This study used a mixed model design to demonstrate technical feasibility; patient engagement; and the acceptability, usability, and utility of KIOS with individuals receiving methadone maintenance for OUD.

The focus group protocol was approved by BioMed IRB (approval number 18-1-100-OUD). Participants provided informed written consent and were paid for their participation. In addition, the KIOS use study protocol was approved by BioMed IRB and participants provided signed informed consent and were enrolled in a 4-week, single-group, pre-post evaluation of the KIOS app. Participants were paid US $75 after completing orientation and training. They were paid an additional US $75 for participating in a debriefing or focus group. All patient responses were stored in a confidential fashion in a password-protected computer database and only available to approved study personnel. All data were deidentified prior to analysis.

Contextual design [22], an industry-standard, user-centered methodology, was used to design the user experience. Individuals receiving methadone maintenance were recruited for focus groups at a private treatment clinic (CleanSlate Addiction Treatment Center, n=5) and the local public treatment clinic (The Center for Healthcare Services, n=9). Participants provided informed written consent for their participation. Feedback from participants was solicited about the perceived usefulness of the tool, functionality, benefits, and incentives for using the software.

Based on the focus group feedback, a functional user interface (UI) was created. The web-based KIOS UI includes assessment screens for each of the clinical variables (craving, depressed mood, anxiety, irritability, pain, agitation or restlessness, difficulty sleeping, absenteeism, difficulty with usual activities, and conflicts with others; each rated on a 1-7 scale from none to severe). The determination of the clinical variables is described in the Results section. Inside KIOS, the patient rates the severity of these 10 symptoms and logs self-reported opioid use, alcohol use, and recreational drug use in an important behavior checklist. Participants also reported regular exercise, avoidance of triggers or high-risk situations, and sleep habits (not reported here). The KIOS app analyzes the changes in symptoms during the time between patient self-reports and then delivers advice specific to the patient’s evolving symptoms. On the advice screen of the app, KIOS contextually suggests behaviors that are associated with improving self-report variables, including reaching out to support people or treatment providers, if indicated. In these circumstances, the app encourages the integrated use of comprehensive treatment resources (eg, for increasing depression, patients are directed to consult a medical provider). The results of both the symptom changes and the reporting on important behaviors could also be reviewed by the patient’s individual counselor. This offers the counselor access to real-time data over time on both symptoms and behaviors. It offers the ability to reinforce behavioral treatment planning and maximize the limited counseling time available at each appointment. KIOS also contains a home and menu screen, 6 graphs for tracking symptoms over time, a primary advice screen, a second supplemental advice screen if patients want additional advice, a calendar-based journal, and help screens for technical support, feedback, and emergency contacts for help or crisis situations.

The primary outcome was user satisfaction as measured by acceptability, usability, and utility. Secondary outcomes included self-reporting of drug and alcohol use. The study was conducted exclusively on a web-based interface and via telephone. Participants were recruited from Community Medical Services, a private opioid treatment organization. Staff at Community Medical Services provided potential participants with study information and individuals contacted study staff if they wanted to participate.

Inclusion criteria were (1) interest in study participation; (2) 18 years or older; (3) receiving methadone maintenance for OUD for ≥4 weeks; (4) ability to access KIOS via computer, smartphone, or tablet; and (5) no unmanaged major psychiatric illness or suicidality.

Participants attended a web-based KIOS training session. They were asked to complete KIOS assessments on a web-based interface at least 3 times per week but no more than once daily. KIOS could still be accessed by participants as often as desired to review advice and graphs or access the journal. All assessments were logged on the KIOS server to tabulate the frequency of use.

After the trial, participants completed an 18-question KIOS survey. Participants also completed the System Usability Scale (SUS)-Modified, a single-factor 10-item self-report scale that was used to evaluate participants’ subjective experience using KIOS [23,24].

User data were logged on the KIOS server for the 10 variables from each assessment. At the end of the study, a final web-based debriefing focus group was conducted to gather user feedback not otherwise collected during the study. Participants described their experience in the study and provided their impressions, suggestions, and critiques of KIOS in an open dialogue.

Means, medians, and SDs were calculated for usage, SUS, and patient questionnaire data. KIOS variable scores were averaged over each week for each participant to account for differing numbers of assessments. Calculations were similarly made for an overall KIOS index score (sum of all variables except craving) and for emotional (depressed mood, anxiety, and irritability), physical (pain, agitation or restlessness, and difficulty sleeping), and social (absenteeism, difficulty with usual activities, and conflicts with others) subscales. The craving variable was kept separate due to a consistent body of literature linking it to substance use outcomes [25-27]. The changes in mean values for each week were calculated relative to baseline measures, and 2-tailed Student t tests were conducted for significance between baseline and the weekly means. A Pearson correlation was performed by comparing the KIOS index score to the craving scores averaged over the study period.

