The initial step in the project was to conduct a representative survey of students in which a blood alcohol concentration (BAC) measurement would be obtained in conjunction with self-report information. The purpose of this initial survey was to (1) establish a baseline of alcohol use against which future, post-program measurements could be compared and (2) identify the nature and scope of student drinking, looking for particular problem areas that might productively be targeted by an intervention. As mentioned above, relatively little is known about students’ behaviors in association with drinking. It was our hope to gain a better understanding of this phenomenon on this campus prior to developing a program to ameliorate problems.
The sampling procedure was designed to obtain a representative sample of students who were returning home in the evening. It was reasoned that the one place we would most likely be able to sample from the entire student population would be as they returned home since that is the one place that virtually all students go at some point during an evening.
The campus was divided into four geographically distinct routes along which interview teams walked. These routes passed by every residence hall, fraternity and sorority on or near the campus. The sampling plan was developed such that every residential location was visited at least six times: once both before and after midnight on (1) a week night (Sunday through Wednesday), (2) a Thursday night and (3) a weekend night (Friday and Saturday). Most locations were visited several times within each of these time blocks. On each night that data were collected, three interview teams covered separate routes. Respondents were sampled near entrances to student residences.
To avoid attracting too much attention, interview teams moved continuously between locations, typically collecting data at a single location for no more than 10 minutes on a single visit. As teams approached residences they randomly sampled individuals or groups who were either approaching the residence or who were standing in front of the residence and apparently belonged there (i.e., they were not merely passing by).
Figure 2.1 shows a diagram of the on-campus routes the interview teams covered. Starting points on each route were randomly selected to ensure that no particular location was more likely than others to be used for interviewing. Approximately half of the undergraduate student body lives on campus.
Figure 2.1
Diagram of on-campus routes followed by data collection teams
In order to obtain information about those who live in private residences we also conducted interviews at five large apartment complexes that, according to university records, housed a large concentration of university students. At these locations, the interview procedure resembled that typically used in roadside surveys (Foss, Beirness & Sprattler, 1993).
Data were collected on 18 nights between October 2, 1997 and November 14, 1997, between 10 pm and 3 am, including all nights of the week. Three separate interview teams (consisting of a supervisor and 3 interviewers) worked on each night that data were collected. Groups of individuals were randomly selected (using a die) then approached by a single member of the interview team and asked if they would participate. Once consent was obtained, all members of two- or three-person groups were interviewed. For larger groups, the die was used to randomly sample three individuals for the interview. Interviewers and their respondents stepped away from any other individuals in the area to preserve confidentiality of the interviews and to avoid contamination of responses due to the presence of friends or acquaintances.
The interview took approximately 4 minutes and requested information about students’ activities during the night, drinking (where, when, what, how much), perceptions about alcohol use among students, mode of transportation and ways used to avoid drinking-driving. Those who reported drinking during the evening were asked whether they felt any effects of the alcohol and to estimate their BAC. To obtain the BAC estimate, we used a technique developed for experimental research and adapted to field interviews with drinkers (Beirness et al., 1995). Respondents were shown an analog scale representing BAC values from .00 to .24, with .08 marked as the illegal limit for drivers.1 They were asked to indicate where on that scale they thought their own BAC was at the time. Upon completing the interview, respondents were asked to provide a breath sample, which was taken and analyzed using a portable breath test device, the Lion S-D2 Intoxilyzer™. Persons who declined to be interviewed were asked if they would simply provide a breath sample. Interviewers showed all respondents their BAC readings and provided a brief explanation of the risks associated with that level BAC.
At off-campus residences, vehicles entering the parking lot for a complex were motioned to stop. Interviewers introduced themselves and asked if anyone in the vehicle was a UNC-CH student. If a student was present, that person(s) as well as the driver (if a non-student) was interviewed as described above while they remained in their vehicles.
