2002).
Fortunately, the alcohol-involved crash rate has declined in recent years (Moskowitz,
2002). Since 1982, the percentage of US fatal crashes involving alcohol has declined
nearly 50% (Jones & Lacey, 2001). Several efforts have been credited with the progress
to date, including:
(a) Efforts of organizations such as Mothers Against Drunk Driving and Remove
Intoxicated Drivers;
(b) Federal, state and regional programs aimed at reducing the incidence of impaired
driving;
(c) A growing social intolerance of the act of driving while impaired (Jones &
Lacey, 2001);
(d) Use of technology to determine drivers’ alcohol levels (Harding, 1996;
Dubowski, 1992).
While the incidence of impaired driving has declined over the last two decades,
progress seems to have leveled off over the past few years. It has been noted that the
number of alcohol-related crashes has reached a plateau, with little change over the past
several years (Jones & Lacey, 2001; “Progress Against,” 2002).
5
There is consensus that the most effective way to prevent alcohol-impaired driving
behavior is to avoid driving after consuming any alcohol (Burns & Fiorentino, 2002;
Muhammad, 2000). Unfortunately, some people do choose to, and even plan to, drive
after consuming alcohol, with sometimes catastrophic results (Jones & Lacey, 2001).
Several objective methods of determining blood alcohol levels have been developed,
including measuring saliva, vapors emanating from the eye, blood, urine, tissue, spinal
fluid and deep-lung breath (U.S. Department of Transportation, 1994; Caplan, 1996,
National Highway Traffic Safety Administration, 1982). Developed in the 1940s,
measurement of blood alcohol levels from breath samples was originally designed for
law enforcement forensic purposes, but has since spread to other areas, including the
medical, aviation, trucking and other transportation and non-transportation oriented
industries (Mason & Dubowski, 1996; Harding, 1996; Freudenrich, 2002). Breath
testing involves measuring the amount of alcohol captured in expired deep-lung air. Use
of breath alcohol concentration (BrAC) has become a very common method of
determining blood alcohol concentration (BAC), as it does not require using trained
medical personnel to obtain and analyze blood samples (Mason & Dubowski, 1996;
Harding, 1996; CMI, 2002a).
Two main types of breath alcohol testing devices exist: disposable and reusable.
Disposable devices are inexpensive and typically involve the user exhaling through a
clear plastic cylinder approximately the size of a cigarette. The tube contains a mixture
of chemicals that reacts as breathe-borne ethanol flows through. Users interpret any
6
resulting change in color of the mixture to assess their blood alcohol level. Such devices
are usually set to react at specific alcohol thresholds, such as .04 or.08 (AlcoPro, 2003).
Several types of reusable BrAC measurement devices have been developed. The most
often used devices are large fixed based units primarily used for law enforcement
evidentiary purposes. One of the first of these devices was the “Breathalyzer,” invented
in 1954 by Dr. Robert Borkenstein (Ezelle, 2002). Known as evidentiary breath testers
(EBTs), these devices, such as the Intoxilyzer 5000, represent the most accurate breath
alcohol measurement instruments available (CMI, Inc., 2002a). These BrAC devices
generally remain at one location, require regular calibration and necessitate thorough
training of their operators (Taylor & Hodgson, 1995; Dubowski & Essary, 1992). EBTs
provide a digital readout to the 1000th of one percent BrAC.
Preliminary breath testers (PBTs) are approximately the size of a VHS cassette.
These hand-held, battery powered screening devices are used in the field to supplement a
law enforcement officer’s observations in determining whether a suspected alcoholimpaired
driver should be arrested (Olson, 1986; Forrester, 1997). Results from PBTs
may or may not be introduced as court-reported evidence. These instruments also
provide a digital readout of BrAC, either two or three digits to the right of the decimal.
Passive alcohol sensors (PASs) surreptitiously collect normally exhaled breath from
drivers during an interaction with law enforcement personnel. Designed to help officers
screen potentially impaired drivers, these devices require no action by motorists and are
built into innocuous-appearing devices such as flashlights and clipboards. PAS
instruments help determine whether alcohol is present and, if so designed, approximately
7
how much. An effective PAS will minimize false positive readings (where low BrACs
are incorrectly identified as high BrACs) and maximize the likelihood that high BrACs
are detected (Lestina & Lund, 1992). Results are displayed either numerically or by a
series of lights.
Breath alcohol ignition interlock devices (IIDs) are designed to prevent drivers who
have consumed even small amounts of alcohol from starting their automobiles. About
the size of an electric razor, these breath analysis devices are hard-wired into a vehicle
and will not permit engine ignition if the driver’s breath has a breath alcohol
concentration higher than a predetermined threshold, usually .025 (Voas, Blackman,
Tippetts and Marques, 2002). IID users are generally DWI offenders who have received
permission to resume driving after having lost all such privileges for some period of time
(Frank, 1997). Modern IIDs log all start attempts and violations and also mandate
“rolling retests,” which require the driver to periodically provide additional breath
samples in order for the vehicle engine to continue running (Smart Start, 2003; Comeau,
2000; Marques, Voas, Tippetts & Bierness, 1999).
Coin operated breath measuring devices permit users to self-test their alcohol level.
Slightly smaller than cash register machines, these instruments are designed for
establishments that serve alcohol, including hotels and bars (Wundersitz, 2002). For
each single use of the instrument, users pay a small fee, which generally includes a fresh
mouthpiece (The Alcohol Alert System, 2002). Currently, coin operated breath testers
(COBTs) are not widely available in the US.
8
The sixth group of breath test devices consists of a relatively new class of testing
devices. Yet to acquire a commonly used label or acronym, these portable units are
designed more for personal/civilian use, rather than for law enforcement applications.
These battery-powered devices are even smaller than PBTs, with some being quite thin
and not much larger than a pack of chewing gum (Stellin, 2001). They generally do not
permit user calibration, require no training other than reading the manufacturer’s
operating instructions and, at a cost of $25 to $150, are far less expensive than both
EBTs and PBTs.
This group of devices provides information about BrAC in one of two ways. First,
some devices provide a qualitative readout, generally using a system of lights to provide
information to the user. This can take the form of a binomial system (alcohol present or
alcohol not present) or a system of ranges (e.g., BrAC ranges of .00-.04%, .04-.08%,
.08-.12%, .12-.15% and .15% and higher). Second, other devices provide a quantitative
readout, generally in 100th percent of BrAC. This paper shall refer to these devices as
“pocket-model breath testers” (PMBTs). Images of the PMBTs tested in this study are
provided in Appendix A.
EBTs, PBTs, PASs, IIDs and COBTs have all undergone rigorous laboratory and
field-based analyses to evaluate their performance. However, there is a lack of
evaluation of the newer, lower-cost PMBTs. Given that drivers are generally poor
estimators of their own alcohol level (Silverstein, Nathan & Taylor, 1974; Van Tassel &
Manser, 2000), the best option for any drinker is to not drive after consuming any
alcohol. Sadly, people do all too often elect to drive after drinking. In situations where
9
the optimal rule of no-driving-after-drinking fails, the next best option might be to
facilitate feedback of drinkers’ own alcohol levels. Such feedback might result in better
decisions about whether or not to drive. PMBTs have the potential to fulfill this
function, and researchers have identified the need to develop and validate alternate
methods of informing drinkers about the alcohol levels they have achieved (Dubowski,
1985). This study seeks to evaluate the performance of small-scale breath testers.
1.1. Statement of the Problem
Determination of breath alcohol from expired air is a commonly used method to
measure blood alcohol concentration. The traditional devices used to obtain
measurements of BrAC (EBTs, PASs, PBTs, CODs and IIDs) have been evaluated
thoroughly to assess their performance. These units have been shown to demonstrate
sufficient precision and accuracy to be used for their intended purposes.
A new class of measurement devices, pocket-model breath testers, has been released
for public use within the past few years that may have the potential to help reduce the
incidence of alcohol-impaired driving. These units have not yet undergone rigorous
evaluation of their performance and an exploratory evaluation is needed prior to any
widespread use.
1.2. Purpose of Study
The purpose of this exploratory study is to evaluate the performance of commonly
available quantitative pocket-model breath testers (PMBTs). The evaluation consisted of
10
two experiments. The first was designed to assess the accuracy and precision of
measurements made under laboratory conditions. The second was performed to assess
the accuracy and precision under simulated field conditions.
1.3. Value of Study
Individuals sometimes make important decisions after consuming alcohol, including
decisions about driving. Poor decisions can result because decision-making ability is the
first human function to be affected by alcohol (Texas DWI Education Program, 2001).
Thus, not only do drinkers tend to make poor decisions, but because of alcohol
impairment, they do not recognize that their decision-making ability has been affected.
In most drinking situations, the sole input drinkers have about their current BAC is
their subjective estimate; there is rarely an available method for them to obtain a
quantitative measure of their BAC. Perhaps not surprisingly, drinkers’ subjective
estimates of BAC, made in the absence of accurate feedback, have been found to be of
low accuracy (Silverstein, Nathan & Taylor, 1974).
It has been recommended for some time that new and better means of providing
drinkers information about their current alcohol level be validated (Dubowski, 1985).
As most states’ laws involve a numerical expression of intoxication (e.g., .08% BAC),
drinkers’ decisions might benefit from numerical input regarding their current BrAC.
The PMBT devices to be examined in this study have the potential to provide such
quantitative input, possibly aiding drinkers in making better decisions after alcohol is
11
consumed. Better decisions regarding driving by people who have consumed alcohol
could lead to fewer alcohol-related injuries and deaths.
1.4. Research Hypotheses
1.4.1. Experiment One
1.4.1.1. Hypothesis One
Under laboratory conditions, each PMBT device will be less accurate than the
National Highway Traffic Safety Administration (NHTSA) criteria at each alcohol level
(will yield a systematic error greater than ± .005).
HO: Systematic error ≤ ± .005 at each alcohol level
HA: Systematic error > ± .005 at each alcohol level
1.4.1.2. Hypothesis Two
Under laboratory conditions, each PMBT device will be less precise (more variable)
that the NHTSA criteria at each alcohol level (will yield a standard deviation greater
than .0042).
