Background
Substance use disorders (SUD) are prevalent, undertreated, and costly to society [
1‐
5]. Approximately 265 people a day died from drug overdose in the United States in 2021 [
6]. However, just 8-14% of people with past-year SUD receive treatment [
1,
2]. Most individuals with SUD prefer to receive treatment in primary care [
7], and it is believed that providing SUD care in primary care would be less stigmatizing [
8]. Consequently, to increase access to SUD treatment, the National Academies of Medicine, Science, and Engineering and others have called for increased integration of substance use care in primary care [
9‐
15].
New ways to address SUD in primary care are needed. Health systems have taken steps to prevent and treat SUD in primary care through screening, brief intervention, referral to specialty care, and medication treatments that can be prescribed by primary care providers (PCPs) [
13‐
21]. However, health systems have been unable to effectively provide treatment to the high volume of patients with SUD who visit primary care [
22,
23]. Implementation of SUD treatments is often hindered by feasibility problems, lack of capacity, and discomfort in treating SUD in primary care [
24‐
26]. For instance, traditional treatments such as cognitive behavioral therapy and contingency management are among the most proven psychosocial treatments for SUD [
27‐
31], but may require extensive resources to implement and deliver [
32,
33]. Buprenorphine, a life-saving treatment for opioid use disorder (OUD), can be prescribed in office-based settings, but most do not receive this medication [
34‐
36]. There is a clear need to identify treatments that are both effective and feasible to implement in primary care.
The adoption of digital therapeutics in primary care is potentially one way to increase access to evidence-based treatments. Several digital therapeutics for SUD are supported by evidence for their efficacy or effectiveness [
37‐
39]. Digital therapeutics may deliver intervention content such as assessments, treatment modules, and normed feedback to patients via websites or smartphone apps, often under the guidance of a clinician [
40]. The use of digital therapeutics could potentially help overcome common barriers in primary care, while providing access to an effective treatment [
41,
42]. For instance, studies have shown that digital therapeutics for SUD can produce beneficial effects while reducing the amount of time that clinicians need to spend with patients [
39,
43], they are acceptable to patients [
44,
45], they may improve clinical outcomes when delivered in real-world care [
46‐
49], and they can be effective when added to usual care approaches that lack an evidence-base [
39,
43].
At least two logistical challenges must be addressed in implementation research to determine how to achieve a far-reaching, sustainable implementation of digital therapeutics in primary care. First, health systems must solve barriers to getting clinicians to offer these treatments to patients. For instance, clinicians encounter difficulty with executing novel workflow processes [
50‐
54], such as creating login accounts for patients and “teaching” them to use apps [
40,
55,
56], often impacting adoption of these treatments [
52,
57,
58]. A second challenge is that patients often need human support to effectively engage in digital therapeutics [
53,
54,
59‐
62]. Successful implementations must provide support to patients to help them engage in use of apps, without overburdening primary care teams [
51,
53,
54,
59,
63]. Given the known time constraints and competing demands in primary care [
25,
64], teams may find it infeasible to offer adequate support for engagement in digital therapeutics. It is unknown whether clinicians in primary care can add additional tasks to their already demanding workload to encourage digital therapeutic engagement, or if they alternatively need dedicated staff to ensure engagement.
The DIGITS Trial
The Digital Therapeutics for Opioids and other SUDs trial (DIGITS Trial) seeks to identify how to best implement digital therapeutics for SUDs in primary care. The clinical intervention includes two 12-week, smartphone-based prescription digital therapeutics, reSET® and reSET-O® made by Pear Therapeutics, which have been authorized by the United States Food and Drug Administration (FDA) for the treatment of SUD and OUD, respectively. reSET and reSET-O are commercial versions of a computerized cognitive-behavioral treatment, the Therapeutic Educational System, which was shown to improve patient outcomes in four RCTs [
39,
43,
65‐
67]. However, all prior RCTs were conducted in specialty addiction treatment settings, not in primary care.
The DIGITS Trial seeks to evaluate whether practice facilitation and health coaching can improve digital therapeutic deployment in primary care beyond a standard implementation strategy. The standard implementation strategy, which serves as a comparator, is based on a multifaceted implementation strategy previously used by clinical leaders at the participating healthcare system to implement app-based treatments for depression and anxiety. It includes discrete strategies such as clinician training and electronic health record (EHR) tools.
