Impaired driving behaviors among college students: a comparison of web-based daily assessment and retrospective timeline followback.

Abstract

Although impaired driving is a widespread risk behavior among college students, there are limited tools for assessing its frequency and context to inform effective interventions. This study involved the development and evaluation of two innovative approaches for assessing impaired driving. The Impaired Driving Assessment (IDA), a modified Timeline Followback assessment, was compared to the Impaired Driving Electronic Assessment (IDEA), a web-based daily diary, among a sample of 15 college students over a 5-week period. We found high correlations between the IDA and the IDEA on all impaired driving outcome variables. IDEA respondents had high survey response rates but response rates were somewhat lower on weekends. Overall, this pilot study revealed that both the traditional and the electronic assessment approaches show promise for assessing impaired driving among college students, and additional research is needed to further explore their relative strengths and limitations.

INTRODUCTION

Alcohol consumption among students is one of the primary health concerns for U.S. colleges and universities. Heavy consumption of alcohol is common among undergraduate students with an estimated 48.6% of men and 40.9% of women classified as binge drinkers according to the 2001 Harvard College Alcohol Study (Wechsler, Lee, Kuo, Seibring, Nelson, & Lee, 2002). This behavior is particularly problematic because research has demonstrated a direct positive correlation between quantifies of alcoholic beverages consumed and the frequency and severity of reported alcohol-related problems (Wechsler & Nelson, 2001).

Of the many negative behaviors associated with heavy alcohol use, impaired driving may be viewed as one of the most serious because of the severity of its potential consequences, which can include criminal penalties, injuries, and death. This high-risk behavior is widespread among college students. A recent national study found that more than one quarter of college students reported driving following alcohol consumption during the previous month (Wechsler et al., 2002). Another recent study estimated that more than two million U.S. college students drove after drinking during the previous year (Hingson, Heeren, Zakocs, Kopstein, & Wechsler, 2002).

Despite the high estimated prevalence and potential severity of these high-risk behaviors, limited research has explored the patterns and contextual factors related to impaired driving among college students. One reason for this paucity of research has been the lack of existing instruments to objectively measure impaired driving behaviors (Hingson, Heeren, Levenson, Jamanka, & Voas, 2002). This study sought to fill this gap by developing two innovative approaches for assessing impaired driving and evaluating their relative strengths and limitations among college students.

Our first assessment approach uses the Impaired Driving Assessment (IDA), a semi-structured interview that retrospectively collects detailed information on the frequency and nature of impaired driving (Usdan, Schumacher, McNamara, & Bellis, 2002). It also calculates blood alcohol concentration estimates per drinking episode and assesses the circumstances surrounding the impaired driving episode. The IDA was originally developed to inform personalized feedback and goal setting interventions for respondents and was found to be consistent with self-reported admissions of drinking and driving episodes (Schumacher, Usdan, McNamara, & Bellis, 2002).

The Timeline Followback (TLFB), on which the IDA is based, is one of the most widely used alcohol assessment tools and has been shown to be a reliable and valid method for obtaining retrospective estimates of alcohol consumption (Sobell, Maisto, Sobell, & Cooper, 1979; Sobell, Toneatto, & Sobell, 1994). The TLFB uses a calendar method and other recall-enhancing techniques to assist individuals with their reconstruction of events. Previous research validating the TLFB has found good agreement between reports of alcohol use from the TLFB and collateral's reports from the same period (O'Farrell, Cutter, Bayog, Dentch, & Fortang, 1984; Zweben, 1986). In addition to the measurement of alcohol use, the TLFB approach has been successfully used to obtain information on illicit drugs such as cocaine and heroin, as well as HIV-related sexual behaviors (Ehrman & Robbins, 1994; Weinhardt, Carey, Maisto, Carey, Cohen, & Wickramasinghe, 1998).

The other approach for assessing impaired driving for this research is the Impaired Driving Electronic Assessment (IDEA), a web-based daily diary. Daily monitoring has been used to collect real-time information on a wide range of health-related behaviors and is often used as the standard by which to compare retrospective assessments (Leigh, 2000). For example, a recent study found no differences between paper-based and web-based alcohol assessment methods and suggested web-based methods as an alternative to traditional assessment methods (Miller et al., 2002).