HO: Standard deviation ≤ .0042 at each alcohol level
HA: Standard deviation > .0042 at each alcohol level
12
1.4.1.3. Hypothesis Three
Under laboratory conditions, each PMBT device will become less accurate as test
BrAC increases (systematic error will increase when measured at .02, .04, .06, .08, .10
and .16).
HO: Systematic error at .02 ≤ .04 ≤ .06 ≤ .08 ≤ .10 ≤ .16
HA: Systematic error at .02 > .04 > .06 > .08 > .10 > .16
1.4.1.4. Hypothesis Four
Under laboratory conditions, each PMBT device will become less precise (more
variable) as test BrAC increases (standard deviation will increase when measured at
.02, .04, .06, .08, .10 and .16).
HO: Standard deviation at .02 ≤ .04 ≤ .06 ≤ .08 ≤ .10 ≤ .16
HA: Standard deviation at .02 > .04 > .06 > .08 > .10 > .16
1.4.2. Experiment Two
1.4.2.1. Hypothesis Five
Under simulated field conditions, each PMBT device will yield results significantly
different than results from a calibrated Intoxilyzer 5000 breath alcohol test instrument.
HO: Intoxilyzer = A = B = C = D = E = F = G = H at each alcohol level
HA: Intoxilyzer ≠ A ≠ B ≠ C ≠ D ≠ E ≠ F ≠ G ≠ H at each alcohol level
13
1.5. Independent Variables
IV1: Simulator solution alcohol concentration (Experiment One).
IV2: PMBT used by each participant (Experiment Two).
IV3: Amount of alcohol consumed by each participant
(Experiment Two).
1.6. Dependent Variable
DV3: BrAC measurement result (Experiments One and Two).
1.7. Operational Definitions
(a) Accuracy- A measure of the closeness of agreement between the result of
analysis and the true value of the quantity being measured; the proximity of a
quantified measurement result to the true value of the property being measured.
(b) Blood Alcohol Concentration (BAC)- Grams of alcohol per 100 milliliters of
blood. This is equivalent to the metric used to measure breath alcohol
concentration (BrAC), grams of alcohol per 210 liters of breath.
(c) Breath Alcohol Concentration (BrAC)- Grams of alcohol per 210 liters of breath.
This is equivalent to the metric used to measure blood alcohol concentration
(BAC), grams of alcohol per 100 milliliters of blood.
(d) Precision- Closeness of agreement between independent results of measurements
obtained by a procedure under prescribed conditions; the variation or scatter of
14
the measurements about the mean; the degree to which replicate measurement
results agree amongst themselves.
(e) Alcohol Solution Simulator- A device containing approximately 500 ml of an
ethanol/water solution heated to a known and constant temperature and designed
to provide a known vapor concentration of ethanol for calibration and testing of
instruments.
Unless otherwise stated, all measurements results in this study will be expressed in terms
of BrAC, grams of alcohol per 210 liters of breath.
15
CHAPTER II
REVIEW OF LITERATURE
2.1. Modern Breath Alcohol Testing
With references to the methodology made as early as 1874, breath alcohol analysis
has developed into a primary method of measuring the concentration of alcohol in the
body (Lucas, 2000; Deveaux & Gosset, 2000). Several advantages have led to its
increased use throughout the world. First, unlike measuring alcohol directly from blood,
medical personnel are not required to collect a sample. Second, no laboratory services
are necessary for sample analysis. Third, it offers immediate results. Fourth, it
minimizes the time between the event or arrest and the subsequent testing (Mason &
Dubowski, 1996; Harding, 1996; National Highway Traffic Safety Administration,
1982). These advantages have combined to move breath testing to the forefront of
alcohol measurement. Modern breath testing instruments have developed to the point
that when used by people with limited or no scientific training, they can provide reliable
results under non-laboratory conditions (Harding, 1996).
The basic process of breath alcohol testing can be divided into three components
(Dubowski, 2002):
1. Input Phase. The participant provides a breath sample into a measurement
instrument.
16
2. Analysis Phase. Any ethanol present in the breath sample is identified and
quantitated.
3. Output Phase. The measurement instrument displays the results of the test.
Secondary phases would include the interpretation and use of the results, and quality
assurance efforts.
Breath alcohol measurement is based on the principle of equilibrium. This principle
asserts that the ratio of alcohol concentrations between a blood sample and a breath
sample is a constant value (National Highway Traffic Safety Administration, 1982).
That is, arterial blood is in equilibrium with deep lung (alveolar) air (Hlastala, 1998).
Not surprisingly, the concentration of alcohol in blood is much higher than that of
alcohol in alveolar air. A ratio of 2100:1 has traditionally been used to describe the
relationship between alcohol in deep lung breath and blood, respectively (Mason &
Dubowski, 1996; Hlastala, 2002; CMI, 2002a). This ratio is generally referred to as the
“partition ratio” (Melethil, 2002).
As states began to adopt per se intoxication laws, their statutes frequently and
logically included a metric commonly used to specify the amount of alcohol in the
blood: grams of alcohol per 100 milliliters of blood (Dubowski, 2002; Gullberg, 1990a).
However, as the use of breath alcohol testing spread, it became the norm to express
breath alcohol results in a metric more closely aligned with its gaseous-form sample
source: grams of alcohol per 210 liters of breath (Jones, 2002). Thus, each method
17
dictated its most appropriate and scientifically sound metric. Most states’ statutes
included only the original metric used for blood alcohol measurement, however.
In order to provide any sort of meaningful comparison between BrAC and BAC, it
became necessary, using the partition ratio, to convert BrAC results to BAC results.
This conversion has traditionally been a substantial source of contention. The originally
applied ratio of 2100:1 may have been somewhat arbitrarily employed (Mason &
Dubowski, 1996). In addition, contention over the ratio has arisen because of claims that
the partition ratio may not be constant (Hlastala, 1998); it may vary depending on a
number of factors (Jones & Andersson, 1996). Factors said to affect the partition ratio
include breath temperature, breathing technique just prior to providing a sample (hypoor
hyperventilation), hematocrit value, alcohol loss to the airway mucosa and
atmospheric pressure (Melethil, 2002; Hlastala, 2002).
Today, many states have amended legislation to include both blood and breath metrics
in their definition of intoxication. Thus in those states, if a suspect provides a blood
sample that exceeds a specified number of grams of alcohol per 100 milliliters of blood,
or provides a breath sample that exceeds a specified number of grams of alcohol per 210
liters of breath, he or she is considered legally intoxicated. By including both
definitions, the matter of converting breath alcohol results to blood alcohol results
became moot (Jones, 2002).
Not every interested party believes that breath alcohol testing is completely accurate
and precise. Detractors counter with claims that too many factors can interfere with such
testing to permit its use in evidentiary circumstances. It has been claimed that asthma
18
inhalant and nasal spray chemicals can inflate test results (Logan, Distefano & Case,
1998). Residual alcohol or mouthwash in the mouth has also been said to inflate test
results (Harding, 2002; Spector, 1971). In addition, interference from radio frequencies
has been cited as affecting test results (Gullberg, 2002a). Some detractors remain quite
vocal in their opposition to breath testing. One such detractor, Tucson defense attorney
James Nesci, proclaimed “breath testing in general is a load of crap, just pseudoscience
that they try to pull off” (Joseph, 2002, p. 3).
Effective quality control measures adequately address most, if not all, these
challenges. For example, concerns about inhalant and nasal sprays and residual mouth
alcohol can be addressed simply by employing a 15 minute deprivation period, where
the subject is not permitted to ingest any material for 15 minutes prior to providing a
breath sample. During the deprivation period, these potential interferents will have
dissipated (Logan, Distefano & Case, 1998; Brown, 1994). Claims that dentures and
mouth jewelry, such as tongue piercings, retain alcohol in the mouth and thus inflate
results have also been scientifically refuted (Harding, McMurray, Laessig, Simley,
Correll & Tsunehiro, 1992; Logan & Gullberg, 1998). Regarding radio frequency
interference, most modern evidential breath test devices feature shielding specifically
designed to prevent such interference (Gullberg, 2002a).
Non-invasiveness, advances in technology, immediate results and other factors have
led breath alcohol analysis to become more accepted worldwide. It has become the
standard measurement system used in the prosecution of impaired driving cases
(Gullberg, 2000; National College for DUI Defense, 2002).
19
2.2. Breath Alcohol Detection Technology
2.2.1. Infrared Spectroscopy
Analysis of breath alcohol through infrared (IR) spectroscopy has become
widespread; it is currently the most common method of measuring breath alcohol
(Intoximeters, 2002; Dubowski, 1992). IR analysis is based on measuring the amount of
IR that is absorbed by a substance (Drug Library, 2003). In fact, specific molecules can
actually be identified by the way they absorb light (Freudenrich, 2002), similar to the
way fingerprints can be used to identify specific humans (Fiandach, 2002).
Infrared devices have breath sample chambers, into which a subject provide a breath
sample. IR light of a specific frequency is then passed through the chamber. Ethanol
absorbs some of the IR light (Gullberg & Zettl, 2002). A photocell at the receiving end
of the chamber measures the residual amount of IR received and compares it to the
amount originally emitted (Freudenrich, 2002). The alcohol concentration of the sample
is proportional to the amount of infrared light that is absorbed (Harding, 1996).
The foundational principle of this type of analysis is the Beer-Lambert Law. In this
context, it posits that the concentration of any alcohol present is directly proportional to
the amount of IR absorbed by alcohol dissolved in alveolar air. Basically, if the amount
of IR that has been absorbed is known, the concentration of the alcohol can be computed
(Fiandach, 2002; Gullberg & Zettl, 2002).