Practice facilitation is designed to overcome workflow challenges by supporting clinicians to tailor implementation to their local context [
68‐
70]. Facilitation methods have been used to implement addiction interventions and experts describe it as one of the most successful implementation strategies [
58,
71‐
75]. It is both a process and a set of strategies designed to build relationships, identify and overcome barriers, and help the clinical team with implementation. Meta-analysis has demonstrated that facilitation increases the odds of evidence-based primary care [
76].
Health coaching employs a centralized mid-level provider, namely, a medical assistant (MA), to coach patients to engage in the digital therapeutic while reducing burden on primary care teams. The health coach encourages patient engagement and use of the apps, reinforces learning and skill practice, and encourages completion of healthcare visits with the primary care team. The strategy’s conceptual targets are informed by the literature on patient-mediated implementation strategies, such that health coaching is designed to inform and educate patients, activate them in healthcare, and promote collaboration between patients and healthcare teams [
77]. Coaching is an effective strategy for engaging patients in digital therapeutics [
53,
54,
59‐
61].
The DIGITS Trial is a parallel group, factorial, cluster-randomized trial. Cluster-randomization is at the clinic level because the implementation strategies were assigned to primary care clinics. The strong evidence for reSET and reSET-O from specialty care settings provides sufficient inferential evidence to conduct a trial in primary care that uses a hybrid effectiveness-implementation design with a main focus of evaluating implementation outcomes and a secondary focus of evaluating population-level effectiveness and cost-effectiveness of implementation strategies (“hybrid type 3”) [
78].
Specific Aims
The first aim is to estimate the effect of practice facilitation and health coaching in increasing the reach and fidelity [
79,
80] of digital therapeutics. We hypothesize significantly higher reach among clinics randomized to practice facilitation (hypothesis 1) and significantly higher fidelity among clinics randomized to health coaching (hypothesis 2), compared to clinics that did not implement with these respective strategies.
The second aim is to compare the population-level cost-effectiveness (PLCEA) [
81] of the implementation strategies in improving reach, fidelity, and substance use. This analysis will inform the economic value of the additional implementation strategies relative to the standard implementation strategy. PLCEA methods consider that real-world implementations of an evidence-based practice may yield different effectiveness or cost-effectiveness results than tightly controlled RCTs of the same intervention.
Other study objectives are to: 1) conduct a formative evaluation to provide feedback to the healthcare system, monitor implementation fidelity, and record adaptations, 2) evaluate additional secondary and other outcome measures, including sustainment of the implementation, to provide a comprehensive assessment of implementation success, and 3) evaluate patient-level moderators of reach.
Quantitative Evaluation
Study Sample and Eligibility Criteria
Sites are the unit of analysis. The study uses an open-cohort design, identifying patients following randomization because the digital therapeutic is offered when patients have a clinical encounter. For the statistical analyses, patients will be attributed to the site where they had their first qualifying visit. Patient eligibility criteria for automatic inclusion are determined with EHR data. Patient inclusion criteria are (1) had a primary care visit in a participating clinic from 2 weeks before through the active implementation period (for primary outcome analyses) or sustainment period (for sustainment analyses), (2) screened positive for substance use on the day of the visit or in the prior year, and (3) adult 18 years of age or older at the time of the visit. Positive screens are indicated by patient self-report of daily cannabis use or any drug use in the past year on instruments described previously [
91,
92]. Patients are excluded if they have previously requested to opt out of research studies.
Outcome Measures
Table
4 outlines the study’s primary and secondary outcomes, other pre-specified outcomes, and other explanatory and sensitivity analysis measures. Outcomes are conceptualized via the RE-AIM framework (reach, effectiveness, adoption, implementation fidelity, maintenance/sustainment) and measured at the site-level [
93,
94].
