Standardized Field Sobriety Test: False Positive Test Rate among Sober Subjects

Abstract

Standardized Field Sobriety Test: False Positive Test Rate among Sober Subjects
The Standardized Field Sobriety Test (SFST) is a series of exercises that a law enforcement officer gives to a driver suspected of driving under the influence of alcohol. The original research that demonstrated a high correlation between failure of the SFST and a high blood alcohol concentration did not utilize a standard control group to validate that the failure of the SFST was not a characteristic of the population at large. This study examined a series of drug naive subjects to determine the rate of failure of the SFST to accurately distinguish a suspect with high blood alcohol content from the general public. Of the 185 subjects tested, 26% of the drug naive subjects failed the SFST.

Full text

a SciTechnol journal
Research Article
Yoshizuka et al., J Forensic Toxicol Pharmacol 2014, 3:2
http://dx.doi.org/10.4172/2325-9841.1000120 Journal of Forensic
Toxicology & Pharmacology
All articles published in Journal of Forensic Toxicology & Pharmacology are the property of SciTechnol, and is protected by
copyright laws. Copyright © 2014, SciTechnol, All Rights Reserved.
International Publisher of Science,
Technology and Medicine
Standardized Field Sobriety
Test: False Positive Test Rate
among Sober Subjects
Keith Yoshizuka*, Paul J Perry, Greta Upton, Ingrid Lopes and
Eric J Ip
Abstract
The Standardized Field Sobriety Test (SFST) is a series of
exercises that a law enforcement of¿cer gives to a driver suspected
of driving under the inÀuence of alcohol. The original research
that demonstrated a high correlation between failure of the SFST
and a high blood alcohol concentration did not utilize a standard
control group to validate that the failure of the SFST was not a
characteristic of the population at large. This study examined a
series of drug naive subMects to determine the rate of failure of the
SFST to accurately distinguish a suspect with high blood alcohol
content from the general public. 2f the 185 subMects tested, 2 of
the drug nawve subMects failed the SFST. Since the SFST is used as
evidence of probable cause to Mustify an arrest, a 2 false positive
rate in the SFST may imply that the SFST may be only a minor factor
in combination with other articulated evidence to Mustify suf¿cient
probable cause for an arrest for driving under the inÀuence, and
may affect the weight of the evidence given to the SFST.
Keywords
Standardized ¿eld sobriety test False positives %lood alcohol
concentration
*Corresponding author: 'r. .eith Yoshizuka, Pharm', M%A, J', Associate
Professor, College of Pharmacy, Touro 8niversity, California, 8SA, Tel: 707-
38-5992 Fax: 707-38-5953 (-mail: keith.yoshizuka#tu.edu
Received: January 08, 2014 Accepted: April 09, 2014 Published: April 1,
2014
laboratory to 297 drinking volunteers with BACs ranging from 0 to
0.18%. e ocers were able to accurately categorize 81% of subjects
on the basis of the SFST test results as to being a BAC <0.10% or a BAC
≥ 0.10%. e inter-rater and test-retest reliabilities kappa coecient
for the test battery ranged from a statistically acceptable 0.60 to 0.80.
Burns reviewed ten studies assessing the validity of sobriety tests and
concluded that a failed SFST was a reliable indicator regarding the
presence of alcohol but not other CNS depressant drugs [5].
