Assessing DUI risk: Examination of the Behaviors & Attitudes Drinking & Driving Scale (BADDS)

Available online at www.sciencedirect.com

Addictive Behaviors 33 (2008) 853 – 865

Assessing DUI risk: Examination of the Behaviors & Attitudes Drinking & Driving Scale (BADDS) Jeremy D. Jewell ⁎, Stephen D.A. Hupp 1 , Daniel J. Segrist 2 Southern Illinois University Edwardsville, Department of Psychology, Box 1121, Edwardsville, IL 62026, United States

Abstract Despite research findings indicating attitudinal differences among drivers with and without a history of driving under the influence (DUI) offenses, there are no well-established instruments specifically designed to clinically assess drinking and driving attitudes and behaviors among adults. The purpose of this current series of three studies was to investigate the psychometric properties of the Behaviors & Attitudes Drinking & Driving Scale (BADDS). The BADDS was developed in previous studies by the authors and assesses respondents' rationalizations for drinking and driving, likelihood of future drinking and driving, drinking and driving behaviors, and riding with a drinking driver behavior in the previous month. Study 1 (N = 179) and Study 2 (N = 338) assessed college participants, while Study 3 gathered data from adult DUI offenders (N = 160) and non-DUI offenders (N = 166). Results indicate good to excellent test–retest reliability and internal consistency estimates for BADDS scale scores. Support for the construct validity as well as concurrent and predictive criterion validity of BADDS scores was also demonstrated. Potential applications for the measure, as well as need for future research are described. © 2008 Elsevier Ltd. All rights reserved. Keywords: Alcohol drinking attitudes; Drinking behavior; Driving under the influence; Risk assessment; BADDS; Behaviors & Attitudes Drinking & Driving Scale

With the National Highway Traffic Safety Administration (NHTSA) reporting approximately 221 million vehicles on the road in the United States (NHTSA, n.d.), drinking and driving continues to be one of the greatest preventable health risks nationwide. In fact, vehicular accidents are a leading cause of death ⁎ Corresponding author. Tel.: +1 618 650 3734; fax: +1 618 650 5087. E-mail addresses: [email protected] (J.D. Jewell), [email protected] (S.D.A. Hupp), [email protected] (D.J. Segrist). 1 Tel.: +1 618 650 3280. 2 Tel.: +1 618 650 5391. 0306-4603/$ - see front matter © 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.addbeh.2008.02.002

854

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

in the United States, with only cancer and heart disease responsible for more years of life lost (NHTSA, 2006). In 2004, accidents involving alcohol accounted for 39% of all traffic fatalities, resulting in 16,694 deaths, with most of these fatalities (86%) involving a legally intoxicated driver with a BAC meeting or exceeding .08 g/dL (NHTSA, 2004). Compounding these issues is the fact that despite the pervasiveness of drinking and driving, only one driving under the influence (DUI) arrest is made for every 300 to 1000 drunk driving episodes (Yalisove, 2004). Research on drinking-related cognitions as they relate to Drinking and Driving Behaviors has become a topic of increasing interest. Schell, Chan, and Morral (2006), for example, found that positive alcohol expectancies (e.g., “I feel relaxed when drinking") were related to drinking and driving among DUI recidivists, such that those with stronger expectations for the positive consequences of drinking engaged in drinking and driving more frequently. In examining motivations specifically for drinking and driving behavior, Kulick and Rosenberg (1999) found that while the majority of their college student sample regularly drove after drinking, participants rationalized this behavior in a variety of ways. The most commonly reported rationales for driving under the influence of alcohol were the desire to go to a particular destination (e.g., home), perceiving oneself as only slightly intoxicated, perceiving that friends and others are too intoxicated to drive, and perceiving oneself as having had less to drink than friends and others. McCarthy, Pedersen, and Leuty (2005) found that drinking and driving attitudes were predicted by participants' history of drinking and driving consequences (e.g., car accident). In short, those who had experienced consequences from drinking and driving held more lenient attitudes toward driving after having one, three, five, or more drinks. Perceived risk has also been shown to relate to drinking and driving behavior. In one study, for example, adolescents' perceptions of the dangers of drinking and driving, as well as perceived likelihood of getting into an accident while driving after drinking, was negatively correlated with drinking and driving (Gibbons, Lane, Gerrard, Pomery, & Lautrup, 2002). Mannering, Bottiger, and Black (1987) argue that attitudes toward drinking and driving, while affecting drinking and driving decisions, are not well understood. Despite research findings indicating differences among non-DUI drivers and DUI drivers, most research in this area either relies on one or two attitudinal items (Stacy, Bentler, & Flay, 1994) with little evidence regarding reliability and validity of the questions, or uses a general driving behaviors questionnaire that includes a drinking and driving item (e.g., Driving Behaviour Questionnaire; Ozkan, Lajunen, & Summala, 2006). Additionally, there are no well-established instruments specifically designed to assess drinking and driving attitudes and behaviors among adults. Specifically assessing drinking and driving attitudes and behaviors may provide an effective means for identifying potential DUI offenders. Measuring and understanding attitudes toward drinking and driving is also critical because attitudes have the potential to change consequent to experience, prevention, and intervention. Given the pervasive problem posed by drinking and driving as well as the seemingly intractable nature of drinking and driving as exemplified by its high recidivism rate (Greenberg, Morral, & Jain, 2005; McCarthy et al., 2005; Wiliszowski, Murphy, Jones, & Lacey, 1996) more extensive tools for investigating attitudes and behaviors toward drinking and driving are crucial in order to more fully understand driving after drinking as well as designing effective universal, selective, and indicated prevention efforts. Comprehensive, reliable, and valid measures of DUI attitudes and behaviors would also equip researchers and practitioners with a means of assessing the effectiveness of such interventions. McCarthy, Pederson, Thompson, and Leuty (2006) recently developed the Positive Expectancies for Drinking and Driving for Youth (PEDD-Y) in order to meet this described need. This measure includes