Infrared measurement instruments also feature the ability to detect for the presence of
residual mouth alcohol. This is possible due to the instrument’s capacity to continuously
20
measure the alcohol in a subject’s breath as he or she exhales into the breath sample
chamber (Harding, 1996). When plotted with time (in seconds) on the x-axis and BrAC
on the y-axis, subject breath samples with and without residual mouth alcohol will yield
very different breath exhalation profiles. Samples with residual mouth alcohol will yield
a high alcohol level from the very beginning of a long breath, then taper off. Samples
without residual mouth alcohol will slowly build in alcohol amount, peaking at the end
portion of a long breath. The capability of infrared devices to detect these differences is
called “slope detection.” Devices so equipped can be programmed to not provide a
result under residual mouth alcohol conditions and to notify the operator that an error
has occurred (Harding, 1996). As previously mentioned, the employment of a 15 minute
waiting period prior to any breath test should serve to effectively eliminate concerns
about residual mouth alcohol. Current devices employing infrared spectroscopy include
the Intoxilyzer 5000 series (CMI, 2003a) and the BAC Datamaster series (National
Patent Analytical Systems, 2003).
2.2.2. Chemical Oxidation/Photometry
Breath alcohol measurement through chemical oxidation is the oldest testing
technique still in use (Dubowski, 1992). It is the system that moved breath alcohol
measurement into widespread use among law enforcement (Harding, 1996). Chemical
oxidation involves noting the change in color resulting from a chemical reaction between
alcohol in breath and normally inert detection chemicals (Freudenrich, 2002).
21
The technique involves directing a subject’s breath sample into a vial or ampule
containing oxidizing chemicals that react with ethanol. The most common chemicals
used in these ampules include sulfuric acid, silver nitrate, potassium dichromate and
water (Freudenrich, 2002). After the breath sample is introduced to the chemicals, any
alcohol is oxidized (burned) to acetic acid (National Highway Traffic Safety
Administration, 1982). This results in a proportional change in the color of the original
chemicals, generally from yellow to shades of green. This change in color occurs due to
a decrease in the amount of ultraviolet light absorbed by the chemicals (Harding, 1996).
The color change is then measured by a photometer, the result of which is revealed via
analog or digital display (Dubowski, 1992).
Chemical oxidation of breath alcohol is a very precise and accurate technique. In
addition, it is selective and sensitive for alcohol, and is effective in ignoring the presence
of other volatile substances (Dubowski, 1992).
2.2.3. Fuel Cell/Electrical Oxidation
Originally developed to provide power for the aerospace industry, fuel cell technology
was adapted to the measurement of breath alcohol. Discovered in the 1800s, fuel cell
technology was first shown to be capable of specifically identifying alcohol in the 1960s
by researchers at the University of Vienna (Intoximeters, 2002). A fuel cell is basically
an electromechanical device capable of converting an oxidant and a fuel into direct
current (Harding, 1996). For breath alcohol measurement, atmospheric air is the oxidant
and ethanol is the fuel.
22
A fuel cell generally consists of two platinum electrodes, between which is
sandwiched an electrolyte material capable of conducting ions (CMI, 2002b; Dubowski,
1992). As a subject’s expired air flows through the fuel cell, the alcohol is oxidized,
resulting in the creation of an electrical current (Harding, 1996). As more alcohol is
converted to water through oxidation, the current grows stronger. Thus the current
created is proportional to the amount of alcohol that is exposed to the fuel cell. This
current flows by wire from the electrode to a microprocessor that calculates and displays
the resulting concentration of alcohol in the breath sample (Freudenrich, 2002).
Fuel cells have shown to be highly resistant to interference from other chemicals
(Dubowski, 1992). However, the sensitivity of the devices changes over time,
necessitating more frequent calibration than some devices employing other technology
(Harding, 1996). Fuel cell technology continues to develop, as does the number of
applications in which it is used. For example, fuel cell technology has since expanded
into automobile applications (Autoweek Online, 2002). Current breath alcohol testing
devices employing fuel cell technology include the Alco-Sensor (Intoximeters, 2003)
and the Intoxilyzer 400PA (CMI, 2003b).
2.2.4. Solid State Semiconductor (Taguchi) Gas Sensor
Patented in the US in 1973, Taguchi semiconductor sensors are solid state devices
capable of measuring alcohol (Dubowski, 1992). These units generally require little
power to operate and are inexpensive and small. The sensor itself is an N-type
(negative) semiconductor, comprised of a stannic oxide bead placed in a ceramic
23
cylinder. The porous sensor requires being heated to operational temperature prior to
use (Harding, 1996).
As breath is passed through the cylinder, the bead absorbs alcohol, which causes an
increase in electrical conductivity. This rise in conductivity is in proportion to the
concentration of the alcohol in the breath, which is measured electronically and
converted to direct current voltage (Harding, 1996; Dubowski, 1992). Results are
displayed through a series of lights or a digital readout (Dubowski, 1992).
Taguchi sensor detectors are not specific for alcohol; such sensors will respond to
almost any combustible gas. While quite sensitive, Taguchi sensors lack inherent
stability, and thus require more frequent recalibration (Harding, 1996; Dubowski, 1992).
Current devices employing Taguchi sensor technology include the A.L.E.R.T. Model J4
(Columbia Laboratory Services, 2003) and ignition interlock devices (Harding, 1996).
2.3 Breath Alcohol Sample Simulators
In order to assure the proper functioning of BrAC measuring devices and to ensure
that their operators are properly trained as required, it is necessary to have access to a
method of introducing alcohol vapor into BrAC testing instruments. One option is to
have humans who have consumed alcohol provide breath samples. While such a
procedure might be optimal, especially in terms of testing under simulated field
conditions in which devices would be expected to be used, it is impractical to expect to
have volunteer drinkers on hand for every testing and training procedure.
24
For these reasons, the breath alcohol sample simulator was created. Generally known
as “simulators,” these devices deliver alcohol vapor specimens of known concentration
to BrAC measurement instruments (Dubowski, 1994). Originally designed to provide
simulated breath alcohol vapor specimens for use in operator training, simulators are
now also used to aid in calibrating and assessing the performance of BrAC analyzing
devices (Dubowski, 1992).
There are two types of simulators: dry gas and wet bath. Dry gas simulators use an
inert gas, such as nitrogen, to represent specific alcohol concentrations. The gas is
contained in a pressurized cylinder which, when depleted, must be refilled. Increasingly
used in the US, dry gas simulators have been found to demonstrate acceptable
performance to be used for forensic and other purposes (Dubowski & Essary, 1996).
About the size of a coffee can, closed system wet bath simulators contain an aqueous
solution, through which air can be directed (see Appendix B). This solution is mixed
from precise amounts of water and ethanol; thus, the exact concentration of the solution
is known (Harding, 1996). Generally, a simulator will contain 500 ml of the aqueous
solution and will be heated to a constant temperature of 34 degrees Celsius (Speck,
McElroy & Gullberg, 1991; Dubowski & Essary, 1991). This temperature is used
because it approximates the temperature of human breath (Gullberg & Zettl, 2002).
As air is passed through the simulator, it takes on the alcohol properties of the
mixture, then flows into a BrAC measurement device. This transference of alcohol
properties to the introduced air is based on Henry’s Law, which states that at a given
temperature in a closed system, the alcohol concentration of the air will be proportional
25
to the alcohol concentration of the solution (Gullberg & Zettl, 2002). This law allows
the alcohol concentration of the simulator headspace vapor to be of known quantity
(Gullberg, 2000).
Wet bath simulators provide breath test users with several advantages. First, solutions
can be prepared to virtually any alcohol concentration of interest (Harding, 1996). By
precisely varying the mix of water and ethanol, the user can create any alcohol level
desired. Second, simulators provide samples with properties similar to that of human
breath, in that they flow dynamically. This type of system is superior to fixed-volume
static samples (Dubowski, 1992)
However, wet bath breath alcohol simulators are limited in performance by the fact
that the water/ethanol solution will decrease in alcohol concentration as air is passed
through the mixture. The rate of depletion is relatively slow, with a 1% depletion
resulting after approximately 25 tests. A one percent depletion limit has been noted to
be acceptable for research and calibration purposes (Dubowski, 1979). Common
practice is to discard simulator solutions after a maximum of 25 tests.
Support for the use of simulators in breath alcohol testing is widespread among
researchers. In their research, Dubowski and Essary (1991, 1992) have concluded that
simulators are capable of providing satisfactory and appropriate samples for the testing
of BrAC measurement devices. Simulators have also been found to be very reliable for
this function as well, capable of providing, over multiple tests, consistent and uniform
breath alcohol samples (Gullberg 1989, 2000).
26
2.4 Breath Alcohol Measurement Devices
Numerous studies have assessed the performance of breath alcohol measurement
instruments. Generally, these studies have used for comparison either blood results or
results from an EBT or both. The literature dealing with the performance of each type of
device will be discussed.
2.4.1. Evidential Breath Testers
The largest BrAC testing devices, evidential breath testers (EBTs) have been found to
be sufficiently accurate and precise for their main purpose of providing evidence for use
in the adjudication of criminal proceedings. The potential for substantial impact of
breath alcohol testing on the outcome of impaired driving cases has prompted the
National Highway Traffic Safety Administration (NHTSA) to develop performance
standards for EBTs. Specific standards have been set for precision, accuracy, acetone
interference and blank readings.
For precision and accuracy, all potentially eligible EBTs are tested by NHTSA 10
times at four levels of alcohol concentration: .02, .04, .08 and .16. At each level, the
systematic error (the measure used to assess accuracy) must be ≤ ± .005 and the standard
deviation (the measure used to assess precision) must be ≤ .0042. The sole exception to
these standards occurs with the test at the .16 level; here, the systematic error must be ≤
± .008. The same ≤ ± .005 and ≤ .0042 standards apply when testing with acetone,
which is tested at .02. Blank readings, using alcohol-free human breath, must result in
systematic error ≤ ± .005, with no single result exceeding .005 (U.S. Department of
27
Transportation, 1993). The EBTs that do meet the specifications can be placed on
NHTSA’s Conforming Products List (U.S. Department of Transportation, 2002b).
In a study of retrospective data, Harding, Laessig & Field (1990) compared the
performance of EBTs against blood test results. The researchers examined 395 pairs of
blood and breath alcohol test results, each pair of which were obtained within 60 minutes
of each other. The EBT used was an Intoxilyzer 5000. The analysis revealed that the
breath and blood results were in close agreement, demonstrated by a correlation
coefficient (r value) of .94. Further, it was found that the EBT results systematically
underestimated blood tests results by a mean of 11.5%. BrAC results were lower than
BAC results for 67% of the cases, with BAC results exceeding BrAC results only 2% of
the time. Thus, the EBTs tended to underestimate participants’ actual blood alcohol
concentrations, with the bias falling in favor of the suspect.