In 1981, the NHTSA implemented the SFST that enabled law
enforcement ocers to determine driver impairment and establish
cause for arrest [6]. In October 2000, Congress passed and President
Clinton signed into a law a bill that lowered the national standard for
impaired driving to a BAC of ≥ 0.08%. States that did not enforce this
federal provision by October 2003 would progressively lose federal
highway funding, thus currently establishing a BAC of 0.08% as
the legal per se limit in all states [7]. Before the passage of the new
federal BAC limit, the NHTSA sponsored a study that established the
creditability of the 0.08% limit. e study found that the SFST could
accurately dierentiate drivers above or below the more sensitive
BAC ≥ 0.08% in 91% of subjects tested [1]. Data analysis among
297 suspected motorists found the SFST to be extremely accurate in
discriminating between BACs above and below 0.08% threshold. e
mean estimated and measured BACs of the 297 motorists tested were
0.117% and 0.122%, respectively. Further, analyses found the HGN
test to be the most predictive of the three components of the SFST
battery (r=0.65); a higher correlation was obtained when the results
of all three tests were combined (r=0.69). However, a major limitation
of this study was that only drivers suspected of being under the
inuence of alcohol were evaluated. us, the study did not include a
control group to evaluate the SFST pass rate [1]. erefore, it was not
established whether a person could fail the SFST without any form of
drug impairment or a BAC of 0.0%.
e objective of this manuscript is to report baseline (sober)
failure rates from three studies that perturbed the SFST by drug-
induced challenges to the study subjects [8-10].
Methodology
e three experimental studies that collected baseline control data
on sober test subjects included trazodone (sedating antidepressant)
versus acetaminophen (over-the counter non-narcotic analgesic
[8]; diphenhydramine (sedating antihistamine) versus fexofenadine
(non-sedating antihistamine) [9]; and dextromethorphan (over-the
counter opioid cough suppressant) versus docusate sodium (stool
soener) [10].
Only researchers identied in the institutional review board
proposals had full access to the data, and all researchers completed
the National Institutes of Health human subjects training program.
e study procedures were performed in accordance with the
ethical standards of the Declaration of Helsinki, and the protocol
was approved by the Touro University institutional review board.
Sample size decisions for the three studies were based upon power
analysis calculations using data from prior studies that demonstrated
that diphenhydramine, trazodone, and dextromethorphan impaired
psychomotor performance. A summary of the methodology of each
of the studies is presented below.
Introduction
In 1975, the National Highway Trac Safety Administration
(NHTSA) endorsed research that correlated the Standardized Field
Sobriety Test (SFST) with blood alcohol concentrations (BACs)
[1,2]. In an eort to standardize and reliably assess impaired drivers,
researchers evaluated tests that were most commonly used by ocers
at the time, including the one-leg stand (OLS), walk-and-turn
(WAT), nger-to-nose, nger count, horizontal gaze nystagmus
(HGN), tracing, and alternate tests (Romberg body sway, subtraction,
counting backward, letter cancellation). All these tests were
perturbed by alcohol. However, statistical analyses concluded that
the combined score of three tests, the HGN and two divided attention
task tests (OLS and WAT), were the “best test set” to assess sobriety.
Using discriminant analysis, the combined scores of the three tests
predicted that law enforcement ocers could correctly classify 83%
of test subjects as either sober or intoxicated, i.e., a BAC of 0.10% or
greater [3]. A conrmatory study funded by the NHTSA supported
the previous ndings [4]. Ten ocers administered the SFST in a

Citation: Yoshizuka K, Perry PJ, Upton G, Lopes I, Eric J Ip (2014) Standardized Field Sobriety Test: False Positive Test Rate among Sober Subjects. J Forensic
Toxicol Pharmacol 3:2.
xPage 2 of 3 x
doi:http://dx.doi.org/10.4172/2325-9841.1000120
9olume 3  Issue 2  1000120
Dextromethorphan versus docusate sodium (DM/DSS) [10]
is experiment was a randomized, double-blinded, repeated
measures design involving 80 healthy adult participants. e
study determined the failure rates on the SFST aer a single dose
of dextromethorphan 30mg or docusate sodium (DSS) 200mg.
e SFST, administered by two physician-trained evaluators, was
evaluated before drug ingestion (i.e. at baseline) and two hours
post dextromethorphan or DSS ingestion. In this study, there was
no statistical dierence between the experimental and control test
groups in subject age, gender, ethnicity, height, and weight or body
mass index.
