OUP user menu

The Impact of Screening, Brief Intervention and Referral for Treatment in Emergency Department Patients’ Alcohol Use: A 3-, 6- and 12-month Follow-up

DOI: http://dx.doi.org/10.1093/alcalc/agq058 514-519 First published online: 27 September 2010

Abstract

Aims: This study aims to determine the impact of Screening, Brief Intervention and Referral for Treatment (SBIRT) in reducing alcohol consumption in emergency department (ED) patients at 3, 6, and 12 months following exposure to the intervention. Methods: Patients drinking above the low-risk limits (at-risk to dependence), as defined by National Institute of Alcohol Abuse and Alcoholism (NIAAA), were recruited from 14 sites nationwide from April to August 2004. A quasi-experimental comparison group design included sequential recruitment of intervention and control patients at each site. Control patients received a written handout. The Intervention group received the handout and participated in a brief negotiated interview with direct referral for treatment if indicated. Follow-up surveys were conducted at 3, 6, and 12 months by telephone using an Interactive Voice Response (IVR) system. Results: Of the 1132 eligible patients consented and enrolled (581 control, 551 intervention), 699 (63%), 575 (52%) and 433 (38%) completed follow-up surveys via IVR at 3, 6, and 12 months, respectively. Regression analysis adjusting for the clustered sampling design and using multiple imputation procedures to account for subject attrition revealed that those receiving SBIRT reported roughly three drinks less per week than controls (B = −3.00, SE = 1.06, < 0.05) and the level of maximum drinks per occasion was approximately three-fourths of a drink less than controls (B = -0.76, SE = 0.29, < 0.05) at 3 months. At 6 and 12 months post-intervention, these effects had weakened considerably and were no longer statistically or substantively significant. Conclusion: SBIRT delivered by ED providers appears to have short-term effectiveness in reducing at-risk drinking, but multi-contact interventions or booster programs may be necessary to maintain long-term reductions in risky drinking.

INTRODUCTION

There is substantial evidence that brief interventions (BIs) by providers in primary-care settings are an efficacious and cost-effective modality for eliminating or reducing harmful health behaviors related to alcohol abuse (Bertholet et al., 2005; Dunn et al., 2001; Fleming et al., 2002; Wilk et al., 1997). Despite the substantial numbers of emergency department (ED) patients with alcohol problems (Bernstein et al., 1996; Cherpitel, 1999; Lowenstein et al., 1998; Ly and McCaig, 2002; Whiteman et al., 2000), these techniques are used infrequently by ED providers, who typically lack the knowledge, skills and time to engage patients in health-promoting behavior change (Geller et al., 1989; Graham et al., 2000). Recent studies suggest that BI may be effective in reducing alcohol use and associated harms when used in the ED (Bazargan-Hejazi et al., 2005; Bernstein et al., 1997; Blow et al., 2006; Havard et al., 2008; Longabaugh et al., 2001; Monti et al., 1999, 2007; Walton et al., 2008). On the other hand, two well-designed and rigorously conducted ED trials failed to show intervention effects (Daeppen et al., 2007; D'Onofrio et al., 2008). Despite the preponderance of evidence suggesting some efficacy associated with brief alcohol interventions in the ED, the majority of interventions provided in these studies were not performed under real-life ED conditions or by ED providers (save for D'Onofrio et al., 2008), but by personnel—psychologists, social workers or peer advocates—not indigenous to the ED. Although it may be feasible and perhaps desirable to have extra personnel to provide the BI, ‘extenders’ are not standard practice in the ED.

A recent national study involving 14 academic EDs across the USA indicates that BIs administered by ED providers have short-term effectiveness in reducing alcohol use among ED patients drinking at risky levels (Academic ED SBIRT Research Collaborative, 2007b). Patients participating in a brief negotiated interview (BNI) with an ED provider reported consuming 3.25 less drinks per week than controls (< 0.05) and reported a level of maximum drinks per occasion of almost three-fourths of a drink less than controls (< 0.05) 3 months following enrollment. Further analysis indicated that the clinical significance of reductions in alcohol use was confined to at-risk as opposed to dependent drinkers.

