The study was prospective, randomized, and controlled. Patients were randomized 1:1 to the intervention or control. A prospective study is one in which patients are enrolled as they become known to be eligible, in contrast to a retrospective study of records or databases. Randomization is a process for assigning study patients to a group. A 1:1 randomization is similar to flipping a coin. A controlled study is one in which the group on which the intervention is tested is matched against a group on which the intervention is not tested, to serve as a control.


Patients were enrolled at two emergency departments, Carolinas Medical Center, located in a urban setting in Charlotte, North Carolina, and Pitt County Memorial Hospital in Greenville, North Carolina, a small city - rural setting. The two EDs have a combined annual census of approximately 150,000. Patients of driving age constitute 75-80% of that total. An analysis prior to the beginning of the study estimated that 9,000 to 12,000 MVC related injuries were likely to be treated in the two EDs during the study period.

Carolinas Medical Center

Carolinas Medical Center (CMC) is an 843-bed academic medical center located in Charlotte, North Carolina's largest city with a metropolitan area of 1.2 million. It is the region's designated trauma center, with over 2,000 admissions yearly to its trauma service. The CMC Department of Emergency Medicine is a department of 23 full-time academic faculty, 30 residents, and 4 fellows in a 50 bed ED facility with an annual census of approximately 100,000. Virtually all severely injured patients from motor vehicle crashes in the region are taken to CMC. A large volume of ambulatory patients is cared for as well. All ED visits are recorded on a computer database with codes to indicate external cause of injury for all injured patients, enabling an accurate measure of the population baseline.

CMC is the central hospital of Carolinas Health Care System (CHS). CHS is a multi-hospital network, including the Center for Behavioral Health, which operates inpatient and outpatient treatment centers for alcohol disorders, other substance abuse disorders, and dual diagnosis. Nurses and counselors staff an emergency phone line 24 hours per day to assist with referral of patients for substance abuse treatment. This phone line was used at the CMC site for this study. Patients were referred to one of two evaluation centers, depending upon their third party payer requirements.

Pitt County Memorial Hospital

Pitt County Memorial Hospital (PCMH) is a 740-bed tertiary academic medical center located in Greenville, North Carolina, in the middle of the most rural part of North Carolina. It is its region's designated Level I trauma center, with 1,200 annual trauma admissions. The PCMH Emergency Department has an annual census of approximately 52,000. The medical center is the only hospital in Pitt County, North Carolina, and the only tertiary medical center for a 100-mile radius. The medical center serves a population base of both semi-urban and rural communities in eastern North Carolina, including all severely injured patients as well as many others transported by emergency medical services (EMS).

Substance abuse services in Pitt County were provided by the county mental health system and by private practitioners.

University of North Carolina at Chapel Hill

The fundamental mission of the UNC Highway Safety Research Center (HSRC) is to conduct basic and applied research that increases knowledge and contributes to reducing death, injury, and the related societal costs. HSRC strives to translate developed knowledge into practical interventions that can be applied at local, state, national and international levels. The expertise of the HSRC was therefore employed to take advantage of research planning experience in the area of highway traffic safety, as well as extensive experience in the management of large research databases from multiple sites. Investigators from the UNC School of Medicine, Department of Emergency Medicine provided the clinical research monitoring function, as well as assistance in data maintenance.


Consecutive patients sustaining motor vehicle injury presenting at the two EDs between the hours of 10 a.m. and 10 p.m. during a 1-year period were studied. Motor vehicle injury was defined as an injury occurring to persons injured as motor vehicle drivers or occupants, including motorcycles and mopeds, as well as pedestrians and bicyclists. Thus, patients enrolled in the study were not necessarily drivers in the crash that just occurred, but because of age, all were potential drivers, and therefore at the same risk of driving after drinking, whether they had been the driver in this particular occurrence. We also did not account for culpability in the crash, since that information was not known to the interviewers and is a matter for the courts.

Excluded were patients who were admitted to the hospital outside of the ED Observation unit, those who were unconscious or too impaired to cooperate with the questions, and those who could not communicate in English sufficiently to comprehend the consent process or the questions being posed by the research assistants.

The hours of 10:00 AM to 10:00 PM were chosen because of the cost limitations of 24-hour staffing for the study with respect to the relatively low volume of eligible patients presenting from 10:00 PM to 10:00 AM. A histogram of presentation times for eligible patients created during the planning phase revealed a small peak in crash frequency during the morning rush hour followed by a steady increase in crash frequency into the evening hours. The data collection schedule was designed to capture the later end of the morning rush hour crashes prior to patients' disposition from the ED, as well as the evening peak. Although it was anticipated that the proportion of alcohol-related crashes would increase as the day progresses, morning crashes should be just as likely to yield high risk patients appropriate for randomization, since a large proportion of patients involved in daytime crashes, sober or not, will be alcohol dependent.4,5 It would therefore not be useful to concentrate solely on alcohol-related crashes for this study. Rather, the high risk patient who is not currently intoxicated is more likely to be a candidate for intervention, and should be the subject of systematic prevention efforts.