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

855

four scales: drinking and driving because of convenience, feeling control over drinking and driving, avoiding consequences, and seeking excitement through drinking and driving. While the authors present preliminary evidence of the scale's reliability and validity, they point out several limitations. For example, no adults were used in the study, limiting its generalizability beyond an adolescent population. This is especially important given that adolescents and adults differ in their perception of risk (Cohn, Macfarlane, Yanez, & Imai, 1995), particularly with regard to driving behaviors (Finn & Bragg, 1986). McCarthy et al. (2006) also relied entirely on self-report measures. Finally, a major limitation of the scale itself is that it does not include any questions measuring actual drinking and driving behavior. Given that health behaviors and attitudes appear to have a reciprocal relationship in which attitudes influence behaviors and behaviors influence attitudes (Gerrard, Gibbons, Benthin, & Hessling, 1996), assessing both attitudes and behaviors regarding drinking and driving is critical. Cognitive theory (see Sayette, 1999) posits that risky drinkers have maladaptive thoughts that are antecedents to drinking and driving. For example, the more commonplace drinking and driving is considered to be, the less risky it is perceived to be (Gibbons et al., 2002). Additionally, people may be more likely to drink and drive if they believe drinking and driving is more acceptable when it is just a short distance. Although driving may be impaired by moderate alcohol consumption, people may not consider driving impaired unless BAC is at or above .10 (Higson, Heeren, & Winter, 1999). While the previously described PEDD-Y may have promise in the assessment of consequence-based reasoning related to drinking and driving, another recently developed scale (the Behaviors & Attitudes Drinking & Driving Scale (BADDS)) also focuses on the attitudes people have about drinking and driving, as well as the related behaviors, and is the focus of the current series of studies. An initial study focused on assessing the effectiveness of a drinking and driving prevention program tool with college students was conducted by Jewell, Hupp, and Luttrell (2004). In order to assess attitudinal change after the program, the authors developed the Attitudes toward Drinking and Driving Scale (ADDS), as no similar measure could be found in the existing literature. The ADDS asked participants in this study to respond to a list of common Rationalizations for Drinking and Driving as well as to rate the likelihood they would drive based on varying amounts of alcohol consumption and driving distance. In a follow-up study on this prevention program tool, Jewell and Hupp (2005) added behavior items and termed the instrument the BADDS. Both studies demonstrated initial evidence for the internal consistency and validity of the measure with college students. 1. Purpose The general purpose of this current series of studies was to validate psychometric properties of the BADDS observed in earlier studies (Jewell et al., 2004; Jewell & Hupp, 2005) and to expand the body of evidence regarding its reliability and validity with additional measures and samples. The specific psychometric indices investigated in Study 1 were internal reliability of BADDS scale scores and their concurrent criterion validity. Study 2 also investigated internal reliability, as well as test–retest stability, construct validity, and predictive criterion validity of BADDS scores. Construct validity of BADDS scale scores was estimated by correlating them with three scales from the Substance Abuse Subtle Screening Inventory, 3rd Edition (SASSI-3; Lazowski, Miller, Boye, & Miller, 1998; Miller & Lazowski, 1999). As the SASSI-3 is a screening measure for substance (alcohol or drug) abuse, it was hypothesized that BADDS scale scores, which assess attitudes and behaviors regarding drinking and driving, would be positively correlated with these SASSI-3 scale scores. The ability of BADDS attitudes scores to predict

856

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

future Drinking and Driving Behaviors was examined by regressing participants' Time 1 attitudes scores on their Time 2 Drinking and Driving Behaviors measured 4 weeks later. The purpose of Study 3 was to examine the discriminant validity of the BADDS by obtaining a criterion sample of adults with a recent DUI offense as well as a comparison group of adults with a traffic offense but with no history of DUI arrest or conviction. 2. Methods 2.1. Participants Participants in Study 1 were 179 college students attending a university in the Midwest. Participants in Study 2 were 338 college students drawn from the same Midwestern university. Participants in Studies 1 and 2 were not asked how often they drove a motor vehicle, although currently unpublished data gathered from a similar population at the same university indicated that 94% regularly drive a vehicle (at least 4 days a month). In Study 3 an adult sample comprised of two groups was obtained. The first group (DUI group, n = 160) was comprised of adults who were ticketed and arrested for a DUI offense within the previous 6 weeks. The second group (Traffic Court group, n = 166) was comprised of adults who were ticketed for some other moving violation (e.g. speeding) or some other minor violation in the community (e.g. noise violation) but had no history of DUI ticket or arrest according to self-report. Analyses indicated no significant differences in the composition of the groups in Study 3 in terms of age or ethnicity, although the Traffic Court group consisted of significantly more women than the DUI group (χ 2 (1, 320) = 19.92, p b .01). Demographic characteristics of participants in all three studies are shown in Table 1. It should be noted that participants in each study did not overlap and were separate and distinct samples. The size for the samples of each study were determined by calculating the minimum sample size needed for each analysis presuming a moderate effect size and obtained power at or above .90.