Diphenhydramine versus fexofenadine (DPH/FXF) [9]
e goal of this experiment was to evaluate if the SFST could
dierentiate individuals who took diphenhydramine versus those
who took fexofenadine. Using a randomized, double-blinded,
repeated measures study design; 60 healthy adult individuals ingested
either a single dose of diphenhydramine 50 mg or fexofenadine 60
mg. e SFST administered by two physician-trained evaluators, was
evaluated before drug ingestion (i.e. at baseline) and two hours post
dextromethorphan or fexofenadine ingestion.
Trazodone versus acetaminophen (TZD/APAP) [8]
e goal of this study was to evaluate the passage of the SFST
aer a single dose of trazodone 100 mg or acetaminophen 650 mg.
A randomized, double-blinded, repeated-measures design was
employed. Forty-ve healthy adult subjects were administered the
SFST by two NHTSA-manual trained evaluators at baseline and two
hours post trazodone (30 subjects) or acetaminophen (15 subjects)
ingestion. SFSTs were conducted and evaluated by two trained
individuals.
For the dextromethorphan and diphenhydramine studies, the
SFST was administered at baseline and 2 hours post drug ingestion.
SFST ratings were scored by two physician-trained evaluators. e
physician, an experienced college professor, studied and then adapted
the 161-page DWI Testing and Standardized Field Sobriety Testing
student manual to train the evaluators [11]. Emphasis was placed
on the mechanics of the SFST, the physiology and pathophysiology
gauged by the test, and the nuances that enable testing to pass scrutiny
in a court of law. Training videos assembled by law enforcement were
reviewed multiple times. Each evaluator met with the physician in a
workshop in order to practice each maneuver with both the verbal
description and physical demonstration required by law. For the
trazodone study, the raters were trained by two clinically experienced
clinical pharmacists using the same DWI Testing and Standardized
Sobriety Field Testing student manual.
Results
Table 1 presents a summary of the baseline failure rates for
the studies. In the dextromethorphan versus DSS study, 1 subject
failed the HGN, 18 subjects failed the WAT, and 5 subjects failed
the OLS. Overall, 22 of 80 (28%) subjects failed the SFST. In the
diphenhydramine versus fexofenadine study, 11 failed the WAT
and 5 failed the OLS, resulting in an overall failure rate of 14 of 6o
subjects (23%). Finally, in the trazodone versus acetaminophen study,
3 subjects failed the HGN, 11 failed the WAT, and 3 failed the OLS.
Overall 13 out of 45 (29%) of the subjects failed. Considering the total
control and intervention study populations from the three studies,
49 out of 185 subjects failed some aspect of the SFST at baseline.
In other words, there was a 26% failure rate of the SFST when no
form of pharmacologic perturbation to the study subjects occurred.
Figure 1 presents the pass-fail rates for the three independent studies.
A chi-square test of independence was performed on these data to
determine if there were dierences pass-fail frequency (count) data
between the 3 studies. e p-value of 0.46 indicates that the pass-fail
rates between the studies did not dier. us the SFST failure rate was
reproducible between dierent groups of subjects.
Discussion
e SFST is commonly used throughout the country to test for
impairment while driving. e original research demonstrated a
high correlation between failure of the SFST and a BAC of ≥ 0.08%.
e limitation of the Stuster and Burns [1] data is that there was no
control group to compare the results. However, this deciency can
be remedied by using the data of Stuster and Burns [1] as a historical
experimental group. e authors acknowledge that use of such a
control group is a limitation as it may introduce a bias. us, a 2×2
contingency table was constructed. e control group consisted of the
Figure 1: %aseline (sober) pass-fail rates for SFST for 3 studies (chi-
square=0.48, p=0.79).
Table 1: %aseline SFSTa failure (pre-drug administration) rates for 3 studies.
Study HGNb n
(%) WATc n
(%) OLSd n
(%) Overall combined failure rate
n (%)
Dextromethorphan vs.
DSS Study
Dextromethorphan 30mg
(n=40) 0
(0)
 (15.0) 2 (5.0)
22 (28%)
Docusate Sodium 200mg
(n=40) 1
(2.5)
12 (30.0) 3 (7.5)
Diphenhydramine vs.