Managing recovery from OUD is a complex process characterized by constantly changing intra- and interpersonal circumstances. It was essential to identify a limited number of relevant clinical variables that would both effectively reflect an individual’s subjective behavioral state as well as provide sufficient information to permit the assignment of appropriate behavioral advice. Deidentified data from large-scale randomized OUD clinical trials (the POATS trial: n=653 and the X:BOT trial: n=570) [28,29] were analyzed to provide relevant clinical variables. The KIOS software requires clinical variables that change frequently to operate optimally. Therefore, data obtained from instruments in these studies were analyzed to determine how frequently variables changed from one assessment to the next.

An expert panel (Table 1) selected a subset of 32 variables that were most relevant to OUD recovery. The members of the panel were selected for their expertise in behavioral interventions in OUD and to impart a multicenter perspective to the study. To achieve panel consensus, a modified Delphi process was used [30]. A web-based survey tool and web-based meetings were used to expedite the process. The Delphi process was iterated until consensus was achieved. The decision rule for the inclusion of variables into the system called for the variable to be selected by a majority of the panel members.

Independently, each panel member selected variables with the criteria that a change in the variable is important in determining the patient’s current clinical state. If consensus was not achieved, the panel was queried until the decision rule was met. The variables proposed by the expert panel were then evaluated in a correlation matrix to eliminate redundant variables. Ten individual variables (craving, depressed mood, anxiety, irritability, pain, absenteeism, agitation or restlessness, difficulty sleeping, difficulty with usual activities, and conflicts with others) were selected to capture a person’s clinical state most comprehensively while minimizing the number of questions required of the patient at any assessment.

Next, the panelists proposed combinations of 2 or more variables that provided the most valuable information for describing the current state of the patient and that would be particularly sensitive to identifying transitional worsening or improvement in OUD. This grouping of variables created the underlying structure that mapped the patient trajectory and enabled the development of interventions. Reports were generated by the KIOS software describing all the possible changes for the selected set of interacting variables. The panel supplemented these descriptions with interpretations and recommended interventions.

Appropriate behavioral interventions incorporating cognitive behavioral therapy (CBT) and mindfulness, using the Behavior Change Technique Taxonomy [31] as a reference, were applied to each patient state. These change techniques have been shown effective in multiple studies [32] (Table 2). All recommendations were phrased in patient-friendly language and subjected to a rubric specifying reading level and text length, and that the advice is consistent with a CBT approach. In a previous study, the research team successfully followed a similar approach to chart trajectories and develop self-management interventions for bipolar disorder [21].

In total, 19 individuals signed informed consent. Three did not complete orientation. One stopped participating after 11 days and was considered lost to follow-up. The median time in methadone treatment at the beginning of the study was 9 months: 4 patients had been in treatment for ≤6 months, 6 patients had been in treatment for 6-12 months, and 5 had been in treatment for >12 months. Participants (N=19) were 47% (n=9) White (non-Hispanic), 21% (n=4) White (Hispanic), 11% (n=2) more than 1 race, 11% (n=2) American Indian or Alaska Native, 5% (n=1) Black or African American, and 5% (n=1) Asian. The mean age of all participants was 34.4 (SD 7.3) years, and 63% (n=12) were women.

In total, 15 of 16 participants who logged in to KIOS at least once finished the trial. Those 15 participants completed 191 total assessments, averaging 12.7 (SD 3.8) per user (median 13, IQR 11-14). During the poststudy debriefings, several participants requested continued access to KIOS after the trial ended without further compensation; 6 participants used KIOS poststudy and 1 participant used the app for over 10 months.

The final debriefing focus group generated many comments that described the individualized experience of using the app. All participants who attended the group stated it was helpful and worth using. For example, focus group attendees noted that KIOS gave timely and responsive advice, helped with adherence to treatment goals, and promoted reflection and motivation. Some commented on how natural the advice seemed, almost like a counselor was responding in real time.

Some participants mentioned features that could improve KIOS such as a customized reminder system to specify the time of day to use the app, features to connect directly with counseling staff or peer support, gamified content (such as daily challenges, more content that included pictures), positive reinforcement (such as digital trophies, badges, or monetary rewards), and expanded and more specific description of exercise or meditation practices beyond those currently included in the app.

The mean SUS-Modified score was 89.5 (SD 9.2; median 92.5, IQR 85-95). Higher scores indicate greater usability. The psychometric properties of the SUS have been validated and replicated, and the score for KIOS falls slightly below the top score, “best imaginable=90.9” and above “excellent=85.5” on an adjective rating of the SUS [23,24] (Table 3).