Table 4
Primary, secondary, and other outcomes in the DIGITS Trial1
Primary Outcome Measures |
Reach. Reach of the digital therapeutic to patients in the primary care clinic, measured as the proportion of patients who initiate the digital therapeutic, defined by instances in which patients open the app, enter the prescription code, and use a treatment module Fidelity. Fidelity of patients' use of the digital therapeutic to clinical recommendations, measured as the mean number of weeks during patients’ 12-week prescription in which patients use 4 app modules/week and have visited a clinician in the past 30 days2,3 |
Secondary Outcome Measures |
Engagement (patient engagement in substance use care). Mean number of months in which patients make ≥1 visit for substance use disorder2,3 |
Economic costs. Costs from the perspective of a health system and payer including implementation, direct intervention, operating, and other indirect healthcare costs. This measure will be used to calculate the population-level cost effectiveness of increasing reach, fidelity, and engagement. |
Other Pre-Specified Outcome Measures |
Reach-2. The proportion of patients prescribed the digital therapeutic |
Fidelity-2. Mean number of weeks in which patients use at least 1 module/week3 |
Substance use. The proportion of patients who are reached and reduce their substance use4 |
Abstinence. The proportion of patients who are reached and are abstinent from substances4 |
Sustainment. The proportion of patients who are reached during the sustainment period. |
Other Explanatory and Sensitivity Analysis Measures |
Adoption. The proportion of healthcare provider prescribing the digital therapeutic, overall and by provider type |
Adoption-2. The mean number of months in which providers access clinician dashboards |
Reach-3. Proportion of patients who download and unlock the digital therapeutic |
Fidelity-3: Mean number of weeks in which the patients use at least 1 module/week3 |
Fidelity-4: Mean number of weeks in which the patients use 4 modules per week but without the requirement that they visit a clinician3 |
Fidelity-5: Mean number of modules completed over the 12-week prescription3 |
Substance use-2. The proportion of patients who are reached and reduce their substance use, as measured by self-report data collected by the digital therapeutic |
Abstinence-2. The proportion of patients who are reached and are abstinent from substances, as measured by self-report data collected by the digital therapeutic |
Abstinence-3. Abstinence verified by urine drug screens among patients prescribed the digital therapeutic for opioid use disorder, based on EHR data |
Reach and fidelity are primary outcomes. Reach is the site-level proportion of patients who initiate reSET or reSET-O. For this outcome to occur, a clinician must prescribe the digital therapeutic and the patient must complete at least one treatment module. Fidelity is the site-level mean number of weeks in which patients use the digital therapeutic as recommended. This includes completing a recommended 4 or more modules per week while under the care of a clinician [
39,
95,
96].
Patient engagement in SUD care, a secondary outcome, is operationalized as the mean number of months in which patients make at least one visit for SUD in any setting. Engagement is conceptualized as an effectiveness outcome. Economic costs are another secondary outcome operationalized using PLCEA methodology (see Economic Evaluation).
Other pre-specified outcome measures include an additional measure of reach quantifying the proportion of patients who are prescribed reSET or reSET-O by a clinician, irrespective of whether a patient initiates its use. An additional fidelity measure will estimate the number of weeks in which patients complete at least one module per week. Two effectiveness outcomes include substance use and abstinence, measured as the proportion of patients who are reached and reduce their substance use, or who use no substances, respectively.
Sustainment will be operationalized as the proportion of patients who are reached (same definition as above) during the sustainment period of the study.
Exploratory measures will be analyzed to comprehensively assess implementation and effectiveness. One adoption measure is operationalized as the proportion of healthcare providers prescribing reSET or reSET-O, overall and by provider type. Another adoption measure is the mean number of months in which providers access clinician dashboards. An additional reach variable will indicate the proportion of patients who download and unlock the digital therapeutics. Two additional fidelity variables indicate the mean number of weeks in which patients use 4 modules per week regardless of whether they see a clinician, and the mean number of modules completed over the 12-week prescription. Additional substance use and abstinence measures will use self-report timeline follow-back data collected by the app every four days during the 12-week prescription [
75]. We will also measure the proportion of patients who are reached by reSET-O and achieve abstinence from all substances, as evidenced by results of routine urine drug screens administered as part of clinical care (often administered among patients with OUD).
Statistical Analysis
Following procedures for factorial trials, we will analyze main effects on primary and secondary outcomes by examining the mean responses at one factor level and at a contrasting factor level, collapsed across all levels of the other factors (e.g., practice facilitation versus no practice facilitation) [
97‐
99].