Fexofenadine
Diphenhydramine (n=40) 0
(0)
7
(17)
4 (10)
14 (23%)
Fexofenadine (n=20) 0
(0)
4
(20)
1
(5)
Trazodone vs.
Acetaminophen
Trazodone 100mg (n=30) 3 (10.0) 7 (23.3) 1 (3.3)
13 (29%)
Acetaminophen 650mg
(n=15) 0
(0.0)
4 (2.7) 2 (13.3)
Overall Failure Rates n=185 4 (2.2) 40 (22)13 (7.0)49 (26%)
SFST: Standardized Field Sobriety Test +*1: +orizontal *aze 1ystagmus :AT: :alk-And-Turn 2/S: 2ne-/eg Stand

Citation: Yoshizuka K, Perry PJ, Upton G, Lopes I, Eric J Ip (2014) Standardized Field Sobriety Test: False Positive Test Rate among Sober Subjects. J Forensic
Toxicol Pharmacol 3:2.
xPage 3 of 3 x
doi:http://dx.doi.org/10.4172/2325-9841.1000120
9olume 3  Issue 2  1000120
present study’s data of 185 subjects of whom 49 who had a positive
or failed SFST. e alcohol consumer group consisted of the Stuster
and Burns 234 subjects who tested positive for a BAC of >0.08% BAC
of whom 210 had a positive or failed SFST. As illustrated in Figure 2,
a signicant dierence was obvious between the two groups (Fisher’s
exact test p<0.0001, sensitivity=0.90, specicity=0.74, positive
predictive value 0.81 and negative predictive value=0.85). More
importantly, the data demonstrate that there is a false positive rate
of 0.26 and false negative rate of only 0.10 that is problematic from a
legal standpoint. Additionally, examination of the entire Stuster and
Burns [1] data set of 297 subjects nds that of 83 individuals with
BAC <0.08%, 24 (29%) were characterized by the ocers as having a
BAC of ≥ 0.08%. Stated in another way, their false positive rates are
nearly identical to our sober or drug naïve subject 26% SFST failure
rate. A control group is used to establish that the correlation has a
meaningful relationship to the outcome, and is not just a characteristic
of the population at large.
Harris identied 23 viable visual clues that are used by police
ocers to detect drivers driving under the inuence [12]. He
generated a Drunk Driver Detection Guide as an aid for use by
ocers. He found that there was an association between the number
of clues and the probability of the driver having a BAC of either
>0.1% or >0.05%. As an example, if a driver had been noted to be
following a car too closely, the probability of a 0.1% BAC was 55%. If
two additional clues of straddling the center line and slow response
to trac signals were observed, the probability of a DUI increased to
65%. e same three clues generated an 85% probability of a 0.05%
BAC. e Harris system accuracy nearly approaches the SFST 91%
accuracy rate reported by Stuster and Burns [1].
e observation that visible clues of Harris [12] and the SFST
data of Stuster and Burns [1] generate similar accuracy rates provokes
the question as to whether some component(s) of the SFST have a
negative eect on the accuracy rates. Inspection of Table 1 notes a
HGN failure rate of only 2.2% versus 21.6% and 7.0% failure rates for
the WAT and OLS respectively. Stuster and Burns had only a 1.9%
(4/209) failure rate on the HGN [1]. ey have referred to the HGN as
contributory evidence that “provides indisputable evidence of alcohol
in a motorist’s system” [1]. e explanation for this “dogma” is that
experienced drinkers can perform the voluntary physical divided
attention task tests, i.e., OLS and WAT correctly even with a BAC
>0.01% but they cannot pass the HGN because it is an involuntary
reaction over which they have no voluntary control.
With a false positive rate of 26%, the authors submit that the
SFST can only be contributing evidence to justify an arrest for driving
under the inuence, in combination with other articulated evidence
such as the ocer’s observation of erratic driving behavior, the odor
of alcohol on the driver’s breath, or other such evidence in order to
have sucient probable cause to justify an arrest.