Taylor and Hodgson (1995) compared three different EBT devices against blood
results from 18 male and female volunteers. The participants consumed alcohol, then
provided breath and blood samples. Strong relationships were found between the EBTs
and the blood results. Correlation coefficients ranged from .971 to .989, indicating
strong relationships between results obtained directly from blood and results obtained
through deep lung breath samples. In addition, the results showed that all three EBTs’
measurements fell below those of the blood samples, thus underestimating blood alcohol
concentration.
28
In their study of low BACs, Dubowski and Essary (1999) collected pairs of breath
samples from 62 law enforcement breath testing sites using Intoxilyzer 5000-D EBTs.
The data were from drivers who had been suspected of driving while impaired by
alcohol, with their breath alcohol concentrations ranging from 0.00 to .059. It was
concluded that EBTs can provide accurate measurements of low BrACs. The
researchers further noted that the EBTs’ performance, in terms of precision, accuracy
and sensitivity, was quite adequate for use in forensic, research and clinical applications.
In a comparison of blood alcohol concentrations and EBT BrAC results, Italian
researchers examined results from tests performed on weekend nights between 1997 and
1999 (Zancaner, Giorgetti, Cavazeran, Snenghi, Castagna & Ferrara, 2000). The study
involved 278 pairs of breath and blood tests, where each pair of tests was performed
within 10 minutes of each other. Breath test results were obtained at the roadside, using
an EBT powered by a vehicle engine. The results indicated good agreement between the
breath and blood results. The median difference between the two types of results was
5.2% and the results exhibited a strong correlation of .96.
The researchers also found that the relationship between breath and blood results was
related to the alcohol concentration. BrAC results were found to be higher than BAC
results at alcohol concentrations below .10. BAC results were higher than BrAC results
at alcohol concentrations at and above .10. The authors concluded by recommending
conducting breath tests in controlled conditions, such as at a law enforcement site, to
confirm breath test results performed at the roadside.
29
Method, Reed, Kamendulis and Klaunig (2002) performed a study of the stability of
Datamaster EBTs. Over a 3 year period, simulators were used to deliver surrogate
samples, resulting in the collection of 771 data points. Results indicated that over time,
there was a tendency for a slight decrease in breath test results. Further, all EBT test
results were equal to or below the alcohol concentration sample provided by the
simulators; no overestimation by the EBTs occurred. The authors conclude that EBT
results under these conditions would be biased in favor of the suspect.
In sum, EBTs’ performance has been repeatedly demonstrated to be of adequate
precision and accuracy for its purpose. In fact, it has been claimed that the performance
of EBTs for forensic purposes is better than it needs to be (Gullberg, 2002b).
Additionally, there have been calls for the development of roadside evidential breath test
devices (Reckers & Breen, 2002; Scott & Breen, 2000). EBTs continue to be the most
widely used breath alcohol testing instruments.
2.4.2. Preliminary Breath Testers
Preliminary breath testers (PBTs) are designed to serve as pre-arrest alcohol screening
devices to aid field officers in determining whether a suspected impaired driver should
be arrested. Compared to officers having to make such decisions without this
technology, PBTs can help identify intoxicated drivers who might be able to mask
traditional signs of impairment and can help identify drivers for whom a medical
condition, as opposed to an ingested drug, might be the cause of suspect behavior
(National Highway Traffic Safety Administration, 1982).
30
According to the National Highway Traffic Safety Administration (1982), the proper
use of PBTs can result in the following:
(a) An increase in the number of DWI arrests;
(b) A decrease in the mean BACs of those arrested;
(c) General acceptance of PBTs by law enforcement officers.
However, in his review of portable breath testing devices, Olson (1986) differs
slightly in his assessment of expected results of the use of PBTs. While he agrees that
the number of DWI arrests should rise, he argues that the mean BAC of arrestees will
not necessarily decrease. Olson cites the common problem of subjects not blowing long
or hard enough into PBTs as partially responsible for the uncertainly of effect upon
mean BAC of those arrested.
In its PBT instruction manual, NHTSA cites the experiences of five states using PBTs.
It noted that employment of PBTs could be attributed to a mean increase in arrests of
53% and a mean decrease in average BACs of those arrested of 17% (National Highway
Traffic Safety Administration, 1982). In another year-long study of six states’ results of
over 3,600 preliminary breath tests, the agency noted that the mean BAC of those
arrested decreased from .201 to .172, a 14% drop. In addition, feedback about PBTs was
obtained from law enforcement officers, with 75% stating that PBTs were a good idea.
In the late 1970s, Jones & Goldberg (1978) began a four-part study on an early PBT
model, the Alcolmeter Pocket Model. Their first research tested the device using
simulators, at five alcohol levels. Results were positive, as the researchers noted a mean
correlation between the simulated alcohol samples and the PBT of .967. The Alcolmeter
31
demonstrated good precision, yielding a standard deviation of .0175, or 1.91% of the
mean alcohol concentration. Their assessment of accuracy revealed that the PBT
systematically underestimated BAC by 3-12%. They also noted that the Alcometer was
very stable, with only a slight downward trend in results over repeated use.
Jones’ (1978) second study involved human drinkers. Thirty-nine male participants
first provided a total of 120 alcohol-free breath samples. All PBT results were negative
for alcohol. Participants then consumed alcohol and provided breath and blood samples.
Jones found that the relationship between the PBT BrAC results and the blood test
results was dependent on whether the participant was in the absorptive or elimination
phase. During the absorptive phase, the PBT results were higher than BAC results;
during the elimination phase, the PBT results were lower than BAC results. Precision
was found to be highest during the elimination phase, however the instrument became
less precise as the alcohol concentration increased. Jones also emphasized that the
standard error estimate (Syx) is a good estimate of overall error associated with breath
test results.
In the third study, Jones performed a controlled field trial of the Alcolmeter (Jones,
1985a). This involved 10 police officers consuming one of two doses of alcohol and
providing blood samples and breath samples with a PBT. The overall relationship
between the two types of results was strong, with a correlation coefficient of .95. As in
the second study, he noted that the standard deviation of the PBT measurements
increased with increasing alcohol concentration, indicating that precision was a function
of alcohol level. BrAC measurements were found to underestimate actual BAC by
32
5.1%. Jones concluded that the Alcolmeter’s precision and accuracy was satisfactory
and thus it would be practical and useful for use as an alcohol screening device.
The final study involved using 84 Alcolmeters at roadblock checkpoint, traffic crash
and traffic offense events throughout Sweden (Jones 1985b). Breath and blood samples
obtained from 333 drivers were pooled. Since the blood tests were performed up to 220
minutes after the PBT tests, the blood test results were adjusted to reflect for alcohol
eliminated during the delay between tests. The rate of .015 g/ml per hour was used for
this adjustment. The relationship between the blood and breath tests was statistically
significant, yielding a correlation coefficient of .84. At alcohol levels below
approximately .08, BrAC results were found to exceed BAC results. At alcohol levels at
or above .08, BAC results exceeded BrAC results. Further, the Alcolmeter’s false
positive rate was relatively low, at 5% of all tests. Overall, each of the four tests of the
Alcolmeter supported its continued use.
In his discussion about the advantages of PBTs capable of collecting evidential data,
Forrester (1997) described two studies examining PBT performance. The first study
used three participant drinkers who provided breath samples through an Intoxilyzer EBT
and an Alco-Sensor IV PBT, along with blood samples. Results indicated that the
devices’ results agreed with each other to within a mean of .004. The PBT results were
found to be approximately 9% below the blood results. The second study involved 412
participants under field conditions, who provided breath results with a PBT and blood
results. Breath results were again found to be slightly below blood test results. The
study concluded that PBTs demonstrate acceptable consistency for use in the field.
33
Reckers and Breen (2002) also examined the performance of PBTs for evidential
applications. Two Alco-Sensor IV-XL PBTs were used, as was a Datamaster EBT.
After consuming known amounts of alcohol, three volunteer participants provided breath
samples with both types of devices. The differences in performance between the two
types of instruments were quite small. Mean PBT breath results were found to be within
.005% of mean EBT breath results. There was a small overall mean difference between
the device types of .0018%. The authors concluded that the Alco-Sensor IV-XL PBT
shows promise as an evidential breath testing instrument.
PBT instruments continue to be employed a variety of testing applications, including
law enforcement, drug abuse treatment centers and operators of large motor vehicles and
aircraft (National Commission Against Drunk Driving, 2002a). Their performance and
ability to evaluate drivers’ alcohol levels close to the time of driving make them a useful
tool in the fight against impaired driving (Gullberg, 1991).
2.4.3. Passive Alcohol Sensors
The least intrusive of breath alcohol test devices, passive alcohol sensors (PASs)
capture drivers’ breath without their knowledge or active participation; hence the
“passive” descriptor (National Commission Against Drunk Driving, 2002b). As with
PBTs, PAS instruments do not capture evidential test results. Rather, they are intended
to aid law enforcement officers in their initial screening of suspected impaired drivers
(Wells, Preusser & Williams, 1992).
34
In an early study of PAS devices, Jones and Lund (1986) examined the performance
of PASs used in sobriety checkpoints. The data were collected from checkpoints
performed in Charlottesville, Virginia on weekend nights over a period of two months.
Officers stopped all motorists arriving at the checkpoint and examine their drivers’
licenses. It was at this point that officers used PASs to check each driver’s breath for the
presence of alcohol. If as a result of that interaction an officer suspected that a driver
was impaired, he or she would ask the driver to provide a breath sample through an
Alco-sensor II PBT. The officer would then take the appropriate action, based on the
result of the PBT test.
Data from 1644 drivers were used in this study. At checkpoints where PAS devices
were used, the detection rate of impaired and intoxicated drivers improved significantly,
compared to checkpoints at which PAS devices were not used. For drivers with BrACs
between .050 and .099, detection rates increased from 24% to 45%. For drivers with
BrACs at or above .10, detection rates increased from 45% to 68%. These correspond to
percentage increases of 88% and 51%, respectively. In addition, the number of drivers
with BrACs between .020 and .049 who were unnecessarily detained decreased by 56%.