Table 1 Respondent demographic characteristics by study Demographic variable

Study (and group) Study 1

Number of participants Age Gender (%) Male Female Ethnicity (%) Caucasian African American Other Hispanic origin

Study 2

Study 3 DUI

Traffic court

179 22.9 (4.9)

338 19.7 (3.1)

160 29.8 (10.1)

166 30.1 (12.9)

24 76

28 72

79 21

55 45

88 9 3 1

87 9 4 4

87 10 3 5

89 11 0 2

Note: DUI = participants with a recent DUI arrest. Traffic court = participants with no self-reported history of DUI arrest.

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

857

2.2. Measures The BADDS is a self-administered questionnaire that is designed to be completed in 10 min or less. The BADDS is mainly comprised of the following four scales: Rationalizations for Drinking and Driving, Likelihood of Drinking and Driving, Drinking and Driving Behaviors, and Riding Behaviors with a Drinking Driver (readers should contact the first author of this study for more information on obtaining the BADDS for research or clinical use or visit www.thebadds.com). The Rationalizations for Drinking and Driving scale is comprised of 12 items rated on a fivepoint Likert scale ranging from “disagree” to “agree.” All items begin with the stem “I believe it is okay to drink and drive if.” An example of an item is “I believe it is okay to drink and drive if… nobody else is in the car.” Items are scored from 0 to 4, with the range of possible scores on this scale from 0 to 48. The Likelihood of Drinking and Driving scale is comprised of 15 items rated on a five-point Likert scale ranging from “very unlikely” to “very likely.” The first five items ask respondents how likely they would be to drive a short distance (a few blocks to a mile) after having a particular number of drinks, beginning at one drink and increasing per item to more than 6 drinks. “Drinks” are defined as one beer, one glass of wine, or one “shot” of liquor. The next five items have the same format except respondents are asked how likely they would be to drive a medium distance (about 10 miles), while the last five items are regarding a long distance (over 20 miles). Items are scored from 0 to 4, with the range of possible scores on this scale from 0 to 60. The Drinking and Driving Behaviors scale is comprised of two items that ask respondents to identify the number of times they have driven in the past month after drinking one or two drinks, and three or more drinks within the previous hour. The Riding Behaviors with a Drinking Driver scale is comprised of two items that ask respondents to identify the number of times in the past month they have been a passenger of a driver who drank one or two drinks, and three or more drinks in the previous hour. On these two scales the minimum possible scores are 0, while there are no maximum possible scores. 2.3. Procedures 2.3.1. Study 1 Participants were recruited from a variety of psychology courses and given credit for their participation. The BADDS was administered once to each participant, with most participants taking less than 10 min to complete the measure. Written informed consent was obtained from each participant. Consent forms were keyed with unique ID numbers to the BADDS. Participants for Study 2 also were recruited from a variety of psychology courses and given credit for their participation. Again, written informed consent was obtained from each participant and consent forms were keyed with unique ID numbers to all measures. In order to estimate the test–retest reliability of BADDS scores, a subsample (N = 190) of Study 2 participants was randomly chosen to be administered the BADDS a second time. The time between the first and second administration of the BADDS was 4 weeks. Of the 190 participants asked to complete a second BADDS 4 weeks after the first administration, 29 were lost to attrition, with a remaining sample of 161 participants. Independent samples t-tests indicated no significant differences on any BADDS scales between those who agreed to complete a second administration versus those who did not. Additionally, there were no differences between these groups in terms of age.

