Discussion

We have shown that high risk patients who receive an intervention for alcohol problems in the emergency department following a motor vehicle crash are more likely to seek a formal evaluation for alcohol problems than those patients treated under the current standard of care who receive no ED intervention.

As seen in Figure 2, of those receiving the intervention, 25 out of 130 (19.2%) received a formal evaluation, compared to 7 out of 57 (4.5%) in the control group. In fact, there was 5 times great chance that they would seek a formal evaluation [OR = 5.1, 95%, CI = 2.128 – 12.235]. It should be noted, that for those persons who agreed to an evaluation, 21 out of 43 (48.8%) actually showed up.

The protocol used in this project is consistent with the role of emergency medicine in the health care system. Emergency medicine is the entry point into the health care system following a traumatic event or medical emergency. It is tasked with disease detection and intervention, and depends on referral for definitive care of chronic medical problems.

There were several components of this protocol that were likely essential to the differences between the groups. The use of a structured screening instrument embedded in questions regarding future risk of injury was non-threatening and provided reliable, validated evidence of alcohol problems. The simplicity and brevity of the "ED DIRECT" intervention was important in that it did not present an impediment to other care rendered in the ED. The availability of a referral destination was also essential, in that it provided a seamless way for patients to receive help, and the barrier of the patient's needing to negotiate the health care maze to receive follow up was removed.

Studies of injured, impaired drivers admitted to the hospital or the ED are numerous. All have come to the unanimous conclusion that alcohol impaired driving in this population is under-reported, patients are infrequently charged and even less frequently convicted, and that opportunity for intervention is lost.4,5,11-14 This population is likely to drink and drive again and therefore likely to cause further injury or death. The National Commission Against Drunk Driving has recently published a position paper on the issue, with suggested state legislative remedies as well as concerted efforts by medical and law enforcement professionals.15 The American College of Emergency Physicians has recently developed policy on the issue, calling for greater detection of potentially impaired drivers and greater awareness and provision for treatment of alcohol abuse and alcohol dependency.16 Recognizing that those who abuse alcohol may also experience crash injury when sober, the net should be widened to detect alcohol abusers when they interface with the health care system, not only when they exhibit impairment after a crash. In the study of screening tools by Cherpitel, 17% of "current drinkers" seen in the ED met standard criteria for harmful drinking and 19% met criteria for alcohol dependence, whether or not they had been drinking at the time.9 Maio and colleagues found that 23% of MVC patients at their ED met criteria for AA/AD, although nearly half of those patients had no evidence of alcohol ingestion at the time they were seen in the ED.3 We expected therefore that screening for AA/AD among patients treated for MVC injury would be a high yield test, worthy of the added time and effort.

Large numbers of patients were required for screening in order to detect those most at risk for harmful drinking. It is a disease with an approximate 15% prevalence in our population, which certainly justifies screening and detection. We chose to test the hypothesis using a 1:1 randomization scheme. In retrospect, a 2:1 or 3:1 randomization of intervention to control would have improved our number of patients getting the intervention and referral. This may have allowed for better discrimination of the factors predicting patients amenable to intervention and referral. Unfortunately, the numbers of patients accepting the intervention and subsequently receiving treatment were too small for a multivariate analysis.

There were differences between the two sites with respect to numbers of patients receiving treatment after having received the intervention. Although the reasons for this are not fully understood, there were inherent differences between the sites that may have implications for implementation of a clinical protocol. First, at CMC, the referral destination and the emergency department are part of the same health care organization. There is a 24-hour call center staffed by mental health nurses or counselors experienced in substance abuse counseling. Follow-up data were therefore much easier to acquire, since many patients were referred within the same hospital system. Patients who needed to be referred outside the system were referred to a single destination that had a protocol to report whether or not a patient kept the referral appointment. At PCMH, the patients were referred into a mental health system not specifically dedicated to substance abuse treatment. In smaller communities where patient volume is insufficient to support dedicated substance abuse centers, the general mental health system would often be the referral destination. Also, patients in that system who were referred to private substance abuse therapists were lost to follow up if they were not available for self reporting at three or six months. Second, the populations of the communities at the research sites differ. CMC is located in a large urban area, with crash injury patients widely distributed among different ages and races. PCMH is located in a small city with a high proportion comprised of students and faculty of East Carolina University. The extent to which site bias was introduced because of college students’ different drinking habits and beliefs about drinking is not known. An important feature of this research that may have accounted for some of the difference between centers was the use of a standard intervention instrument by all research staff, irrespective of patient characteristics. It is possible that a more individualized intervention tailored to the needs of the patient and the style and experience of the research assistant may have resulted in greater compliance at PCMH.

Much of the follow-up data were limited by the inability of the research assistants to reach the patient by phone. The inherent limitations of self-reporting are well known and will not be reiterated, except to point out that some patients who agreed to evaluation and treatment may have been hesitant to report non-compliance to the same person that performed the intervention. It should also be noted that we were unable to detect the positive effects the intervention may have had, even if the patient refused to go for alcohol treatment. Many diseases such as hypertension often require multiple attempts to get patients to comply with treatment, and the disease of AA/AD is no exception.

Ideally all interviewers would have been multi-lingual due to an unexpected rise in the number of Spanish-only speaking patients in our emergency departments. There is insufficient data on the ethnology of alcohol-related vehicle injuries, but evidence indicated that drinking patterns are highly influenced by cultural norms, and awareness of driving laws is lacking in this population.17,18 Our study would have been stronger if we could have been able to detect differences in risk, including ethnicity, but the language barrier precluded the use of the screening instrument by our research technicians.

