Strategies to increase implementation of pharmacotherapy for alcohol use disorders: a structured review of care delivery and implementation interventions

Background Effective medications for treating alcohol use disorders (AUD) are available but underutilized. Multiple barriers to their provision have been identified, and optimal strategies for addressing and overcoming barriers to use of medications for AUD treatment remain elusive. We conducted a structured review of published care delivery and implementation studies evaluating interventions that aimed to increase medication treatment for patients with AUD to identify interventions and component strategies that were most effective. Methods We reviewed literature through May 2018 and used networking to identify intervention studies with AUD medication receipt reported as a primary or secondary outcome. Studies were identified as care delivery studies, characterized by patient-level recruitment and willingness to be randomized to candidate treatment options, and implementation studies, characterized by inclusion of all patients treated at sites involved in the study. Each identified study was independently coded by two investigators for strategies used, guided by a published taxonomy of implementation strategies. All authors reviewed coding discrepancies and revised codes based on consensus. After reaching internal consensus, we solicited feedback from lead investigators on studies to code additional strategies. We reviewed implementation strategies used across studies to assess their relationship with medication receipt, as well as alcohol use outcomes, as available. Results Nine studies were identified: four RCTs of care delivery interventions, four quasi-experimental evaluations of large-scale implementation interventions, and one quasi-experimental evaluation of a targeted single-site implementation intervention. Implementation strategies used were variable across studies; no strategy was universally used. Effects of the interventions on receipt of AUD pharmacotherapy and alcohol use outcomes also varied. Three of four care delivery interventions resulted in increased receipt of AUD medications, but only one of these three improved alcohol use outcomes. One large-scale and one single-site implementation intervention were associated with increased AUD medication receipt, and these studies did not assess alcohol use outcomes. Patterns of implementation strategies did not clearly distinguish studies that successfully increased use of pharmacotherapy versus those that did not. Conclusions Our review did not reveal strategies most effective for implementing AUD medications. Interventions designed to overcome identified barriers may have missed the mark, or differences in the intensity or targets of strategies may matter more than differences in strategies. Further research is needed to understand effective implementation methods and to better understand patient-level perspective, preferences and barriers to receipt of medications.


Introduction
Alcohol use disorders (AUD) are common and associated with significant morbidity and mortality [1][2][3], but are substantially undertreated. In 2013, 16.6 million U.S. adults met diagnostic criteria for an AUD, but research suggests only 7.8% received any formal treatment [4]. One of the major gaps in treatment for AUD is the significant under-utilization of medications that are effective for treating AUD [1,5,6]. Three medications-disulfiram, acamprosate, and naltrexone (both oral and injectable)-have FDA approval specifically for the treatment of AUD, and topiramate has strong meta-analytic support [7]. Efforts to increase treatment of AUD with medications is motivated in part because the modality may address many reported barriers to receiving any formal AUD treatment [4,8]. For instance, psychosocial treatments are often offered in group settings, heightening stigma-related issues for some patients, whereas medications can be provided on an individual basis [9]. In addition, patients may not be ready to abstain [8,10]. Further, though this may be shifting over time [11,12], many treatment programs view abstinence as the ultimate goal [8], whereas abstinence is not required with all medications and reduced drinking can be a goal of medication treatment [9]. Finally, AUD medications can be offered across healthcare settings, including primary care, which has been highlighted as an optimal setting for expansion of care for AUD [8,13,14].
Despite the promise of medication treatment for addressing several known barriers to AUD treatment and national recommendations encouraging medications be made available to all patients with AUD [15,16], rates of pharmacotherapy for AUD remain extremely low. Among patients with AUD, 4-12% are treated pharmacologically [1,6,[17][18][19][20][21]. Among subsets of patients with AUD and co-occurring schizophrenic, bipolar, posttraumatic stress or major depressive disorder, receipt of medications for AUD ranged from 7 to 11%, whereas receipt of medications for the comorbid disorder ranged from 69 to 82% [19]. This gap in the quality of AUD treatment is well known, and the substantial barriers to provision of AUD medications in diverse contexts have been described [22][23][24][25][26][27]. However, the optimal strategies for addressing these barriers and increasing use of medications for AUD treatment remain elusive.
In recent years, two related lines of research have contributed to knowledge regarding strategies to increase use of medications to treat AUD: evaluations of care delivery interventions and evaluations of implementation interventions. Care delivery interventions typically focus on improving patient-level clinical outcomes (e.g., reduction in heavy drinking days or abstinence from alcohol use), but often secondarily assess patient-or clinician-level process outcomes focused on treatment receipt (e.g., engagement in pharmacotherapy for AUD). Implementation interventions are typically designed to improve patient-or clinician-level process outcomes, but sometimes secondarily include patient-level clinical outcomes when the evidence for the effects of the underlying practice is weak (so called Hybrid I studies) [28]. Other key differences exist between these types of research that may influence both clinical and process outcomes. Most importantly, care delivery interventions typically involve recruitment of patients who are willing to be randomized to the treatment arms contained within the new care delivery model. Thus, these trials may be restricted to patients who are at least open to, if not actively interested in, treatment for AUD. On the other hand, evaluations of implementation interventions typically recruit and intervene on clinical entities (e.g., providers, clinics, hospitals) who serve large groups of patients who likely have more variable interest in treatment. Further, evaluations of care delivery interventions are typically designed to establish the effectiveness (or lack thereof ) of particular care delivery models. Thus, these studies generally put significant effort and resources into ensuring fidelity to the care delivery model. On the other hand, implementation evaluations are often trying to establish the effectiveness of bundles of strategies (interventions) to increase uptake of practices that do not depend on external research resources. Thus, evaluations of implementation interventions may measure fidelity as a process outcome but typically exert less direct control [29].
Even though care delivery and implementation interventions differ in terms of methodology, patient inclusion criteria, and primary outcomes, they may evaluate the effectiveness of the same underlying implementation Conclusions: Our review did not reveal strategies most effective for implementing AUD medications. Interventions designed to overcome identified barriers may have missed the mark, or differences in the intensity or targets of strategies may matter more than differences in strategies. Further research is needed to understand effective implementation methods and to better understand patient-level perspective, preferences and barriers to receipt of medications.
Keywords: Alcohol use disorders, Medication assisted treatment, Pharmacotherapy, Naltrexone, Acamprosate, Disulfiram, Implementation strategies, Alcohol treatment strategies, such as reorganizing, supplementing, or intervening on existing models of care [29]. The fact that the same component implementation strategies (e.g., audit and feedback) have been evaluated by these different research designs with very different patient populations affords an opportunity to take stock of the effectiveness of these interventions, and to distill insights into which designs, contexts, and component strategies appear to drive outcomes. Therefore, our goal was to conduct a structured review of published evaluations of care delivery and implementation interventions that have either primarily or secondarily aimed to increase use of pharmacotherapy for patients with AUD, with the goal of identifying component strategies that may be effective in increasing pharmacologic treatment of AUD. Our review was guided by an existing taxonomy of implementation strategies and terms identified via a three-round modified-Delphi process [30]. The purpose of our review was to learn which components have been tried most commonly and which strategies might be associated with larger effects. Also, due to the fact that evaluations of care delivery interventions exert greater efforts to ensure fidelity and include patients willing to be randomized, we hypothesized that higher adoption of medications for AUD will be observed in those contexts compared to implementation interventions, which typically aim to intervene on clinician and patient populations with greater variability in treatment motivation, knowledge, and preferences.