Compared to the Alco-sensor II results, the PAS units were found to underestimate
BrAC at levels .02 and higher by a factor of two. Overall, detection rates at sobriety
checkpoints increased and the detention of drivers with low BrACs decreased, indicating
support for PAS devices in these enforcement circumstances.
Lestina and Lund (1992) tested two different brands of PASs under laboratory
conditions. Twelve volunteer drinkers provided breath samples for 12 models of each
35
brand of PAS device. These results were compared to breath samples collected with an
Alcolmeter PBT. A major variable under examination was the distance between
participants’ mouths and the PAS devices; each drinker provided breath samples with
their mouths at 12.7, 19.1 and 25.4 cm from the PAS units. Results showed that at
distances of 19.1 and 25.4 cm, neither device performed well; both models performed
best at 12.7 cm from drinkers’ mouths. False positive results at the .02 BrAC level
ranged from 13% to 20%. The authors concluded that the PAS units tested performed
with sufficient reliability to be used as roadside alcohol screening devices and that their
performance would be most optimal in the detection of drivers with high BACs.
In their study of PAS devices, Foss, Voas and Beirness (1993) conducted interviews
with 1,145 drivers in Minnesota parking lots between 10:00 p.m. and 3:00 a.m. Drivers
voluntarily submitted to the interviews and provided breath samples through an Alcosensor
III and a PAS instrument. Results indicated that the two devices’ performance
was similar, as evinced by a correlation coefficient of .87, with the PAS results falling
consistently below the PBT BrAC measurements. It was found that PASs resulted in
decision accuracy levels of at least 95% when analyzed at discrete alcohol
concentrations of .02, .05, .08 and .10. Further, the PAS units demonstrated low rates of
false positive outcomes, with less than 4% of drivers being erroneously judged to exceed
.10 BrAC. Even at low BrACs, PAS performance was good, with 93% of drivers at .02
BrAC being detected. The study supported the use of PAS devices and predicted that
widespread application of these instruments would improve the ability of law
enforcement to reduce the incidence of impaired driving.
36
Ferguson, Wells and Lund (1995) also examined the performance of PAS instruments
employed at sobriety checkpoints. As standardized field sobriety tests (SFSTs) are
commonly employed by law enforcement at checkpoints to help identify suspected
drinking drivers, this study sought to determine the effects of adding PAS analysis to the
performance of SFSTs. At six sobriety checkpoints performed in Fairfax County,
Virginia in 1993, 5,192 drivers were interviewed. Approximately half of the drivers
were evaluated with both SFSTs and PAS devices; the other half were evaluated using
only SFSTs.
The study found that the combination of SFSTs and PAS units resulted in improved
identification of impaired drivers than the use of SFSTs alone. At BrACs between .05
and .10, the combination resulted in a 77% improvement in identification of impaired
drivers. At BrACs above .10, the combination resulted in a 29% improvement. The
authors thus noted that the use of PAS instruments might be most effective in the
identification of drivers around the moderate BrAC level of .05 to .08. In addition, the
authors did caution that because PAS instruments draw in ambient air in addition to
drivers’ breath, they are incapable of providing accurate numerical estimates of BAC,
thus reemphasizing these devices’ use as qualitative screening tools, rather than
quantitative measurement instruments.
Research indicates that PASs can be effective in improving the identification of
impaired drivers by law enforcement officers. Originally contained within innocuously
appearing flashlights, other versions are now available, including a model built into a
clipboard (PAS Systems International, 1999).
37
2.4.4. Ignition Interlock Devices
Ignition interlock devices (IIDs) are designed to prevent drivers convicted of DWI
from starting their motor vehicles if they have alcohol in their bloodstream (Coben &
Larkin, 1999). In the US, this threshold level is generally set at .025 (Voas, Blackman,
Tippetts and Marques, 2002). The idea of preventing drivers from driving after
consuming alcohol first surfaced in the late 1960s. Introduced to the US in the mid
1980s, IIDs have spread in application, with over 43 states having adopted some form of
legislation addressing the use of these devices (Frank, 1997; Governors Highway Safety
Association, 2003).
IIDs consist of two components. First is the head unit, which serves to collect a
sample of the driver’s breath. Second is the control module, which is securely connected
to the vehicle. It performs the analysis of the breath sample and, if warranted, prevents
ignition of the vehicle’s engine (Comeau, 2000). Modern IIDs are capable of recording
all attempts to start a vehicle and can require vehicle operators provide rolling retests.
These latter tests involve drivers having to perform additional alcohol-free breath tests
while driving in order to keep the vehicle’s engine running (Marques, Voas, Tippetts &
Beirness, 1999).
Most research into IIDs has focused on the devices’ impact on DWI recidivism, rather
than their accuracy and precision. Every identified study of IIDs’ impacts on recidivism
found that the devices are, when installed on a vehicle, effective in reducing recidivism
(Tippetts & Voas, 1997; Beck, Rauch & Baker, 1997; Voas, Marques, Tippetts &
38
Beirness, 1999; Weinrath, 1997). Other research involving IIDs has examined less
impact-oriented issues, including:
(a) The use of IID-recorded start attempts to identify drivers at highest risk for DWI
recidivism (Marques, Tippetts, Voas & Beirness, 2001);
(b) The ability to and result of efforts to motivate DWI offenders to enter an IID
program (Voas, Blackman, Tippetts & Marques, 2002);
(c) The use of global positional satellite (GPS) technology to precisely monitor the
location of a violator’s vehicle (Comeau, 2000);
(d) The impact of combining adding human services intervention efforts to IID
programs (Marques, Voas, Tippetts & Beirness, 1999);
Ignition interlock devices continue to be the subject of administrative and impact
evaluation. While past research indicates IIDs can have positive impact while installed,
their long term behavioral effects remain undetermined.
2.4.5. Coin Operated Breath Testers
Designed for point-of-purchase breath alcohol testing, coin operated breath testers
(COBTs) permit drinkers to self-test their BrAC. These counter- or wall-mounted
devices hold the potential to earn profits for establishments that offer them for their
patrons’ use. Several other potential benefits have been noted by COBT distributors.
First, it is claimed that COBTs serve to reduce impaired driving. Second, they can serve
to educate consumers, thereby encouraging them to drink moderately and at an
appropriate pace. Third, COBTs offer establishments an objective and tactful way to
39
cease service to specific individuals and thus prevent service of additional alcohol to an
intoxicated patron (The Alcohol Alert System, 2002). While some of the distributors’
claims may come across as too-good-to-be-true, what little research exists on COBTs
tends to support the performance of the devices.
As with IIDs, most research into COBTs has focused on social impacts. Identified
studies have supported the use of COBTs as part of an overall impaired driving
prevention strategy (Haworth and Bowland, 1995; Wundersitz, 2002). However, the
availability of these devices in the US remains limited.
2.4.6. Pocket-Model Breath Testers
There is a lack of evaluation of the performance of pocket-model breath testing
devices. Published works concerning PMBTs generally consist of newspaper or Internet
articles introducing and describing the devices. Only one scientific study examining a
PMBT could be identified. NHTSA tested one such model, sold as the ABI, and found
that it met the Federal performance standards for alcohol screening devices (U.S.
Department of Transportation, 2002b).
In a New York Times article, Stellin (2001) described the units’ technologies,
accuracy, costs and sizes. A member of the New York Highway Patrol who was
interviewed for the article stated that using PMBTs was better than guessing about how
much alcohol is too much.
In his article in The Courier-Journal, Muhammad (2000) wrote that such devices have
sales appeal that is politically correct, but raised questions about the devices’ initial
40
calibration. He also posited that PMBTs could be used to persuade party guests to
engage a taxi rather than drive home themselves and/or to allow the body to learn
individual cues associated with intoxication. Muhammad further suggested that, if
wisely used, the devices could help prevent a DWI arrest or an alcohol-related motor
vehicle crash. However, he cautioned that if used unwisely, PMBTs could cause a
drinker to think he or she is more sober than he or she really is.
CNN interviewed a law enforcement officer involved in DUI training about PMBTs
(CNN, 2002). The officer agreed that such devices were needed because of alcohol’s
effects on the brain and the resulting inability to think clearly. He also emphasized that
the decision about whether or not to drive should be made prior to the consumption of
any alcohol.
WCCO-TV (2003) tested two PMBTs using human drinkers, comparing the results to
Minnesota’s state-approved EBT. They found that one device did not work at all and the
second device, the ABI Personal Breath Alcohol Screener, read higher than the state’s
EBT, at multiple BrACs.
The relative lack of experimental research on PMBTs has been repeatedly confirmed,
indicating a need to examine these devices’ performance (A.W. Jones, personal
communication, December 6, 2002; M. Cowan, personal communication, April 1, 2002;
M. Parker, personal communication, April 1, 2002; R. G. Gullberg, April 1, 2002; J. F.
Frank, personal communication, April 2, 2002). Parties who might be interested in an
analysis of the performance of these devices could include drinkers, alcohol-serving
41
establishments, emergency room personnel, probation officers, workplace testing
personnel, law enforcement and the devices’ manufacturers.
42
CHAPTER III
METHODOLOGY
3.1. Test Devices
Eight small scale, reusable breath alcohol testing devices were procured for this study.
All were readily available; one device was obtained through a local retailer and all others
were obtained via Internet-based retailers. Per device costs ranged from $40 to $104,
excluding shipping charges. All the devices provided numerical readouts of estimated
BrAC, to the hundredth of one percent (two digits to the right of the decimal). One of
the devices tested, the ABI Professional Breath Alcohol Screener, is on NHTSA’s
Conforming Products List.
The option of requesting the manufacturers and/or merchants of the devices to provide
the instruments free of charge for testing was considered and discarded. This decision
helped keep the study as pure from potential contamination as possible, serving to
maximize study integrity. Had the manufacturers been aware of the study and provided
the devices, the chance of obtaining a device whose performance would be substantially
different from the population of all devices of that model would have increased,
weakening the study.