It should be emphasized that a TWEAK score of three was felt to be insensitive to detect AA/AD in the ED following a crash. After the first three months of data collection, is became obvious to the interviewers that these patients were becoming apprehensive after the first two or three questions about their drinking behavior. The first two questions of the TWEAK test are meant to detect tolerance, which by itself is a good predictor of excessive drinking. Although many patients were forthright in answering the questions subsequent to the tolerance questions, it became clear that having a score of two, rather than three, should be considered a positive TWEAK score. Therefore, after consultation with our addiction psychiatrist consultant and with NHTSA, we decided to re-define a "positive" TWEAK score as two, allowing for the patient to be randomized even if only the tolerance questions were positive. This is internally consistent with the protocol, in that a patient could be randomized based on presence of physiologic tolerance alone using a high breath alcohol and lucidity as evidence.

The interview containing the TWEAK may itself have provided some motivation to change drinking behavior. It should be noted that seven patients randomized the control group reported receiving help with their drinking subsequent to the ED visit. This suggests that in some people with preexisting insight or readiness to change, simply bringing up the issue of risk behaviors may be helpful.

The study was designed to gather data over a 12-month period. There were personnel issues that occurred during the year at both research sites that mandated a change in protocol. At one site, the lead Research Assistant left the employ of the institution for personal reasons. At the other site, the lead Research Assistant sustained a medical condition necessitating the hiring of a replacement. Because of institutional hiring protocols there was a hiatus in data acquisition. Therefore, in consultation with the statisticians and with NHTSA, the lost days were replaced with identical days of the week and times of the day at the end of the original twelve-month period. We believe this to be valid and no bias of results occurred.

The passive sensor was not helpful in this study. Of all the patients who agreed to the interview, not one entered the randomization solely because of breath alcohol. The study protocol called for entrance into the randomization if the BAC was 0.12 or greater. Although there were patients who were probably higher than this level, they either could not consent because of heavy intoxication or elected not to do so, and were therefore not randomized. Were this a clinical protocol and not a study protocol, those patients would have been observed in the emergency department until they were competent to receive an intervention and referral. Thus a passive sensor may very well be useful to detect tolerance in harmful drinkers prior to a screening interview.

The ideal outcome measure for a study such as this is the proportion of patients receiving the intervention that reduced their drinking over a year and reported fewer episodes of driving after drinking. Due to the time constraint of the project, one-year outcome data were not measurable in the time allotted. It is particularly obvious that, as a starting point, sobriety or reduced drinking begins with an initial interface with the treatment community. Our intent was to demonstrate that a protocol for screening detection and referral of high-risk patients results in a significant increase in the proportion of patients who receive a formal evaluation. Our results demonstrate that such a protocol indeed dramatically increases the opportunity for people to interface with the treatment community and therefore, they have the opportunity for sobriety, fewer drunk driving episodes, and safer highways.

Physicians, nurses, or administrators who are considering the implementation of a screening and intervention protocol in the emergency department should not be concerned that our percentage yield of patients who agreed to treatment was low. First, due to the nature of the study and the need for a consent, there were over 1,400 patients who would have been screened under a clinical protocol who were not, because of the need for consent to participate in research. We do not know why patients refused to participate, but anecdotally it was noticed that many patients who refused appeared to be high-risk by various indicators, including alcohol on their breath and fear of investigating police officers in the emergency department. Moreover, an experienced interviewer, not operating from a script, may be able to detect AA/AD with higher sensitivity than we were able to do with our standard scripted interview. In fact, some of our more highly trained research associates with experience in alcohol treatment became frustrated with being prohibited by the experimental design from providing needed counseling for those who refused to consent or were randomized to the control.

Emergency physicians and administrators of EDs who are not screening for alcohol use problems should be aware of the large volume of patients who go untreated in emergency departments throughout the country. Although it may not be practical to have every physician and nurse trained in alcohol screening and intervention, the use of a simple screening tool as part of the history taking has value. In emergency departments with large patient volumes, it may be cost effective to employ an individual, who is not a part of the physician-nurse treatment team, with the primary focus on detecting AA/AD and facilitating referral. Many hospitals with substance abuse and behavioral health services may find that the ED population is an untapped source of patient revenue. For those systems, there is most certainly a break-even point at which a dedicated individual in the emergency department to detect the disease and arrange for treatment would be cost effective.

It is well documented that emergency physicians fail to detect alcohol abuse and refer. The reasons for this are unknown, but they likely have to do with more pressing problems, unrelated diagnoses, difficulty of referral, and a recalcitrant population. Perhaps the likelihood of future injury to self or others is not viewed as a health problem worthy of attention in the emergency setting. When put into the perspective that the ED treatment for injury may be those patients’ only interface with the health care system, the importance of screening for the disease among those at highest risk is obvious. Detection and referral of AA/AD patients is not, however, a foreign concept in the ED. Patients with liver disease, gastrointestinal bleeding, poor nutrition and other disorders directly related to alcohol consumption are frequently referred or transferred to a structured situation that incorporates treatment for substance abuse. But because of the widespread lack of recognition of injury as a disease, the same aggressive approach to substance related injury has not traditionally been employed.

We realize that the study was performed under ideal conditions, in that we employed trained dedicated personnel to screen, perform intervention and refer. Whether or not our results can be duplicated in settings where nurses, physicians, or social workers with other duties must perform the intervention is not known. Clearly, if we had shown no difference under ideal conditions, it would have been unlikely to work in any setting. Fortunately, screening, referral and intervention does have benefit, and implementing it in different settings awaits further analysis.

Continue to Conclusion Return to Contents