The KIOS questionnaire was developed by the research team to evaluate the utility of the KIOS app. The mean response to all KIOS questionnaire data was 6.32 out of a maximum of 7.0 (SD 0.25), indicating an overall strong agreement with most statements. The highest rated statement was “I see benefits using KIOS,” which averaged 6.73 out of 7 (SD 0.46); each respondent rated this item as either 6 or 7. The lowest rated item was “KIOS helped identify some personal triggers,” which was still generally well agreed upon and rated 5.73 (SD 1.62; Table 4).

The Healthy Behavior Checklist allows patients to check yes or no for engaging in the behavior at each assessment. The behaviors include opioid use, recreational drug use, alcohol use, exercising, good sleep habits, and avoiding triggers and high-risk situations. Only 1 participant self-reported 1 occurrence of unprescribed opioid use. This occurred at the baseline assessment, and no other illicit opioid use was reported by any participant during the study. Four participants reported 18 occurrences of alcohol use and 3 others reported 24 instances of recreational drug use (nonspecified); no participant reported both alcohol and recreational drug use.

Figures 1 and 2 demonstrate how KIOS data can be used to track patient reporting over time. In Figure 1, the KIOS score and craving score were highly correlated, indicating that the symptom-reporting scales appear to be closely related to self-reported craving scores over time (Pearson r=0.550; P<.001). Two KIOS subscales were significantly lower at the end of the trial. Emotional (week 4; P=.01) and social (week 4; P=.02) subscales were significantly lower compared to the baseline (Figure 2).

The KIOS app is an adjunctive tool to OUD treatment that gives patients 24/7 access to behavioral intervention strategies and supportive and reinforcing behavioral guidance that potentially could augment and improve response to routine clinic-based counseling interventions. In this study, participants rated the primary outcomes of acceptability, usability, and utility very highly. The scores on the SUS-modified were in the excellent range. The mean response to all KIOS questionnaire data was 6.32 out of a maximum of 7.0 (SD 0.25), indicating an overall strong agreement. Participants used KIOS on average about 3 times weekly, which is a very good indication of sufficient use of the app to generate relevant advice.

Participants on average used KIOS 3 times per week, which is the minimum requested in the study. This allowed sufficient time for relevant behavioral change between self-reports to generate pertinent recovery advice to the user. However, this may indicate a need to make the KIOS app more engaging, since participants could use the app daily if desired. Focusing the content on participants earlier in OUD treatment who may want more recovery advice or improving the experience by augmenting personalization could make the app more engaging [33,34]. Revisions to the app will include more explicit recommendations to continue in OUD treatment and more varied suggestions about mindfulness-oriented and positive self-care activities. Patients commented favorably on the individualized and specific nature of the advice and how this motivated and kept them on track in recovery. Some even commented that it seemed a counselor was responding to their self-report assessments in real time. This sense of copresence, of someone concurrently participating in the intervention with a psychological connection to the app user, was associated with improved satisfaction and efficacy in studies of other software apps [35,36].

The SUS-Modified is a well-validated instrument [23,24]. KIOS was rated as very usable for these patients who were recruited from a community-based methadone maintenance treatment program. One item stood out as less positive: how cumbersome the app was to use. Since ease of use is clearly important, revisions to the app will involve attention to this aspect of use (eg, ease of moving between the home screen and different app functions). The revised KIOS app will be used on a smartphone platform to improve this aspect of usability.

Utility was measured by a questionnaire specifically developed to evaluate the KIOS app. Participants strongly agreed on most questions related to usefulness, applicability, and relevance to recovery. Items that were less strongly endorsed were related to identifying personal triggers and improved health status. Since about one-third of the participants had been in treatment over 12 months, it is possible that identifying personal triggers was less of a treatment concern, and therefore, the mean score for this item was somewhat lower. In addition, since the study was only 4 weeks in duration, there was not much time to notice changes in general health status. Self-report of drug and alcohol use was modest in this sample, where only a minority of the participants was in treatment for less than 6 months. Much of the content of the app is aimed at changes in symptoms related to substance use problems; yet, participants who had minimal drug use problems and few mental health concerns rated the experience as helpful and relevant to their recovery. The randomized controlled study of KIOS will have a duration of 6 months and only focus on patients who are early in the treatment process. This will result in greater changes in drug use behavior and a longer period of time to capture potential changes in personal triggers and general health status. Revisions to the app will include more focus on positive behavioral changes to improve patient self-management and social support for recovery.