Technological advances, including the growth of computer power, development of user-friendly interfaces, and decreased cost have led to the proliferation of computer-based data collection, now commonly called computer-assisted interviewing. When computer-assisted interviews are completed directly by participants without the active intervention of a human interviewer, the process is called computer-assisted self-interviews (CASI). CASI systems have many strengths, including the ability to present questions in a consistent manner, reduce potential interviewer effects, and allow users to proceed at their own pace (Lapham, Kring, & Skipper, 1991; Lapham, Henley, & Kleyboecker, 1993). In addition, people are more likely to disclose sensitive information about themselves to a computer than on surveys or to a human interviewer (Wright, Aquilino, & Supple, 1998). The web offers many advantages for administering CASI including high levels of access among selected populations, the freedom to complete surveys at user-selected times and locations, and the ability to use survey functions that are commonly encountered in web-based forms.

Perhaps the most important strength of web-based CASI for assessing substance use and abuse is the ability to administer daily surveys easily and affordably. McCabe and colleagues (2002), for example, found web-based surveys to be an effective mode for assessing alcohol and drug use data from undergraduate students, who tend to have high rates of access to the Internet. Miller et al. (2002) suggest that web-based alcohol assessments can be cost-effective by increasing survey accessibility while minimizing data collection and entry errors. Other assessment approaches, including TLFB, require respondents to recall drinking episodes many days or weeks after they occurred. The potential recall bias associated with retrospective assessments also may limit memories of the details and contextual factors of each episode that are essential for developing effective interventions. Although many studies have found daily monitoring effective for assessing alcohol consumption (Carney, Tenne, Affleck, del Boca, & Kranzler, 1998; Hilton, 1989), no known studies have appfied this assessment approach to measuring impaired driving.

The purposes of this study are therefore to compare these two approaches for assessing alcohol use and subsequent drinking and driving among college students. If successful, this research can be used as a framework for internet-based educational programs to reduce impaired driving in this high-risk population.

METHODS

Participants

Undergraduate college students were recruited from a large, public university in the Southeastern U.S. for participation in the study, following approval from the campus Institutional Review Board. Students who reported two or more drinking and driving episodes during the past month were invited to participate. A total of 20 students who met this eligibility criterion consented to participate. Participants were white, 56.3% male, with an average age of 21.1 (SD = 2.3) years. Of those students who enrolled, 16 (80.0%) students completed the study, but one participant was removed prior to analyses because he failed to complete the required 25% of the IDEA daily surveys.

Design and Procedure

This study used a cohort design with participants completing the web-based daily-retrospective IDEA each day for 35 days and the retrospective IDA interview at the completion of the 35 days. Upon enrollment, participants provided informed consent and were given instructions for completing the daily IDEA, unique user identification numbers, and passwords. Participants then were instructed to go to the IDEA website from any computer with web access before 2:00 p.m. each day for the next 35 days to answer questions about their alcohol use and impaired driving behaviors over the past 24 hours. Incentives were distributed according to a graduated pay scale for completion of daily IDEA that started at $1.00 per day for the first 20 entries and went up to $8.00 per day for the 31st through 35th entries.

After five weeks, participants returned at a scheduled time to meet with a researcher who administered the IDA. Participants received an additional $10.00 for completing the IDA interview. A total of $110 could be earned for perfect adherence. At the completion of the final interview, participants were debriefed about the study, gave feedback about using the IDEA, and those who reported an excessive amount of impaired driving were given referrals to the University Health Center for alcohol-related counseling.

Measures and Variables

IDEA. The Impaired Driving Electronic Assessment measures daily alcohol use and impaired driving behaviors with survey questions posted on a secure website that can only be accessed with assigned identification numbers and passwords. Completed surveys were automatically downloaded into a secure database along with identification numbers, date, and time of completion. Questions assessed behaviors experienced by participants during the previous 24 hours including use of alcohol, impaired driving, and riding in a car with a driver who had been drinking. For the purposes of this study, one standard drink referred to 12 ounces of beer, or a 5 ounce glass of wine, or 1 1/2 ounces of liquor (Bailey, 1993). Automatic branching of questions was used to avoid administration of unnecessary items. Participants also were asked about the amount of time they spent on the Internet and how often they used e-mail. The time needed to complete the IDEA ranged from one to five minutes. Data collected on the IDEA produced the following variables: Driving self after drinking days (DS days); Driving self after drinking episodes (DS episodes); Drinks per DS episode; Driving with other impaired driver days (DO days); Driving with other impaired driver episodes (DO episodes); and Number of days used any alcohol (Drink days).

