Excessive alcohol use is associated with severe acute injury and is a significant risk factor for injury and trauma recidivism (Rivara et al., 1993). In addition, alcohol use disorders and other psychiatric disorders contribute to the significant mental health burden following acute injury (Bryant et al., 2010). Service delivery models that incorporate screening, assessment, tailored treatment, and ongoing monitoring of mental health needs are required to address the complex needs of acute injury survivors (O'Donnell et al., 2008). Acute medical settings are well placed to implement alcohol screening and brief intervention (SBI) for at-risk and problem drinkers to further reduce the burden of both alcohol use and injury. Importantly, brief interventions have been shown to be an effective and cost-effective intervention to reduce alcohol consumption and injury among acute injury patients (Gentilello et al., 1999, 2005). The available evidence has led to significant changes to service delivery policy. Significantly, the American College of Surgeons Committee on Trauma has recommended that all trauma centers should incorporate SBI for alcohol as part of routine trauma care (American College of Surgeons, 2006).
The availability of a valid alcohol screening test is essential to implement SBI among acute injury patients. Ideally, a test will be brief and acceptable to both staff and patients and will accurately identify individuals at risk of problem drinking who require brief or more intensive intervention. The Alcohol Use Disorders Identification Test (AUDIT) (Saunders et al., 1993) has been recommended for use as part of an SBI protocol for use with acute injury patients (U.S. Department of Health and Human Services, 2007). However, the time taken to administer the 10-item AUDIT is a potential barrier to incorporating the full screen into routine screening procedures. As a result, studies have assessed the performance of the AUDIT-Consumption (AUDIT-C), which is a brief alcohol screening test comprising the first 3 consumption items of the AUDIT (Bradley et al., 2007; Bush et al., 1998). The AUDIT-C has been used in quality improvement initiatives in large healthcare systems (Rose et al., 2008; United States Department of Veterans Affairs, 2012).
Most studies to date that have examined the performance of the AUDIT-C to screen for problem drinking have relied upon the areas under receiver operating characteristic curves (AUROCs) to estimate the optimal cut off score based on a comparison of the sensitivity and specificity at each possible score threshold (or cut-point) on the test. Although it can prove difficult to determine the optimal cutoff score (e.g., sensitivity is sometimes more important than specificity, or vice versa, depending on the cost of a false-positive versus a false-negative screen), a total score of 4 or more for males and 3 or more for females on the AUDIT-C has been recommended as an indicator of an alcohol use disorder and/or hazardous alcohol use (Bradley et al., 2007). An alternative approach to a single cutoff score is to calculate stratum-specific likelihood ratios (SSLRs) that provide a summary measure of sensitivity and specificity for strata, or risk zones, of a screening measure. An advantage of likelihood ratios is that they are based on a ratio of sensitivity and specificity so they should be relatively stable across settings with varying prevalence of the target disorder. In addition, for acute injury (and other) patients, multilevel risk zones of the AUDIT-C based on SSLR analysis (rather than the use of a dichotomous cut-point) retains more information provided by a patient's score and thus provides a more accurate estimate of a patient's probability of an alcohol use disorder. A practical advantage of this approach is that it can help practitioners to determine the probability that any given patient has an alcohol use disorder based on their result on the AUDIT-C (Attia, 2003; Peirce and Cornell, 1993).
The focus on screening for alcohol use disorders in the current study is consistent with current practice guidelines (National Institute on Alcohol Abuse and Alcoholism, 2007), and alcohol use disorders are a priority target in acutely injured patients using evidence-based brief interventions (Zatzick et al., 2004). To date, only 2 published studies (Johnson et al., 2013; Rubinsky et al., 2010) have used SSLR analysis to examine the performance of the AUDIT-C, and in both studies the target disorder was alcohol dependence (rather than any alcohol use disorder) in a primary care sample. The objective of the current study was to use SSLR analysis to examine the performance of previously recommended AUDIT-C risk zones (based on a single dichotomous cut-point 0 to 3, 4 to 12) and derived AUDIT-C risk zones (based on SSLR analysis) to risk stratify injury patients at the time of their hospital admission with regard to their probability of having an alcohol use disorder.
