Improving the identification and management of alcohol misuse is a VA priority. Alcohol misuse includes the spectrum of alcohol consumption ranging from hazardous drinking to alcohol use disorders (alcohol abuse and dependence). The VA recommends using a set of clinical strategies referred to as Screening, Brief Intervention, and Referral to Treatment (SBIRT) to identify and address alcohol misuse in primary care settings. Brief intervention (BI), a core component of SBIRT, significantly reduces alcohol consumption, morbidity, and healthcare utilization in hazardous drinkers, but its efficacy is not well-established outside of outpatient settings. In the hospital setting, nurses are well-positioned to deliver BI, but research is needed to determine the efficacy of inpatient nurse-delivered BI, particularly with hazardous drinkers. The few previous trials of BI in the inpatient setting demonstrated limited effects on alcohol consumption and alcohol-related problems, potentially due to assessment reactivity--extensive patient assessment that inadvertently raises patient awareness about drinking in both groups, mimicking the effect of BI and thus driving findings towards the null. Additionally, very few of these trials involved nurse delivery of the intervention and many included patients with alcohol use disorders, patients believed to be beyond the "therapeutic reach" of BI.
The primary goal of this 3-arm randomized controlled trial is to examine the efficacy of nurse-delivered alcohol BI with hospitalized patients who are hazardous drinkers and to identify barriers and facilitators to implementation of BI in inpatient settings. Arm 1 (BI) consists of Veterans randomized to nurse-delivered BI, Arm 2 (AC) consists of Veterans randomized to an attention control, and Arm 3 (AC-LA) consists of Veterans randomized to an attention control with limited assessment of readiness to change and adverse consequences of alcohol use so as to reduce and evaluate assessment reactivity.
Specific Aim 1 of this study is to determine the impact of a nurse-led BI on the alcohol screening status, number of drinks/week, number of binge drinking episodes, readiness to change drinking behavior, and adverse consequences of alcohol use in hospitalized hazardous drinkers.
Specific Aim 2 of this study is to formatively evaluate the process of the intervention implementation to inform the design and execution of a future, multi-site randomized effectiveness trial of the intervention.
Specific Aim 3 of this study is to identify factors, issues, and themes related to Veterans' interest and motivation for changing their alcohol consumption.
We recruited 320 hospitalized Veterans admitted to one of the three medical-surgical units at the VA Pittsburgh Healthcare System. Veterans will be included who are >21 years old, able to speak English, and are hazardous, non-dependent drinkers, as defined by criteria established by the National Institute for Alcohol Abuse and Alcoholism and by the Composite International Diagnostic Interview Substance Abuse Module. Patients randomized to Arm 1 received a three-part nurse-delivered BI. Patients randomized to Arms 2 and 3 received usual care plus healthy lifestyle brochures addressing general healthy lifestyle behaviors, such as limited alcohol consumption, tobacco cessation, and weight management.
For Specific Aim 1, we used various multivariable linear and logistic regressions that account for continuous outcomes, dichotomous outcomes, and clustering within medical units. Poisson regressions or negative binomial regressions were used if the continuous outcome measures were not normally distributed.
For Specific Aim 2, we used basic descriptive statistics in order to describe the numbers of deviations and interruptions to intervention delivery as planned. Based on interventionist field notes, we categorized the types of deviations/interruptions. We also used these descriptive statistics to describe duration of the audio-recorded intervention; the presence/absence of various techniques within the BI; the receipt/nonreceipt of additional alcohol feedback, advice, or counseling; and patient responsiveness to/perceptions of the BI. We also coded brief free-text perceptions responses using a modified grounded theory approach.
For Specific Aim 3, descriptive statistics were used to characterize the characteristics of the participants. Transcript data were analyzed using the grounded theory technique of constant comparison. Through initial line by line review of the session transcripts, open coding allowed for identification of basic concepts related to Veterans' interest and motivation for changing their alcohol consumption.
Screening, recruitment, and retention
Of the 11,876 admissions to the three medical-surgical units between December, 2012 and June, 2014, 44% (n= 5228) were pre-screened (by chart review) as eligible for the study. When clinical staff sought permission for study staff to approach the patient regarding research participation, 84% (n=4367) of those individuals declined. After meeting with study staff, 27% (n= 216) of those patients who had agreed to be approached by study staff were screened for study eligibility. Only 47% (n=102) of those screened were actually eligible for study participation; the majority of those screened ineligible (92%, n=114) were ineligible due to substance use, e.g., not meeting criteria for hazardous drinking, meeting DSM-IV criteria for alcohol dependence, illicit drug use, or current involvement in substance use treatment. Eighty percent (n=82) of the eligible patients completed informed consent and were randomized in the previously described 2:1:1 ratio; 37 were randomized to the BI arm, 23 to the AC arm, and 22 to the AC-LA arm.
