|
Article Excerpt
MOST ARRESTS FOR DRIVING WHILE intoxicated (DWI) are made in patrol operations in which officers watch for signs of impaired driving and stop drivers who show evidence of impairment. Currently, however, there is considerable interest by safety advocates and the National Highway Traffic Safety Administration (NHTSA) in persuading police to conduct sobriety checkpoints where all drivers are stopped and interviewed to detect impairment (Fell et al., 2003). Sobriety checkpoints have been shown to be effective in reducing alcohol-related crashes (Lacey et al., 1986, 1999; Levy et al., 1989; Ross, 1992; Shults et al., 2001; Stuster and Blowers, 1995; Voas et al., 1985; Williams and Lurid, 1984). The implementation of this procedure has been resisted by police departments, however, partially because few DWI arrests are made in such enforcement operations (Fell et al., 2004). One reason for the low arrest rate is the short time (30-60 seconds) provided at checkpoints for the officer to detect drinking.
Passive alcohol sensors (PAS), which draw in a mix of expired and environmental air from in front of a person's face, can provide a good estimate of whether a driver has been drinking (Farmer et al., 1999). The PAS appears to be particularly effective when observation time is short; therefore, it has the potential to become a helpful police aid at checkpoints or during regular patrol operations. Furthermore, a series of studies has demonstrated that when officers use passive sensors at a checkpoint, more drinking drivers are detected and the arrest rate increases by approximately 50% (Ferguson et al., 1995; Lund and Jones, 1987; Lund et al., 1991).
Despite their apparent utility, police officers who are provided with PAS units often do not adopt them during the long term (Leaf and Preusser, 1996). Some reasons for not using PAS units are that (1) police officers would have to wear another piece of equipment on their belts, (2) arrests based on the PAS would be susceptible to legal challenges, and (3) the PAS is not accurate enough under less-than-optimal conditions (e.g., applied by unskilled operators; lack of or little cooperation by subjects; used at an excessive distance from the subject's mouth, and/or under windy, humid, or cold conditions) (Wisconsin Department of Transportation, 2002). Arguably, the annoyances associated with using the PAS unit would likely be set aside by police officers if its legality and usefulness were guaranteed.
To be useful, the PAS must be reasonably accurate (i.e., be associated with the blood alcohol concentration [BAC] as measured by an evidential breath-test device strong enough to allow the PAS to serve as a valid screening device). The accuracy of the Sniffer PAS III (PAS Systems International, Fredericksburg, VA; www.sniffalcohol.com) was tested in the 1996 National Roadside Survey (Voas et al., 1998). Each of the nine bars on the PAS flashlight corresponds to a BAC value of .01% to [greater than or equal to] .12% (Figure 1). These BAC values were established through laboratory calibration procedures under highly controlled conditions. In the field, however, conditions under which the measurements are made cannot be controlled. The PAS measures can be affected by the skill of the operator and, somewhat, by the cooperation of the subject. Thus, field measures vary considerably from those of the manufacturer's specified values.
[FIGURE 1 OMITTED]
Recently, a new data set with slightly fewer than 15,000 drivers containing measures of BAC levels collected by both PAS and preliminary breath-test (PBT) devices became available (Blomberg et al., 2005). This study was the most carefully controlled data set of its kind. It achieved a 90.5% response rate from crash-involved motorists (excluding hit-and-run incidents) and a 97.6% response rate from the comparison set of motorists. Further, it is the first survey of its kind to obtain BAC data from a sample of hit-and-run drivers to adjust relative risks for a nonparticipant bias. Thus, maximum use of these data is well justified to address issues and hypotheses that were not fully evaluated in the original study. The PAS data, in particular, provided a basis for conducting secondary analysis on the accuracy of the Sniffer PAS III that was used in the 1996 National Roadside Survey.
The primary objective of this secondary analysis was to determine the optimal PAS threshold for officers to use when deciding whether to initiate an investigation of a driver legally stopped on the highway to determine alcohol impairment. In addition to determining the relationship of PAS measures to BAC measures, a study was conducted of factors that might mediate the relationship, including (1) the influence of gender (women expel less air in speaking), (2) the influence of age (younger drivers may be less cooperative and more difficult to measure accurately), (3) the influence of light conditions (the light scale may be more difficult to read during the day), (4) the influence of testing motorists in a crash compared with randomly stopped noncrash-involved motorists (crash-involved motorists may be less cooperative, making it more difficult to get a good measure), and (5) the ability of officers to rate the value of a PAS measure.
Method
Data collection
The study funded by NHTSA (Blomberg et al., 2005) that created the data set used in this secondary analysis was conducted for 1 year in Long Beach, CA, and for 1 year in Fort Lauderdale, FL. A study team, consisting of an officer and researcher/interviewer, was sent to the site of a reported crash. An officer made the initial contact with the crash-involved drivers and requested their participation in the survey (most crashes involved two drivers). If the driver or drivers were willing, the researcher conducted the interviews. A week later, on the same day and at the same time, the survey crew returned to the same site, and the officer stopped two motorists, at random, traveling in the same direction as each crash-involved driver. (In crashes involving multiple vehicles, two controls were selected as matches for each crash-involved driver.) Thus, the unit of sampling was the crash, and the unit of analysis was the driver.
The PAS measures for the study by Blomberg et al. (2005) were taken by police officers rather than by research...
|