The scores derived from the KIOS self-report generated potentially useful data for both the KIOS user and the treatment team. As expected, the KIOS score and craving score were highly correlated. The correlation between these 2 measures suggests that the assessment variables, which collectively form the KIOS score capture relevant emotional, social, and physical phenomena that correspond with self-reported craving. Figures 1 and 2 demonstrate how KIOS data can usefully track patient reporting over time. These results could also be reviewed by the patient’s counselor to reinforce behavioral treatment planning and maximize the limited counseling time available at each appointment to assist the patient with this aspect of their care. The KIOS journal feature is available to help the patient record and organize pertinent thoughts and reflections, and the behavioral checklist helps to organize and track drug use and positive behaviors associated with improved recovery outcomes.

KIOS advice is designed to track patient self-report items related to OUD recovery over time. The KIOS app modifies the advice given to the user based on the person’s changes in self-report at each assessment. Advice is then directed specifically to the symptoms of most concern to the user. KIOS can draw the patient’s attention to triggers and potential problem behaviors with helpful and supportive advice in real time to reinforce recovery activities and goals at times of vulnerability to drug use. It may advise reaching out to support people or treatment providers depending on the type of a variable or increasing intensity of symptom reporting (eg, for increasing depression and insomnia consulting a medical provider), further encouraging integrated use of comprehensive treatment resources. It gives specific, evidence-informed suggestions for various symptoms including sleep hygiene, stress management, avoiding triggers, and pain reduction behaviors.

Expanded access to digital behavioral interventions has the potential to bridge a significant treatment gap due to the lack of counseling resources in many OUD treatment settings. However, there are very few digital behavioral apps that specifically address OUD [13,17,37,38]. Due to the nearly uniformly serious nature of this disorder, it is unlikely that a digital app by itself would be potent enough to help a person manage this disorder without professional treatment participation. The available apps focus primarily on connecting users to treatment services and peer support, making available CBT educational modules, using self-report check-ins, and using contingency management interventions. Their use has demonstrated significant improvements in some treatment outcomes [39]. For example, studies of reSET-O showed efficacy regarding opioid use or abstinence and in reinforcing and increasing treatment activity frequency in controlled trials [40]. The web-based CBT4CBT app improved retention and reduced drug use when combined with office-based treatment as usual [41]. It is one of the few digital interventions that has shown improved efficacy compared to in-person CBT in randomized controlled trials and demonstrates the potential power of these types of treatment augmentations [41,42]. The results of a randomized controlled trial of the A-CHESS digital app for use as an adjunct treatment in OUD did not demonstrate significant benefits compared to the control condition [16,43]. However, some results indicated that for specific subpopulations there could be benefits.

Other apps use a therapeutic relational approach using a conversational agent, such as those using a chatbot platform [12,44]. Some of these interventions have demonstrated efficacy [45]. They are currently used to help patients with a variety of mental health issues, although few have an evidence base [13,45-47]. The use of the W-SUDs modification of the Woebot app [12,13] has been associated with self-reported reductions in drug use and urges to use drugs and improvement in anxiety in users with substance use concerns [13]. The Woebot gives the impression of a therapeutic coach in the conversation generated by the chatbot. There is evidence from prior studies that some users are more likely to disclose information [48] to nonhuman apps, and that a strong therapeutic alliance can form [49-51]. Although not measured in this study, some participants commented on how KIOS responses seemed like real-time exchanges with a therapist. Future studies of KIOS should examine this aspect of the UI.

The study limitations include the short study duration in this trial to primarily test the acceptability, utility, and usability of the software. There was no control group, and it was not compared to another treatment condition or substance use treatment app to determine the specific benefits of KIOS compared to other apps or interventions for persons with substance use problems. The small sample size limits generalizability in several ways: to other opioid agonist treatment clinics or office-based OUD treatment, other patient samples that may be less interested in using digital apps, and in the ability to detect statistical differences between study variables. The patient focus group comments reflected the views of persons who were interested in using the app and potentially predisposed to having a positive user experience. KIOS relies on patient self-reporting to generate advice, so if a patient did not report accurate information, then they would not get accurate advice. This could be a problem if the patient’s counselor had access to the patient’s data. For example, if the patient was not ready to reveal their substance use, they might not give an accurate self-report. This should also be examined in future studies of the app.

The KIOS app offers a different approach to the digital app for augmenting outpatient treatment for OUD. Potentially, the advice generated from the app may be more specific to the user because self-reported symptom clusters are used to identify appropriate behavioral interventions selected from evidence-based literature. It also has the potential advantage of using the therapeutic relational coaching approach that many users find more engaging.

Current plans for the development of KIOS include adding varied advice features, incorporating and reinforcing more positive recovery behaviors (eg, positive social support), and pursuing Food and Drug Administration approval as a prescription medical device. Although KIOS was studied in methadone maintenance clinics, it may be a very helpful tool in office-based buprenorphine addiction medicine treatment settings that often also have limited counseling availability. A National Institute on Drug Abuse–funded randomized controlled trial of KIOS is currently being conducted to evaluate its efficacy.