858

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

In addition to the BADDS, participants in Study 2 were administered the SASSI-3, which is a screening instrument for substance use disorders. Internal reliability estimates reported for the three direct measures of substance misuse on the SASSI-3 employed in the current study are: Face Valid Alcohol (FVA) .93, Face Valid Other Drugs (FVOD) .95 and Symptoms of Substance Misuse (SYM) .79 (Lazowski et al., 1998). Additionally, stability and discriminant validity estimates range from good to excellent for the entire measure as well as individual scales (see Lazowski et al., 1998 for more detailed information). In Study 3, participants were approached outside of a county courthouse in the Midwest immediately prior to their court appearance. For both groups, this was their first court appearance for the current offense. Groups were obtained separately, as the court docket was scheduled for DUI-only offenses on days of the week separate from other offenses. A short interview was conducted with each participant in order to verify that participants in the DUI group were indeed called to court for a recent DUI offense, and that participants in the Traffic Court group did not have any history of a DUI offense for which they had been ticketed and/or arrested. Any potential participant in the Traffic Court group with a previous history of a DUI offense was excluded from participation in this study. All participants signed a written informed consent form that was keyed to a unique ID number that matched the BADDS each participant completed. While there are certainly many people not arrested for DUI who nevertheless have a history of drinking and driving, these two groups were selected on the basis of having been identified by the legal system for Drinking and Driving Behaviors. In addition we selected the traffic offenders as a comparison group rather than nonoffender participants in an attempt to control variance attributable simply to breaking the law. 3. Results 3.1. Study 1 Means and standard deviations for BADDS scale scores are displayed in Table 2. In order to estimate the internal reliability of the BADDS, Cronbach's alpha was calculated on each scale. Findings indicated the following alpha coefficients: Rationalizations for Drinking and Driving α = .91, Likelihood of Drinking and Driving α = .94, Drinking and Driving Behaviors α = .80, and Riding with a Drinking Driver α = .79.

Table 2 Means and standard deviations of BADDS scales by study and group Study and group

BADDS scale RD

LD

DB

RB

Study 1 Study 2 Study 3 DUI Traffic court

12.80 (9.43) 13.70 (10.94)

21.24 (13.36) 18.84 (14.65)

3.93 (5.30) 3.05 (5.93)

3.56 (5.11) 3.18 (5.72)

17.49 (12.22) 10.07 (10.47)

22.86 (15.56) 12.56 (15.18)

3.32 (6.51) 1.29 (3.40)

4.72 (7.97) 1.99 (4.22)

Note: Standard deviations are provided within parentheses. DUI = DUI offender participant group. Traffic Court = group of participants with no DUI arrest history. RD = Rationalizations for Drinking and Driving scale. LD = Likelihood of Drinking and Driving scale. DB = Drinking and Driving Behaviors scale. RB = Riding Behaviors with a Drinker Driver scale. For Study 2, scale scores for Time 1 are reported.

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

859

The concurrent criterion-related validity of the BADDS scale scores was also examined by calculating the correlations between the reported attitudes toward drinking and driving (Rationalizations for Drinking and Driving and Likelihood of Drinking and Driving scale scores) and the BADDS measures of behaviors in the previous month (Drinking and Driving Behaviors score and the Riding Behaviors with a Drinking Driver score). Significant positive correlations with Drinking and Driving Behaviors scores were observed for both the Rationalizations for Drinking and Driving scores (r = .37, p b .01) and for the Likelihood of Drinking and Driving scores (r = .57, p b .01). The BADDS attitudes scores also showed significant positive correlations with the Riding Behaviors score: Rationalizations for Drinking and Driving (r = .29, p b .01), Likelihood of Drinking and Driving score (r = .41, p b .01). 3.2. Study 2 Means and standard deviations for BADDS scale scores are displayed in Table 2. Internal reliability of the BADDS scales was examined in this sample as well. Findings indicated the following Cronbach's alpha coefficients: Rationalizations for Drinking and Driving α = .92, Likelihood of Drinking and Driving α = .95, Drinking and Driving Behaviors α = .86, and for the Riding Behaviors with a Drinking Driver α = . 90. Test–retest reliability of the BADDS scales was estimated by calculating the zero order correlations between participants' scores on the first and second administrations. Results indicate that the test–retest reliability for the Rationalizations for Drinking and Driving (r = .87), Likelihood of Drinking and Driving (r = .89), Drinking and Driving Behaviors (r = .69) and Riding Behaviors with a Drinking Driver (r = .83) ranged from satisfactory to excellent. In order to determine whether the internal and test–retest reliabilities calculated in Studies 1 and 2 were possibly inflated due to some participants being under the legal drinking age, we examined alpha coefficients and test–retest correlations for responses provided by participants who were less than 21 years old and for those who were 21 or older in both studies (see Table 3). Findings were comparable for both groups for the internal reliability of the subscales. In addition, retest coefficients in study 2 were

Table 3 Cronbach's alpha and test–retest correlations for Studies 1 and 2 by age group Study and age group

Study 1 Cronbach's alpha Under 21 years 21 years and older Study 2 Cronbach's alpha Under 21 years 21 years and older Study 2 test–retest Under 21 years 21 years and older

BADDS scale RD

LD

DB

RB

.87 .92

.91 .95

.66 .81

.85 .78

.92 .93

.95 .95

.84 .92

.92 .82

.89 .73

.87 .91

.52 .91

.82 .88

Note: RD = Rationalizations for Drinking and Driving scale. LD = Likelihood of Drinking and Driving scale. DB = Drinking and Driving Behaviors scale. RB = Riding Behaviors with a Drinker Driver scale.