Primary Outcome Analyses
Analyses will follow intent-to-treat principles, with site analyzed according to their assigned treatment group regardless of the amount of implementation strategy delivered. We will fit a linear regression model to estimate the main effect of each factor level (practice facilitation, health coach) [
97,
99]. For the primary outcomes of reach and fidelity, we will apply linear regression to model the proportion of patients reached within a site and the site-specific mean number of weeks with fidelity, respectively. We will test hypotheses 1 and 2 by testing the appropriate contrast from the regression model and use the Holm procedure to control the familywise type 1 error rate of the two primary hypotheses at 0.05. In addition to the hypothesis above (see Specific Aims), we secondarily hypothesize that practice facilitation is superior to health coaching in increasing reach, and health coaching is superior to practice facilitation in increasing fidelity. We also will examine whether the two enhanced strategies together are superior to standard implementation by comparing clinics with both enhanced strategies to clinics with neither. We will also examine interaction effects between the two enhanced strategies.
Secondary and Other Pre-Specified Outcome Analyses
We hypothesize significantly higher engagement among sites randomized to practice facilitation or health coaching. Analyses will follow the same general modeling approach as the primary outcomes. Secondary outcome analyses will apply a traditional two-sided type 1 error rate of 0.05; considering issues of multiple comparisons, findings will be interpreted with caution. See Economic Evaluation for a description of cost analyses.
Additional pre-specified reach and fidelity measures and other outcomes will follow the same analytic procedures as the primary outcomes (e.g., linear regression of the site-level measures).
Sustainment analyses will describe reach over time during the sustainment period graphically by study arm. If reach is greater than 5% during sustainment (overall or in any study arm), we will conduct secondary analyses allowing the intervention effect to vary over time. Specifically, we will subdivide the sustainment period into discrete time intervals (e.g., 4-month windows) and include interaction terms with intervention group; repeated measures over time within a clinic will be accounted for using a mixed-effects model with site-specific random intercepts.
Patient-level moderators of reach and fidelity
We will conduct exploratory patient-level analyses stratified by sex and by SUD type when estimating reach, fidelity, abstinence, and substance use reductions, allowing us to provide data about implementation effectiveness in specific subgroups.
Sensitivity Analyses
We will conduct several sensitivity analyses to examine the robustness of trial results. If, due to random chance, factor levels are imbalanced in any baseline characteristics (e.g., site size), we will perform sensitivity analyses where we include these characteristics in the regression models. Following reporting guidelines [
100,
101], we will examine whether the proportion of screen-positive patients (the measure defining the study population) differs across arms. To assess for identification bias [
102,
103], we plan to conduct sensitivity analyses in which we consider alternative population denominators: all patients with a SUD diagnosis, and all patients with visits regardless of whether they screened positive for substance use.
Statistical Power
Minimal detectable differences for fixed 80% power were estimated based on 27 clinics (number of clinics available during the study pilot phase) and a two-sided type 1 error rate of 0.025 for the two primary outcomes of reach and fidelity (to control the familywise type 1 error at 0.05). We estimated that we will have >0.80 power to detect an increase of 2 percentage points in site-level reach among screen-positive patients in sites with versus without practice facilitation. We estimated that we will have >0.80 power to detect an increase in the site-level mean number of weeks fidelity of 0.088 among screen-positive patients in sites with versus without a health coach. Power analysis assumptions and other details are in Additional file
4.
Economic evaluation
Aim 2 will conduct PLCEA to measure the cost of each implementation strategy in increasing reach, fidelity, and engagement.
To operationalize PLCEA, we will measure the following ratio:
$$\frac{\textrm{Incremental}\ \textrm{Population}\ \textrm{level}\ \textrm{Costs}}{\textrm{Incremental}\ \textrm{Population}\ \textrm{level}\ \textrm{effectiveness}}$$
where incremental population-level costs are the difference in costs between population in sites targeted by each implementation strategy. Population-level costs include implementation costs, direct intervention costs and indirect healthcare costs. While the primary outcome analyses above test specific hypotheses regarding the main effects followed by exploratory tests of interactions, the PLCEA will estimate the difference in the outcomes of reach, fidelity and engagement (see Outcome Measures) between respective populations targeted by one of the enhanced implementation strategies and the standard strategy. This produces information about incremental costs of each implementation approach while accounting for potential multiplicative effects on implementation costs [
104]. Analyses will be performed from the perspective of a healthcare system and payer.