HSRC researchers have conducted studies in which more than 40,000 individuals throughout North America have been interviewed and breath-tested during the past decade. Based on this experience, we did not anticipate more than an occasional minor problem with students who may have had so much to drink that they were at risk for alcohol poisoning or other medical consequences. Nonetheless, as is always the case, extensive precautions were taken for dealing with any person who was a risk to him/herself or others by virtue of risky alcohol consumption. For the first few nights of data collection, a substance abuse counselor from the Student Health Service accompanied the research teams to assist any student who might need help due to alcohol impairment (after no problems were detected during the first few nights, the procedure was modified so that a counselor was always on call at the Student Health Service while data were being collected). All research team supervisors carried mobile telephones so they were able to contact other teams, the project supervisor, the university student health service, the local emergency department and the campus or local police at any time.
Interviewers were trained to attend to several sources of information that would indicate there was reason for concern about a student’s welfare, and especially whether there was any possibility of alcohol poisoning.2 Any time an interviewer had a concern, the team supervisor was called over to speak to the student. The supervisor then assessed the situation and decided what, if any, protective action should be taken. Three types of information were available to interviewers that could indicate a student had consumed enough alcohol, or had done so in such a manner, that he or she might experience a serious medical problem. The main sources of information were the student’s behavior (e.g., difficulty walking, standing or talking; signs of feeling sick; disorientation or confusion) and statements made during the interview (e.g., mentioning very recent consumption of a large amount of alcohol that might not yet be reflected in either their behavior or BAC measurement). The final source of information was the measured BAC. Because the effects of alcohol vary substantially across individuals, we were reluctant to identify a BAC level that was itself a cause for concern. Rather, the BAC measurement was considered in conjunction with the other two sources For example, a moderate BAC measurement would not be cause for concern, but in conjunction with a statement about having just consumed several shots would suggest a potential problem as the recently-consumed alcohol entered the bloodstream. In such a case, the team supervisor would be notified and could take a second BAC measurement to determine whether the student’s BAC was rising rapidly.
It is necessary to keep interview times to a minimum when collecting data outdoors late at night. Accordingly, we asked a relatively small number of questions in the interview.3 To guide selection of questions from among the large number that might have been asked, we settled on the principle that, to be included, a question should either (1) ask for information that can meaningfully be provided only when asked in situ, that is, at a particular location or time for which it is appropriate (e.g., "How do you feel right now," to assess subjective experience of alcohol impairment) or (2) be uniquely valuable in conjunction with a BAC measurement (e.g., "How many drinks have you had tonight?" or "How did you get home tonight?"). In addition to soliciting basic demographic information, a few questions were asked about drinking behavior during the evening, one pertained to drinking during the past two weeks, others dealt with subjective feelings/perceptions, transportation and activities during the evening other than drinking. Interviews typically lasted from 3 to 5 minutes. A copy of the questionnaire is included in Appendix 2.a.
Because this was an unusual kind of study for a college campus, we held a news conference prior to beginning data collection to announce the survey. The goal was to ensure that students knew in advance they might encounter interview teams and what the teams were doing. It was important for students to recognize that this was not part of an enforcement effort. Such a belief would have dramatically reduced cooperation. The Director of the Governor’s Highway Safety Program spoke briefly to explain why his office was supporting the project. The president of the student body also spoke, giving his strong endorsement of the study. Following these speakers, several members of the project team explained the study, its purpose and demonstrated survey procedures for collecting a breath sample. The story was well covered by most local print and electronic media, including the student daily newspaper which is read by a very large percent of the undergraduate student body.
Although university officials approached this study with some trepidation, they were extremely interested in having an additional, objective, source of information about student drinking. The greatest concern was that the interview team would encounter students who were extremely intoxicated and that the university would experience increased liability should they subsequently injure themselves or someone else. It was made clear to all university officials that research teams doing BAC surveys always take extensive precautions to ensure that anyone whose well-being might be in question as a result of drinking is taken care of (as described above). This is a typical reaction to BAC surveys and is based, at least in part, on the misperception by nearly everyone, that drinking is a great deal more common and extreme than is actually the case. This general misperception probably results, in part, from the tendency for the news media to focus attention on extreme cases.
