1 Drunk Driving: An Assessment of “What Works” in the Areas of Classification, Treatment, Prevention and Control Prepared for: The Council of Productivity and Management State of Maryland Department of Parole and Probation Drinking Driver Monitor Program Prepared by: James M. Byrne, Professor Department of Criminal Justice University of Massachusetts Lowell With the Assistance of: Andrea Vey, Heather Madruga, Jennifer Teagarden, and Danielle Sparks Graduate Program in Criminal Justice University of Massachusetts Lowell 2 Table of Contents Introduction and Overview: Searching for Solutions to the Drunk Driving Problem…….1 1. Treatment Programs for Drunk Drivers………………………………………………..2 1A. What Works in the Area of Treatment……………………………………….4 1B. What Does Not Work in the Area of Treatment……………………………..4 1C. Promising Developments Vis-à-vis Drunk Driver Treatment………………..6 1D. What is “Unknown” About the Treatment of Drunk Drivers: Directions for Future Research……………………………………………..11 2. Prevention Programs for Drunk Drivers: Individual and CommunityLevel Strategies……………………………………………………………………….15 3. The Identification and Classification of Drunk Drivers……………………………...20 4. Control-Oriented Strategies for Addressing the Drunk Driver Problem……………..25 4A. What Works in Controlling Drunk Drivers…………………………………26 4B. What Does Not Work in Controlling Drunk Drivers…………….…………26 4C. Promising Developments Vis-à-vis Drunk Driving Control………………..27 4D. What is “Unknown” About Control-Oriented Sanctions: Directions for Future Research……………………………………………..29 Concluding Comments: An Application of our “What Works” Review to MADD’s Rating the States 2002 Survey……………………………………………...34 Recommendations for Change…………………………………………………………...36 Appendix A………………………………………………………………………………38 3 Introduction and Overview: Searching for Solutions to the Drunk Driving Problem The purpose of the following review is to identify promising solutions to the drunk driving problem in this country, focusing on four broad areas of inquiry: (1) Can effective treatment programs for drunk drivers be identified? (2) Can effective prevention strategies for drunk drivers be identified? (3) Can effective identification strategies for drunk drivers (utilizing classification and prediction techniques) be identified? (4) Can effective control strategies for drunk drivers (utilizing administrative and/or criminal Justice interventions) be identified? To complete this analysis, we conducted a review of all the available published research on drunk driving released since 1990. Each of the articles we identify in our review is included in the “what works” compendium accompanying this report. Not surprisingly, the quality of the empirical research published during our review period (1990-2002) is generally weak, with few experimental studies conducted on this important problem. To make sense of this body of research, we have adopted a strategy recommended by Welsh and Farrington (2003) in their recent article, “Toward an Evidence–Based Approach to Preventing Crime”. First, we have ranked the “quality” of each of the research studies included in this review on a scale of 1 (weakest) to 5 (highest), utilizing the same scaling criteria used in the recent research reviews completed for the National Institute of Justice by the Campbell Collaboration Crime and Justice Group (see Appendix A for a description of review criteria). Next, we used these rankings to organize our assessment of what works, what does not work, what is promising, and what is unknown in the area of drunk driving (see Appendix A for a definition of key review terminology). Finally, we have attempted to examine the State of Maryland’s drunk driving strategy by providing a critical review of MADD’s “Rating the States 2002” report on drunk driving, which includes an overall grade for Maryland’s drunk driving strategy (C) along with a breakdown of grading in each of the following areas: (1) state political leadership, (2) statistics and records, (3) law enforcement, (4) 4 administrative measures/criminal sanctions, (5) regulatory control/availability, (6) youth legislation, prevention, and education, (7) victim issues, (8) laws, and (9) fatality trends. (Note: the full text of the MADD report, along with a breakdown of what each state in the country is currently doing to address the drunk driving problem, is included in our “what works” compendium). Utilizing the results of our review of “what works” in the areas of treatment, classification, prevention, and control, we offer our recommendations for changes in MADD’s grading criteria and then discuss the implications of our findings for drunk driver policies and practices in Maryland. 1. Treatment Programs for Drunk Drivers According to a recent Bureau of Justice Statistics, Special Report: DWI Offenders Under Correctional Supervision (June 1999:1). “In 1997, an estimated 513,200 offenders were on probation or in jail or prison for driving while intoxicated by alcohol (DWI): 454,500 on probation, 41,100 in Jail, and 17,600 in state prison”. It would be a mistake to assume that every offender convicted of a DWI will receive some form of treatment (e.g. education, self-help via Alcoholics Anonymous (AA), outpatient counseling, detox, and/or residential treatment). In fact, a significant proportion of DWI offenders in prison and on probation do not receive any form of treatment. According to the BJS special report on this subject (June 1999:1), 46% of DWI offenders on probation reported receiving alcohol treatment, while 62% reported that they had participated in an alcohol-related self-help program. For DWI offenders in jail, only 4% reported receiving alcohol treatment, while 17% reported participation in a self-help program. One possible explanation for this “lack of treatment” finding is that probation and jail administrators are taking a “triage” approach to limited treatment resources, reserving their availability to a subgroup of “problem drinkers”. If this is the case, then the development of reliable (and valid) classification tools that distinguish problem from social drinkers is critical. Of course, another possible explanation for the failure to provide alcohol treatment to all convicted drunk drivers is that probation and jail administrators simply do not believe that existing alcohol treatment programs are effective recidivism reduction strategies. Finally, it is certainly possible that probation and jail administrators are considering a 5 combination of factors, (including treatment type, availability, and cost, treatment effectiveness, and/or DWI offender amenability to treatment) when they develop specific DWI policies and practices. In terms of the offender amenability (or readiness for change) issue, research that compares the relative impact of voluntary versus mandatory treatment options for DWI offenders would certainly be important to review. Unfortunately, none of the research studies we reviewed empirically evaluated this issue. Table 1 provides an overview of the available evaluation research on the effectiveness of various forms of treatment for convicted DWI offenders. For the purposes of this review, we have defined treatment broadly, including in our review any planned intervention designed to reduce the offender’s reliance on alcohol and, in the process, to reduce DWI recidivism. The National Academy of Sciences relied on the same definition to organize their review of the effectiveness of offender rehabilitation programs two decades ago. Among the ten program evaluations and four research reviews included in table 1 are two evaluations of inpatient alcohol treatment (Kelly, Myers and Brown, 2002; Shapiro, 2001), six evaluations focusing on outpatient treatment and/or education (Taxman and Piquero, 1998;Wells-Parker, 1994; Pratt, Holsinger and Latessa, 2000; Tonigan, Miller, Juarez, and Villanvera, 2002; Deyoung, 1997; VanderWalde, Urgenson and Weltz, 2002), Three evaluations of victim impact panels (C’DeBaca, Lapham, Liang, and Skipper, 2001, Polacsek, Rogers, Woodall, Delaney, Wheeler and Rao, 2001; Fors and Rojack, 1999), one evaluation that combined electronic monitoring and mandatory participation in outpatient treatment (Courtright, Berg, and Mutchnick, 2000) and one evaluation that examined the importance of matching offender problems to appropriate impatient and outpatient treatment programs (Project Match Research Group, 1996). Although the overall quality of the evaluation research studies included in this review of treatment effectiveness is quite low, it is possible to offer the following assessment of what works, what doesn’t work, what is promising and what is unknown about the treatment of drunk drivers. 