860

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

also examined for both age groups. Overall the findings do not indicate greater consistency for the older than younger participants. Construct validity of the BADDS scale scores was investigated by correlating them with measures of substance misuse on the SASSI-3. Table 4 displays the correlations between the BADDS scales and the SASSI-3 FVA, FVOD, and SYM scales. These three SASSI-3 scales were chosen for analysis because they are the most directly related to substance misuse and the items are phrased in a face valid manner. The SASSI-3 Manual goes on to describe these scales as “FVA, FVOD, and SYM address substance misuse in a direct manner; they reflect common consequences, causes, and correlates of substance dependence” (Miller & Lazowski, 1999, p. 16). As can be seen in Table 4, each of the BADDS scales showed significant positive correlations with the three SASSI-3 scales (alpha levels were corrected using the Bonferroni procedure, with alpha required for significance set at α = .006). As one would expect, positive correlations between the BADDS scales and the SASSI-3 measure of misuse of drugs other than alcohol (FVOD) were significant but generally slightly lower. The associations between BADDS attitudes scores at Time 1 (T1) with subsequent Drinking and Driving Behaviors was investigated by calculating the correlations between the two attitude scales at T1 (Rationalizations for Drinking and Driving and Likelihood of Drinking and Driving) and the impaired driving and riding behaviors reported 4 weeks later on the second administration of the BADDS (T2). Significant positive correlations were observed between Rationalizations for Drinking and Driving (T1) and Drinking and Driving Behaviors (T2) (r = .50, p b .01), and between Likelihood of Drinking and Driving (T1) and Drinking and Driving Behaviors (T2) (r = .50, p b .01). Significant positive correlations also were found between Rationalizations for Drinking and Driving (T1) and Riding Behaviors with a Drinking Driver (T2) (r = .48, p b .01), and between Likelihood of Drinking and Driving (T1) and Riding Behaviors with a Drinking Driver (T2) (r = .46, p b .01). Together these findings provide evidence of the predictive validity of BADDS scale scores. The ability of BADDS attitudes scores to predict subsequent Drinking and Driving Behaviors, over and above a measure of problem drinking, was also investigated by conducting a sequential regression analysis with Drinking and Driving Behaviors at T2 as the criterion variable. FVA scores from the SASSI-3 at T1 were entered in a single block as a predictor variable, with participants' Rationalizations for Drinking and Driving and Likelihood of Drinking and Driving at T1 entered together into the equation in a second block. This analysis was conducted to address whether the BADDS attitude scores provide unique predictive utility over and above self-reported problematic drinking behaviors as assessed by the FVA on the SASSI-3. For the criterion variable of Drinking and Driving Behaviors at T2, findings showed the final regression model was significant, F(3, 154) = 21.13, p b .01, R2 = .29, with an effect size of R2 change = .12 Table 4 Correlations between BADDS scale scores and selected SASSI-3 scale scores BADDS scale

Rationalizations for Drinking and Driving Likelihood of Drinking and Driving Drinking and Driving Behaviors Riding Behaviors with a Drinking Driver

SASSI-3 scale FVA

FVOD

SYM

.39 .52 .47 .41

.22 .30 .28 .24

.39 .45 .49 .43

Note: SASSI-3 = Substance Abuse Subtle Screening Inventory, 3rd Edition. FVA = Face Valid Alcohol. FVOD = Face Valid Other Drugs. SYM = Symptoms of Substance Misuse. All correlations were statistically significant beyond p b .006.

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

861

Table 5 Sequential regression of the Time 1 Rationalizations for Drinking and Driving, Likelihood of Drinking and Driving, and Face Valid Alcohol Scores on the Time 2 Drinking and Driving Behaviors Scores for Study 2 Variable Block 1 Face Valid Alcohol Block 2 Rationalizations for Drinking and Driving Likelihood of Drinking and Driving

B

SE B

Β

.15

.11

.10

.14 .10

.05 .04

.28** .24*

sr2 .12** .17**

Note: * = p b .05, ** = p b .01. Face Valid Alcohol entered into the model in a first block while Rationalizations and Likelihood of Drinking and Driving were entered together into the model in the second block. sr2 = incremental R square change.

for the first block (Face Valid Alcohol), and R 2 change = .18 for the second block (Rationalizations for Drinking and Driving and Likelihood of Drinking and Driving). It should be noted that both BADDS attitude scores were significant predictors (p b .01) of subsequent Drinking and Driving Behaviors in this model, beyond the predictive utility afforded by a measure of acknowledged problem drinking (FVA). A similar regression analysis was conducted with Riding with a Drinking Driver scores at T2 as the criterion variable. Again, findings showed the final regression model was significant, F(3, 152) = 16.98, p b .01, R2 = .25, with an effect size of R2 change = .08 for the first block (Face Valid Alcohol), and R2 change = .18 for the second block (Rationalizations for Drinking and Driving and Likelihood of Drinking and Driving). In this model, Rationalizations for Drinking and Driving was a significant predictor, while Likelihood of Drinking and Driving approached significance (p = .06). Tables 5 and 6 display findings for these regression analyses. In sum, results indicate that the BADDS attitude scores are predictive of future Drinking and Driving Behaviors as well as Riding with a Drinker Driver, over and above a measure of problematic drinking behavior in and of itself, as measured by the SASSI-3 Face Valid Alcohol scale. 3.3. Study 3 To investigate internal reliability of the BADDS scale scores in adult samples of DUI and traffic offenders, Cronbach's alpha coefficients were again computed. Findings indicated the following alpha coefficients: Rationalizations for Drinking and Driving α = .91, Likelihood of Drinking and Driving α = .96, Drinking and Driving Behaviors α = .87, and Riding with a Drinking Driver α = .87, replicating earlier findings of high internal consistency for the BADDS scales. Table 6 Sequential regression of the Time 1 Rationalizations for Drinking and Driving, Likelihood of Drinking and Driving, and Face Valid Alcohol Scores on the Time 2 Riding with a Drinker Driver Scores for Study 2 Variable Block 1 Face Valid Alcohol Block 2 Rationalizations for Drinking and Driving Likelihood of Drinking and Driving