IDA. The Impaired Driving Assessment (Schumacher et al., 2002; Usdan et al., 2002) uses retrospective TLFB methodology to assess drinking and driving episodes over the past 5 weeks. The IDA attempts to enhance participant recall through anchoring (identifying important events such as holidays, parties, exams, etc.) and a calendar to document each episode of driving after drinking and riding with an impaired driver. Detail about time, duration of consumption, location, and quantity of alcohol consumed over isolated or patterned episodes of impaired driving are assessed using the IDA. The amount of time needed to complete the IDA varies based on the extent of impaired driving, and takes approximately 20 minutes on average. Reliability of the IDA has been demonstrated with test-retest correlation coefficients ranging from .76 to .96 and inter-rater correlations from .73 to .94 (Usdan et al., 2002). Data collected on the IDA produced the same variables as those collected from the IDEA.

Data Analysis

We hypothesized that the IDEA would demonstrate agreement with the IDA on selected drinking and impaired driving variables. To assess the levels of consistency on the variables of interest, impaired driving variables measured on each assessment were compared using paired t-tests and Pearson correlation coefficients. Examinations of the feasibility of the IDEA as a daily monitoring strategy utilized chi-square analysis to compare adherence rates by days of the week. Analyses were conducted using SPSS software (SPSS, 2002).

RESULTS

IDEA Adherence

Participants completed the IDEA on 430 days out of an overall possible 525 days (i.e., 35 days x 16 participants) for an 81.9% adherence rate. The average number of completion days for participants was 28.7 out of 35 (SD = 5.45) and the range was from 17 days to 35 days. When IDEA adherence was analyzed by day of the week, we found that average adherence rates ranged from 88.8% on Sunday through Thursday and 64.7% for Friday and Saturday (See Table 1). Grouping week and weekend days revealed similar differences in adherence rates with significantly less IDEA adherence on weekends ([chi square]= 42.1,p < .001).

IDEA vs. IDA

Results on the IDEA and the IDA over the same time period were compared to determine the level of consistency between the two assessments. Analyses revealed high correlation coefficients ([greater than or equal to] .72, p < .01) between the IDEA and the IDA on all outcome variables (See Table 2). For each of these variables, the reported values were higher on the IDEA than on the IDA, and this difference was significant for impaired driving days (t = 2.16, p < .05), as well as both days (t = 2.51, p < .05) and episodes (t = 3.84, p < .01) of riding with an impaired driver.

We calculated level of agreement between the IDEA and the IDA by coding identical responses on both instruments as "agreements" and all other responses as "disagreements" for each drinking and driving variable (See Table 3). Our first, more conservative, comparison treated IDEA non-compliant days (days on which the participant did not complete the IDEA) as "disagreements" and found levels of agreement ranging from 62.1% (drinking days) to 70.1% (riding with an impaired driver days). However, when including only those days reported on both instruments, we found agreement rates ranging from 79.9% (drinking days) to 90.2% (riding with an impaired driver days).

DISCUSSION

Despite the recent attention given to alcohol use and associated problems among college students in recent years, research shows that at best, the proportion of students classified as heavy episodic drinkers has remained constant since 1993 (Wechsler et al., 2002). Unfortunately, the same can be said for those consequences associated with heavy alcohol use, including driving after drinking and riding with an impaired driver. Recent research indicates that brief, motivational interventions offering personalized feedback offer effective strategies to reduce negative consequences associated with heavy alcohol use (Latimer & Cronce, 2002). The use of daily, web-based assessment methods, such as those described in this paper, offer a medium through which to deliver such prevention messages.

The concurrent validity of the IDEA was examined by comparing it to reports from the IDA. Agreement with the IDA was demonstrated by the findings from three analyses. First, although the two measures were significantly correlated across all outcome variables, the IDEA yielded higher levels of impaired driving behaviors on several variables. The second analysis revealed strong correlations between the two instruments above .70 for all outcome variables, indicating a high degree of agreement despite the differences in the magnitude of the values by instrument. Finally, the level of agreement between the two instruments as assessed by each variable demonstrated an average agreement of 81.7% when using available matches only. When assessing agreement more conservatively, the average rate of agreement was considerably lower (66.9%).

It is important to note that these differences exist despite the lower adherence rate of the IDEA. One would assume that the levels of agreement and degree of correlation between the two measures would be even higher with greater IDEA adherence. Since the IDEA might be expected to field significantly more drinking and driving behaviors than the IDA because of its strengths as a web-based, daily assessment method, this analysis alone is insufficient to demonstrate concurrent validity.

