A Mobile Health Intervention to Improve Hepatitis C Outcomes Among People With Opioid Use Disorder: Protocol for a Randomized Controlled Trial

Background People who inject drugs are at a disproportionate risk for contracting hepatitis C virus (HCV). However, use of HCV prevention and treatment services remains suboptimal among people with substance use disorders due to various health system, societal, and individual barriers. Mobile health applications offer promising strategies to support people in recovery from substance use disorders. We sought to determine whether the Addiction-Comprehensive Health Enhancement Support System (A-CHESS), an existing mobile health application for opioid use disorder, could be adapted to improve HCV screening and treatment. Objective The goals of this paper are to describe: (1) the components and functionality of an HCV intervention incorporated into the existing A-CHESS system; and (2) how data are collected and will be used to evaluate HCV testing, linkage to care, and treatment. Methods People with recent opioid use were enrolled in a randomized controlled trial to test whether A-CHESS reduced relapse. We developed and implemented HCV intervention content within the A-CHESS platform to simultaneously evaluate whether A-CHESS improved secondary outcomes related to HCV care. All A-CHESS users received the HCV intervention content, which includes educational information, private messages tailored to an individual’s stage of HCV care, and a public discussion forum. Data on patients’ HCV risk behaviors and stage of care were collected through quarterly telephone interviews and weekly surveys delivered through A-CHESS. The proportion of people with opioid use disorder who are HCV untested, HCV-negative, HCV antibody-positive, or HCV RNA–positive, as well as linked to care, treated and cured at baseline is described here. The 24-month follow-up is ongoing and will be completed in April 2020. Survey data will then be used to assess whether individuals who received the HCV-enhanced A-CHESS intervention were more likely to reduce risky injection behaviors, receive HCV testing, link to medical care, initiate treatment, and be cured of HCV compared to the control group. Results Between April 2016 and April 2018, 416 individuals were enrolled and completed the baseline interview. Of these individuals, 207 were then randomly assigned to the control arm and 209 were assigned to the intervention arm. At baseline, 202 individuals (49%) self-reported ever testing HCV antibody-positive. Of those, 179 (89%) reported receiving HCV RNA confirmatory testing, 134 (66%) tested HCV RNA–positive, 125 (62%) were linked to medical care and 27 (13%) were treated and cured of HCV. Of the remaining 214 individuals who had never tested HCV antibody–positive, 129 (31%) had tested HCV antibody–negative within the past year and 85 (20%) had not been tested within the past year. Conclusions The A-CHESS mobile health system allows for the implementation of a bundle of services as well as the collection of longitudinal data related to drug use and HCV care among people with opioid use disorders. This study will provide preliminary evidence to determine whether HCV-specific services embedded into the A-CHESS program can improve HCV outcomes for people engaged in addiction treatment. Trial Registration ClinicalTrials.gov NCT02712034; https://clinicaltrials.gov/ct2/show/NCT02712034 International Registered Report Identifier (IRRID) DERR1-10.2196/12620

In this representation, the number of aims has been reduced by 1/3 rd . Note that all of the secondary aims differ from one another in both substantive concern as well as analytic method. Also, while we do plan to characterize the struggles we encounter in implementing and sustaining the interventions, we no longer list this as a formal aim.
Concern. The investigative team lacks an experienced qualitative methodologist. Dr. Nora Jacobson, Senior Scientist and Qualitative Methodologistat our Institute for Clinical and Translational Research will mentor Dr. Alagoz and guide the qualitative study design, data collection, and data analysis process. Attached you can find her biosketch. Dr. Jacobson is an interpretive social scientist with more than 15 years of health service research experience who uses qualitative and community-based participatory methods to study the delivery of health services and the development of health policy. Over the last decade much of her research focused on how healthcare organizations implement innovative practices. In addition, she has been involved in several studies that examined contextual factors that promote knowledge transfer and exchange between applied health services, researchers and health system decision-makers. As the qualitative methodologist for the Institute for Clinical and Translational Research (ICTR) at the University of Wisconsin-Madison, she provides consultative research support to ICTR investigators and we believe is well qualified to mentor Dr. Alagoz. It should be noted that Dr. Alagoz has been promoted to the position of Assistant Scientist in our Center.