The devices were of two types: those that employed a mouthpiece to facilitate direct
insertion of breath to the unit’s sensor (4), and those that did not feature a mouthpiece
(4). The latter devices featured a breath port into which the user expires his or her
43
breath. Descriptions and specifications of each device tested can be found in Appendix
A. Upon acquisition, each device was randomly assigned a letter, ranging from A to H.
Only one of each model of device was tested. The author recognizes the possibility
that any device’s performance could have been affected by handling prior to arrival for
testing. Each device did arrive apparently undamaged, with all packing materials intact
and unblemished. Great care was exercised in the storage and handling of each device
upon arrival, so as to minimize the effects upon performance. All devices were stored
and transported together in the same container, such that all devices would be subject to
identical conditions (temperature, movement, etc.). In addition, no test device was used
other than during the testing procedures, reducing the possibility of performance
differences resulting from differential use.
3.2. Pilot Testing
All pilot and additional laboratory testing was conducted at the Texas Department of
Public Safety (DPS) building at 1540 East Highway 6 Bypass, Bryan, Texas. This
location’s breath testing technical supervisor, Margaret Parker, oversees such testing
operations throughout nine counties and facilitated the testing of the instruments. All
data were collected in the facility’s conference room, under fluorescent lighting
conditions. There were no nearby sources of radio frequency interference.
3.2.1. Phase One Pilot Testing
This testing had several goals:
44
(a) Determine the appropriate testing time interval, partially dependent upon the
devices’ recovery times;
(b) Create a method of consistently delivering alcohol samples to the devices;
(c) Develop the apparatus to deliver an alcohol sample to the non-mouthpiece
devices.
Based on this testing, it was determined that the time interval between successive tests
could not be less than two minutes. Thus the minimum testing interval was set at two
minutes.
Further, an alcohol sample delivery system was developed for both types of devices.
The system for mouthpiece devices used surgical tubing to direct the alcohol sample. At
the input (human) end of the tube, a standard DPS mouthpiece was attached; the other
end of the tube was connected directly to the solution simulator. The simulator’s exit
tube was connected directly to the device’s mouthpiece. Plastic, funnel-shaped reducers
were used as needed to ensure proper mating between the connections. This provided
direct input with a flexible tube through which the alcohol sample could flow without
contamination or dilution.
The delivery system for non-mouthpiece devices also used surgical tubing with a DPS
mouthpiece at the input end. Most of the non-mouthpiece devices’ instructions stated a
recommended distance between the user’s mouth and the device’s input port. This
distance ranged from 1.3 cm to 3.8 cm. If a non-mouthpiece device’s instructions did
not include such a recommended distance, the distance was set at a default of 1.3 cm.
45
As several of the non-mouthpiece devices were quite small, a small vice was used to
hold the smaller devices steady for all tests. Cardboard or thin rigid plastic was used to
form a stable mounting point for the tubing; the mounting extended perpendicularly
from each non-mouthpiece device, allowing the tube to be pointed directly at the
devices’ input ports.
It was also determined that wet bath breath alcohol solution simulators could be
effectively used to deliver alcohol samples to both mouthpiece and non-mouthpiece
devices. National Draeger, Inc. Mark IIA simulators (see Appendix B) were used for
this purpose (Draeger, 2003). The alcohol mixtures consisted of a combination of
distilled water and a predetermined amount of 200 proof alcohol designed to produce
certain equivalent measures of BrAC. The alcohol came from the DPS stock. The DPS
technical supervisor prepared all the solution sample mixtures according to DPS
standards.
For each group of tests at each alcohol level, 500ml of mixture was inserted into the
simulator. The solution was then warmed by the simulator’s integral heating element to
the proper temperature, 34ºC, ±.5ºC. Temperature was verified at the start of each test
run with an NIST-certified thermometer. The simulator’s integral agitating propeller
served to maintain a properly blended solution.
Because a given simulator sample’s alcohol strength will diminish as breath is blown
through the mixture, only 20 tests were conducted with each sample. After 20 tests, the
solutions were discarded. Further, solutions were changed to different alcohol strengths
only on an increasing basis. That is, only the next stronger solution was permitted to be
46
inserted into a given simulator. This was done to avoid a situation where any residual
alcohol in a simulator would be at a strength higher than the subsequent mixture,
possibly contaminating the next mixture.
3.2.2. Phase Two Pilot Testing
This testing had two goals:
(a) Test the devices’ abilities to resist yielding false positive readings;
(b) Test the devices’ abilities to resist yielding false negative readings.
Eight devices were tested in this experiment. To assess their ability to minimize false
positive readings, each device was tested 20 times at an alcohol level of .00. This is the
same concentration at which NHTSA tests PBT devices. To create this non-alcohol
sample, only pure distilled water was inserted into the simulator, so as to employ the
same procedure of blowing through a simulator at all alcohol levels. To assess the
device’ ability to minimize false negative readings, each device was tested 20 times at an
alcohol level of .032. NHTSA also specifies this test level in its testing protocol.
After installing fresh batteries in all devices and prior to collecting data at either level,
two “warm-up” tests were performed, but data were not recorded. This was done in
order to:
(a) Ensure that each device’s sensor had reached operating temperature;
(b) Determine that each device was functioning properly.
Tests were conducted at two minute intervals. At the halfway mark (after 10 tests), the
tubing was temporarily disconnected between the simulator and device and shaken to
47
remove any condensation that might have accumulated in the system. To enhance
consistency among samples, a single human provided all breath samples for all alcohol
positive tests.
To advance to Experiment One, each device was required to meet the following
performance criteria:
(a) Yield no more than one positive result in 20 trials at an alcohol level of .00
(positive equaling .02 or higher);
(b) Yield no more than one non-positive (below .02) result in 20 trials at an alcohol
level of .032.
Data were recorded on pre-prepared data forms, along with the temperature of the
solution.
For all pilot testing and Experiment One, human breath was expired through the
simulator mixture into each device. Prior to providing breath samples through the
simulators, the human breath provider’s BrAC was measured using a calibrated
Intoxilyzer 5000 (Intoxilyzer) to ensure that the provider’s breath was free of alcohol.
This model Intoxilyzer is the latest version used by DPS. The Intoxilyzer instruments
are the only evidential breath testing equipment used in Texas; it is the standard used
throughout the state.
In addition, each solution mixture was tested 20 times, in recirculation mode, by a
calibrated Intoxilyzer in order to confirm the targeted strength of the mixture. No data
were recorded from the pilot tests, all of which were performed at .08 alcohol
48
concentration. All breath samples were provided by one individual, who was trained a
priori by Department of Public Safety personnel to provide adequate breath samples.
3.3. Experiment One
The goal of this experiment was to assess the accuracy and precision of the devices at
multiple alcohol levels, under laboratory conditions. In its assessment of these
measures, NHTSA tests each device at the following alcohol levels: .02, .04, .08 and
.16. In order to maintain a full range of .02 increments between .02 and .10, the devices
were also tested at the .06 and .10 alcohol levels. Each device was tested 20 times at
each alcohol level, resulting in a total of 120 tests per device.
Alcoholic simulator solutions were created by the method given in the Pilot Testing.
As in the Pilot Testing, two warm-up tests were performed with each device at each
alcohol level, prior to collecting data. Also, the same procedure for removing any
condensation was employed. Tests were conducted at two minute intervals. Data were
recorded on pre-prepared data forms, along with the temperature of the solution.
3.4. Experiment Two
The goal of this experiment was to assess the devices’ performance under actual
drinking (in vivo) conditions. As such, volunteer participants agreed to consume alcohol
and provide numerous breath samples. For such tests to be performed, it was necessary
to obtain approval from the Texas A&M University Institutional Review Board (IRB),
49
which oversees all research involving humans. All aspects of this study were approved
by the IRB.
3.4.1. Participants
3.4.1.1. Participant Eligibility
Participants were limited to those between the ages of 21 and 34. This limitation was
imposed for several reasons. First, this age range represents the group of drivers that are
most involved in fatal DWI behavior in Texas (Texas Department of Public Safety,
1999). Second, the age range was limited to restrict the effect of large age variation on
the breath test results. A total of eleven (11) participants from the local community were
included in the study.
3.4.1.2. Participant Screening
Potential participants were pre-screened to exclude:
(a) Pregnant females
(b) Non drinkers
(c) Heavy/problem drinkers
(d) Alcoholics
(e) Diabetics
(f) Those allergic to alcohol.
50
The Alcohol Use Disorder Inventory Test (AUDIT) questionnaire developed by the
World Health Organization in 1987 and the Numerical Drinking Profile were used as the
first-line screening instruments (see Appendixes C and D). Potential participants scoring
a 6 or below on the AUDIT were eligible for participation. Alternatively, potential
participants with an NDP score of 3 or less were eligible to participate.
Several other questions were also presented along with the alcohol abuse screening
instruments, to ascertain whether participants were diabetic, allergic to alcohol, and/or in
poor health. Potential participants who answered affirmative to any of these questions
were excluded from participation. Prospective female participants were required to
administer a portable pregnancy self-test on the day of the study to exclude all who
tested positive.
All potential participants were informed that they would consume alcoholic
beverages, provide multiple breath tests, could withdraw at any time, but must remain at
the testing site until their BrAC returned to 00. They were further informed that the
target peak BrAC would be .09.
3.4.2. Testing Location
Experiment Three was conducted at the Texas Transportation Institute Gibb Gilchrist
Building on the West Campus of Texas A&M University. A large, first-floor classroom
was used, access to which was facilitated by Dave Willis, Director of the Center for
Transportation Safety (personal communication, October 20, 2002). To reduce possible
complications associated with too many participants present at one time, two drinking
51
sessions were performed, with five and six participants in the first and second sessions,
respectively.
3.4.3. Materials
3.4.3.1. Measurement
3.4.3.1.1. BrAC Instrumentation
One Intoxilyzer unit was used at the testing location with a second unit immediately
available for backup. All eight PMBT devices were present as well, along with all
necessary tubing and connecting apparatus used for the collection of breath samples.
Drinking straws cut in half in length were used as mouthpieces. This mouthpiece was
economical, and easily replaceable should participants chew on or lose them.