Methods
For this structured literature review, we sought to identify published evaluations of care delivery and implementation interventions reporting effects on receipt of medication treatments for patients with AUD. We reviewed literature through May 2018. Studies were identified via searching PubMed, Google Scholar, and PsychInfo with relevant search terms (e.g., pharmacotherapy, alcohol use disorder medications, AUD medications, naltrexone, Acamprosate, disulfiram, medication-assisted treatment). We also reviewed reference lists from identified studies to identify additional studies that may have been missed by our search. Finally, because we have personally conducted and/or served as co-investigators on related studies, additional studies were also identified via networking. Once identified, each individual article was coded for implementation strategies used, as guided by Powell et al. 's refined compilation of implementation strategies resulting from the Expert Recommendations for Implementing Change (ERIC) project [30]. All articles were independently reviewed and coded by two investigators (EW and TM). When multiple articles and/or published protocols or commentaries were identified that described a single intervention and/or implementation effort, these articles were aggregated to the level of the intervention (e.g., three studies had adjoining published protocol papers, which were coded under the umbrella of a single study). Once coded, all authors met to review coding discrepancies, discuss interpretation of codes, arrive at consensus, and revise individual codes based on consensus.
After reaching internal consensus on coding, we reached out to the lead or senior author of each study to ask whether our codes aligned with their understanding/interpretation of their study and associated report. We shared Powell et al's description of strategies and asked them to review our coding to see if they thought we had missed or miscoded anything. Finally, process (e.g., rates of prescribed AUD pharmacotherapy) and alcohol use outcome data were extracted from each study and described. All authors reviewed the coding of implementation strategies against study outcomes data to qualitatively identify sets of implementation strategies that might have been be most effective for increasing provision of AUD medications and report whether interventions that increased AUD pharmacotherapy also improved alcohol use outcomes.