Concern. The aims were "underdeveloped methodologically". We were somewhat unclear about this concern. However, we interpreted that characterization to mean that the panel wished for greater specificity regarding particular aspects of the proposed analyses. In particular, we believe that the reviewers were interested in more details about the nature of the dependent variable to be analyzed for the Primary Aim, about the qualitative assessment and analyses, and the nature of the mediational analyses.
1. Dependent variable. With regard to the nature of the dependent variable in the Primary Aim analysis, we state on p. 7 of the application that this analysis will "Detect the difference in illicit opioid use between patients who have MAT + A-CHESS vs. MAT alone (primary aim). Hypothesis: Patients with MAT + A-CHESS will use fewer illicit opioids." However, we can understand that the panel may have wanted clarification of how we would measure the dependent variable. Self-reported drug use days will be analyzed in 30-day periods. For Baseline, the TimeLine FollowBack 1 (TLFB) for the last 30 days prior to admission will be obtained and a urine drug screen (CTN-approved drug use outcome measures) 1 . For follow-up assessments the TLFB for the previous120 days will be obtained, corroborated by a urine screen. The TLFB has been successfully used to obtain drug use data for extended periods of time and with poly drug using patients 2 .
2. Mediational analyses, we believe that review panel might have wanted more information about the timing of the mediator assessments relative to the outcome assessments (will there be overlap in the assessment of the mediators and outcome?) and whether the mediation analytic plan takes into account the fact that the various mediators might overlap with one another in their relations with treatment and outcome. To clarify, our mediational models will involve no temporal overlap in the collection of mediators and the outcome. The mediational variables will all be collected in the first 2 visits (at the 4 & 8 month visits) while the outcome (illicit drug use) will collected at the 12, 16, 20, & 24 month visits. Moreover, because mediator-outcome relations might reflect the effects of drug use while the mediator is being assessed (e.g., drug use might suppress ratings of competence), drug use that occurs during the mediator assessment period will be covaried out of models to examine and control its influence. Moreover, in order to assess the nonorthogonality of the mediators, (which seems quite likely with the self-determination variables), we will use multiple mediator analyses based on a Bayesian approach illustrated in Yuan and MacKinnon (2009) (Krippendorf, 2013;Schreier, 2012 that we propose to employ for our data analysis follows a consistent set of steps for each set of data (interviews and case study data) and with each coder. These steps and our implementation of them include: a) Deciding on the research questions (the questions are stated in the proposal), b) Selecting the data collection tools (either interview data or case study data based on which research questions we would like to address), c) Building a coding scheme (our coding scheme will be data driven using a method built on grounded theory; hence we will not use an a-priori scheme), d) Dividing the data into units of coding (in order to refine our unit of analysis, we will use sentences as our unit of analysis. Sentences will be useful in calculating inter-rater reliability and in keeping our coding consistent among coders.), e) Testing the coding scheme (we will pilot test and modify our coding scheme on a subset of data so that our codes are detailed enough to capture our research questions), f) Main analysis (coders will code the data independently using the modified coding scheme.) and g) Interpreting and presenting findings.
Concern. Data collection burden on patients was too high. The reviewers are correct. The burden from each follow-up interview was 285 items. We reduced the burden by 50% (to 144 items) by eliminating multiple scales for one concept and using revised (reduced) scales instead of original scales. All scales are still validated. Specifically, our changes include:  Using the SF-12 quality of life scale rather than the 26 items WHOQO-BREF scale;  Using the Revised Dyadic Adjustment Scale (14 items) instead of the original (24 items);  Using the Adjective Rating Scale of Withdrawal (16 items) but not the 16 item Distress Tolerance Scale.  Measuring each of the four mediators with one instrument rather than two , with o Relatedness measured by our 6-item bonding scale (drop the Important People and Activities scale -19 items); o Intrinsic motivation, measured by the intrinsic motivation subscale (4 items) of the Client Motivation for Therapy scale and drop the Treatment Self Regulation Questionnaire (15 items) and the CBI (35 items) o Competence, measured by the revised 8-item Drug Taking Confidence Scale rather than the 50 item original.