Patients younger than 18 years of age were not included in the study for two reasons. First, there was the practical limitation imposed by the need for, and difficulty of, obtaining consent from parents prior to entering minors into the study. Second, the investigators believe that any patient under 18 years of age who is injured in a crash with a positive breath alcohol should receive the maximum intervention available in the ED, without being subjected to the control group.

Prior to starting the study, the target sample size was estimated to be approximately 5,000. This sample size was derived through the following rationale: Recent research conducted in Michigan by Maio and colleagues reported that 19% (222) of the 1,161 motor vehicle injury victims screened for AA/AD were positive for current AA/AD.3 An additional 3% (30) of the patients who screened negative for current AA/AD tested positive for elevated levels of ethanol in blood, breath, or serum (22 mmol/L, 0.10 g/dl). Based on these results, we assumed that 1,000 (20%) of the 5,000 initial screens would be eligible for referral to AA/AD counseling. Under the research protocol, persons eligible for referral to AA/AD counseling were to be randomly assigned to either the treatment group (intervention and referral) or to the control group (no intervention or referral other than that made under existing standard of care, by the court system, or self-referral). Thus, 500 persons each would be assigned to the treatment and control groups. Past research suggests that 12% - 19% of the persons in the control group may seek counseling on their own or be referred by the court system.6 For purposes of establishing target sample size, we used the figure of 15%. The goal of this demonstration project was to increase the proportion of persons seeking AA/AD counseling by at least 50% (p0 = .15, p1 = .225). At the .05 level of significance for a one-sided test, a sample size of 500 in each group would be able to detect this difference at a power of 0.92. We assumed that some portion of the cases would result in incomplete data and would not be available for analysis in both treatment and control groups. Even if 20% of the cases are lost, sample sizes of 400 in each of the groups would be able to detect at difference of 50% (p0 = .15, p1 = .225) at a power of 0.86.

Research Interventions

Each site employed dedicated research assistants to screen patients and collect data. The two lead Research assistants were trained in counseling at the Master's or PhD level. The other Research assistants were either experienced counselors at the Bachelor's degree level or were studying for a Master's degree. All Research assistants were trained together in the protocol and the principles of screening and intervention in a training session prior to the study. The principal investigator and a psychiatrist with a sub-specialty in addictions conducted the training. During the study hours, the Research assistants were present in the ED and monitored the ED log and status boards to identify all patients aged 18 and older being treated for a motor vehicle related injury. Once identified, the Research Assistant kept a record of each patient eligible for screening, consisting of name and medical history. The Research Assistant approached each patient to be screened, made an introduction, and informed the patient verbally about the research project with patients sustaining motor vehicle related injury. Patients were assured of confidentiality, and verbal consent was obtained. Patients were made aware that they could withdraw at any time. Patients were not made aware of the objective of the study, since that knowledge would bias the answers and confound the study.

The Research assistants were responsible for the entire patient interaction and data entry, and their activities were totally independent of the ED nurses and physicians. The findings of screening or the results of interventions were not communicated to the physicians or nurses to avoid altering their interactions with the patients.

The Research Assistant asked the patient a series of questions, lasting less than five minutes, about their future risk of being in a MVC, including standardized questions about speed of driving, seat belt use and alcohol use (Appendix 1). The alcohol questions posed were actually the TWEAK screening tool (Appendix 2). Patients were then informed that the alcohol in the air was being measured. If serum alcohol content was obtained by the treating physician for clinical reasons not related to the study and available to the Research Assistant, it was also recorded in the case report form.

Patients who were A) TWEAK positive (score > 2) or B) had a quantitative breath alcohol (BAC) of 0.12 or higher were randomized to intervention or control. Patients randomized to the control were thanked for their cooperation, and were reminded that we would call them in 3 months and 6 months to follow up to see how well they were doing. Patients randomized to the intervention received a brief intervention at that point administered by the Research Assistant using a standard protocol called ED DIRECT (Appendix 3). This intervention approach was created for the study by a psychiatrist specializing in addictions. ED DIRECT is an adaptation of the FRAMES methodology employed in primary care and behavioral heath settings and takes only a few minutes to perform. 7 Concluding the intervention is the recommendation that the patient undergo a formal assessment by a specialist to see if the problem requires treatment.

Following the intervention, patients who agreed to further evaluation were offered an appointment on the spot. Patients who refused the evaluation were regarded as treatment failures. These data were recorded contemporaneous with the patient interaction. Other data were recorded from the medical record, including demographic data and injury diagnoses. All data were entered into the database, and uploaded at convenient intervals to the central site.

Selection of the Screening Tool

The use of screening tools to detect alcohol dependency in a health care setting has been thoroughly described. There are several screening tests for alcohol problems that are used in the general health care setting, including CAGE, AUDIT, Michigan Alcohol Screening Test (MAST), brief MAST, and TWEAK. A minimally trained individual can perform these rapid screening tools in only a few minutes. If a screening instrument for any medical condition is to be successful in the ED, it must be able to be administered quickly and have high sensitivity, even if specificity must be established over time.