Evaluation goals and methods
Formative evaluation is used to continuously identify and document modifications, adaptations, and implementation determinants (i.e., barriers and facilitators) as they pertain to each DSF construct. Evaluators communicate salient findings with the implementation team [
70]. Evaluators also monitor for contamination across implementation strategy conditions and observe how implementation outcomes change in correspondence to changes in DSF constructs. The Framework for Reporting Adaptations and Modifications to Evidence-based Implementation Strategies (FRAME-IS) guides tracking of modifications, adaptations, processes of and reasons for change [
107,
108].
Formative evaluation analyses follow a qualitative rapid assessment process [
109,
110]. Using templated forms (Additional file
6), evaluators take field notes at implementation meetings and review secondary data sources (e.g., meeting notes, communications, and implementation documents). Analysis procedures consist of structured data reduction to categorize and summarize fieldnotes and secondary data sources according to predetermined domains [
90,
107,
111], while also capturing emergent findings (e.g., decision points). Summary data are entered into a strategy-by-domain matrix to systematically extract themes, detect trends, and answer main evaluation questions (Additional file
7). Synthetized findings and recommended adaptations are regularly presented to the study team and healthcare system partners, who decide by consensus whether to execute adaptations.
Sustainment
Following the end of the active implementation period, funding and support for implementation strategies by the research grant is scaled back to study sustainment and institutionalization of the digital therapeutics. The grant will provide additional prescriptions after the active implementation period ends to allow time for the health system to decide whether to continue offering reSET and reSET-O, and to allow for the continued measurement of reach and fidelity and qualitative measures of sustainment.
Trial Status
No outcome data have been analyzed yet; the last site completes active implementation on 2/10/2023.
Discussion
This trial is designed to overcome several limitations of prior research on digital treatments for SUD. Real-world implementation studies in health systems may provide more actionable information to decision makers such as health system leaders than traditional effectiveness trials [
81,
84]. For instance, other trials may be conducted within selected patient populations who consent to participate in studies of treatment for SUD, who may not be representative of typical patients. Moreover, other trials may rely on researchers to deliver digital treatments whose workloads and primary responsibilities are very different from those who are delivering clinical care.
This study fills a gap in implementation science, where evidence is needed on the comparative effectiveness and cost-effectiveness of different implementation strategies on improving care. Findings regarding the population-level cost effectiveness of implementation strategies will further provide information to decisions makers about the financial implications of this study’s strategies. For instance, while an implementation strategy may result in greater effectiveness, it may not necessarily be more cost-effective. Other aspects of this study, such as the workflows created and the insights into sustainment ascertained through qualitative research methods, will help create a roadmap for other health care organizations wishing to care more effectively for SUDs.
Strengths and Limitations
Significant strengths of this trial include that 1) the health system has high universal screening rates for substance use that can assist with comprehensively identifying the population targeted (i.e., the reach denominator); 2) the study has access to a diverse set of data enabling the detailed analysis of clinical and economic outcomes, and 3); the study is built on principles of preserving real-world conditions. However, the study design also presents limitations. Notably, the study is being conducted after the COVID pandemic began; in the context of significant healthcare system staffing shortages, fewer clinics were recruited, and the intervention is delivered in clinics with reduced capacity. The substance use and abstinence measures collected from the EHR will be available only for an estimated ~70% of patients because of reliance on follow-up screening data. Findings may not generalize into systems that primarily serve uninsured populations.
Acknowledgments
The authors are grateful to the patients and clinicians of Kaiser Permanente Washington. We thank Rebecca Parrish, MSW, LICSW of Kaiser Permanente Washington for her important contributions to the study’s design and her continued implementation support. We also wish to express gratitude to scientific members of the DIGITS Trial Steering Committee, Mark McGovern, Geoffrey Curran, and Sarah Becker for their feedback on the study design and conduct. The authors thank the faculty and fellows of the National Institutes of Health Implementation Research Institute, the Summer Institute on Randomized Behavioral Clinical Trials, and the Mixed Methods Research Training Programs for their feedback on early conceptualizations of this grant proposal.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.