Background
Over 29% of U.S. adults have had an alcohol use disorder (AUD) at some time in their lives [
1,
2]. Most will never seek alcohol treatment, but many will receive health care in primary care settings [
1,
2]. Experts have recommended patient-centered approaches to management of AUDs in primary care for over 25 years [
3‐
5], but remarkably little research has addressed how practically to achieve this and whether it improves patient outcomes [
6‐
10].
In 2009, we designed the Considering Healthier drinking Options in primary CarE (CHOICE) intervention to manage AUDs as a chronic condition in primary care patients. The CHOICE model of care is a year-long intervention offered by nurses and based on (1) the Chronic Care Model, (2) principles of patient-centered care [
11], and (3) promising approaches to addressing AUDs in primary care [
12‐
23]. We hypothesized that, although most primary care patients with AUDs do not engage in alcohol treatment [
2], many would be interested in addressing alcohol use as part of patient-centered care from a registered nurse. We further hypothesized that the resulting increased engagement in AUD care would lead to reductions in drinking and related symptoms, and thereby improve health outcomes. In 2010, the National Institute on Alcohol Abuse and Alcoholism (NIAAA) funded a trial to test the effectiveness of the CHOICE intervention in primary care patients at high risk for AUDs in the Veterans Affairs (VA) health care system.
This report has several purposes. First, it summarizes the conceptual and scientific foundations of the CHOICE model of patient-centered care for AUDs. Second, it describes the design of the CHOICE trial and intervention, including all elements needed for replication of the intervention and comparison to other primary care counseling interventions for AUDs, as advocated by experts [
24,
25]. Finally, it reports results of recruitment (completed October 2014) and patient-reported characteristics of the enrolled sample at baseline. Main outcomes of the CHOICE trial were under analysis at the time this report was submitted and will be reported in a follow-up paper.
Discussion
The purpose of this report was to provide a rationale and description of the CHOICE intervention and trial as recommended by experts [
24]. The CHOICE intervention was designed to improve outcomes of primary care patients at high risk for AUDs by offering patient-centered chronic care management using MI skills and shared decision-making over 12 months. Patients were recruited and engaged in the intervention regardless of their recognition of problems due to drinking, severity of unhealthy alcohol use, or readiness to change drinking. The CHOICE trial will evaluate the effectiveness of the nurse care management intervention for resolving high risk drinking and alcohol-related symptoms. Baseline data show that the trial recruited a sample of patients with varying severity of unhealthy alcohol use, readiness to change, and a significant burden of mental health and other substance use conditions.
Randomized controlled trials of interventions to manage AUDs in primary care patients have had varying designs and results [
6,
7,
9,
14,
59]. Understanding the implications of their varying results requires a detailed knowledge of the interventions, populations studied, and trial designs. Unfortunately, all too often, neither is adequately characterized in relatively brief reports in medical journals [
24]. As a result, experts have recommended several frameworks to improve reporting and interpretation of the literature on behavioral interventions that can be offered in primary care [
24,
25,
121,
122]. One recommendation is that a checklist, the Template for Intervention Description and Replication (TIDieR), be used to ensure adequate description of behavioral interventions and the design of trials conducted in primary care [
25]. This report includes those TIDieR elements—intervention components, materials used in the intervention, procedures and processes used, intervention providers and their expertise, modes of delivery, locations, whether the intervention was personalized or adapted, and how adherence was assessed—that are known before blinded analyses of main and secondary outcomes are completed.
Randomized controlled trials of behavioral interventions in primary care samples fall on a continuum from purely “explanatory,” efficacy trials to more “pragmatic” effectiveness trials, and different dimensions of a trial can vary in the extent to which they are explanatory or pragmatic [
123]. The CHOICE trial is considered a blended pragmatic effectiveness-efficacy trial based on the 10 indicators of the “Pragmatic-Explanatory Continuum Indicator Summary (PRECIS) [
123]. Pragmatic elements of the CHOICE trial include: non-stringent criteria for inclusion/exclusion, the intervention which was flexible and not monitored systematically for fidelity, and use of telephone interviews for outcomes, all designed to maximize flexibility and thereby external validity. Exclusion criteria were limited, and the primary inclusion criterion relied on the frequency of heavy drinking. Frequent heavy drinking was used as an inclusion criterion because it is strongly associated with AUDs [
70,
124] commonly assessed by brief alcohol screens in primary care [
125], and because AUDs are often under-recognized in primary care [
69], and patients under-report AUD symptoms due to stigma [
126]. Consistent with shared decision-making principles, the CHOICE intervention was adapted to patient preferences regarding drinking goals, frequency of contacts, setting (telephone vs in person), and treatment options. The comparison condition was usual primary care with no additions. Measurements were pragmatic in that main outcomes were measured by telephone at diverse times of day, seven days a week, to increase flexibility and acceptability to patients. Only limited secondary outcomes were collected from patients at three months, with EHR data used to measure other secondary outcomes to decrease participant burden and assessment reactivity. Finally, embedding the trial in the VA health care system enabled future use of annual alcohol screening and the EHR to evaluate longer term outcomes after the main patient reported outcomes at 12 months. While these pragmatic elements of the trial were expected to limit its explanatory potential, especially if the intervention was not effective, explanatory elements below were designed to counter those limitations.