3.4.3.1.2. Participant Data Collection Instrumentation
Several measurement instruments were used to collect data from the participants
throughout each session, including:
(a) Body weight scale
(b) Pregnancy tests (Equate brand, procured from WalMart)
(c) Ruler (for measuring height)
(d) Oral thermometer
52
(e) Body water content device (model BIA 3000)
(f) Stopwatches.
3.4.3.2. Dosing
To ensure the precise administration of alcohol to each participant, specialized dosing
equipment was present, including a graduated cylinder scaled in milligrams, a calculator
to compute doses to be administered, and an alcoholic beverage. The beverage served to
each participant was a mixture of vodka and orange juice, served over ice. The vodka
was 100 proof Smirnoff Number 57, procured from a local liquor merchant.
3.4.3.3. Administrative
Several administrative materials were used to facilitate the sessions. A large-readout
digital clock was positioned in the testing room to record the time of each sample
collected. To ensure seamless operation of the instruments to be tested, extra batteries
were present for each device. In addition, a first-aid kit was present during all testing.
3.4.3.4. Participant Accommodation
In order to assure a minimum level of comfort for participants during the sessions, a
controlled amount of food and beverages were on hand, including:
(a) Breakfast foods- bagels, bananas, raisin bread
(b) Miscellaneous snacks, including pretzels and party mix
(c) Bottled water.
53
The point during the sessions that each participant was given access to these food was
strictly controlled. Various forms of entertainment were provides as well, including:
(a) Games- cards, board games
(b) Music- radio/CD player
(c) TV/VCR with assorted videos.
3.4.3.5. Additional Materials
To further support the smooth conduct of each session, additional materials were on
hand. A box was provided for car keys, as any participant who drove to the testing
location had to relinquish his or her car keys. A portable folding cot, borrowed from the
Texas Transportation Institute, was present to support the acquisition of data regarding
each participant’s current body water content. Both cotton and paper towels were on
hand for any use required.
3.4.4. Personnel
In addition to the researcher, the DPS technical supervisor was present for the
duration of each session. This person assumed total responsibility for the operation of
the Intoxilyzer. Further, two sober volunteers were present for each session. These
personnel assisted with the acquisition of data and monitoring of participants.
54
3.4.5. Procedure
3.4.5.1. Initial Setup
The afternoon prior to each session, the classroom tables and chairs were arranged to
facilitate testing and participant comfort. The evening prior to each session, final contact
was made with each participant to review procedures and to maximize the chances that
each participant would be present on time and at the proper location. Participants were
asked to refrain from consuming alcohol that evening, to get a full night’s rest and to
avoid eating any breakfast foods prior to arrival at the testing site.
Upon arrival of volunteer personnel the mornings of the sessions, the Intoxilyzer and
PMBT devices were placed into their respective testing locations. At that time, fresh
batteries were installed in all PMBT devices. Breakfast foods were available for
consumption.
3.4.5.2. Upon Arrival of Participants
As participants arrived, introductions were made and each was thanked for their
participation and informed that he or she could partake of the breakfast foods. A light
breakfast was provided to control, to the degree possible, how much food was in each
participant’s stomach prior to consumption of alcohol. This helped ensure that all
participants had consumed at least some food that morning, thus minimizing any
differences in absorption time among participants. In addition, this step was taken to
55
prevent participant discomfort that could result from consuming alcohol on an empty
stomach.
After all participants had arrived, the researcher thanked them as a group and
informed them about how the day was to proceed. They were reminded that any
participant could withdraw from the experiment at any time, but once alcohol was
consumed, participants would have to remain at the testing site until their BrACs
returned to 0.00, as measured on the Intoxilyzer.
At this point, participants completed the Informed Consent forms (see Appendix E).
Each participant also agreed not to drive for 12 hours following the conclusion of the
experiment. All were assured that they would receive transportation home, should they
be unable to secure rides themselves.
During this meeting, participants were informed that the sober volunteers would guide
them through all the testing and would watch for any signs of discomfort or any other
problems on the part of the participants. The locations of the bathrooms were identified
and participants were informed that whenever they needed to use the bathroom, a sober
volunteer would accompany them to the bathroom door. This was done to ensure that all
participants were supervised at all times, and that no participants with a positive BrAC
left the testing facility.
The possibility of becoming ill due to the consumption of alcohol was discussed. A
trashcan was present in the event of regurgitation and towels were available for any
necessary cleanup. They were also informed that any participant who became sick
would no longer be able to consume alcohol nor would they be allowed to provide breath
56
samples. Any ill participant would be removed from the study, but would be required to
remain at the test site, unless medical attention became necessary.
Following this meeting, female participants completed the portable pregnancy tests.
The results of the tests were visually confirmed by a female sober volunteer. Each
participant’s body weight was then measured. No participant’s weight was made
available to the other participants.
To help maximize the consistency among the breath samples to be obtained,
participants received training in providing breath samples. Each participant provided
several breath samples into the Intoxilyzer, monitored by the DPS breath test technician.
Once the technician was satisfied that each participant had reached the required level of
competency, participants were permitted to provide samples into the test devices. Table
1 contains information on participant variables.
Table 1
Participant Variables
Participant Sex Age Weight (kg) Height (cm)
1 M 23 100.9 173
2 M 22 105.9 190
3 M 23 75.9 177
4 F 22 69.1 166
5 M 23 125.9 196
6 F 22 69.1 168
7 M 23 97.7 182
8 F 32 77.3 172
9 F 22 56.4 169
10 F 22 50.0 164
57
3.4.5.3. Dosing
The amount of alcohol that each participant would consume to reach a target peak
BrAC of .09 was computed as a function of body weight. This target peak BrAC was
chosen to maximize the chances of being able to capture data from each participant
during the post-absorptive phase as he or she “passed” through the .08 BrAC level, the
first level at which it was intended participants be tested. The DPS formula for dosing
participants was used: .9 ml per pound of body weight. Each participant was to
consume three alcoholic beverages, each of equal strength. Thus each participant’s total
amount of alcohol to be consumed was divided by three.
The breath test technician then mixed the first round of drinks, mixing the ice, vodka
and orange juice. In accordance with DPS research procedures, participants were given
15 minutes to consume each drink, for a total consumption period of 45 minutes. As
consumption began, a stopwatch was started to monitor the timing of consumption.
Participants were given a timed countdown during each drinking segment.
At the end of the first and second 15 minute consumption periods, fresh drinks were
prepared using the aforementioned procedure. Participants were closely monitored for
signs of discomfort or other problems.
3.4.5.4. Waiting Period
Following the consumption period, a 15 minute waiting period was induced. The
purpose of this period was to allow time for any residual mouth alcohol to dissipate and
to ensure that nothing else was ingested during this time. This waiting period is
58
recommended as part of any breath testing program (Dubowski, 1994; Gullberg, 2000)
and is standard procedure for Texas DPS. Again participants were closely monitored for
discomfort or other problems.
3.4.5.5. Confirmation of Post-Absorptive Phase
At the conclusion of the waiting period, each participant was tested using the
Intoxilyzer. The goal was to identify the point at which participants’ absorption had
ended. Participants provided BrAC samples approximately every five to ten minutes
during this monitoring phase. Having two successive downward BrAC readings was
used as the criteria for a participant to be considered in the post-absorptive phase.
3.4.5.6. BrAC Data Collection
Once a participant had been identified as being in the post-absorptive phase, he or she
was monitored so that a reading could be obtained at the .08 level. The goal was to test
each participant at four distinct declining levels: .08, .06, .04 and .02. As it would be
very difficult, due to random fluctuations and error, to capture a participant at exactly the
.08/.06/.04/.02 level, a range of acceptable BrACs was used. Participants’ measurements
could fall between ±.005 of the target level. Table 2 shows the target levels and the
associated acceptable range parameters.
When a participant was confirmed to be in the post-absorptive phase and within the
first range of testing (.075-.085), he or she would then provide a second sample with the
Intoxilyzer and then would provide duplicate samples with all test devices. Duplicate
59
Table 2
Acceptable Range for
Individual Testing
Target Level Acceptable Range
.08 .075-.085
.06 .055-.065
.04 .035-.045
.02 .015-.025
samples, which have been rated quite adequate for forensic uses, were collected at every
test, on every device (Gullberg, 1989). All data, including the time of each test, were
recorded on pre-prepared data forms.
In order to reduce any effects of the order of treatment, the order in which participants
provided samples with the devices was counterbalanced. A Latin Square system was
used to create different specific orders so that every participant used a different order,
with all eight devices (Bordens & Abbott, 1996). Latin Square treatment ordering
systems are appropriate when the researcher is willing to set the number of treatment
orders equal to the number of treatments, in this case eight. This involved creating eight
distinct, randomly generated treatment orders. These orders were then randomly
assigned to participants at the beginning of each data collection session. Appendix F
provides the specific counterbalanced orders of treatment.
Because the number of participants exceeded the number of treatment orders, eleven
and eight, respectively, once all eight treatment orders had been assigned once, the
remaining three participants were assigned the first three sequential testing orders.
Participants one through eight were assigned orders one through eight, respectively, and
participants nine through eleven were assigned treatment orders one through three.
60
Thus, treatment orders one through three were used twice and treatment orders four
through eight were assigned only once.
After each participant had provided duplicate samples using each test device, they
returned to the Intoxilyzer to provide a final pair of samples. Thus, a full “round” of
testing involved:
1. Providing initial (Pre) duplicate samples on the Intoxilyzer
2. Providing duplicate samples on each test device, according to the individually
assigned treatment orders
3. Providing final (Post) duplicate samples on the Intoxilyzer.
Once a participant had undergone the first round of testing at .08, he or she was
allowed access to the light snacks and bottled water. These materials were withheld
until that point to ensure that each participant was in the post-absorptive phase so as to
eliminate the possibility of food delaying any further absorption of alcohol. Participants
were periodically tested with the Intoxilyzer to identify the point at which they passed
into the next lower testing range. In between providing breath samples, participants had
access to games, Fatal Vision® impairment-simulating goggles, music and assorted
videos.
As each participant’s BrAC was found to be within the next lower test range, he or she
again provided duplicate breath samples for each device, according to his or her assigned
treatment order. After completing the round of tests, participants again provided two
final samples for the Intoxilyzer. This sequence continued through all four test ranges.