Feasibility of the IDA and the IDEA

The IDA has previously shown high estimates of both test-retest and inter-rater reliability for collecting information on impaired driving outcome variables. Despite fielding lower levels of impaired driving behaviors, the IDA was also found to have a high level of correlation with a web-based daily measure of alcohol use and impaired driving behaviors, the IDEA. This research suggests that although the IDA is suitable for obtaining detailed information on alcohol use and impaired driving, this instrument may underestimate incidence of specific episodes of this behavior.

Findings on college alcohol use and associated problems from large-scale surveys are limited by their poor response rates. Using adherence as a measure of feasibility, IDEA participants completed 81.9% of all possible reporting days. Further analysis of IDEA adherence by day of the week, however, revealed a problem with the IDEA by lower rates of adherence on weekend days as compared to weekdays. If not overcome, this will significantly interfere with the feasibility of this instrument.

Interviews with participants at the end of the study shed important light on this problem that offer some potential solutions. First, participants reported less access to on-campus computers on weekends that were commonly used by the students when classes were in session during the week. Second, participants reported that they were more likely to forget to complete the IDEA on the weekends and were less likely to complete the assessment past the deadline time of 2:00 p.m. due to the belief that a late entry would not earn the incentives. Despite these plausible explanations, there is also the possibility of social desirability as a reason for the lower adherence rates on weekends. As the majority of college drinking (and subsequent negative outcomes) takes place on Friday and Saturday nights, it is possible that students intentionally failed to complete the IDEA on the weekends because of their impaired driving behaviors during the previous 24 hours.

There were several limitations to this research that should be addressed. Although all participants had to meet eligibility criteria to enter the study, the small sample size limits the generalizability of these findings, even to those college students at high-risk of impaired driving. Future studies of the IDEA should utilize a larger sample representing a more diverse group of individuals. This research was also limited by relying exclusively on self-reported information. In this research, self-report could bias the data through participants' intentional or unintentional misrepresentation of their alcohol use. As not all participants in this study were of the legal drinking age, in addition to the stigma associated with drinking and driving, information obtained from both the IDEA and IDA are susceptible to a social desirability bias. However, numerous studies have found self-reports to be valid, acceptable measures of alcohol use (Midanik, 1988).

CONCLUSION

Data from this pilot study suggest that both the IDA and IDEA can be used to collect detailed information on impaired driving among college students, but that the web-based IDEA seems to capture greater incidence of such behaviors. This study of the feasibility and validity of a web-based daily assessment of drinking and driving behaviors has demonstrated both strengths and limitations with this approach.

The literature suggests that web-based assessment methods may possess important advantages over more traditional assessment approaches of alcohol use, including the TLFB technique (McCabe et al., 2002; Miller et al., 2002). This study demonstrated that the IDEA is potentially both feasible and valid as a measure of drinking and driving behaviors if limitations of weekend missing data can be resolved. Web-based daily assessments such as the IDEA are cheap, easy to administer, and allow for the collection of information on impaired driving behaviors and contextual factors. The systematic pattern of weekend non-adherence, however, is a real concern for the use of this instrument and if not overcome, will prevent the meaningful assessment of drinking and driving behaviors for measuring prevalence, nature, or for intervention use.

During interviews conducted at the end of the study, participants indicated that the IDEA was accessible and easy to complete. Other consistencies from the interviews were that participants found the IDEA to be easy to use, secure, and non-threatening. Some participants also suggested that e-mail reminders would have increased adherence. Upon entry into the study, all of the students reported that they check their email every week, and the vast majority (86.7%) reported that they check their e-mail at least 5 days a week. Although the methodology employed in this research was able to collect information on over 80% of all reporting days, we feel this adherence rate could be even greater.

The following recommendations are offered to overcome the limitation of the IDEA as a viable measure of drinking and driving behaviors. First, researchers must assure that participants have daily access to the Internet. Second, investigators should make it clear that completion of the IDEA prior to the daily deadline is encouraged, but late admissions are acceptable. Finally, follow-up studies of daily web-based assessment should include e-mail reminders to increase rates of daily adherence. These reminders would probably be most effective if sent prior to the weekends, during which participants were more likely to be noncompliant.