Despite the above pragmatic elements of the CHOICE trial design, other elements were less pragmatic. The recruitment procedures included at least seven interactions between research staff and all patients (including the control group) before the 12 month assessment: an invitation letter; a telephone screen; an appointment reminder letter with consent form; an in-person baseline assessment; a letter with results of randomization; a 3-month reminder letter; and a telephone assessment at 3 months, with each contact often requiring many attempts to reach or schedule the patient. These could act as a barrier to participation, but also as a catalyst to change by increasing self-awareness of alcohol-related problems in both groups [
76,
77,
122,
127]. The 13-page written informed consent document, patient-level randomization, and measurement of main outcomes with patient interviews were also on the explanatory end of the continuum. The randomly selected No Contact Control Group of men will allow evaluation of the potential impact of recruitment and assessment at baseline and three months, allowing explanatory analyses of changes in drinking in the usual care group—albeit with pragmatic measures. Elements of the intervention, which were also less pragmatic, included having CHOICE nurses manage only AUDs while referring other medical or mental health issues to primary care teams, not requiring that primary care providers provide any of the alcohol-related care including prescription of AUD medications, location of the nurses outside primary care clinics, and collaboration between CHOICE RNs and primary care providers limited largely to medications or medical issues.
The CHOICE trial included 304 primary care patients, a low proportion of those who were initially identified from the EHR for screening and recruitment for the trial. Of the 3952 patients initially identified, 1807 (46%) either declined before eligibility was determined (1382) or were never reached (425). Of the 571 patients potentially eligible based on phone screening for the trial, 53% enrolled. Most patients who were ineligible did not drink above levels required for inclusion. Given all the demands of screening, recruitment, and measurement above, the low enrollment rate among all VA patients eligible for screening and recruitment was as expected and similar to the proportion of patients who engaged in prior effectiveness-efficacy trials of collaborative care [
100]. CHOICE was not designed to maximize reach of the intervention unlike many pragmatic trials [
128‐
131]. At the same time, our analyses show that the 304 enrolled patients were similar in age and gender to the 2074 who declined or were never reached but had higher mean AUDIT-C scores indicative of more severe unhealthy alcohol use [
132]. Given the positive association between severity of unhealthy alcohol use and readiness to change [
133], we suspect recruited patients also had greater readiness to change than those who were not recruited. Despite efforts to oversample women, the CHOICE trial did not enroll adequate numbers of women to generalize trial findings to women.
Two prior approaches to chronic care management of AUDs in primary care, tested in recent randomized controlled trials, had markedly differing outcomes from each other: one showing a benefit of chronic care management and the other finding no benefit [
6,
7]. The two trials both blended pragmatic elements (intended to increase generalizability) with explanatory elements (intended to improve internal validity) but the samples, interventions, and comparisons in those studies differed in important ways from the CHOICE trial. A trial in the VA conducted by Oslin et al. [
6] recruited primary care patients who met criteria for DSM-IV alcohol dependence irrespective of their readiness to change (excluding patients with other substance use and psychotic disorders). In contrast, the AHEAD trial by Saitz and colleagues included patients with DSM-IV AUD or other substance use disorders recruited outside primary care without exclusions, and many were seeking treatment. The median readiness to change was 10 (on a scale of 1–10) [
7]. CHOICE, in contrast, recruited primary care patients with frequent heavy drinking, irrespective of comorbidity or symptoms of AUDs, and although most patients met criteria for AUDs, the median readiness to change was 5. The intervention in the Oslin trial was delivered by behavioral health specialists (nurse or psychologist) [
6], and the intervention in AHEAD was delivered by an interdisciplinary team (nurse, internist, social worker and psychiatrist) [
7], while the CHOICE intervention was delivered by non-specialist nurses, supported weekly by an interdisciplinary team. The comparators of the three studies also differed; the Oslin study compared the intervention to referral to treatment [
6], whereas AHEAD compared the intervention to referral to specialty addictions treatment and augmented primary care [
7]. In contrast, the comparator in the CHOICE trial was usual primary care within a health care system that provided access to the full continuum of addiction specialty care. When the CHOICE trial results are available, thorough examination of the differences in the three studies may help identify promising elements of patient-centered care management of AUDs.
Authors’ contributions
The following co-authors contributed to design of the CHOICE trial: KAB, EJL, JOM, AJS ECW, EH, and DRK. The following co-authors contributed to implementation, delivery, and innovation regarding the intervention: KAB, EJL, SJR, CEA, JOM, AJS, RMC, DMG. The following co-authors contributed to integrity of the data and analyses: LJC, JFB, AKL, JER, RMT, GL. All co-authors contributed to the design and editing of this manuscript.