Although promising, the 3-month follow-up period for this study was too brief to demonstrate the long-term effectiveness of BNI with this patient population. This study extends these results by comparing the quantity and frequency of drinking among those receiving BNI with minimally treated controls at 6 and 12 months post-intervention.

METHODS

Study design

The design of this study has been described previously (Academic ED SBIRT Research Collaborative, 2007b). Participants included 1132 adults from 14 academic EDs in the United States. The quasi-experimental comparison group design called for the recruitment of ED patients from each site in two sequential enrollment periods beginning in April 2004—the first representing the control group and the second the intervention group. The control group received screening for at-risk drinking and a written list of referral resources. Following exposure of ED providers (attending physicians, residents, nurses and nurse practitioners, emergency medical technicians and social workers) to a standardized Screening, Brief Intervention and Referral for Treatment (SBIRT) curriculum (Academic ED SBIRT Research Collaborative, 2007a), the intervention group received a BNI and a referral for treatment if indicated. Patients who agreed to participate were screened using the National Alcohol Screening Day (NASD) Primary Care Screening Form adapted for Emergency Medicine. A positive screen for at-risk drinking was defined using National Institute of Alcohol Abuse and Alcoholism (NIAAA) low-risk drinking criteria (National Institute on Alcohol Abuse and Alcoholism, 2005). Each site screened patients until a target of 40 patients per enrollment period was attained. Following enrollment, all participants received a $10 honorarium. The project was approved by the Institutional Review Boards at all participating sites. In addition, a Certificate of Confidentiality covering all sites was obtained from the NIAAA.

The intervention

The BNI, based on research on the efficacy of motivational interviewing (Burke et al., 2003; Dunn et al., 2001; Hettema et al., 2005; Miller, 1999; Miller and Rollnick, 2002; Vasilaki et al., 2006), was adapted for the ED setting (www.ed.bmc.org/sbirt; Academic ED SBIRT Research Collaborative, 2007a; Bernstein et al., 1997; D'Onofrio et al., 2005). The BNI consists of a four-step process: (a) engage the patient and seek permission to discuss alcohol use, (b) provide feedback on current drinking and consequences in relation to the NIAAA low-risk drinking guidelines, (c) assess decisional balance (by weighing the pros and cons of drinking) and readiness to change, (d) provide a menu of options and a written prescription for behavioral change with assistance in obtaining appointments or placements if desired. A more detailed description of this very structured intervention is reported elsewhere (Academic ED SBIRT Research Collaborative, 2007a).

Follow-up assessment

Three, 6 and 12 months following the initial assessment, participants completed follow-up surveys by telephone using an Interactive Voice Response (IVR) system at the University of Connecticut Health Center. Of the total patients enrolled, 699 (63%), 575 (52%) and 433 (38%)  completed follow-up surveys via IVR at 3, 6 and 12 months, respectively. Patients received $20 compensation for each follow-up survey completed.

Measures

The NASD screening form is a one-page questionnaire that includes three standard questions regarding aspects of alcohol use. These include: the frequency of alcohol use (e.g. the number of days per week the respondent consumes alcohol), the quantity of alcohol use on a typical day during the past 12 months and the maximum number of drinks on any given day during the past month. Patients' were considered ‘at-risk’ if, by self-report, they exceeded the NIAAA guidelines for low-risk drinking (National Institute on Alcohol Abuse and Alcoholism, 2005). For comparability with other studies of drinking interventions, the frequency of alcohol use was multiplied by the quantity of alcohol use on a typical day to yield an estimate of the average number of drinks per week. The NASD form also includes the CAGE, with two positive responses to this four-item measure indicating possible alcohol-dependency (Cherpitel, 1997; Ewing, 1984; O'Brien, 2008). Once enrolled, participants completed a self-report intake form which required 5–10 min to complete. The intake form captured basic demographic information (e.g. marital status, education, language spoken at home, work status, living situation and health insurance status).