AUDIT was developed for the primary care setting, and was shown to be sensitive for harmful drinking in a six-nation validation study, and has been recognized by the World Health Organization. {Saunders JB, Aasland OG, Babor TF, de la Fuente JR, Grant M. Development of the Alcohol Use Disorders Identification Test (AUDIT): WHO collaborative project on early detection of persons with harmful alcohol consumption. Addiction 1993; 88: 791-804 1992} CAGE is favored by many because of its brevity, it is easily remembered, and it can be worked in easily to the medical history. (Ewing JA. Detecting alcoholism: the CAGE questionnaire. JAMA 1984; 252:1905) TWEAK is similar to CAGE, but adds detection of physiologic tolerance, and asks about current usage, as opposed to life experience (Appendix 2). It has been found to be of high sensitivity and specificity in both primary care and general populations.

Each of these tests has its advantages, and comparison studies are few. A recently released report to NHTSA compared some of these tests for sensitivity and specificity, but did not include TWEAK.8 More recently, Cherpitel assessed the sensitivity and specificity of rapid alcohol screening tools that can be applied in the ED in a study performed in a level 1 trauma center in the southern United States.9 The screening tools were compared to a lengthy "gold standard" interview for the diagnosis of harmful drinking or alcohol dependence. Two screening tools--the TWEAK and AUDIT-- demonstrated sensitivities of 83% and 81% respectively. Sensitivity was highest in those individuals who were injured, male, and nonwhite, and was much higher than that for breath alcohol analysis and self-reported drinking (sensitivity 20% and 29% respectively).

The TWEAK was therefore chosen for this study in view of its demonstrated high sensitivity in a similar setting to our EDs, its speed and ease of use, and questions that seem less likely to be threatening following involvement in a MVC.

Breath Alcohol Content

Although breath alcohol content (BAC) itself is not sensitive for detecting alcohol dependency, breath levels greater than 0.14 in a patient who is coherent and able to ambulate is an unequivocal indication of tolerance to the drug, and is therefore a specific indicator of chronic intake.1 Detection of a high BAC in a coherent patient would detect false negative self-reporting of tolerance in the TWEAK. We therefore postulated that the combination of TWEAK screening tool and the physiologic marker of high BAC would be more sensitive and specific for AA/AD than either test alone.

We employed the use of a passive alcohol sensor (PAS Vr. System, Public Service Technologies, Fredricksburg, VA) for the study. This device is a breath analyzer that analyzes a 10 cc sample of ambient air "inhaled" by the machine in the passive mode, or it can be used in the active mode by having the subject blow into the device. No fixed correlation exists between the active and passive modes due to inconsistency in ambient air characteristics among measurements, but any error in the passive mode would tend to underestimate the BAC relative to the active mode. This particular instrument was graduated in increments of .04 g/dl, so that is was necessary to choose between BAC .12 and .16 as an entry criterion. Because of the likelihood of falsely low readings in the passive mode, we chose to include patients with BAC greater than or equal to .12 who were alert and oriented in an attempt to include all patients with alcohol tolerance.

For the purposes of our study, the manufacturer mounted the PAS into aluminum clipboards that also held the study documents. This enabled the research assistants to collect the air sample easily and efficiently without fumbling with devices and documents. Patients were informed of the purpose of the device, and were allowed to refuse or withdraw from the study at any time.

End Points and Follow Up

The primary end point of the study is whether or not the patient received the recommended evaluation. At one site, the endpoint determination was by self-report at follow up, and at the other, the agencies to which the patient was referred were contacted following the appointment date to determine whether the patient kept the appointment, in addition to self-report.

At three months and six months from the date of ED treatment, all reasonable attempts were made to contact all consenting patients by phone to obtain the following data regarding the period between the ED visit and follow up: If the patient

  1. Received treatment for AA/AD,
  2. Was involved in a subsequent MVC,
  3. Was arrested for an alcohol related incident, including driving while impaired,
  4. Received treatment for another injury.

The file for each patient was closed after the six-month follow up. Patients who could not be located were pursued by the Research Assistant until the end of the study. Those who could not be located were designated "lost to follow-up."

Six months after all subjects were entered into the database, HSRC cross-referenced the study records with DMV driver files, court records, North Carolina Hospital Discharge Database, and North Carolina Ambulance Call Report data to validate and augment the self-report data.

Data Analysis

Data were entered locally into a database written in Microsoft Access, and uploaded to the University of North Carolina at Chapel Hill for maintenance and reporting. In addition, the UNC investigators also served as clinical monitors for the study to ensure consistency in data reporting between the two sites.

At the six-month point in data collection, the UNC investigators made an audit visit to each site to ensure consistency in data collection and study conduct. This coincided with a change in the lead research assistant at PCMH.

Data were analyzed with the assistance of the Institute for Health Services Research (IHSR) at CMC using standard summary and analytical statistics. The database was transferred to IHSR in December 1999.

Continue to Results Return to Contents