B

SE B

Β

.06

.15

.03

.20 .11

.07 .06

.30** .22

sr2 .08** .18**

Note: * = p b .05, ** = p b .01. Face Valid Alcohol entered into the model in a first block while Rationalizations and Likelihood of Drinking and Driving were entered together into the model in the second block. sr2 = incremental R square change.

862

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

To test the discriminant validity of the BADDS, a MANOVA was calculated with group (DUI versus Traffic Court) as the independent variable and the four BADDS scales (Rationalizations for Drinking and Driving, Likelihood of Drinking and Driving, Drinking and Driving Behaviors, and Riding Behaviors with a Drinking Driver) as the dependent variables. Gender was also added as another independent variable due to the observed differences in gender composition of the DUI and Traffic Court groups, in order to investigate possible gender by group interaction effects on any of the dependent variables. Means and standard deviations on the BADDS scales are reported for each group separately in Table 2. Due to missing items on some of the dependent variables the valid N for this analysis is 304. Results indicated that the overall MANOVA was significant for the independent variables of group (Lambda(4, 297) = .91, p b .01) as well as gender (Lambda(4, 297) = .96, p b .01), but the interaction between group and gender was not significant (Lambda(4, 297) = .99, p = .67). Follow-up univariate ANOVAs indicated that the DUI and Traffic Court groups differed on the Rationalizations for Drinking and Driving scale scores, F(1, 300) = 22.19, p b .01, the Likelihood of Drinking and Driving scale scores, F(1, 300) = 24.64, p b .01, the Drinking and Driving Behaviors scale scores, F(1, 300) = 5.27, p b .05, and the Riding Behaviors with a Drinking Driver scale scores, F(1, 300) = 10.79, p b .01. Specifically, the DUI group displayed higher means on all four scales compared to the Traffic Court group, indicating greater endorsement of rationalizations and Likelihood of Drinking and Driving and higher incidence of drinking and driving and riding with a drinker behaviors. Effect sizes for the Rationalizations for Drinking and Driving, Likelihood of Drinking and Driving, Drinking and Driving Behaviors, and Riding Behaviors with a Drinking Driver scale scores were calculated as well (partial eta squared = .07, .08, .02, and .04 respectively). In fact, scores for all four scales were approximately double for the DUI group when compared to the Traffic Court group. Therefore, BADDS scale scores appear to discriminate between recent DUI offenders and those who have no self-reported history of DUI arrest. To further examine the role that level of drinking plays in defining Drinking and Driving Behaviors, and to test whether those arrested for DUI offenses differ from traffic offenders on both levels of alcohol consumption specified in the criterion measures on the BADDS, an additional MANOVA was calculated. In this analysis the DUI versus Traffic Court groups and gender were the independent variables and the individual items on the BADDS driving and riding criterion measures were the dependent variables. Recall that the items in the dependent measures ask for the number of driving and riding behaviors after consumption of ‘one or two’ or ‘three or more’ drinks in the previous hour. Findings showed the overall MANOVA was significant for the DUI versus Traffic Court group variable (Lambda(4, 304) = .96, p b .05), but not for gender (Lambda(4, 304) = .98, p = .14) or the interaction between gender and group (Lambda(4, 304) = .99, p = .71). Further, univariate ANOVAs indicated significantly higher incidence of both driving and riding with a driver after consumption of ‘one or two’ and ‘three or more’ drinks for the DUI group as compared to the Traffic Court group (all p values b .05). Thus questions on the BADDS related to driving after having consumed even moderate amounts of alcohol (one or two drinks) discriminates DUI offenders from those with no history of DUI, as does driving after higher levels of alcohol consumption. 4. Discussion Drinking and driving behavior continues to be a nationwide problem that has significant consequences to society in terms of both human life and financial cost (NHTSA, 2006). It is therefore even more notable that to date there has been very little research focused on developing a measure assessing attitudes and