6 (1A) What Works in the Area of Treatment The systematic review criteria for determining program effectiveness employed by Sherman, et al (1997) requires that “for a program to be classified as working, there must be a minimum of two level 3 studies with significance tests demonstrating effectiveness and the preponderance of evidence in support of the same conclusion” (Sherman, et. al, 1997 as summarized in Welsh and Farrington (2003:170)). Examination of table 1 reveals that there were two level 5 studies, seven level 2 studies, and 1 level 1 study chosen for review. The two level 5 studies were randomized field experiments on the effectiveness of two quite different drunk driving “treatment” strategies. In the first level 5 study, Polacsek et al. (2001) conducted a randomized field experiment on the effect of victim impact panels and DWI schools on both offender movement through “stages of change” in their attitudes toward drinking and driving and on subsequent DWI recidivism. In the second level 5 study, Project Match evaluators (1996) examined the effectiveness of three separate treatment modalities in both outpatient and aftercare settings. The three treatment modalities included: (1) cognitive behavioral coping skills therapy; (2) motivational enhancement therapy; and (3) twelve-step facilitation. Since only one of these field experiments reported positive overall findings, we are left one “good” evaluation, plus several weak level 2 studies on which to base our assessment of what works in the area of drunk driving treatment. For this reason, we conclude that there is no sound body of recent empirical evidence (1990-2002) available to determine the effectiveness of various forms of offender treatment for drunk drivers. (1B) What Does Not Work in the Area of Treatment Further review of table 1 reveals that three evaluations of the effectiveness of victim impact panels have been conducted in the past decade (actually 1990-2002). As we noted earlier, one of these evaluations was a randomized field experiment (Polacsek, et al, 2001), while the other two evaluations employed (level 2) quasi-experimental designs. The authors of the randomized field experiment concluded that “participation by DWI offenders in a Victim Impact Panel [following attendance in a DWI school] in 7 comparison with participants in a DWI school only did not (1) more effectively move them though the stages-of-change toward maintenance of drunk driving prevention, or (2) result in lower rates of DWI recidivism (Polacsek, et. al., 2001:354). A similar finding was reported by C’deBaca, Lapham, Liang and Skipper (2001) in their assessment of 5,238 first-time and 1,464 repeat convicted DWI offenders who were convicted of a DWI offense in New Mexico between 1989 and 1994 and who were subsequently interviewed by the Lovelace Comprehensive Screening Program in Bernalillo, New Mexico. Comparisons were made between those DWI offenders who reported (self-report) being referred to a victim impact panel (67% of all respondents) and those DWI offenders who were not referred to a VIP (33%). The authors of this report concluded that referral to a victim impact panel (VIP) was not associated with drunk driving recidivism for either male or female first-time offenders or male repeat offenders. Among female repeat offenders, it was found that referral to a VIP actually doubled the probability of re-arrest. A different view of the importance of victim impact panels is offered by Fors and Roject (1999), in their examination of DWI recidivism patterns (12 month follow-up) among two cohorts of DWI offenders convicted in Athens/Clarke County, Georgia. Members of one cohort of offenders (n=404) were required to attend a VIP presentation as part of their sentence (i.e. those convicted between February and July 1994), while offenders included in the comparison group (n=431) were arrested for drunk driving in the same county prior to the initiation of victim impact panels in this jurisdiction. Although questions can certainly be raised about this type of pre/post (no control group) comparison, the differences between the recidivism rates of the VIP and pre-VIP groups (5.94% vs. 15.08%) do suggest a possible intervention effect. Overall, however, the weight of the empirical evidence certainly favors the conclusion that victim impact panels – as currently designed – may not have the desired effect on subsequent offender behavior; in some cases, attendance at a victim impact 8 panel may actually exacerbate the problem behavior. C’deBaca, et al. (2001:620) offer an explanation for this finding: “VIPs are thought to influence DWI offenders on an emotional level, by means of confronting them with the consequences of drinking and driving… There is evidence that confrontational approaches are ineffective in the treatment of alcohol problems (Hodgson, 1994; Miller, et.al.1993, 1995) and conceivably could result in the opposite of the intended effect. VIP participation may exacerbate feelings of guilt and shame, which may in turn lead to increased alcohol use and, ultimately, re-offense”. Although it is certainly possible that Victim Impact Panels could be redesigned in ways that would result in more positive findings (e.g. changes in panel composition length, format, timing, etc.), it appears that other intervention strategies offer much more promising results. We discuss these treatment-based strategies in the following section. (1C) Promising Developments Vis-À-Vis Drunk Driver Treatment According to the criteria developed by Sherman and his colleagues (and summarized in Welsh and Farrington, 2003), “promising” strategies are defined in the following manner: “For the classification of promising, at least one level 3 study is required with significance tests showing effectiveness and the preponderance of evidence in support of the same conclusion” (170). Our review of the evaluation research identified one level 5 study, which reported positive treatment-related outcomes for three separate treatment modalities in both outpatient and aftercare settings (see Project Match, 1996), along with several level 2 studies reporting positive treatment outcomes (see Taxman and Piquero (1998), Pratt, Holsinger and Latessa (2000), and Deyoung, 1997). The results from this latest wave of research on drunk driving are consistent with the results of meta-analyses of treatment effectiveness conducted by Wells-Parker (1994) and 9 Wells-Parker and Bangert-Drowns (1995), which examined program evaluations completed before 1992. Overall, it appears that some form of mandatory treatment delivered in an outpatient setting, combined with license restrictions (i.e. suspension or revocation) represents the most promising strategy for addressing our drunk driving problem. Deyoung (1997) offers a succinct summary of the current research on both license restrictions and treatment: “License curtailment reduces driving exposure and prompts more cautious driving (Ross and Gonsales, 1988; Sadler, et al., 1991), while alcohol education/treatment programs allow offenders to address their alcohol problems. There are both logical and empirical reasons for combining the two” (995). Wells-Parker and Bangers-Drowns (1995) have estimated that the impact of providing some form of alcohol education/treatment to offenders was an 8-9 percent reduction in drinking/driving recidivism. They go on to observe that the somewhat modest recidivism reduction effect identified in their meta-analysis of 215 drunk driving evaluations published between 1955-1992 is probably a conservative estimate, due to the large number of “treatment’ programs that relied on alcohol education as the sole treatment modality. If their assessment is correct, jurisdictions that offer both alcohol education and treatment to offenders should experience even greater reductions in both drunk driving recidivism and alcohol-related crashes/fatalities. The evaluations included in table 1 offer a preliminary assessment of the likely impact of a broad range of treatment-based drunk driver intervention strategies. Pratt, Holsinger, and Latessa (2000) conducted a long-term (10 year) follow-up study of the recidivism patterns of a cohort of chronic drunk driving offenders who participated in Turning Point, a residential treatment program that combined both cognitive and behavioral treatment modalities, and an aftercare component. They found that participation in this alcohol treatment program resulted in reductions in recidivism between 10 and 30 percent. Similarly, Shapiro (2001) examined the research on the therapeutic community (TC) model’s use with drunk drivers and substance abusers in both residential and outpatient settings and found that the residential TC model is the 10 “most effective treatment available for long-term criminal alcohol and substance abusers” (28). Unfortunately, most jurisdictions appear to spend little time and only a small proportion of drunk driving treatment resources on the residential treatment of the chronic drunk driver. For many jurisdictions across the country, alcohol treatment comes in one of the following three generic forms: (1) alcohol education programs, (2) self-help programs, (such as AA), and (3) alcohol treatment programs. In most instances, participation in one or more of these programs is mandatory (Wells-Parker, 1994). By far the most common sanction for first time convicted drunk drivers is participation in an alcohol education program. Taxman and Piquero (1998) found that for offenders in the Maryland program they evaluated, offenders who completed an alcohol education program had a recidivism rate that was 22 percent lower than for comparable offenders who did not attend these programs. Deyoung’s (1997) evaluation revealed that for first