Until the development of practical, effective solutions to the problem of excessive alcohol use among college students, it is imperative that researchers focus on reducing the harms associated with heavy drinking for this population. As the rate of drinking and driving among college students persists, it is imperative for researchers to develop and implement prevention programs to reduce the magnitude of this problem. Web-based interventions offer drug and alcohol educators an effective, feasible medium through which to deliver such programs to the college student population. The research discussed in this paper identifies the IDEA as a tool to measure alcohol use and impaired driving behaviors on a daily basis. Computerized instruments such as the IDEA could be used to provide information to be delivered in the form of personalized feedback as part of an individualized prevention program. To ensure that prevention efforts targeting high-risk individuals benefit not only such students, but also the entire university community, it is also necessary to adopt and enforce appropriate alcohol policies to change the environment in which drinking and driving is so pervasive.

Table 1

Number of IDEA days missed, percent days missed,
and percent of adherence out of total of 35 assessment days
by weekday and combinations of weekdays

Day of week           Days     Percent     Percent
                     missed    missed     adherence

Sunday                  9       12.0%       88.0%
Monday                  7        9.3%       90.7%
Tuesday                 6        8.0%       92.0%
Wednesday               9       12.0%       88.0%
Thursday               11       14.7%       85.3%
Friday                 28       37.3%       62.7%
Saturday               25       33.3%       66.7%
Sunday-Thursday        42       11.2%       88.8%
Friday-Saturday *      53       35.3%       64.7%

* [chi square] = 42.10,p < .001

Table 2

Correlation Coefficients and Summary Statistics for
Impaired Driving Variables (n = 15) from the IDA and IDEA

                                      IDEA           IDA
Variable                           mean [SD]      mean [SD]     r (a)

Impaired driving (DS) days *       4.73 [3.79]    3.60 [3.11]    .85
Impaired driving (DS) episodes     6.87 [7.44]    5.13 [4.82]    .83
Drinks per DS episode              4.61 [3.25]    4.24 [3.96]    .73
Impaired riding (DO) days *        4.13 [3.25]    2.67 [2.23]    .72
Impaired riding (DO) episodes *    8.00 [6.11]    4.47 [3.94]    .83
Drinking days                     11.07 [5.99]   10.13 [7.79]    .80

(a) For all Pearson correlation coefficients, r, p < 0.01.

* p < .05

Table 3

Agreement between IDEA and IDA instruments
by impaired driving variable

       Impaired driving
           variable               Agreement *    Agreement **

Impaired driving (DS) days           69.3%          89.2%
Impaired driving (DS) episodes       68.4%          86.6%
Drinks per DS episode                62.7%          80.7%
Impaired riding (DO) days            70.1%          90.2%
Impaired riding (DO) episodes        69.0%          88.8%
Drinking days                        62.1%          79.9%

* IDEA and IDA combinations were considered to disagree if IDEA
day was missing.

** Missing days on the IDEA were excluded from the analysis.

References

Bailey, W.J. (1993). Drug Use in American Society (3rd ed.). Minneapolis: Burgess Publishing Group.

Carney, M.A., Tennen, H., Affleck, G., del Boca, F.K., & Kranzler, H.R. (1998). Levels and patterns of alcohol consumption using timeline followback, daily diaries and real-time "electronic interviews." Journal of Studies on Alcohol, 59(4), 447-454.

Ehrman, R.N. & Robbins, S.J. (1994). Reliability and validity of six-month timeline reports of cocaine and heroin use in a methadone population. Journal of Consulting and Clinical Psychology, 62, 843-850.

Hilton, M.E. (1989). A comparison of a prospective diary and two summary recall techniques for recording alcohol consumption. British Journal of Addiction, 84, 1085-1092.

Hingson, R.W., Heeren, T., Levenson, S., Jamanka, A., & Voas, IL (2002). Age of drinking onset, driving after drinking, and involvement in alcohol related motor-vehicle crashes. Accident Analysis and Prevention, 34, 85-92.

Hingson, R.W., Heeren, T., Zakocs, R.C., Kopstein, A., & Wechsler, H. (2002). Magnitude of alcohol-related mortality and morbidity among U.S. college students ages 18-24. Journal of Studies on Alcohol, 14 (Suppl.), 136-144.

Lapham, S.C., Henley, E., & Kleyboecker, K. (1993). Prenatal behavior risk screening by computer among Native Americans. Family Medicine, 25, 197-202.

Lapham, S.C., Kring, M.K., & Skipper, B. (1991). Prenatal behavior risk screening by computer in a health maintenance organization-based prenatal care clinic. American Journal of Obstetrics and Gynecology, 165, 506-514.