The 3-, 6- and 12-month follow-up surveys contained the same baseline questions regarding drinking behavior in the past 30 days (e.g. quantity, frequency and maximum use).

Data analysis

To account for the clustered sampling design in which patients were nested within sites, we used SUDAAN 9.0.1 software (Research Triangle Institute, 2005) to perform regression analyses of intervention effects. To account for both subject attrition and item non-response in our analysis, we performed multiple imputation (Rubin, 1987), a simulation-based approach that generates multiple plausible values for each missing element in order to represent the inherent uncertainty in the missing data (Schafer and Graham, 2002). We used the Markov Chain Monte Carlo (MCMC) method in SAS, under the assumption that the data were missing at random (i.e. where missingness can be accounted for by variables measured in the study) to produce 20 imputed datasets. Diagnostic plots indicated that the MCMC converged well. With 20 iterations, the efficiency of estimates was 98%.

RESULTS

Data on enrollment, retention and demographic characteristics of the sample are presented in Fig. 1 and in Tables 1 and 2. Of the 1132 patients who consented and were enrolled in the study (551 intervention, 581 control), 32% (n = 357) were women, 37% (n = 416) were Black, 20% (n = 225) were Hispanic and 39% (n = 431) were White. Average age was 35.8 (SD = 12.3), with an age range of 18–86 years old, and 17% (n = 191) were married. Bivariate analyses were conducted to assess the comparability of the intervention group and control group in terms of gender, race/ethnicity, age, education, marital status, employment status, living situation, health insurance status and drinking characteristics. Except for homelessness (χ2 = 11.6, df = 1, < 0.05) and race (χ2 = 12.0, df = 3, < 0.05), there were no other differences in the baseline characteristics of the two groups. Intervention and control groups did not differ in the baseline measures of the typical number of drinks per week and maximum number of drinks per occasion.

View this table:
Table 1.

Baseline demographic characteristics (n = 1132)

Total (n = 1132)Control (n = 581)Intervention (n = 551)
Gender (% male)68%68%69%
Racea
 Black37%35%40%
 White39%40%37%
 Hispanic20%22%18%
AgeEmbedded ImageEmbedded ImageEmbedded Image
Health insurance (% yes)52%55%49%
Education
 Not High School Grad27%30%23%
 High School Grad33%31%36%
 Some college/Tech29%29%30%
 College Grad11%10%11%
Married17%15%19%
Employment status
 Full time (35+ h)35%34%36%
 Part time (<35 h)16%15%17%
 Not working49%52%47%
Homelessa11%14%9%
Past Alcoholic Treatment22%23%21%
  • Notes: Embedded Image denotes mean value.

  • aControl and Intervention groups differ (< 0.05).

View this table:
Table 2.

Baseline drinking characteristics (n = 1132)

CharacteristicMeanConfidence intervalMinimumMaximum
Days drink
 Total3.4(3.2, 3.5)07
 Control3.4(3.2, 3.6)07
 Intervention3.3(3.1, 3.5)07
Typical weekly drinks
 Total23.5(22.0, 24.9)084
 Control24.1(22.1, 26.1)084
 Intervention22.8(20.8, 24.9)084
Maximum drinks
 Total8.6(8.4, 8.8)012
 Control8.5(8.2, 8.8)012
 Intervention8.7(8.4, 8.9)012
CAGE score
 Total1.8(1.8, 1.9)04
 Control1.9(1.8, 2.0)04
 Intervention1.8(1.7, 1.9)04
  • Note: Treatment and controls were not statistically different on any of these measures (< 0.05).

Figure 1.