time drunk drivers in California, a combination of mandatory education (10 hours education on the effects of alcohol on the body, the ability to drive, and applicable DWI laws) and counseling (10 hours of individual counseling and 10 hours of (group) education/counseling) is effective as a recidivism reduction strategy, especially when used in conjunction with license restrictions. Similar findings were reported for second-time DWI offenders who were “sentenced by the court to attend an 18 month SB38…program. SB38 programs are 18 months in length and require at least 12 hours of education, 52 hours of counseling and bi-weekly face-to-face interviews” (Deyoung, 1997:990). These findings are consistent with the results reported in Wells-Parker’s meta-analysis of treatment programs (WellsParker, 1994), where multimodal treatment programs were found to be more effective recidivism reduction strategies (by 10 percent or more) than any one of the following “stand-alone” sanctions: (1) alcohol education, (2) psychotherapy or counseling, (3) probation monitoring and follow-up by treatment providers, or (4) aftercare by treatment providers. In addition to alcohol education, many jurisdictions mandate participation in “self-help” programs, such as alcoholics anonymous (AA). This type of mandatory 11 referral is not surprising, especially given the resource constraints facing most corrections administrators today. AA programs are inexpensive to run and they are readily available in most jurisdictions. Kelly, Myers and Brown (2002) report that “AA is the most commonly accessed source of help for an alcohol-related problem, with an estimated 9% of the U.S. population having attended an AA meeting, 3% for help with their own problems” (293). However, even a cursory review of the AA literature reveals that it is one thing for an offender to attend an AA meeting; it is quite another for an offender to affiliate him/herself with the 12-step recovery process. Stated simply, we can mandate attendance, but not affiliation. As we highlight in table 1, the affiliation issue is critical to understand, especially when it is viewed in the broader context of matching different types of offenders to appropriate treatment programs. For example, Kelly, Myers and Brown (2002) interviewed adolescents (14-18) who participated in an inpatient alcohol abuse treatment program and at 3 and 6 month intervals after discharge in order to examine the relative effects of 12-step attendance/affiliation on subsequent alcohol problems. They concluded that “more severely substance-involved youth were more motivated for abstinence and more likely to attend and affiliate with 12 step groups (2002:293). They recommend that, “for youths with less severe problems, focus could be placed on increasing motivation for abstinence and 12-step group attendance” (304), and on developing alternative strategies for addressing alcohol-related problems that do not involve attendance at an AA meeting. Importantly, Kelly, Myers and Brown (2002) have also reviewed the adult-based research on the impact of AA programs and their observations are worth noting here. First, many researchers make the same distinctions between AA attendance and AA affiliation when discussing the impact of AA on adults with alcohol problems, although there are certainly mixed results on this topic (see, e.g. Snow, et. al., 1994; Tonigan, et. al., 1996, 2000). Second, research on adults with alcohol problems also suggests that the more severe substance abusers are the most likely to affiliate with the 12-step AA process. And third, “motivation for change” is critical to the success of AA programs for both adults and juveniles. 12 Focusing on the application of these findings to statewide drunk driving initiatives, it seems obvious to us that probation departments play a critical role in the development of effective drunk driving interventions for two basic reasons: (1) it is probation personnel who will be responsible for assessing the extent of the offender’s drinking problem and making the appropriate referral, which will not include attendance at self-help programs for a significant number of offenders convicted of drunk driving in most states; and (2) it is probation personnel who are the most likely to have an impact on an individual offender’s motivation for change, utilizing a combination of formal and informal control techniques (Taxman and Byrne,2001). In Maryland, probation officers are currently being trained on the application of motivational interviewing techniques to the probation supervision and offender change process. This strategy certainly makes sense for offenders convicted of drunk driving. However, for probation-based drunk driving sanctions to be effective, it is essential that probation departments focus less on offender “processing” strategies (e.g. counting the number of times and offender attends AA) and more on the real work of offender change (e.g., helping the offender recognize (1) the extent of his/her drinking problem, (2) the need to address this problem, and (3) the ability to change). One final caveat on the effectiveness of AA programs needs to be considered. Our review suggests that AA programs may not be particularly effective for many female offenders convicted of drunk driving, because women with drinking problems are often categorized as multiple problem offenders (alcohol, drugs, mental illness), requiring a multimodal treatment setting (VanderWalde, Urgenson and Weltz, 2002). Tonigan, Miller, Juarez and Villanueva (2002) have found that Hispanic clients respond differently to twelve-step AA programs than other groups. For Hispanic clients, attendance at AA meetings did not predict either degree of affiliation with the AA program or subsequent drinking behavior. This finding highlights the futility of monitoring an offender’s treatment progress by simply collecting AA attendance slips, since for some groups of offenders, attendance may not be related to affiliation. Despite these limitations, it appears that AA and other self-help programs represent one of several “treatment” options that need to be considered for DWI offenders with drinking problems. 13 In addition to alcohol education and participation in self-help groups, our research review identified several “promising strategies” in the area of alcohol treatment. Taxman and Piquero (1998) found that offenders receiving alcohol treatment conditions (including either inpatient or outpatient treatment) had a 17 percent lower risk of recidivism than similar offenders who did not receive this condition (of probation). Pratt, Holsinger and Latessa (2000) also reported significant, long term reductions in recidivism among a cohort of chronic DUI offenders who completed a residential treatment program (Turning Point) that incorporated both cognitive and behavioral treatment modalities to each phase of treatment, including aftercare. These findings are consistent with the metaanalyses conducted by Wells-Parker (1994-1995), who concluded that jurisdictions interested in reducing drunk driving recidivism need to develop multimodal treatment strategies (e.g. education, counseling, probation monitoring and follow-up, aftercare) utilizing both residential and outpatient programs. (1D) What is “Unknown” About the Treatment of Drunk Drivers: Directions for Future Research Our review of the research on treatment strategies vis-à-vis the drunk driver reveal only a handful of comparatively weak evaluations of specific treatment programs. At minimum, future treatment-focused research needs to examine a number of basic drunk driving program design issues, such as treatment type, treatment length, and treatment setting/size. In addition, we need much better evaluation research on the individual and/or combined effects of participation in alcohol education, self-help groups (such as AA), and various forms of inpatient and outpatient alcohol treatment. Finally, we know remarkably little about the ability of local probation officers, in conjunction with treatment providers, to develop/affect an offender’s motivation to change (Taxman and Byrne, 2001). It is within this broad context of offender change that proactive community supervision strategies need to be systematically evaluated. 