Latimer, M.E., & Cronce, J.M. (2002). Identification, prevention and treatment: A review of individual-focused strategies to reduce problematic alcohol consumption by college students. Journal of Studies on Alcohol, 14 (Suppl.), 148-163.

Leigh, B.C. (2002). Using daily reports to measure drinking and drinking patterns. Journal of Substance Abuse, 12, 51-65.

McCabe, S.E., Boyd, C.J., Couper, M.P., Crawford, S., & D'Arcy, H. (2002). Mode effects for collecting alcohol and other drug use data: Web and U.S. mail. Journal of Studies on Alcohol, 63, 6, 755-761.

Midanik, L. (1988). Validity of self-reported alcohol use: A literature review and assessment. British Journal of Addictions, 83, 1019-1030.

Miller, E.T., Neal, D.J., Roberts, L.J., Baer, J.S., Cressler, S.O., Metrik, J., & Marlatt, G.A. (2002). Test-Retest Reliability of Alcohol Measures: Is There a Difference Between Internet-Based Assessment and Traditional Methods? Psychology of Addictive Behaviors, 16(1), 56-63.

O'Farrell, T.J., Cutter, H.S., Bayog, R.D., Dentch, G., & Fortang, J. (1984). Correspondence between one-year retrospective reports of pretreatment drinking by alcoholics and their wives. Behavioral Assessment, 6 (3), 263-274.

Schumacher, J.E., Usdan, S.L., McNamara, C.M., & Bellis, J.M. (2002). Screening for impaired driving risk among college students. College Student Journal, 36,(2), 180-187.

Sobell, L.C., Maisto, S.A., Sobell, M.B., & Cooper, A.M. (1979). Reliability of alcohol abusers' self-reports of drinking behavior. Behaviour Research and Therapy, 17, 157-160.

Sobell, L.C., Toneatto, T., & Sobell, M.B. (1994). Behavioral assessment and treatment planning for alcohol, tobacco, and other drug problems: Current status with an emphasis on clinical applications. Behavior Therapy, 25, 533-580.

SPSS, version 11.0 (2002). Applications guide and computer software. Chicago: Author.

Usdan, S.L., Schumacher, J.E., McNamara, C., & Bellis, J. (2002). Measuring impaired driving behaviors among college students: Development and reliability of the Impaired Driving Assessment. American Journal of Health Studies, 18(1), 10-18.

Wechsler, H., Lee, J.E., Kuo, M., Seibring, M., Nelson, T.F., & Lee, H. (2002). Trends in college binge drinking during a period of increased prevention efforts. Journal of American College Health, 50(5), 203-217.

Wechsler, H., & Nelson, T.F. (2001). Binge drinking and the American college student: What's five drinks? Psychology of Addictive Behaviors, 15(4). Special Issue: Understanding binge drinking, 287-291.

Weinhardt, L.S., Carey, M.P., Maisto, S.A., Carey, K.B., Cohen, M.M., & Wickramasinghe, S.M. (1998). Reliability of the timeline folowback sexual behavior interview. Annals of Behavioral Medicine, 20, 25-30.

Wright, D.L., Aquilino, W.S., & Supple, A.J. (1998). A comparison of computer-assisted and paper-and-pencil self-administered questionnaires in a survey on smoking, alcohol, and drug use. Public Opinion Quarterly, 62, 331-353.

Zweben, A. (1986). Problem drinking and marital adjustment. Journal of Studies on Alcohol, 47(2), 167-172.

Author Note

Stuart L. Usdan, Assistant Professor of Public Health, University of South Carolina, Arnold School of Public Health; Joseph E. Schumacher, Associate Professor of Medicine, University of Alabama at Birmingham, School of Medicine, Department of Medicine, Division of Preventive Medicine; Jay M. Bernhardt, Assistant Professor of Public Health, Emory University, Rollins School of Public Health, Department of Behavioral Sciences and Health Education.

The authors would like to thank the Social Science Research Center at Mississippi State University and MASEP for supporting this research. To our colleague and good friend Jeff Bellis whose intellectual spirit will continue to inspire and drive forward this research.

Correspondence concerning this article should be addressed to Stuart L. Usdan, Assistant Professor of Public Health, University of South Carolina, Arnold School of Public Health; P [phone]; F [phone]; [e-mail]; USC, Department of Health Promotion, Education and Behavior, 800 Sumter St., #216, Columbia, South Carolina 29208.