Concern : Dropout rates may be underestimated.. In our RCT, 88 of the patients were using opioids as well as alcohol; 261 were not using opioids. We compared the post-test interview response rate of opioid using patients to the patients that did not use opioids. The non-opioid-using the patients' response rates were: 94.3% at 4 months; 90.6% at 8 months and 86.7% at 12 months. The opioid using patients' response rates were 91.2%, 86%, and 79.1%. The response rates declined in a relatively linear fashion in both groups, with reductions of about 5% in each period. We assumed a 65% response rate at 24 months by continuing the drop off at a 5% rate for each of the 3 succeeding periods from 79% to 74% to 69% to 64%. Hence we believe it is likely that by the end of the study we will still be able to reach 65% of patients originally enrolled.
Concern: The incorporation of services related to HIV and hepatitis C was a "distracting" addition to the proposal, We believe there are two ways the HIV/HCV component adds value to the intervention. 1) The prevalence of HIV/HCV infection is high among opioid using populations, yet most addiction treatment centers do not perform any routine testing. Facilitating protocol-driven approaches to infectious disease screening and linkage to care is a promising function of mHealth interventions. In the present study, bundling HIV/HCV services with A-CHESS could help the addiction treatment centers fulfill an obligation to screen a high risk population for two serious but highly-treatable conditions. 2) Screening for HIV-HCV is consistent with the project's overall goal of improving access to comprehensive health services for opioid-dependent patients, rather than focusing narrowly on promoting abstinence from opioids. We recognize that despite availability of evidence-based interventions, many patients who have injected opioids will relapse. The bundled intervention seeks to meet a public health goal of reducing the number of people who are infected with HIV or HCV but are unaware, and therefore continue to place others at risk.

Concern. The HIV/HCV component was not completely weaved into the entire application.
We acknowledge that as the newest components of the A-CHESS system, the services related to HIV/HCV testing and linkage to care are the least well-integrated into the existing application, and therefore additional programming resources will be required in year 1 of the study. We have, however, accounted for this need in our work plan and associated budget. Since the proposal was submitted, Dr. Westergaard's team has continued to accumulate experience with mHealth approaches to HIV/HCV screening and risk reduction through other NIH-funded research projects. This project will continue to strengthen the collaboration among the CHESS team, Dr. Westergaard and the UW Division of Infectious Diseases. We believe that the way that we have reframed the aims has allowed us to better demonstrate how the HIV/HCV outcomes fit into our study design.
Concern. The cost analysis was underdeveloped regarding the measurement and analysis of healthcare utilization. We appreciate the opportunity to provide further details on our plan. Our proposed analysis of health services utilization was motivated by the potential A-CHESS has shown to reduce the kinds of costly, unscheduled health services utilization associated with relapse; in a field test with U.S. military veterans, A-CHESS users decreased re-hospitalizations due to relapse by 71%. Our approach to measuring and analyzing health utilization data is adapted from McCollister & French's 2003 analysis of the economic benefit of addiction interventions 8 . It defined several categories for healthcare utilization, including: Therapeutic Community Treatment (day), Emergency room (visit), Hospital detox (day), Short term residential treatment (day), Non-residential treatment (visit), Outpatient treatment (visit), Individual psychotherapy (visit), Methadone maintenance treatment (day), Outpatient psychological treatment (visit), and In-patient psychological treatment (day). Of these categories, we will include the following as unscheduled use: Emergency room (visit), Hospital detox (day), and Short term residential treatment (day). To these categories, we will also add a category for urgent care visits and apply cost estimates derived from a national survey of urgent care clinics (Weinick et al., 2009) 9 . We will include use of hospitals, emergency rooms, and urgent care for any reason (i.e., possibly but not necessarily related to substance use). It is worth noting that many of the sources for cost estimates used in McCollister & French's analysis are dated (some going back as far as 1996); we will derive up-to-date cost estimates for ER visits, hospitalizations, and residential treatment using data from the American Hospital Association and American Medical Association. Using patient surveys, we will also assess patients' use of outpatient addiction treatment services following relapse, using the categories of outpatient addiction care outlined by McCollister and French (above). Costs for outpatient addiction treatment associated with relapse will be estimated using service cost estimates provided in French et al.'s 2008 national survey of 110 substance abuse treatment programs 10 adjusted for inflation.