14 Table 1: Evaluations of Treatment Programs for Drunk Drivers 1990-2002 Author (year) (1) Taxman and Piquero (1998) Scientific Method Score 2 (2) Kelly, Meyers, and Brown (2002) 1 (3) VanderWalde, Urgenson and Weltz (2002) * (4) Tonigan, Miller, Juarez, and Villanueva (2002) 2 (5) C’DeBaca, Lapham, Liang and Skipper (2001) 2 Key Findings For offenders receiving alcohol education, the risk of recidivism was 22 percent less (than for offenders without alcohol education). Offenders receiving alcohol treatment had a 17 percent lower risk of recidivism than offenders without this condition. Among adolescents (14-18) in an inpatient treatment program for alcohol abuse, motivation (to change) was found to influence the relationship between 12-step affiliation and future substance use outcome. Women with alcoholism/drinking problems are often influenced by a range of biological and psychological factors that must be addressed in a multimodal treatment setting (e.g. individual counseling, group counseling, support groups). Hispanic clients assigned to twelve-step affiliation (TSA) reported significantly less AA attendance 6 months after treatment than non-Hispanic whites Overall, specific AA-related practices were not associated with increased abstinence. Victim Impact Panel (VIP) referral was not associated with drunk driving recidivism for either male or female first time 15 (6) Polacsek, Rogers, Woodall, Delaney, Wheeler, and Rao (2001) 5 (7) Coutright, Berg, and Mutchnick (2000) 2 (8) Fors and Rojack (1999) 2 (9) Wells-Parker (1994) * offenders. Female repeat offenders referred to VIP’s were twice as likely to be rearrested as female repeat offenders who were not referred. Participation by DWI offenders in a Victim Impact Panel (VIP) in comparison with participants in a DWI school only did not (1) more effectively move them through stages of change toward maintenance of drunk driving prevention, or (2) result in lower rates of DWI recidivism. Among offenders on electronic monitoring, attendance at treatment was not related to overall parole success, although it was related to a lower rate of technical violations. DUI offenders who attended a Victim Impact Panel (VIP) presentation had lower rearrest rates at 3 month, 6 month, and 12 month follow-up intervals than a comparison group of DUI offenders who did not attend a VIP presentation. Results of meta-analysis of DUI treatment programs revealed that (across all types of treatment) treatment had a consistently small but positive effect on reducing DUI recidivism and involvement in alcohol related crashes. In the studies reviewed , “treatment” was compared with no treatment, punishment (e.g. fines or jail), or licensing sanctions (e.g. suspension). Multimodal treatments were more effective (by at least 10 16 (10) Shapiro (2001) * (11) Project Match Research Group (1996) 5 (12) Pratt, Holsinger and Latessa (2000) 2 (13) Wells-Perker and BangertDowns (1995) * percent) in reducing DUI offender recidivism than any one of the following methods alone: (1) education, (2) psychotherapy or counseling, (3) probation monitoring and follow-up by treatment providers, and (4) treatment provider aftercare. The evaluation research on therapeutic communities (TC) cited in this review supports the continued development on the TC model for drunk drivers. The “Matching hypothesis” is tested in two randomized field experiments: one focusing on inpatient and the other on outpatient treatment. With the sole exception of psychiatric safety and outpatient treatment assignments, it appears that client characteristics do not need to be considered when matching offenders to the three. treatment modalities examined: (1) cognitive behavioral coping skills therapy, (2) motivational enhancement therapy, and (3) twelve step facilitation therapy, Clients in all three groups demonstrated “significant and sustained” improvements in drinking outcomes (1 year post treatment).“ Ten-year follow-up on a cohort of turning point participants revealed a consistently strong (between 10 and 30 percent) treatment effect for the program, which combines both cognitive and behavioral treatment modalities, and an aftercare component. Meta-analysis of 215 drunkdriving evaluation studies 17 (14) Deyoung (1997) 2 conducted between 1955 and 1992 revealed that “the average effect of remediation on drinking/driving recidivism was an 8-9% reduction over no remediation” (1), with a similar effect size for alcohol - involved crashes. Quasi-experimental design used to examine the effectiveness of alcohol treatment, driver license actions, and jail terms in reducing drunk driving recidivism among 1st time and repeat drunk drivers in California. For both groups, a combination of license curtailment and participation in an alcohol treatment program resulted in the lowest recidivism rate of any sanctions. Jail terms were found to be the most ineffective sanction for both groups. * These research/literature reviews were included for informational purposes; they do not represent original research studies. 2. Prevention Programs for Drunk Drivers: Individual and Community-Level Strategies Up to this point we have focused our review on research that evaluates the effectiveness of treatment strategies targeting convicted drunk drivers. In this section we examine the available research on individual and community-level prevention strategies. We begin by asking a deceptively simple question: why do individuals drink and drive? For adults over the minimum drinking age, the answer appears to be two-fold: (1) it is legal, and (2) it is socially acceptable, to drink and drive in this country. In order to prevent adults from drinking and driving, it seems logical to suggest that we need to change the law and/or change the general public’s attitudes towards drinking and driving. In a recent review of the research on alcohol prevention programs, Holder (2000:844) 18 observed that “For the most part, local prevention strategies have been programs that target individual problem drinkers, not the total community. Results from evaluations of these individual-focused strategies have not been particularly encouraging (see Casswell, 1995)”. Our examination of the research on prevention-based research on drunk drivers revealed only two evaluations conducted during our review period (1990-2002). For this reason, we conclude that much is currently unknown about the prevention of drunk driving, both at the individual level (e.g., strategies targeting individuals with drinking problems, school-based education programs, media campaigns, etc.) and at the community level (e.g., programs targeting changes in community structure). One example of a prevention strategy targeting “problem” drinkers is Newcomb, Rybow, Hernandez, and Monto’s (997) study of the influence of peers on the drunk driving behavior of their friends. As we have highlighted in table 2, this study found that when “friends” intervene to stop a potential drunk driving incident, they are usually successful. Of the 388 UCLA college students who participated in the (self-report) study, 303 (78%) indicated that they had witnessed a potential drunk driving incident/situation in the past year. About two-thirds of these “observers” (N=206,68%) actually decided to get involved. Interestingly, 63% of those who decided to get involved were women, which may be an indicator of possible gender differences in either the perceived risk or the behavior, and/or the perceived need for intervention. Since respondents indicated that 73% of all interventions were successful, it certainly appears that peer intervention represents an important, front-line drunk driving prevention strategy. Unfortunately, we know very little – outside of laboratory research – about the dynamics of this type of drunk driving intervention. Given the limited scope and design of the one individual-level research study cited here, it seems apparent that much more basic research on individual decisions to drink and drive, conducted in conjunction with research on societal reaction to drivers who drink, needs to be conducted. At this point, the saying “friends don’t let friends drive drunk”, appears to be a media-driven myth, in most communities. 19 We do not presume that positive peer influence (and direct peer intervention) will eliminate the drunk driving problem in this country. Consider, for a moment, the number of “problem” drinkers among us who either drink alone or in social contexts (e.g., bars, restaurants), without direct peer observation. Nonetheless, we do suspect that it is exactly this type of informal social control that offers the most promising individual-level strategy for preventing individuals from driving while intoxicated. As Grasmick and his colleagues (1993) have suggested, changes in individual decisions to drink and drive may reflect changes in social values regarding drunk driving, which result in shame and embarrassment for individuals who violate this norm. A review of the recent public opinion research on drunk driving (see, e.g. Knutsson and Kuhlhorn, 1996) reveals that the vast majority (90%) of the general public believe that “drunk driving can never be justified” (217). Consistent with this prevailing social value, several jurisdictions across the country have developed comprehensive, community-wide drunk driving prevention strategies (see Holder, 2000, for an overview of these community-level approaches). According to Holder (2000:844), “The rationale for targeting communities, as opposed to individuals, is compelling. First, substance use occurs largely within community contexts. That is, particularly in the case of alcohol, communities provide structures(e.g. zoning of alcohol establishments) through which alcohol is typically obtained. Second, many of the costs associated with alcohol are borne collectively at the community level in the form of car crashes and alcohol-related violence.” Table 2 includes a summary of the results from Holder’s recent multi-site evaluation of a five part community-level drunk driving prevention program. The 5-year research project included three “experimental” communities (one in Northern California, one in Southern