Of the 2,530 persons contacted, 86%4 cooperated to some degree; 2,023 (80%) completed the interview and provided a breath measurement; 4.3% declined the interview, but did provide a breath sample, and 2.1% did the interview but would not provide a breath sample. Among the latter, 7 (13%) showed evidence of drinking. Among the 14% who refused to participate from the beginning, 64 (19%) exhibited evidence of drinking (typically the smell of alcohol). Although this assessment is based on interviewer observation and judgment, it is consistent with the extent of drinking found among those who did provide a breath measurement. As has been found in late night breath measurement surveys of drivers (Beirness, Foss & Mercer, 1998), the most common reason for declining to participate was being in a hurry, followed by simple lack of interest in being interviewed, that is, not wanting to be bothered, rather than a concern about being discovered to be drinking. Typical reasons given for refusing were the time of night in conjunction with an early morning class or test and being cold (although it was late fall and some nights were quite cold, individuals were often encountered without coats as they rushed from a car to their residence).
Although sampled persons were interviewed regardless of their student status, results presented here represent only those 1,841 individuals explicitly identified as students enrolled at UNC-CH.5 A breath alcohol measurement was obtained for 1,786 (97%) of these individuals. It should be noted that we are unable to determine student status for persons who either refused the entire interview, or who provided only a breath sample. Alcohol use among the latter group, as indicated by measured BAC, was virtually identical to that of known students for whom results are presented here.
To determine whether the sample of students we interviewed are representative of the general student population, we compared demographic characteristics of respondents with university records. Table 2.1 shows characteristics of the sample and of the UNC-CH student body as a whole (during the fall semester of 1997).
It is clear that the sample composition differs somewhat from the university as a whole. The overrepresentation of younger students was to be expected since data collection was concentrated on and near campus, where younger students are more likely to live. The overrepresentation of males probably reflects the fact that data were collected late at night when males are probably more likely than females to be outside. The overrepresentation of members of Greek organizations probably results from the fact that they are more likely to live in smaller residence units (fraternity/sorority houses average about 40 persons, whereas residence halls average several hundred occupants), increasing their likelihood of selection when a team was at their location.
In sum, the demographic characteristics of the sample differs somewhat from that of the UNC-CH student body in that it is somewhat younger and contains more males. The greater concern is whether drinkers were properly represented in the sample. The high response rate, which far exceeds that of most telephone or questionnaire surveys, provides some assurance that the sample is a good representation of the population of interest. There is some question about whether sampling only persons who were outdoors may have resulted in our missing persons who remained inside for the entire evening, especially if those individuals are more likely to be drinkers. Information available from a self-report mailed questionnaire survey administered by the university at the same time of the BAC survey allows us to address this issue. A typical question to assess student drinking is "About how many times in the past two weeks have you had five or more drinks in a row?" In that self-report survey of students at UNC-CH, 38% had done this at least once; in the BAC survey which asked the same question, 49% indicated that they had consumed five or more drinks in a row during the past two weeks. Hence, if anything, the BAC survey may have an overrepresentation of drinkers. That is not necessarily the case, however, as the response rate to the mailed questionnaire was only 50%, raising the equally plausible explanation that drinkers were less likely to take the time to respond to that questionnaire.
To provide an overall estimate of alcohol use for the campus, data were weighted to equalize the contribution of each day of the week, since approximately twice as much time was spent interviewing on Thursday, Friday and Saturday nights as on other nights of the week.
Table 2.2 shows the full BAC distribution for the three times of the week studied, in addition to the overall distribution weighted to adjust for oversampling of weekend and Thursday nights.
*Weighted to adjust for oversampling of Thursday through Saturday nights.
For simplicity of reporting subsequent analyses, various cut points along the BAC continuum are used to create dichotomous variables. Here we use three points: Any non-zero reading, as an indication of drinking; .08, to represent substantial drinking; and .15 as a measure of heavy drinking. The selection of cut points is often fairly arbitrary. Zero vs. a positive reading is clearly not arbitrary. The .08 point was chosen based on the fact that this is the illegal limit for operating a motor vehicle in North Carolina. As such it reflects a legally meaningful point, but beyond that it is a somewhat arbitrary dividing line. To designate a BAC level to indicate heavy drinking, we queried several survey researchers with experience measuring BAC values in non-clinical populations. They were asked what BAC level a "binge" drinker would probably exhibit and they were nearly unanimous in suggesting .15.