Between rounds of testing, participant variables were measured, including:
61
(a) Body water content
(b) Oral temperature
(c) Height
(d) Resting heart rate.
The results of these measurements were recorded on Participant Data Sheets (see
Appendix G).
After each participant completed the round of tests at .02, his or her BrAC was
periodically monitored using the Intoxilyzer. Once a participant’s BrAC reached 0.00,
he or she was permitted to leave the test site, via either by being picked up or a ride
home by the research personnel. No participants were allowed to leave until a 0.00 level
had been reached. Each data collection session took approximately eight hours.
Participants were thanked as they left the test site, and thanked again via email the
following day.
3.5. Limitations of the Study
This study involved several limitations, which should be taken into account when
assessing the study’s value in advancing the literature and generalizing the results to
other populations. First, only devices available in the U.S. were tested. The author
recognizes that, due to the relative ease of acquisition of products from other countries
because of the Internet, several additional numerical readout devices could have been
obtained. However, due to the additional shipping charges that would have been
involved, there would likely be little reason for any U.S.-based user to purchase a device
62
from a non-U.S. retailer. Thus, the devices tested are considered representative of the
devices conveniently available to U.S.-based users.
Second, only one of each model of PMBT was tested. It is possible that any single
device might not be reflective of the model line’s true performance capabilities.
However, it is likely that users will only purchase a single instrument, and thus will have
only one model for their use. In this respect, by testing only one of each model, this
study reflects the likely actual purchase/use scenario.
Third, the age range of participants in Experiment Two was restricted to 21-34 years.
This limitation was imposed to use participants within the age range most likely to be
involved in alcohol-related crashes (Texas Department of Public Safety, 1999). This
could limit the degree to which results could be generalized to other age groups.
Lastly, although Experiment Two was designed to simulate actual drinking
conditions, the participants did not have full control over use of the devices. Instead,
participants were told how and when to use the devices. In an actual drinking scenario,
at a bar, for example, drinkers would have to complete the additional tasks of
determining how to use the devices and make decisions regarding when to use them.
These variables were controlled for the purpose of reducing the influence of factors other
than the independent variables of interest. Thus, Experiment Two’s conditions cannot
be viewed as a totally realistic social drinking scenario.
63
3.6. Delimitations of the Study
The study also necessitated several delimitations to maximize the ability to test the
hypotheses in question. First, all samples were collected under controlled situations,
whether in the DPS laboratory or at the TTI facilities. These arrangements were
designed to help reduce the influence of outside factors affecting the dependent variable
of interest. Second, in Experiment Two, only data collected after participants were
found to be in the post-absorption phase were used for analysis. No data collected
during the absorption or diffusion phases were analyzed, although participants were
measured during these phases to determine the point at which each participant had
moved to the post-absorption phase.
64
CHAPTER IV
RESULTS
4.1. General
A total of eight devices were included in the study. Because of poor function, two
devices’ results had to be discarded. During Experiments One and Two, Device G
displayed its maximum value, .19, on virtually all tests, regardless of the level of alcohol
at which it was tested; it exhibited no ratio scale properties, yielding essentially only
binomial results. Thus Device G’s data were excluded from analysis.
Approximately halfway through Experiment Two, Device H simply stopped
functioning. Although its integral power light indicated it was receiving full power, it
began to show .00 readings at all alcohol levels. Installing fresh batteries did not
alleviate this problem. Thus, Device H’s data were excluded from analysis. Because
full and useful data were not obtained for these two devices, they were both eliminated
from analysis in both experiments. The six remaining instruments yielded complete data
for all tests.
4.2. Pilot Testing
One goal of the pilot testing was to provide a level of initial screening of performance
to determine whether devices would advance to further testing. In the assessment of the
65
devices’ abilities to minimize false positive and false negative readings, 320 total tests
were performed (eight devices, each tested 20 times at .00 and at .032).
In the assessment of false positives using the .00 simulator mixture (distilled water
only), all devices displayed .00 on all 20 tests. That is, no device yielded results above
.00 at any time during this testing.
In the assessment of false negatives using the .032 simulator mixture, all devices
yielded readings at or above the .02 threshold level. That is, no device read below .02
during this testing. Figure 1 shows the mean BrAC results of each device when tested at
.032.
0
0.01
0.02
0.03
0.04
0.05
0.06
A B C D E F
Device
BrAC Result
Figure 1. Mean BrAC results of each device tested at .032.
66
4.3. Experiment One
In the assessment of the devices’ accuracy and precision in this experiment, 960 total
tests were performed: eight devices, each tested 20 times at six different alcohol levels.
As in the pilot testing, the validity of each simulation concentration was verified through
20 tests using the Intoxilyzer.
Table 3 contains the mean values broken out by device at each concentration level.
Figure 2 shows the graphic representation of these results. Figure 3 shows each device’s
accuracy results separately, plotted against the simulator standards. No single device
met the NHTSA criteria for accuracy at all concentrations. The mean results show that
five out of the six devices read higher than the simulator standard.
Table 3
Mean Test Results for
Each Device at Each Concentration
Simulator Device
Conc A B C D E F
.02 .010 .036 .030 .020 .014 .034
.04 .047 .069 .038 .048 .062 .047
.06 .094 .117 .082 .084 .060 .057
.08 .080 .152 .119 .132 .081 .073
.10 .095 .196 .155 .181 .109 .080
.16 .141 .327 .250 .190 .169 .117
To assess precision, the standard deviation (SD) of results was computed for each
device at each concentration level. This yielded a value that reflects the spread of scores
of each device at each concentration (the smaller the SD value, the tighter the dispersion
of the scores around their mean). In certifying devices for its Conforming Products List,
67
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.02 0.04 0.06 0.08 0.1 0.16
Simulator Alcohol Level
Mean BrAC Result
Sim
A
B
C
D
E
F
Figure 2. Mean BrAC results of devices at each simulator concentration.
68
0
0.12
0.24
0.36
0.02 0.04 0.06 0.08 0.1 0.16
Device A
Sim
0
0.12
0.24
0.36
0.02 0.04 0.06 0.08 0.1 0.16
Device C
Sim
0
0.12
0.24
0.36
0.02 0.04 0.06 0.08 0.1 0.16
Device E
Sim
0
0.12
0.24
0.36
0.02 0.04 0.06 0.08 0.1 0.16
Device B
Sim
0
0.12
0.24
0.36
0.02 0.04 0.06 0.08 0.1 0.16
Device D
Sim
0
0.12
0.24
0.36
0.02 0.04 0.06 0.08 0.1 0.16
Device F
Sim
Figure 3. Mean results for each device at each simulator concentration. Simulator test
level is shown on X-axis, test result on Y-axis.
69
NHTSA requires that the standard deviations of results at each of these concentrations be
≤ .0042. Table 4 contains the SDs for each device at each level. Those results marked
with an asterisk meet the NHTSA standard for precision. Figure 4 displays the devices’
precision across the six alcohol levels. No single device met the NHTSA criteria for
precision at all concentrations.
Table 4
Standard Deviations for Each Device at Each Concentration
Simulator Device
Concentration A B C D E F
.02 .0000* .0051 .0000* .0000* .0049 .0052
.04 .0047 .0037 .0044 .0052 .0052 .0042*
.06 .0050 .0066 .0089 .0088 .0000* .0043
.08 .0000* .0049 .0049 .0135 .0031* .0055
.10 .0051 .0083 .0076 .0185 .0031* .0073
.16 .0072 .0109 .0132 .0000* .0059 .0043
Note. * SD ≤ .0042
70
0
0.004
0.008
0.012
0.016
0.02
0.02 0.04 0.06 0.08 0.1 0.16
Device A
0
0.004
0.008
0.012
0.016
0.02
0.02 0.04 0.06 0.08 0.1 0.16
Device C
0
0.004
0.008
0.012
0.016
0.02
0.02 0.04 0.06 0.08 0.1 0.16
Device E
0
0.004
0.008
0.012
0.016
0.02
0.02 0.04 0.06 0.08 0.1 0.16
Device B
0
0.004
0.008
0.012
0.016
0.02
0.02 0.04 0.06 0.08 0.1 0.16
Device D
0
0.004
0.008
0.012
0.016
0.02
0.02 0.04 0.06 0.08 0.1 0.16
Device F
Figure 4. Standard deviations of each device at each simulator concentration. Simulator
test level is shown on X-axis, standard deviation on Y-axis.
71
4.4. Experiment Two
4.4.1. Participant Eligibility
All females performed a portable pregnancy test during the orientation phase and all
results indicated negative. A single sober volunteer confirmed all pregnancy test results
as being negative. Thus, all female and male participants were eligible to continue in
this experiment.
4.4.2. Participant Functioning
Eleven people participated in the third experiment, with five and six participants in
sessions one and two, respectively. During the first session, one female participant
became ill during the consumption phase, after having consumed approximately twothirds
of the total dose. After she regurgitated, she remained at the testing site, and a
sober volunteer was assigned to closely monitor her. The monitor administered cold
towels to the participant’s forehead and neck, with positive results. She continued to
feel better as time passed. Because of this event, this participant was withdrawn from
the experiment and was not permitted to consume additional alcohol. Thus, no data
were collected from this participant, who remained at the testing site until her BrAC
reached .00.
After consuming the full dose assigned, participant four reached a peak BrAC of only
.067. She agreed to consume an additional measured alcoholic beverage to reach the
target peak BrAC of .09. This additional dosing was successful; she subsequently
72
reached the target peak. A total of ten participants completed a total of 640 tests (ten
participants providing duplicate samples at four concentrations with each device).
4.4.3. Data Screening
4.4.3.1. Duplicate Samples
For analysis, the duplicate results from each device at each level were averaged, and
the mean was carried forward into subsequent analysis. The acquisition of duplicate
readings was performed to help reduce the impact of any single measurement. Table 5
displays the means and standard deviations of the means of the duplicate samples from
all ten participants for each device at each concentration.
Table 5
Means and Standard Deviations for Each Device at Each Concentration
Concentration
Device .02 .04 .06 .08
Intox .023 (.001) .039 (.005) .058 (.004) .080 (.004)
A .060 (.016) .083 (.018) .104 (.