California, and one in South Carolina) and three “control” communities that did not employ the five part prevention strategy. The Community Trials Project described by Holder (2000) utilized the following components: (1) community mobilization (utilizing local groups and the local media); (2) 20 responsible beverage service (“to change the serving practices among on-premise alcohol licensees” (2000:849); (3) drinking and driving law enforcement (via changes in police training and new methods of testing for DWI); (4) underage drinking law enforcement (via improved training of retailers, increased enforcement, and media advocacy targeting problem establishments/retailers, and (5) alcohol access restrictions (via zoning law revisions). While the results of the process evaluation revealed community-specific variations in the full implementation of this community-level change strategy, the initial outcome evaluation findings were quite positive. Holder (2000) estimates that there was approximately a 10% annual reduction (1993-1995) in alcohol-involved crashes for the three experimental sites. According to Holder (2000), “The introduction of special and highly visible drink and drive enforcement – with new equipment and special training – produced the significant reduction” (853). While the above findings are consistent with earlier research conducted on the deterrent effects of increased and/or improved law enforcement (see, e.g. Welsh and Farrington, 2003, for an overview), we must caution against making any generalizations from this one study about the impact of communitylevel strategies generally, and on law enforcement in particular, on the decision to drink and drive. In this regard, it is important to consider the fact that most states report a significant decline in drunk driving arrests during the 1990’s, a period of time during which alcohol-related fatalities also declined dramatically (Bureau of Justice Statistics, 1999). Have improvements in law enforcement changed the actual and/or perceived risk of drunk driving apprehension during this period or are other factors (e.g., informal social controls) coming into play that explain these findings (e.g. improved emergency response capabilities by local hospitals, police and fire departments)? Unfortunately, we cannot answer this question based on the research conducted to date. 21 Table 2: Evaluations of Prevention Programs for Drunk Drivers: 1990-2002 Author (year) Scientific Method Key Findings Score (1) Newcomb, Rybow, Hernandez, 1 and Monto (1997) This study examined “personal characteristics and contextual factors among 338 college students who had made an attempt to prevent someone from driving drunk” (1997:1). 73% of all interventions were successful (median 3); 57% of assertive interventions and 47% of passive interventions were successful. (2) Holder (2000) 2 This 5 year study assessed the impact of a community-level alcohol prevention strategy in three separate jurisdictions. Results “show that the project reduced alcohol-involved crashes, lowered sales to minors, increased the responsible alcohol serving practices of bars and restaurants, and increased community support and awareness for alcohol problems (2000:84f3). 22 3. The Identification and Classification of Drunk Drivers Our review of the empirical research on the effectiveness of current classification strategies identified only four studies that directly examined this issue: the Project Match randomized field experiments (a level 5 study) discussed earlier in this report (see project summary in table 1); and three non-experimental (level 1) studies focusing on model development/testing issues (C’deBaca, Miller and Lapham, 2000; Williams, Simmons and Thomas, 2000; Marowitz, 1998). Based on the studies we identified and reviewed, we must conclude that no definitive statements can be offered either about “what works” or “what doesn’t work” in the area of drunk driving classification. However, a review of these three studies does suggest a number of avenues for future research, focusing specifically on the techniques used to match DWI offender problems to specific types of inpatient and/or outpatient treatment. The Project Match Research Group (1997) developed “two parallel but independent matching studies, one with clients recruited at five outpatient sites, the other at five sites with clients who received aftercare treatment following an episode of inpatient or day hospital treatment” (8). The objective of the study was to test a commonly held assumption in the rehabilitation field, that different types of offenders (with alcohol problems) will respond differently to different types of treatment (defined in this study as (1) cognitive behavioral coping skills therapy (CBT), (2) motivational enhancement therapy (MET), and (3) twelve-step facilitation therapy (TSF). With the exception of a measure of psychiatric severity (based on the completion of the Addiction Severity Index), the researchers found that treatment “providers need not take…client characteristics into account when triaging clients to one or the other of these three individually delivered treatment approaches, despite their different treatment philosophies” (1997:7). Our review of this study revealed four possible explanations for this finding. First, it is certainly possible that the three different “types of treatment examined in this study were, in fact, not all that different. Second, it is also possible that the two primary 23 outcome measures used in this evaluation-percent days abstinent and average number of drinks per drinking day – did not adequately capture the actual drinking behavior of participants (due to reliability/validity problems). Third, perhaps the ten part, multiple classification criteria used in the study did not actually distinguish different types of clients with different types of drinking problems. And fourth, the researchers may be correct in concluding that, in general, matching clients to specific treatments doesn’t really matter, because they do well in a variety of treatment modalities. Our inability to rule out the first three explanations makes it difficult, if not impossible, to accept the conclusions of this study without a body of collaborative research challenging the “matching hypothesis”. Unlike the Project Match evaluation, the three studies included in table 3 focus specifically on the classification of convicted drunk drivers. C’deBacca, Miller, and Lapham (2001) conducted a study of the recidivism patterns of close to 1,500 convicted DWI offenders in New Mexico. These offenders, convicted between 1989and 1991, were tracked for up to 4 years by researchers using data from the New Mexico Traffic Safety Bureau. The goal of this study was to devise a simple, reliable method to distinguish each offender’s risk of recidivism, based on the notion that intervention strategies need to be targeted to specific “risk levels”, rather than applied universally to all convicted drunk drivers. The researchers developed a model that combined 5 risk factors in an additive scale (scores 0-5) and reported differences in predicted recidivism ranging from 11% (0 risk factors present) to 53% (5 risk factors present). Risk factors included the following: age (under 29 years old), education level (less than 12 years of education), arrest BAC (200 or higher), AUI receptive-awareness scale score (7 or higher), and The MacAndrew Alcoholism Scale (MAC) score from the MMPI-2 (raw score of 23 or higher). According to the authors (2001:208), the AUI (Horn, Wanburg, and Foster, 1990) is a 228-item self-report instrument that assesses alcohol-related problems. It has been lauded as the strongest comprehensive paper-and-pencil instrument for evaluating alcohol problems (Miller, Westerberg, and Waldron, 1995). Unfortunately, no data were 24 included in this evaluation on either the length of time it takes (on average) to complete the AUI assessment process or the overall reliability of the self-report instrument. Finally, it is unclear from this evaluation why five levels of recidivism risk needed to be isolated and whether different intervention strategies were applied to the offenders included in the current classification study, possibly confounding the results of the classification study itself. The third evaluation included in this section of our review is Williams, Simmons and Thomas’ (2000) evaluation of an “alcohol safety action” program (VASAP) implemented in rural areas of Virginia. According to Williams, et al. (2000:135), “The VASAP evaluates DWI offenders referred by local court systems and assigns these individuals to either education classes, intensive education classes, outpatient counseling, or inpatient counseling programs depending on the offender’s identified need for services”. To assess the need for services among study participants (a sample of 377 mostly male convicted drunk drivers), offenders completed the Michigan Alcohol Screening Test (MAST) “a twenty-four item instrument designed to diagnose alcoholism” (Williams, et al., 2000:135). The two primary conclusions of the study were that (1) the MAST classification instrument was effective in its identification of “high risk” drunk drivers, and (2) the VASAP program reduced DWI recidivism. Although exploratory in design (a level 1, non-experimental study), the authors of this study do raise the “matching hypothesis” once again, which suggests that this may be one of the most important questions for researchers to address in the next wave of drunk driving evaluations. In addition, the simplicity of the twenty-four item classification instrument presented in this evaluation may be attractive to corrections administrators attempting to reduce classification-related workload while increasing contact standards by line staff. Finally, Marowitz (1998) offers a simple alternative to the more complex screening protocols described previously, focusing on three main factors associated with DWI recidivism: BAC level at time of arrest, prior traffic convictions (2 years), and conviction status (1st time vs. repeat offender). All offenders arrested for DUI in California between Jan. 1, 1993 and June 1993 who were subsequently convicted of 25 either DUI or reckless driving were included in the original analyses. The results of this study suggest that a simple classification scheme can be devised in many jurisdictions if the primary goal of the classification scheme is identifying offenders “at risk”, rather than assessing the level/type of an individual offender’s alcohol problem. Table 3: Highlights of Recent Research on Classification of Drinking Problems and the Prediction of Drunk Driving Recidivism Author (year) (1) deBaca, Miller, and Lapham (2001) Scientific Method Score 1 Key Findings Initially a total of six variables were found to predict DWI recidivism: age, education level, arrest BAC, prior DWIs, the receptive-awareness scale from the AUI, and the MAC raw score from the MMPI-2. A model combining 5 of these risk factors (minor prior DWIs) was found to accurately predict offender DWI recidivism. In general, the greater the number of risk factors in a given case, the greater the likelihood of recidivism. Predictions followed the following pattern: # of risk facts 0 1 2 3 4 5 (2) Williams, Simmons and Thomas (2000) 1 Probability of Recidivism 11% 16% 23% 32% 42% 53% The authors reviewed the available research on DUI risk factors and identified the following key risk factors; (1) 26 (3) Marowitz (1998) 1 The degree of the offenders substance abuse problem; (2) “A number of personality, behavioral and attitudinal traits predict drunk drivers”; (3) “Multiple offenders were more likely to have higher blood alcohol concentrations at the time of arrest, more non-traffic arrests, and more accidents than first time offenders” (2000:131). The research presented in this study identified two strong predictors of recidivism (reconviction for DUI) among DUI offenders: (1) failure to complete the VASAP treatment program, and (2) higher scores on the Michigan Alcohol Screening Test (MAST), “A 24 item test designed to diagnose alcoholism” (2000:135). Using data on DUI offenders in California, the probability of DUI recidivism was estimated by a simple 3 variable assessment, including: (1) level of BAC at arrest, (2) prior 2 year traffic convictions, and (3) offender conviction status (1st or repeat offender). According to this research, “first offenders with high BAC levels and prior 2 year traffic convictions are at as high a risk of recidivating as many repeat offenders and might therefore benefit from similar sanctions and/or remedial treatment” (1998:545). 27 4. Control-Oriented Strategies for Addressing The Drunk Driver Problem A wide range of administrative and/or criminal sanctions have been included in recent drunk driving legislation (Note: see MADD’s recent summary of applicable laws, included in our compendium). A staggering number of drunk driving laws have been adopted by legislatures across the country in recent years (MADD, 2002). As a result, drunk drivers are now subject to a variety of control-oriented criminal justice sanctions, including jail terms, house arrest (via electronic monitoring), day reporting/intensive supervision, and mandatory treatment. In many jurisdictions, these criminal sanctions are likely to be combined with one or more administrative sanctions, such as license suspension, license revocation, vehicle impoundment, vehicle immobilization (via “the boot” or some other device), vehicle forfeiture, and/or the use of vehicle interlock devices. While it is difficult to distinguish the effects of any one control strategy when such “sanction-stacking” occurs (Byrne, 1990), a number of recent evaluations of specific control oriented criminal justice and administrative sanctions have been completed during our review period (1990-2002) and are available for review. Table 4 highlights the major findings from 11 separate impact evaluations and 2 research reviews on this topic area, including 3 evaluations of the impact of license suspension/revocation (and per se laws), 1 evaluation of the impact of jail terms in conjunction with other sanctions, 3 evaluations of ignition interlock devices, 3 evaluations of vehicle impoundment/immobilization, and/or forfeiture and 1 evaluation of the implementation of a license plate sticker law targeting drunk drivers. 28 4A) What Works in Controlling Drunk Drivers Examination of table 4 reveals that the evaluation research on control-oriented sanctions is uniformly weak, consisting entirely of Level 1 and Level 2 quality studies. Once again, we must conclude that there is no sound body of recent empirical evidence (1990-2002) on which to base an assessment of which (if any) control-oriented drunk driver intervention strategies are effective, either as a general or specific deterrent to drunk driving. 4B) What Does Not Work in Controlling Drunk Drivers We simply do not have sufficient evidence to offer any firm statements either about “what works” or about “what doesn’t work” in this important policy area. However, our review of the research studies included in Table 4 does suggest that jailbased sanctions are likely to be an ineffective specific deterrent, which we suspect is due to the lack of available treatment for DWI offenders with alcohol-related problems in most jail settings (see, e.g., Hingson, 1993 and Deyoung, 1997), as well as the negative life course changes typically associated with a period of incarceration (e.g., breakdown of family ties, reduced employment prospects, and changes in intimate relationships). Since none of the studies we reviewed attempted to measure the general deterrent effect of jail terms for either first-time or repeat drunk drivers, we simply don’t know about how the threat of a jail term affects the general public in jurisdictions where “jail time” is an option. In general, negative findings reported here for the use of jail-based sanctions with drunk drivers are consistent with the findings from previous research reviews on the impact of control-oriented intermediate sanctions: surveillance and/or control-oriented 29 sanctions do not work as a recidivism reduction strategy (Byrne and Pattavina, 1992). Nonetheless, it is certainly possible that the threat of jail (or even house arrest) may increase the offender’s level of compliance with treatment-oriented conditions/sanctions (see, e.g., Courtright, Berg and Mutchnick, 2000; and Voas, Blackman, Timmits, and Marques, 2000). If this tentative finding is supported by the results of more methodologically rigorous evaluations, then jail terms (and other intermediate sanctions such as house arrest, day reporting, and intensive supervision) may play an important role in “motivating” offenders who refuse to comply with treatment. 4C) Promising Developments Vis-à-vis Drunk Driver Control In order to offer an evidence-based assessment of promising developments in the use of control-based sanctions, at least one of the evaluations included in our review would have to be a Level 3 study and the preponderance of the evidence from the remaining studies would have to support this conclusion (see Welsh and Farrington, 2003). As we noted earlier, none of the studies conducted in this area met this minimum quality threshold. However, we do think it is still possible to identify a number of promising developments regarding the use of control-based sanctions, particularly when these sanctions (both criminal and administrative) are used in conjunction with either individual-level treatment or community-level prevention strategies (see, e.g., WellsParker and Bangert-Drowns, 1995, for an overview of findings from (multimodal) combined treatment and control-oriented interventions with drunk drivers; and see Holder’s 2000 review of research on community crime prevention of alcohol problems). 