Overall, 26% of students reported drinking and 23% registered a non-zero BAC. These were mostly, but not always, the same individuals. Among those who reported drinking during the evening, 12% registered a .00 BAC and another 11% were below .02. Conversely, 1.7% reported not drinking but registered a positive BAC value; most of these were below .05. As Table 2.2 shows, drinking on Thursdays was similar to that on weekends. On week nights, drinking was less common, but among those who did drink, BACs reached similarly high levels. For example, among those 534 respondents who reported drinking during the evening, 41% had a BAC of .08 or higher, and this did not differ by day of week (p > .60).
Heavy drinking among college students has typically been measured as self-reported consumption of five or more drinks on an occasion at least once during the past two weeks. Using this criterion, 49% of respondents were heavy drinkers; 24% qualified as frequent heavy drinkers (five or more drinks on three or more occasions in the past 14 days). On the evening of the interview, however, only 11% qualified as heavy drinkers based on self-reported consumption of five or more drinks. Among those individuals who qualified as heavy drinkers on the evening they were interviewed the mean BAC was .10.
In contrast to the self-report data, measured BACs present a somewhat different picture. Although the self-reported heavy drinking rate in this sample is 49%, there is no day of the week when more than 36% of students had BAC above .00. The proportion with a BAC above .08 never exceeded 16% and the proportion with a really high BAC (.15) was never more than 3%.
Of particular interest is the relationship of BAC to reports of heavy drinking. Examining the BACs of those persons who would be classified as heavy or risky drinkers by the standard measure of that concept among college students (5 in a row for males, 4 for females in the past two weeks), 61% had a BAC of .00 on the night they were interviewed. Nineteen percent were above .08, but only 3.4% had a very high BAC (> .15).
Self reports of drinking on the night of interview were more closely associated with BAC. Table 2.3 shows the BAC distribution for those who reported having five or more drinks for males or four or more for females on the night they were interviewed. Clearly a substantial number of these individuals had dangerously high BACs. On the other hand, there is also a substantial number with relatively low BACs. It is worth noting that 37% of those who qualified as a "binge" drinker on the night they were interviewed had a BAC below the per se illegal BAC limit (.08) for drivers in North Carolina.
Drinking was more common among persons above the legal drinking age (29% vs. 22%, p < .01), but BACs above .08 were equally common for persons below and above the legal drinking age (11% in both cases). Males were more likely than females to have positive BACs (28% vs 20%, p < .01), as well as BACs above .08 (13% vs 8%, p < .01). White students (27%) were more likely than black students (7%), as well as those of other races (11%), to be drinking and to have BACs above .08 (12% white, 3% black, 4% other; p < .01). Members of Greek organizations were more likely to have positive BACs (44% vs. 17%, p < .01) and BACs over .08 (23% vs. 7%, p < .01) than those who did not belong to Greek organizations.
Obtaining data in situ allowed us to examine how drinking and, more importantly, BAC levels change as a function of time of night. Even on weekends, drinking is typically a nighttime activity among students. The average time of first drink on every day was approximately 10 p.m. except Saturday when it was 9:30 p.m. Figure 2.2 shows the proportion of respondents with a BAC above .00, .08 and .15 by time of interview.
Figure 2.2
Proportion of students with BACs above
three thresholds by time of night.
In our efforts to pinpoint the nature and location of problems with student alcohol use, one question was whether celebrations associated with football games might be a particularly important intervention point. On most college campuses, it is believed that sporting events, especially football, and drinking go hand-in-hand. To determine whether drinking, or excessive drinking, is more common in conjunction with home football games, we compared BAC information for the two Saturdays when there was a home football game with Saturdays when there was no game. Table 2.4 shows the proportion of students whose BAC was at or above each of three levels (.00, .08 and .15) on typical Saturday nights and the dates of Homecoming and the "Big Game."7
Although there are clear differences, it is difficult to detect any meaningful pattern. Drinking was substantially more common on the Saturday evening of homecoming, but on the night of what was called the "Game of the Century" locally, drinking was no different from other Saturdays. There are several differences between the two Saturdays that may be related to drinking. The homecoming game was an afternoon game, and the home team won. The Big Game was a night game, and the home team lost. Unfortunately, with only one of each, it is not possible to detect any pattern associated with game days, day vs. night games, or wins vs. losses.