Profile of an ED SBIRT trial through 12-month follow-up

Coefficients and standard errors from regression models predicting self-reported drinking at 3, 6 and 12 months following enrollment are presented in Table 3. Compared with the control group, those receiving the BNI reported significantly lower levels of both typical drinks per week and maximum drinks per occasion at the 3-month follow-up, controlling for baseline drinking and patients' race and homeless status—the only two demographic control variables that were associated with intervention status in preliminary analysis (Academic ED SBIRT Research Collaborative, 2007b). Those receiving the BNI reported consuming roughly three drinks less per week than controls (B = −3.00, SE = 1.06, < 0.05) and a level of maximum drinks per occasion approximately three-fourths of a drink less than controls (B = −0.76, SE = 0.29, < 0.05) at 3 months. As a result of these changes in drinking patterns, 26% of intervention participants no longer exceeded the NIAAA low-risk threshold at the 3-month follow-up, compared with 17% of control participants (data not shown). The magnitude of the intervention effects in this analysis is ∼6–8% lower than that reported in our previous analysis of 3-month outcomes, which did not use multiply imputed data to adjust for subject attrition (Academic ED SBIRT Research Collaborative, 2007b), suggesting that the intervention was less likely to be effective among those lost to follow-up. At 6 months post-intervention, the impact of the BNI on self-reported drinking had almost completely dissipated, as BNI coefficients were of very small magnitude and were no longer statistically significant for either typical weekly drinks (B = −0.18, SE = 1.21) or maximum drinks per occasion (B = −0.31, SE = 0.29). BNI coefficients remained insignificant for typical weekly drinks (B = −0.51, SE = 0.94) and maximum drinks per occasion (B = 0.01, SE = 0.34) at 12 months post-intervention. Tests for a statistical interaction involving baseline CAGE status and the BNI in predicting both typical drinks per week or heavy drinking episodes were not statistically significant, indicating that the efficacy of the BNI did not differ among risky versus possibly dependent drinkers at either 3, 6 or 12 months post-intervention (data not shown).

View this table:
Table 3.

Effects of SBIRT on drinking behavior at 3, 6 and 12 months using multiply imputed data

Typical weekly drinksMaximum drinks per occasion
3 Months6 Months12 Months3 Months6 Months12 Months
BSEBSEBSEBSEBSEBSE
Intercept12.21*1.589.74*1.459.02*1.194.76*0.424.43*0.614.710.54
SBIRT−3.00*1.06−0.181.21−0.510.94−0.76*0.29−0.310.290.010.34
Baseline Drinks0.26*0.030.19*0.030.11*0.030.35*0.040.32*0.050.230.05
Black−0.561.59−0.851.57−1.021.39−1.04*0.33−1.04*0.41−0.810.42
Hispanic−6.04*1.81−3.141.55−2.491.51−1.55*0.43−1.34*0.43−1.180.43
Other0.532.971.233.390.232.75−0.160.65−0.260.93−0.240.70
Homeless3.461.851.851.771.551.790.410.420.260.460.080.46
  • *< 0.05.

LIMITATIONS

The limitations of this study were discussed extensively in our prior work (Academic ED SBIRT Research Collaborative, 2007b). Of particular relevance to this long-term follow-up, however, were the suboptimal retention rates, particularly at 12 months post-intervention. The rate of attrition over time observed in this study is primarily attributable to the transient nature of the ED study population. However, this is also a function of the translational character of this study, which requires real-world conditions for implementation (e.g. homeless patients, dependent drinkers, those without phones). In addition, participation of patients from a diverse group of medical institutions was critical to demonstrating the ability to translate this intervention to real-life ED settings, but did not allow us to achieve efficiencies in data collection that could be obtained in a small group of homogeneous sites. This problem was addressed, however, through the use of multiple imputation procedures, a relatively new and sophisticated approach to the treatment of missing data. Simulation studies suggest that these procedures can be very effective in minimizing the effects of missing data on parameter estimates provided assumptions concerning the nature of the missing data can be supported (Schafer and Graham, 2002).

In addition, although attrition rates in the intervention and control groups were virtually identical at 3 and 6 months, the intervention group had a significantly higher rate of attrition at 12 months (35% retained vs. 42% in the control group). Because we included a term for experimental condition in the imputation model, however, the 12-month imputed data still conform to the MAR assumption, since missingness is dependent upon an observed variable (Schafer and Graham, 2002).