30 First, it does appear that license suspension/revocation may work as a specific deterrent to subsequent drunk driving, particularly when used in conjunction with alcohol treatment. Deyoung (1997) found that license curtailment and participation in an alcohol treatment program resulted in the lowest recidivism rate of any sanction (or combination of sanctions). Importantly, this finding was reported for both first-time and repeat drunk drivers in California. When researchers have evaluated the effects of administrative per se laws alone (see McArthur and Kraus, 1999 for a review) the results were mixed, with three states showing significant reductions in recidivism (North Dakota, Nevada, and California) and two states showing no difference. Second, there is some evidence of a specific deterrent associated with vehicle impoundment/immobilization or forfeiture. Voas and Deyoung (2002:263) reviewed the evaluation research on this type of vehicle program and their analysis “reveals for relatively large recidivism reductions, from denying offenders the use of their vehicles for 1-6 months”. A similar conclusion was reached by Voas, Tippits, and Taylor (1997, 1998) in their evaluation of the use of a vehicle impoundment sanction in Hamilton County and Franklin County, Ohio. Third, the use of ignition interlock devices represents another promising, controloriented strategy, in particular when corrections administrators use the threat of criminal sanctions—such as jail or house arrest—to increase the number of offenders who “volunteer” to participate in the program (Voas, Blackman, Tippits, and Marques, 2002). In fact, Morse and Elliot (1992) found that interlock devices actually were a more effective recidivism reduction strategy for DWI offenders than license suspension. More 31 recently, Weinrath (1997) reported that interlock cases were almost three times less likely than the comparison group to drink and drive again. 4D) What Is “Unknown” About Control-Oriented Sanctions: Directions for Future Research Perhaps the most important avenue for future research on drunk driving interventions is the study of both the formal and informal social control mechanisms affecting the decision to drink and drive. In this regard, the recent research linking recent reductions in (self-reported) drunk driving to an increased threat of “shame or embarrassment” is quite interesting (see, e.g., Grasmick, Bursik, and Arnekley, 1993). According to the authors, “Our surveys revealed a significant increase in the perceived certainty and severity of shame…for drunk driving between 1982 and 1990…The increased threat of shame and its relatively strong deterrent effect appear to be the primary source of reduction in drunk driving in our two surveys” (61). In addition to research examining the relative importance of formal and informal social control mechanisms, it is apparent that the field is in need of randomized experiments testing the individual and combined effects of treatment and control-oriented sanctions. This recommendation is underscored by even a cursory review of the quality of the evaluation research conducted to date on this topic (see, Wells-Parker and BangertDrowns (1995) for a critical review of this literature). 32 Table 4: Evaluations of Control-Oriented Strategies for Reducing Drunk Driving Recidivism Author (year) Scientific Method Score (1) Voas, Blackman, Tippits and Marques (2002) 2 (2) Voas, and Deyoung (2002) * (3) Hingson (1993) * Key Findings Previous research suggests that “two factors limit the effectiveness of the interlock: (A) relatively few offenders volunteer to install interlocks, and (B) the offenders driving behavior while on the interlock does not carryover to postinterlock driving” (2002:449). To address the first of these two problems, program developers in one Indiana county (Hancock) decided to create a greater “incentive” for first time and multiple DUI offenders to volunteer – threat of jail or house arrest for nonparticipants. This strategy resulted in a 62% participation rate and substantial reduction in DUI recidivism for both first time and multiple offenders. A review of the existing evaluation research on vehicle action programs (i.e. including impoundment, immobilization, or forfeiture) reveals “some relatively large recidivism reductions, from denying offenders the use of their vehicles for 1-6 months” (2002:263). Reviews the available research on legislations restricting drinking among minors, legislation to deter impaired driving, enforcement strategies, and informal control 33 (4) Weinrath (1997) 2 (5) Morse and Elliot (1992) 2 mechanisms. It is estimated that raising the drinking age to 21 has resulted in a 12% decline in teenage fatal crashes since 1988. (2) Lower BACs for minors have been linked to a 30% decline in nighttime fatal crashes among teenagers. (3) Mixed results reported for curfews on nighttime driving by teenagers. (4) Negative findings on the impact of jail sentences on first time and repeat drunk drivers. (5) License suspensions/revocations have been found to reduce both DWI recidivism and alcohol-related crashes. It appears that this strategy, when combined with mandatory treatment options, may have a significant impact on both DWI recidivism and crash involvement. (6) Administrative license suspensions have a greater effect on drunk driving recidivism than jail sentences. (7) Mixed results for research on lowering BAC for adults. Ignition interlock cases were almost 3 times less likely than the comparison group to drink and drive again. Program participants were 6.5 times less likely to record a new serious driving violation and about 5.8 times less likely to be involved in an injury collision. The failure rate for ignition interlock cases was 3.4% (after 30 months) compared to 9.8% for a comparison group of DUI offenders who received license suspensions. 34 (6) Deyoung (1997) 2 (7) McArthur and Kraus (1999) * (8) Deyoung, Peck and Helander (1997) 1 Quasi-experimental design used to examine the effectiveness of alcohol treatment, driver license actions, and jail terms in reducing drunk driving recidivism among first time and repeat drunk drivers in California. For both groups, a combination of license curtailment and participation in an alcohol treatment program resulted in the lowest recidivism rate of any sanctions. Jail terms were found to be the most ineffective sanction for both groups. A detailed review of three (level 2 pre-post evaluations of administrative per se laws (Stewart, et. al., 1992; Lacey, et. Al., 1990; and Rogers, 1997) revealed mixed results with three states (North Dakota, Nevada, and California) showing significant reductions in recidivism and two states (Mississippi and Louisiana) showing no difference in recidivism in the period prior to or post law implementation. Using data on fatal passenger car and pickup truck crashes that occurred in California between 1987 and 1992, estimates of the involvement of both suspended/revoked (S/R) drivers and unlicensed drivers in fatal car crashes can be calculated (using the quasiinduced exposure method). Overall, it is estimated that 35.4% of all fatal crashes during the review period involved S/R 35 (9) Voas, Tippits and Taylor (1998) (10) Voas, Tippits and Taylor (1997) 2 (11) Deyoung (2000) 2 and/or unlicensed drivers. Compared to validly licensed drivers, S/R drivers are over involved in fatal crashes by a factor of 3.7:1, while unlicensed drivers are over involved by a factor of 4.9:1. All offenders receiving a vehicle impoundment sanction in Hamilton County, Ohio between 1993 and 1995 under either existing DUI (Driving under the influence) Or DWS (Driving while suspended) laws were compared to a (nonequivalent) comparison group of offenders who were eligible for-but did not receivethe impoundment sanction. Overall, the authors reported a significant reduction is repeat DUI offenses among all groups of DUI offenders (1st time, 2nd time, and 3rd time), but contradictory results for 1st time DWS (no effect) and repeat DWS offenders (strong effect). Using a quasi-experimental, post-test only control group design, researchers compared the recidivism patterns (19931995) of offenders in Franklin County, Ohio who had their vehicles impounded (or immobilized) to a group of similar offenders who did not receive a vehicle sanction. It was found that the recidivism rates of offenders with impounded/immobilized cars were lower than the comparison groups both while the car was impounded and after the car was reclaimed. Using an interrupted time series design, the evaluator compares 36 (12) Deyoung (1999) 2 (13) Voas, Tippits and Lange (1997) 2 crash rates of suspended/revoked drivers prior to the implementation of the impoundment and forfeiture laws in California (Jan. 1, 1995) to crash rates for suspended/revoked drivers after the law went into effect. A control group of California drivers not suspended/revoked was included for the same period (pre/post). The study did not find a general deterrent effect for the vehicle impoundment law in California. This study demonstrates the specific deterrent effect of vehicle impoundment on the subsequent driving behavior of suspended/revoked drivers in California. Compared o similar offenders whose vehicles were not impounded, there were “23.8% fewer driving while suspended/revoked or unlicensed convictions; 18.1% fewer traffic convictions; and 24.7% fewer crashes” (1999:45). The authors evaluated sticker laws for vehicles used by suspended/revoked drivers convicted of drunk driving in two states. Results suggest that the law had both a general and a specific deterrent effect in Oregon, but no effect (specific or general) in Washington. Concluding Comments: An Application of our “What Works” Review to MADD’s Rating the States 