What this study demonstrates is that there are also a signicant
number of persons NOT impaired who cannot successfully pass the
SFST. e SFST is used by law enforcement as evidence of probable
cause to justify an arrest of the subject. e results of this study call into
question the validity of using the SFST as the primary justication for
an arrest for driving under the inuence. Other articulated evidence to
justify the initial trac stop such as an obvious equipment violation,
a moving violation, unusual or suspicious behavior, or almost
anything else that would call attention to the suspect vehicle would
be required [11]. Once the vehicle is stopped, the ocer must observe
and interview the driver face-to-face. At this point, the ocer must
be able to articulate further evidence to justify requiring the driver
to exit the vehicle to administer the SFST. Probable cause is a level
of reasonable belief, based on articulated fact, required to arrest and
prosecute a person in criminal court that a reasonable person would
nd sucient for a conviction. e quantum of evidence required for
arrest is generally a preponderance of the evidence, as compared to
a conviction that requires a quantum of beyond a reasonable doubt.
is study provides information that aects the weight given to the
SFST as evidence. Only when all of these conditions are met is it
reasonable to administer the SFST, and even failing the SFST is not
dispositive of intoxication absent further evidence. Ocers do not
use any single test as a basis to justify arrest, but must evaluate the
totality of the evidence of exhibited behaviors, performance tests, and
other observed evidence as the basis for arrest.
References
1. Stuster J, %urns M (1998) 9alidation of the standardized ¿eld sobriety test
battery at %ACs below 0.10 percent, 8.S. 'epartment of Transportation,
1ational +ighway Traf¿c Safety Administration.
2. Stuster J (200) 9alidation of the standardized ¿eld sobriety test battery at
0.08 blood alcohol concentration. +um Factors 48: 08-14.
3. %urns M, Moskowitz + (1977) Psychophysical Tests for ':I Arrest,
8.S. 'epartment 2f Transportation, 1ational +ighway Traf¿c Safety
Administration, :ashington.
4. Tharp 9, %urns M, Moskowitz + (1981) 'evelopment and ¿eld test of
psychophysical tests for ':I arrest, 8.S. 'epartment 2f Transportation,
1ational +ighway Traf¿c Safety Administration, :ashington.
5. %urns M (2003) An overview of ¿eld sobriety test research. Percept Mot Skills
97:1187-1199.
. 'evelopment of a standardized ¿eld sobriety test (2001) 8.S. 'epartment 2f
Transportation, 1ational +ighway Traf¿c Safety Administration, :ashington.
7. 8.S. 'epartment of Transportation, 1ational +ighway Traf¿c Safety
Administration, Federal +ighway Administration (2003) 8 FR 091-
2peration of motor vehicles by intoxicated persons.
8. Ip (J, %ui 49, %arnett MJ, .azani A, :right R, et al. (2013) The effect of
trazodone on standardized ¿eld sobriety tests. Pharmacotherapy 33: 39-
374.
9. /u ', Ip (, /opes I, %arnett M, Chu (, et al. (ffects of diphenhydramine
versus fexofenadine on the standardized ¿eld sobriety test. Touro 8niversity-
California, 9alleMo, CA.
10. Pal J, Ip (, Trinh ., Yu J, /indfelt T, et al. (2013). (ffects of dextromethorphan
on the standardized ¿eld sobriety test. Touro 8niversity-California, March,
9alleMo, CA.
11. 8.S. 'epartment of Transportation (200) ':I (driving while intoxicated)
Detection and Standardized Field Sobriety Testing.
12. +arris D+ (1980) 9isual detection of driving while intoxicated. +um Factors
22: 725-32.
$XWKRU$I¿OLDWLRQVTop
College of Pharmacy Touro, University-California, USA
Figure 2: SFST pass-fail values for 234 (fail=210) subMects with %A C!
0.08 versus 185 (fail=49) drug-naive subMects (p, 0.0001).