DISCUSSION

Our previous analysis of data from patients receiving care at 14 academic EDs in the USA indicated that BNIs delivered by ED staff were associated with significant reductions in drinking behavior 3 months post-intervention. These effects, the first observed among patients receiving interventions delivered by ED providers, were of comparable magnitude to those observed among primary-care patients exposed to BIs (Whitlock et al., 2004). However, results from a 6- and 12-month follow-up of ED patients enrolled in this study revealed substantial decay in these effects over time. Estimates of the differences between patients in the intervention and control groups showed that reductions in the typical number of drinks per week and the maximum number of drinks per occasion were neither statistically significant nor substantively meaningful at 6 and 12 months post-intervention.

The absence of intervention effects at 12 months, which is consistent with results from other recent ED studies (Bernstein and Bernstein, 2008; Daeppen et al., 2007; D'Onofrio et al., 2008; Roudsari et al., 2009), highlights the importance of multi-contact interventions and/or ‘booster’ sessions for maintaining the impact of BIs on risky drinking over the long term. A review of the substantial literature on BIs in primary-care settings provides strong support for this conclusion, as virtually all of the studies reporting effects persisting beyond 3 months utilized multi-contact interventions (Dunn et al., 2001; Fleming et al., 2002; Hettema et al., 2005; Vasilaki et al., 2006). According to a recent Cochrane database systematic review of alcohol intervention studies (Kaner et al., 2007), effects from all single contact interventions decayed markedly after 3 months. Given the success of BIs in primary-care practice, active referral to primary-care providers offers an opportunity to enhance ED interventions. In light of the fact that half the enrollees in this study were likely to be dependent drinkers, ED system changes to facilitate access to substance abuse treatment are also needed (Bernstein et al., 1997).

It is important to note, however, that given the difficulties we encountered with attrition over time, multi-contact interventions may be too difficult to implement effectively with this population in the absence of very restrictive exclusion criteria for participation. Multiple imputation procedures, although very useful in dealing with problems of non-response, may be of limited utility if non-response is accompanied by or a consequence of the lack of adherence to complex multi-contact interventions.

Funding

Support for this project was provided by the National Institute for Alcohol Abuse and Alcoholism (R21 AA015123, R25AA014957 & RO3 AA01511-14).

STUDY GROUP AUTHORSHIP PAGE

Boston University: Edward Bernstein MD, Judith Bernstein RNC, PhD, James Feldman MD, William Fernandez MD, Melissa Hagan MPH, Patricia Mitchell RN, Clara Safi RNP; Brown University: Robert Woolard MD, Mike Mello MD, Janette Baird PhD, Cristina Lee, PhD; Charles R. Drew University: Shahrzad Bazargan-Hejazi PhD, Brittan A. Durham MD; Denver Health Medical Center: Kerry Broderick MD, Kathryn A. LaPerrier CSW; Emory University: Arthur Kellermann, MD, MPH, Marlena M. Wald MLS, MPH; Howard University: Robert E. Taylor MD, PhD, Kim Walton PhD, Michelle Grant-Ervin MD; Tufts University: Denise Rollinson, MD, David Edwards; University Of California, San Diego: Theodore Chan MD, Dan Davis MD, Jean Buchanan Marshall MS, RN; University Of Connecticut Health Center: Robert H. Aseltine, Jr. PhD, Amy James, PhD; Elizabeth A. Schilling, PhD; Khamis Abu-Hasaballah, PhD; Ofer Harel, PhD; Jun Yan, PhD; University Of Medicine And Dentistry Of New Jersey Robert Wood Johnson Medical School At Camden: Brigitte M. Baumann MD, Edwin D. Boudreaux PhD; University Of Michigan: Ronald Maio DO, MS, Rebecca Cunningham MD, Teresa Murrell RN, CCRC; University Of New Mexico: Michael Bauer, David Doezema, MD; University Of Southern California: Deirdre Anglin MD, MPH, Adriana Eliassen RN; University Of Virginia: Marcus Martin MD, Jesse Pines MD, Leslie Buchanan NP, James Turner MD; Yale University: Gail D'Onofrio MD, Linda C Degutis DrPH, Patricia Owens MS.

REFERENCES

View Abstract