2002 Survey According to a recent survey review by Shults, Sleet, Elder, Ryan, and Sehgal (2002), “The most comprehensive source of information about state level DUI 37 countermeasures is Mothers Against Drunk Driving (MADD’s) Rating the States Survey,” (p. 106). Utilizing 2000 survey data, these authors evaluated the link between MADD’s state level grades (A-F) and the self-reported drinking and driving behavior of residents of each state. They found that “those living in states with a MADD grade of ‘D’ were 60% more likely to report alcohol-impaired driving than those from states with a MADD grade of ‘A’…the association existed for men and women” (Shults, et al., 2002: 106). These findings certainly suggest that MADD’s rating system can be used to gauge a state’s response to its drunk driving problem. For this reason, we conclude our review of the research on drunk driving interventions by examining the empirical basis for the specific drunk driving policy recommendations included in the MADD Rating the States survey. A copy of Maryland’s drunk driving report card is included at the end of this section (Note: A detailed review of all 50 states’ report cards, along with a listing of all applicable DUI laws and legislation, is included in the Research Compendium accompanying this report). Maryland received an overall grade of “C” for its efforts against drunk driving according to the most recent 2002 survey, compared to a 2000 survey grade of B-. According to the MADD grading system: “thirty percent of each state’s MADD grade represented the state’s trend in alcohol related traffic fatalities, 30% presence of DUI laws, 10% to DUI enforcement and 30% to the remaining six components of the survey...To retain a straight “A” aggregate grade, a state had to have an administrative license revocation law, a 0.08 BAC law, and a primary enforcement safety belt law” (Shults, et al., 2002: 107). After examining these review criteria we conclude that the MADD rating system has overestimated the importance of administrative measures (such as license revocations and vehicle impoundment, criminal sanctions (such as jail) and victim impact panels, while underestimating the importance of individual offender classification and treatment. In fact, the MADD survey does not even attempt to evaluate the quantity and quality of “treatment” available to each state’s first-time convicted DUI and repeat DUI offender population. Given the research findings highlighted in this report concerning the combined effect of treatment (i.e., alcohol education, self-help programs, and a variety of 38 inpatient and outpatient treatment modalities) and control-oriented sanctions (in particular, license actions) on subsequent DUI recidivism, it seems obvious that the current MADD grading system is seriously (and fundamentally) flawed. While there is certainly much to learn from reviewing MADD’s assessment of Maryland’s response to its drunk driving problem, it appears that in several critical areas the recommendations for change advocated by MADD are not based on a sober assessment of the drunk driving research literature. Recommendations for Change Based on our review of the recent research on drunk driving (1990-2002), we offer the following recommendations for changes in current policies and practices in Maryland vis-à-vis drunk drivers: 1. Treatment programs for drunk drivers need to be expanded to include a full range of individual and group treatment strategies, including alcohol education, self-help (AA/NA) programs, and a variety of multi-model inpatient and outpatient treatment programs; 2. Participation in treatment programs must be mandatory for both first time and repeat drunk drivers. For noncompliant offenders, a structured hierarchy of sanctions— including jail—should be established and fully implemented; 3. Probation line staff workloads needs to be redefined, allowing significantly more time for staff and DUI offender interaction, emphasizing offender selfassessment, motivation, and change in drinking/driving behavior (via such techniques as motivation interviewing); 4. Existing DUI offender classification systems need to be revised, incorporating the most recent research on (1) alcohol (and substance abuse) problem assessment, (2) the factors associated with subsequent DUI recidivism, and (3) the factors predicting the subgroup of chronic drunk drivers (e.g., prior driving record, BAC concentration); 5. The “matching hypothesis”— linking specific offender problems to specific types of treatment modalities— needs to be field-tested at selected Maryland 39 sites. Research suggests that these field tests should examine such issues as (1) the response of different ethnic groups to different treatment modalities (e.g., Hispanic DUI offenders and AA), (2) age-specific differences in treatment effectiveness, and (3) the unique treatment “challenges” posed by the multiple-problem female DUI offender in Maryland; 6. Prevention strategies need to be developed focusing on “communities at risk” rather than “at-risk” individuals (e.g., binge drinkers, teenagers who drink and drive). Research on community-level prevention strategies (e.g., community mobilization, zoning law revisions, restricting alcohol sales, increased enforcement of underage drinking laws, changes in serving practices in local bars and restaurants) suggests that this approach works better (in terms of all alcohol-related outcomes) than individual-level strategies targeting subgroups of at risk, problem drinkers; 7. A range of control-oriented administrative and criminal sanctions need to be developed and field tested, in conjunction with the individual offender treatment options identified earlier. Promising administrative sanctions include: license suspension/revocation, vehicle impoundment, immobilization, and the use of ignition interlock devices. Although not effective as stand-alone sanctions, both the use of jail terms and house arrest (or other intermediate sanctions) appear to show promise as strategies reinforcing mandatory treatment conditions for DUI offenders; 8. And finally, Maryland’s Drinking Driver Monitor Program needs to be redesigned, emphasizing the goal of offender change in drinking attitudes and behavior (e.g., drunk driving) through the strengthening of informal social controls (e.g., peers, community) and informal sanctions (e.g., shame, embarrassment). 40 Appendix A: Study Inclusion Criteria For Our Review Of The Evaluation Research On Drunk Driving According to the University of Maryland research review protocol, “The scientific methods scale ranks evaluation studies from 1=weakest to 5=highest on overall internal validity: 1. 2. 3. 4. 5. Correlational evidence (low offending correlates with the program at a single point in time); No statistical control for selection bias but some kind of comparison (for example, program group compared with nonequivalent control group; program group measured before and after intervention, with no control group); Moderate statistical control (for example, program group compared with comparable control group, including pre-post and experimental-control comparisons); Strong statistical control (for example, program group compared with control group, with control of extraneous influences on the outcome, by matching, prediction scores, or statistical controls); and Randomized experiment: units assigned at random to program and control groups prior to intervention” (as summarized in Welsh and Farrington, 2003: 169). Definition Of Key Terms Used In The Evaluation Review Welsh and Farrington (2003:169-170) provide the following description of the key terms used in the University of Maryland review of evidence-based crime prevention: “What works. These are programs that the authors (Sherman et al. 1997) were reasonably certain prevent crime or reduce risk factors for crime in the kinds of social contexts in which they have been evaluated and for which the findings can be generalized to similar settings in other places and times. For a program to be classified as working, there must be a minimum of two level 3 studies with significance tests demonstrating effectiveness and the preponderance of evidence in support of the same conclusion. What does not work. These are programs that the authors were reasonable certain fail to prevent crime or reduce risk factors for crime, using the identical scientific criteria used for deciding what works. For the classification of not working, there must be a minimum of two level 3 studies with significance tests showing ineffectiveness and the preponderance of evidence in the same direction. What is promising. These are programs for which the level of certainty from available evidence is too low to support generalizable conclusions but for which there is some empirical basis for predicting that further research could support such conclusions. For 41 the classification of promising, at least one level 3 study is required with significance tests showing effectiveness and the preponderance of evidence in support of the same conclusion. What is unknown. Any program not classified in one of the three above categories is considered to have unknown effects.” Source: Welsh and Farrington, (2003:169-170)