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

863

behaviors related to drinking and driving. The current study offers evidence of the psychometric properties of the BADDS as a promising self-report measure of attitudes and behaviors related to drinking and driving. Three studies were conducted to examine the psychometric properties of the BADDS, including an examination of its internal reliability, test–retest reliability, construct validity, discriminant validity, and criterion-related validity. Overall, evidence was presented which supported the reliability and validity of BADDS scores. Given these results, the BADDS has promise as an instrument for measuring Rationalizations for Drinking and Driving as well as the Likelihood of Drinking and Driving given certain circumstances. Previous research has combined the Rationalizations and Likelihood scales to measure the efficacy of a prevention program for drinking and driving (Jewell et al., 2004; Jewell, & Hupp, 2005). In both studies, participants' scores on these scales showed immediate significant decreases. Clinicians could similarly use the scales to measure attitude change in treatment, and clinicians could also gather useful information by examining responses to individual items. For example, items on the Rationalizations for Drinking and Driving scale assess attitudes related to perceiving oneself as only slightly intoxicated, being the least intoxicated person in a car, and having no other way to get home, which have been shown to affect the decision to drink and drive in other research (Kulick & Rosenberg, 1999). Additionally, the items on the Likelihood of Drinking and Driving scale assess how important driving distance and amount of alcohol are in an individual's decision to drink and drive. Based on each individual's responses, any of these specific attitudes could be directly targeted in a treatment or prevention program. The items on the Drinking and Driving Behaviors scale and on the Riding Behaviors with a Drinking Driver scale would also be valuable for both program evaluation and for clinical applications. These items measure self-reported actual behavior over the past month. Thus, program evaluators could track behavior on a monthly basis, both before and after prevention programs, in order to examine and improve program effectiveness. Similarly, attitudinal change could be measured from pre to post intervention over any amount of time. The questions are also divided based on whether the respondent was the driver or was the rider with someone who had been drinking. Clearly, both behaviors are dangerous; however, some people may not perceive riding with a drinking driver to be as risky as driving under the influence. As drinking and driving continues to be a serious problem, there are many potential applications for a reliable and valid measure of related attitudes and behaviors. Previous studies measuring drinking and driving typically relied on only one or just a few questions. A recent exception is the Positive Expectancies for Drinking and Driving for Youth (PEDD-Y; McCarthy et al., 2006). The PEDD-Y has demonstrated initial reliability and validity as an attitude-based measure focusing on the perceived positive consequences of drinking and driving in youth. However, this measure was developed exclusively for use with youth, and all of the data collected in the study was self-reported. Alternatively, the BADDS focuses on rationalizations and self-reported Likelihood of Drinking and Driving, and many of the items were developed to identify maladaptive thinking patterns related to drinking and driving. In addition to data with college students, data was also presented from adult populations both with and without a history of DUI arrest. Evidence of DUI arrest provides a valuable supplement to the self-report data in this study. Previous research with the BADDS already provided initial evidence of reliability and validity with college students, and the BADDS has been shown to be sensitive to changes in attitudes following a prevention program for drinking and driving (Jewell et al., 2004; Jewell & Hupp, 2005). The current series of studies provide additional evidence of reliability and validity with college students, and the third study in this paper was the first to use the BADDS with adult DUI and traffic offender (not college) populations.

864

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

While these studies indicate the utility of the BADDS, several limitations exist. All of the college student data in the present studies, as well as the two previously published studies using the BADDS (Jewell et al., 2004; Jewell & Hupp, 2005), are from the same Midwestern university, which may limit the ability to generalize these results to other college students. Thus, future research should gather data on the BADDS drawn from a variety of academic and community institutions and geographic areas. Also, while the present studies used college students and adults of a wide range of ages, data from adolescents is needed before the scale can be used in prevention programs in schools. Social desirability could have also affected the responses of some participants in these studies, and future research should attempt to measure the influence of social desirability when completing the BADDS. To date, two studies (Jewell et al., 2004; Jewell & Hupp, 2005) have examined how well the BADDS measures change in attitudes and behaviors following prevention programs, although additional studies utilizing a variety of prevention tools would be helpful. Additional validity data on the predictive validity of the BADDS would also be valuable. Specifically, research focusing on the predictive validity of the instrument, especially as the BADDS may predict DUI recidivism, would lend support to using the BADDS as a screening instrument to link respondents with prevention or intervention programs. Cavaiola, Strohmetz, and Abreo (2007) have stressed the importance of identifying DUI offenders at risk for subsequent DUI offenses. In their study, however, neither the Michigan Alcoholism Screening Test (MAST) nor clinical scales of the Minnesota Multiphasic Personality Inventory-2 (MMPI-2) differentiated recidivists from non-recidivists. Therefore, assessment may also allow clinicians to better understand and identify those at risk for recidivism and develop appropriate interventions (Ouimet et al., 2007). While the BADDS has not yet been used in clinical settings to date, it may have potential as a valuable clinical tool. The BADDS may also have some value for use in drug courts by providing more information in the process in linking DUI offenders to appropriate interventions and predicting recidivism. Specifically, many courts require that DUI offenders undergo assessment prior to sentencing in order to inform the court as to the offender's need for treatment and risk to the community. Use of a measure such as the BADDS could certainly be informative to this process. Given the critical nature of drinking and driving and the impact on human life, research that seeks to provide reliable and valid assessments of DUI risk are sorely needed. The current findings show that the BADDS holds promise as a reliable and valid measure of DUI risk, with many potential applications for impaired driving prevention and intervention programs. References Cavaiola, A. A., Strohmetz, D. B., & Abreo, S. D. (2007). Characteristics of DUI recidivists: A 12-year follow-up study of first time DUI offenders. Addictive Behaviors, 32, 855−861. Cohn, L. D., Macfarlane, S., Yanez, C., & Imai, W. K. (1995). Risk-perception: Differences between adolescents and adults. Health Psychology, 14(3), 217−222. Finn, P., & Bragg, W. E. (1986). Perception of risk of an accident by young and old drivers. Accident Analysis & Prevention, 18, 289−298. Gerrard, M., Gibbons, F. X., Benthin, A. C., & Hessling, R. M. (1996). A longitudinal study of the reciprocal nature of risk behaviors and cognitions in adolescents: What you do shapes what you think, and vice versa. Health Psychology, 5, 344−354. Gibbons, F. X., Lane, D. J., Gerrard, M., Pomery, E. A., & Lautrup, C. L. (2002). Drinking and driving: A prospective assessment of the relation between risk cognitions and risk behavior. Risk, Decision and Policy, 7, 267−283. Greenberg, M. D., Morral, A. R., & Jain, A. K. (2005). Drink-driving and DUI recidivists' attitudes and beliefs: A longitudinal analysis. Journal of Studies on Alcohol, 66(5), 640−648.