Baseline characteristics and alcohol-related measures
The study arm conditions were similar for all demographic characteristics except for employment status where participants in the BI condition were more likely to be employed (p=0.01). For the alcohol related measures at baseline, participants in the BI condition also had higher scores on the Taking Steps subscale of the SOCRATES, i.e., were more likely to be taking active behavioral steps towards reducing their alcohol consumption, compared to those in the Attention Control (AC) condition (p=0.03).
Follow up assessment
Seventy-one out of 82 participants had complete data at both baseline and 6 months (86.6%). By study arm, at the 6 month follow up, 13.5% of the BI condition group and 26.1% of the AC condition group were lost to follow up, but this difference was not significant ( 2(1)=1.50; p=0.22). All participants in the AC-LA study condition were successfully followed at 6 months. Reasons for the lost to follow up were an inability to contact the participant because the participant failed to return calls and respond to written correspondence. In one case, the participant moved out of the country. No differences were seen in the baseline characteristics for those who completed the study compared to those who did not (based on simple logistic regression models).
Effects of the intervention
While within-group mean number of drinks per week and mean AUDIT-C scores significantly decreased over time regardless of condition group (p-values not shown), there were no differences by condition group over time (p=0.71 and p=0.66, respectively). The within-group mean number of binge drinking episodes over the past 30 days also decreased over time but this decrease was not significant. No differences were detected between the condition groups from baseline to the 6 month follow up for number of binge drinking episodes (p=0.79). Taking Steps mean scores increased for the BI and AC condition groups, but the increase was not significant for either condition group, nor were there differences in the rate of increase over time by condition group (p=0.72). Within-group mean scores for SIP-2R scores decreased over time but the decrease was not significant. There was also no difference by study condition (BI vs AC) over time (p=0.69) for the SIP-2R outcome. Cohen's D effect sizes comparing the BI condition to each of the usual care conditions are shown in Table 5. Only the AUDIT-C total score had a moderate to large effect for the BI to AC comparison (ES=0.49). We also fit similar models adjusting for employment status, and results were similar (data not shown). In addition, to assess the sensitivity of the results on the missing follow up data, models were run using only participants with data from both baseline and 6 months. Results of these models were similar to those using all available data.
We were unable to inform the debate about the efficacy of alcohol BI for hospitalized individuals due to a lack of statistical power in our study. However, the fact that our study was one of a growing number of clinical trials unable to meet recruitment targets (40-42) raises questions about the best trial design for informing the question of BI efficacy for hospitalized patients. Because the concept of hazardous alcohol use is relatively unfamiliar to hospitalized patients as well as their clinical caregivers, and because most U.S. hospitals do not yet routinely screen or intervene for hazardous drinking, inpatients who assent to be approached about alcohol BI research participation (and subsequently consent and enroll in the study) may be a select group of individuals already familiar with hazardous drinking as a health-related behavior, already experiencing some effects of their use, and/or already considering or making related behavior changes.
Additionally, prominent alcohol BI researchers have questioned the feasibility of highly controlled efficacy trials in inpatient medical, emergency department, and trauma care settings, citing the myriad of potential moderating variables related to treatment context and patient population. They posit that the mixed results regarding the efficacy of BI in these settings is likely due not only to differences in research design and methods across studies, but also due to heterogeneity in the patient populations, characteristics of the BI intervention itself, and the features of setting and organizational culture in which the BI is being provided, and call instead for hybrid, pragmatic, or practice-based trials that can account for these issues. Until then, the benefit and suitability of BI implementation (e.g., BI-related hospital accreditation and performance measures) will remain in question.
- Broyles LM, Wieland ME, Confer AL, DiNardo MM, Kraemer KL, Hanusa BH, Youk AO, Gordon AJ, Sevick MA. Alcohol brief intervention for hospitalized veterans with hazardous drinking: protocol for a 3-arm randomized controlled efficacy trial. Addiction science & clinical practice. 2015 May 13; 10(1):13.
- Broyles LM, Wieland ME, Confer AL, Youk AO, Gordon AJ. Holes in the pipeline: addressing recruitment challenges for an alcohol brief intervention trial in the acute care setting. [Abstract]. Substance Abuse. 2014 Jan 1; 35(2):208.
- Goplerud E, Broyles LM, Robinson E. Keys to implementing and abstracting the substance use measure set [Centers for Medicare and Medicaid Services Inpatient Psychiatric Facility Quality Reporting (IPFQR) Program Webinar]. [Cyberseminar]. 2015 Sep 1.
- Broyles LM, Wakeman SE. From evidence to action: redesigning substance use care in inpatient settings. Paper presented at: American Society of Addiction Medicine Annual Meeting; 2015 Apr 25; Austin, TX.
- Broyles LM, Wieland ME, Confer AL, Youk AO, Gordon AJ. Holes in the pipeline: addressing recruitment challenges for an alcohol brief intervention trial in the acute care setting. Poster session presented at: Association for Medical Education and Research in Substance Abuse Annual Meeting; 2013 Nov 8; Bethesda, MD.