- Top of page
- Materials and Methods
This study examined the potential benefits of SSLRs to estimate the probability of an alcohol use disorder among acute injury patients using AUDIT-C scores and risk zones. The focus on screening for alcohol use disorders in the current study is consistent with the U.S. National Institute on Alcohol Abuse and Alcoholism (NIAAA) guide that focuses on addressing 2 types of problem drinking: hazardous drinking and an alcohol use disorder, with the latter causing significant distress or impairment, which may require more intensive management and treatment (National Institute on Alcohol Abuse and Alcoholism, 2007).
In this study, AUROC analyses showed that the derived AUDIT-C risk zones performed better than the recommended zones, with the greater number of derived zones increasing the power of the AUDIT-C to discriminate injury patients with and without alcohol use disorder compared with the recommended 2 zones based on a single cut-point.
A practical benefit of SSLR analysis is that it provides an empirical basis for a practitioner to estimate a given patient's risk for an alcohol use disorder based on their AUDIT-C score, which in turn enables more specific feedback to be given to that patient about their alcohol use. For example, based on the derived SSLRs found in this study, the following information could be communicated to a patient with a score of 9 or greater: “Approximately 70% of people with your score on these alcohol questions experience problems linked to their alcohol use. Would it be OK for us to briefly discuss your use of alcohol to see if there's anything I can do to help?” Depending on the availability of resources and the specifics of local protocols, the next steps could include further assessment, brief intervention and follow-up, and referral for specialist treatment.
An additional practical benefit of SSLR analyses is that it can assist with the development of empirically based screening and intervention protocols to guide practitioner practices. The current findings suggest that for any injury patient with an AUDIT-C score of 0 to 3, or 4 to 5, the probability of alcohol use disorder is relatively low (5 and 14%, respectively). For any patient with a higher AUDIT-C score, the probability is 32% for a score of 6, 47% for a score of 7 to 8, and 68% for a score of 9 to 12. Moreover, the majority of patients met criteria for at-risk drinking regardless of derived risk zone, and the large majority of patients in the 3 highest derived zones (scores 6, 7 to 8, 9 to 12) met criteria for heavy drinking. These findings are useful to inform the development of a protocol with decision rules for alcohol screening and intervention among injury patients (Wade et al., 2012). To illustrate, based on the current risk-stratification findings, one might propose the following steps of an SBI protocol. First, patients with an AUDIT-C score of 5 or less (n = 647 or 64.4% of patients in the current study) receive brief health promotion advice and information on the benefits of low-risk drinking. Second, patients with an AUDIT-C score of 6 to 8 (n = 249 or 24.8% of patients) receive a brief intervention on problem drinking with follow-up. Third, patients with a score of 9 to 12 (n = 108 or 10.8% of patients) receive a brief intervention and follow-up and are considered for specialist referral for a full diagnostic assessment and more intensive intervention. These 3 respective groups of patients include 23.7% (54/228), 44.3% (101/228), and 32.0% (73/228) of all patients with an alcohol use disorder, although the patients meeting criteria for an alcohol use disorder in the first group who only receive a health promotion intervention would likely have less severe alcohol problems than those patients in the other groups.
The current study has a number of limitations. First, the relatively small number of female patients meant that gender-specific risk zones and SSLRs could not be calculated. Among female patients, 5 strata derived from each of the AUDIT-C scores did not have at least 1 individual with or without alcohol use disorder, and these “0 scores” resulted in less than an optimal number of derived zones. Second, despite the rules proposed by Peirce and Cornell (1993), there is some subjective judgment required to select cutoff points when determining the optimal number of strata in screening tests. As a result, there will inevitably be some imprecision in the derived AUDIT-C risk zones. Third, hazardous alcohol use was, in part, estimated using AUDIT item 3 which refers to the frequency of having 6 or more drinks on a single occasion, whereas the Australian guidelines (National Health Medical Research Council, 2009) recommend against drinking 5 or more standard drinks on a single occasion to reduce the risk of injury. Despite this discrepancy, the AUDIT item closely approximates the guidelines recommendation, and there is no universally recognized definition of hazardous alcohol use. Fourth, the extent to which the zones derived from the SSLR analysis, and the posttest probabilities, generalize to other injury groups is not clear.
In summary, this study used SSLR analysis to examine the performance of recommended and derived AUDIT-C screening tests to estimate the probability of alcohol use disorder among injury patients. The results of the SSLR analysis provide an opportunity to assess the performance of the AUDIT-C based on multilevel risk zones rather than relying on a single dichotomous cut-point. The risk zones derived from the SSLR analysis can inform staff–patient communications and may prove useful in the development of more efficient and effective SBI protocols in a range of settings.