J.D. Jewell et al. / Addictive Behaviors 33 (2008) 853–865

865

Higson, R. W., Heeren, T., & Winter, M. R. (1999). Preventing impaired driving. Alcohol Research & Health, 23(1), 31−39. Jewell, J., & Hupp, S. D. A. (2005). Examining the effects of fatal vision goggles on changing attitudes and behaviors related to drinking and driving. The Journal of Primary Prevention, 26(6), 553−565. Jewell, J., Hupp, S., & Luttrell, G. (2004). The effectiveness of fatal vision goggles: Disentangling experiential versus onlooker effects. Journal of Drug and Alcohol Education, 48(3), 63−84. Kulick, D., & Rosenberg, H. (1999). Assessment of university students' coping strategies and reasons for driving in high-risk drinking–driving situations. Accident Analysis and Prevention, 32, 85−94. Lazowski, L. E., Miller, F. G., Boye, M. W., & Miller, G. A. (1998). Efficacy of the Substance Abuse Subtle Screening Inventory3 (SASSI-3) in identifying substance dependence disorders in clinical settings. Journal of Personality Assessment, 71, 114−128. Mannering, F. L., Bottiger, W. K., & Black, K. L. (1987). Decisions relating to alcohol-impaired driving: An exploratory analysis. Accident Analysis and Prevention, 19(6), 487−495. McCarthy, D. M., Pedersen, S. L., & Leuty, M. E. (2005). Negative consequences and cognitions about drinking and driving. Journal of Studies on Alcohol, 66(4), 567−570. McCarthy, D. M., Pederson, S. L., Thompson, D. M., & Leuty, M. E. (2006). Development of a measure of drinking and driving expectancies. Psychological Assessment, 18, 155−164. Miller, F. G., & Lazowski, L. E. (1999). The Adult SASSI-3 Manual. The SASSI Institute Springville: IN. National Highway Traffic Safety Administration (n.d.). Our Nation's Highways 2000. Retrieved August 22, 2006 from http:// www.fhwa.dot.gov/ohim/onh00/onh2p1.htm National Highway Traffic Safety Administration (NHTSA, 2004). Traffic Safety Facts, 2004 Data: Alcohol, DOT HS 809 905. National Highway Traffic Safety Administration (NHTSA, 2006). Traffic Safety Facts Research Note: Motor Vehicle Crashes as a Leading Cause of Death in the United States, 2003. DOT HS 809 905. Ouimet, M. C., Brown, T. G., Nadeau, L., Lepage, M., Pelletier, M., Couture, S., et al. (2007). Neurocognitive characteristics of DUI recidivists. Accident Analysis and Prevention, 39(4), 743−750. Ozkan, T., Lajunen, T., & Summala, H. (2006). Driver behaviour questionnaire: A follow-up study. Accident Analysis and Prevention, 38, 386−395. Sayette, M. A. (1999). Cognitive theory and research. In K. E. Leonard, & H. T. Blane (Eds.), Psychological theories of drinking and alcoholism (pp. 247−291)., (2nd ed.) New York: Guilford Press. Schell, T. L., Chan, K. S., & Morral, A. R. (2006). Predicting DUI recidivism: Personality, attitudinal, and behavioral risk factors. Drug and Alcohol Dependence, 82, 33−40. Stacy, A. W., Bentler, P. M., & Flay, B. R. (1994). Attitudes and health behavior in diverse populations: Drunk driving, alcohol use, binge eating, marijuana use, and cigarette use. Health Psychology, 13(1), 73−85. Wiliszowski, C., Murphy, P., Jones, R., & Lacey, J. (1996). Determine reasons for repeat drinking and driving, National Highway Traffic Safety Administration, DOT HS 808 401. Yalisove, D. (2004). Introduction to alcohol research: implications for treatment, prevention, and policy. Boston: Pearson.