Results
Our literature review identified nine studies that evaluated interventions to primarily or secondarily increase utilization of pharmacotherapy for AUD. Four were randomized clinical trials of care delivery interventions designed to improve alcohol-related outcomes [31][32][33][34][35][36][37][38]. Four were quasi-experimental evaluations of large-scale implementation interventions designed to increase medication receipt [39][40][41][42][43], and one was a quasi-experimental evaluation of targeted implementation intervention in a single-site [44]. Two additional studies were identified but not included. The first reported on a largescale implementation intervention designed to increase screening and brief intervention for unhealthy alcohol use and secondarily assessed whether the implementation was associated with increased receipt of AUD medications among those who screened positive [45]. However, it was not clear how many of the patients who screened positive met diagnostic criteria for AUD and thus would have been eligible for medication treatment, and, though findings regarding medication use were summarized, detailed data were not reported. The second report was a description of a demonstration project to implement extended release naltrexone in Los Angeles County, but no evaluation of the program's effect on receipt of medication treatment among patients with AUD was reported [46]. Table 1 presents implementation strategies identified by our internal coding process across each identified   study (labelled with X). All lead or senior authors of studies responded to our request for review of the codes and added additional codes (labelled with an O). Implementation strategies used were variable across the studies, and no strategy was used across all studies ( Table 1). The most frequently used strategies were assessing readiness and identifying barriers and facilitators, distributing educational materials, facilitating relay of clinical data to providers (audit and feedback), and providing ongoing consultation. Strategies less frequently used involved payment and/or incentives or changes in laws and/or credentialing and licensing.
The effects of the interventions on receipt of AUD pharmacotherapy were also variable across studies ( Table 2). In three of the four randomized evaluations of care delivery models [31][32][33], the interventions were associated with varying magnitude of increased receipt of AUD medications. At follow-up, treatment group rates of medication receipt ranged from 13 [36] to almost 70% [31]. The latter study, Oslin's Alcohol Care Management model [31], was the only approach to significantly increase receipt of AUD medications and improve patient-level alcohol use outcomes ( Table 2). Two of the four implementation interventions [40,41] were associated with increased AUD medication receipt. While Ornstein's Practice Partner Research Network-Translating Research Into Practice (PPRNet-TRIP) intervention appeared to have small early effects, proportions of patients receiving medications were so low that continued evaluation over time was not possible [41]. The Veterans Health Administration's (VA) Academic Detailing Program appeared to increase rates of AUD medication receipt from 4.6 to 8.3% among patients with AUD [40]. Receipt of AUD medications also appeared to increase in in a single VA facility after implementation of a group medication management program attended by patients taking and considering medication treatment [44].
Patterns of implementation strategies did not clearly distinguish studies that successfully increased use of pharmacotherapy versus those that did not.

Discussion
Nine studies have evaluated the effects of care delivery or implementation interventions designed to increase active consideration and use of pharmacologic treatment options for patients with AUD. The interventions varied widely in context, intensity, target populations, and the underlying strategies, though many strategies were shared across studies, regardless of design (care delivery or implementation intervention). As hypothesized, care delivery interventions, targeted on patients willing to be randomized, were associated with much larger and more consistent improvements in rates of medication receipt compared to implementation interventions targeted at the overall population of patients with AUD. Among the care delivery interventions evaluated, three out of four increased use of medications. However, of these three, only Oslin's Alcohol Care Management intervention improved initiation of medications for AUD with more than one third of enrolled patients (69%) and improved in patient-level alcohol use. This trial may have been distinct from the others in its recruitment approaches-patients were recruited with the knowledge that the intervention aimed to provide pharmacologic treatment [31].
Among the implementation interventions evaluated, only the VA Academic Detailing Program [40] showed significant promise in increasing rates of medication receipt. It may be noteworthy that, compared to the other implementation interventions, the VA Academic Detailing Program was very labor intensive and targeted to diverse clinical settings with a high density of patients with AUD, not just primary care. The study of group medication management visits, intended as a means to increase prescribing capacity and educate patients who were considering medication treatment, [44] showed signals of effectiveness in one VA facility with a highly motivated champion. Interestingly, group settings have previously been identified as a barrier to receiving treatment for AUD, but appeared to facilitate increased treatment receipt among persons already seeking treatment. This intervention should be more rigorously evaluated in contexts where the primary barrier is low capacity to provide medication management.
A major goal of this review was to identify the underlying implementation strategies that were positively associated with larger effects. We categorized strategies based on published reports, but then solicited feedback from the intervention designers. There was substantial heterogeneity of strategies and some heterogeneity of effects, but no clear mapping of strategies to effectiveness was apparent. This process nonetheless proved informative by highlighting potential limitations of using of Powell et al. 's taxonomy to classify implementation strategies [30]. Specifically, strategies listed in the taxonomy appeared not be hermeneutically distinct, causing frequent difficulty classifying strategies as one or another. Relatedly, strategy definitions are somewhat inexplicit and hard to confidently map onto what was done in the interventions, resulting in different decisions being made by our two independent coders and between our coders and the lead or senior authors of publications. This discordance was greater when our team was not involved with the study and therefore had to rely on the published report to garner information. In all but one case, intervention developers added strategies to those identified by our 2-expert Delphi process. In some cases, the additional strategies          medication treatment for AUD-a substantially underutilized treatment. Unfortunately, our review did not reveal which strategies are most effective for implementing AUD medications. However, we cataloged the use of specific strategies, perhaps suggesting candidates for future study. Further work is needed to understand why rates of medication treatment of AUD continue to be so low, even after patients are enrolled in care management interventions and/or receiving care in a healthcare setting that has been targeted by a multifaceted intervention. It is entirely possible that previous examinations of barriers, and interventions designed to overcome them have missed the mark. To further assess this, research will be needed to better understand patient-level perspective, preferences and barriers to receipt of medications.