The large majority of drinkers (63%) reported drinking in a single location on the night they were interviewed. Another 30% reported drinking in two locations. Persons who reported drinking in two or more locations were more than twice as likely as those who drank in one location to have a BAC above .08 (57% vs. 33%, p < .01) and more than five times as likely (15% vs. 3%; p < .01) as those who drank in one location to have a very high BAC (> .15). It is not possible to determine whether this was due specifically to the number of locations or simply the fact that those who drank in more than one location spent a longer time drinking.
The most common drinking locations were bars and parties. The majority of parties (about 80%) were affiliated with a campus fraternity. Although the amount of drinking did not differ for those over and under the legal drinking age, the location of drinking did differ, in predictable fashion. Underage respondents were much less likely to drink in a bar or restaurant, although substantial percentages still did so. Fifty-four percent of respondents age 21 or older had their last drink in a bar, compared to 12% of those under 21 (p<.01).
Of particular interest in this study was the question of whether, and if so how, students’ BACs varied by the mode of transportation they used to come home. Table 2.5 shows the proportion of individuals with positive (non-zero) BACs and those above .08 for each of four most common modes of transportation.
It appears that UNC-CH students are doing a good job of ensuring that persons who have had too much to drink are not driving. Persons with any alcohol, as well as those with high BACs are much more likely to transport themselves home as a passenger (in a car or bus) or by walking than they are to drive. It is noteworthy that the proportion of drivers with a BAC above .08 is nearly identical to the proportion of the general nighttime driving population in North Carolina (Foss et al., 1995) found to be that high in a 1994 statewide roadside survey (2.4% NC vs. 2.6% UNC-CH).
Nearly two-thirds of students (63%) believe they drink less than the typical UNC student, whereas only 10% believe they drink more. This tendency to think others drink more was also evident when individuals were questioned about their drinking in comparison to those they were with at the time they were interviewed. When asked whether they thought the BAC of others in the group they were with was higher or lower than their own, 38% thought others were higher, 42% thought they were about the same and only 20% thought the others had a lower BAC.
Fifty-seven percent of those persons who reported having 5 or more drinks at least three times within the past two weeks (sometimes called "frequent binge drinkers") believed they drink the same amount as others on campus; 10% think they drink less! In contrast, among those most representative of student drinking (i.e., those who have not had more than five drinks on three or more occasions in the past two weeks, reflecting 76% of the sample we interviewed), 80% believe they drink less than the typical UNC student. It is clear that heavy drinkers believe they are the norm on campus, whereas lighter, less frequent drinkers, despite being in the distinct majority, believe they are atypical.
There is a tendency for first year students to have an even more exaggerated perception of being "out of sync" with drinking on campus. When comparing themselves to other students on campus, first year students are somewhat more likely than older students to say that they drink less than the typical UNC-CH student (66% vs 62%, p < .05). Even more striking is that those freshmen who had consumed enough alcohol on the night they were interviewed to have a BAC greater than .08 were less realistic than older students (with the same high BAC) in assessing how their drinking compares to that of the typical UNC-CH student. Whereas 35% of freshmen with a BAC above .08 thought they drank less than the typical UNC-CH student, only 20% of older students thought they drank less than the typical student (p < .05). Conducting this analysis using other indicators of drinking (positive BAC, self-report of heavy drinking) produced similar evidence that freshmen are even more out of line in their perceptions about student drinking than students who have been on campus for a year or more.
The latter two of these are beyond the scope of the present project. However, university officials are working on both through other avenues and have embraced the program described in this report as an integral part of the overall program to reduce problems associated with students use of alcohol.
The focus of our program was the first two issues listed above: pressures on students to drink resulting from overestimates of the extent of student drinking and the particularly large problem with misperceptions among first year students. Because incoming freshmen are just beginning a new phase in their lives during their first several weeks on campus, university officials are particularly concerned that they begin their college careers on the right foot. In view of the array of serious problems that can result in students’ lives if they become involved with excessive drinking, we decided to put a particular emphasis on enlightening freshmen about the realities of student drinking discovered in the BAC survey.