We conducted a qualitative study in Pune, India, consisting of focus group discussions and in-depth interviews with patients, providers, and stakeholders to evaluate the acceptability, adaptability, and feasibility of a CBI for alcohol reduction. Our semistructured interview guide was developed based on our previous work culturally adapting evidence-based interventions [37]. It was designed to assess and make recommendations regarding (1) virtual counselor acceptability, (2) intervention adaptability, and (3) feasibility. We followed the Consolidated Criteria for Reporting Qualitative Research (COREQ) guidelines [38] to report the study.

The Johns Hopkins School of Medicine institutional review board in Baltimore, Maryland, United States (IRB00174744 / CR00026582), and the Byramjee Jeejeebhoy Government Medical College ethics committee in Pune, India, approved this study.

On one occasion, participants viewed a single session of a US-developed evidence-based CBI designed for people with HIV who consumed alcohol at unhealthy levels [36]. Our CBI used well-established, evidence-based cognitive behavioral techniques, including personalized feedback, pros and cons of drinking, problem solving high-risk situations, and goal setting for reduction or cessation of alcohol use. The software platform was derived from Computer Intervention Authoring Software, an authoring tool to create and enter intervention script into electronic packages featuring a synthetic text-to-speech engine that reads all questions and speaks aloud to the participant (via headphones); synchronous interactivity; natural language reflections; branching logic; and the ability to incorporate specific images, graphs, figures, text, or videos. The flexible and modular platform allows for efficient content modification for a variety of populations [39,40]. Using a motivational interviewing style, a virtual counselor interacts with participants to provide IMB skills that foster alcohol use reduction or abstinence [41,42]. The virtual counselor is a 3D character named Peedy the Parrot that has over 50 specific actions that can be selected to display interest and empathy and increase participant engagement and motivation with the intervention. Although Peedy is an avian character, it appears anthropomorphic through its animations (eye gaze, facial expressions, and body movements) and script (eg, expressive reflections to user answers).

Study participants were selected using convenience sampling and provided informed consent prior to enrolling. For the in-depth interviews, adults aged 18 years and older with AUD were eligible and were referred to us through ongoing clinic research studies at a government tertiary care hospital in Pune. AUD was defined for males as having an Alcohol Use Disorders Identification Test–Concise (AUDIT-C) score of 4 or more and for females having a score of 3 or more on the AUDIT-C [43]. Between June and November 2018, we approached 15 eligible individuals, and 10 agreed to participate. Those who refused (n=5) were all males and were unable to participate for the required hour to complete the interview. We also conducted 3 focus group discussions: 1 with members of an Alcoholics Anonymous (AA) group (n=12) and 2 with HCPs from a tertiary care hospital (n=21), including counselors, nurses, and clinicians. All HCPs who participated in the study work in same hospital where the study was conducted. Focus groups were conducted in the afternoon when their outpatient department work was over. All in-depth interviews and focus groups were conducted in an easily accessible, private conference room at a hospital in Pune, India.

Focus groups were moderated by a study co-investigator (NS), who is an Indian female medical anthropologist, and assisted by the study coordinator (GD), who is also an Indian female medical anthropologist, using an interview guide based on topics reviewed in the CBI and adaptation procedures. Focus groups and in-depth interview participants viewed the intervention, which was displayed screen by screen on a white board using a projector. Both the virtual counselor’s speech and the information on each screen was interpreted from English to the local language, Marathi, by GD and NS. After watching each screen, NS and GD obtained participant feedback using the semistructured field guide. On CBI screens where the virtual counselor asked questions with multiple choice answers, participants of HCP focus groups chose the most common option they encountered in their clinical practice, while AA focus group and in-depth interview participants chose the options applicable or suitable to them. We then discussed the feedback provided by the virtual counselor, Peedy, and reviewed the other answer options and feedback. We sought further suggestions and recommendations from the participants.

The main areas of inquiry expected were as follows:

All in-depth interviews and focus groups were audiorecorded and notes were taken by GD. In-depth interview and focus group texts were then transcribed in the local language and translated into English. A coding guide was prepared based on inductive and deductive codes. All transcripts were coded by GD and reviewed by NS, and data were analyzed in MAXQDA 12 (VERBI GmbH) software using the framework approach [44].

The majority (9/10) of in-depth interview respondents were males with a median age of 41 (IQR 30-56) years. In-depth interview participants included people with TB (6/10), HIV (2/10), HIV/TB co-infection (1/10) and no HIV/TB (1/10; Table 1). Of our 2 HCP focus groups, 1 was conducted with counselors working in HIV and TB research studies who were predominantly female (female to male ratio 12:1). The other focus group was conducted with HIV clinic HCPs and was majority male (female to male ratio 3:6), including 2 counselors, 3 nurses, 2 doctors, 1 pharmacist, and 1 psychologist. Participants of the AA group were all male (12/12) and did not have a TB or HIV diagnosis. Results were organized into 3 domains reflecting the aims of the study: (1) virtual counselor acceptability, (2) intervention adaptability, and (3) feasibility of CBI in clinic settings. Results are illustrated using direct quotations from the respondents.

First, the majority (8/10) of in-depth interview participants found the Peedy the Parrot virtual counselor to be acceptable and felt comfortable honestly answering alcohol-related questions.

Moreover, some respondents stated a preference for a virtual character over a human counselor.

HCP focus group participants also found the virtual counselor to be acceptable and reported that they would expect patients to trust the information from the virtual counselor, noting the advantage of confidentiality of a virtual counselor–guided discussion.

The predominantly male HCP and AA focus group participants felt that a male virtual counselor with a loud, confident voice would be more effective in facilitating disclosure of alcohol use and alcohol use–related consequences.

By contrast, the predominantly female HCP focus group participants preferred a female virtual counselor because a female in the role of caregiver is more acceptable and patients would feel more connected to a human character than to a bird character.

In-depth interview participants had no gender preference; although they liked the bird as counselor, few preferred a human virtual counselor over an animal/avian virtual counselor.

Finally, nearly all respondents agreed that it would be important for a live human counselor to be present in the session. Virtual counselors (whether avian/animal or human characters) were noted to have limitations in communication whereas a human counselor could ensure that the information was more persuasive and compelling.

In-depth interview and focus group participants made several recommendations for adaptations to the CBI key characteristics in the areas of drinking triggers, standard drink definition, and IMB skills needed to reduce alcohol consumption. Table 2 shows these recommendations and associated quotations. For local adaptation, participants advised incorporating triggers such as locality and family background, relief from physical pain and stress from family, and type of work. More information was deemed necessary on types of local alcoholic drinks.

Notably, there was a mixed response to the virtual counselor’s discussion of the effect of alcohol on HIV/TB and on gradual versus immediate cessation of drinking. The in-depth interview participants, although diagnosed with HIV or TB, tended not to see the connection between alcohol use and poor health outcomes, whereas AA and HCP focus group participants agreed with the effects and suggested adding more information. HCP and AA focus group participants were concerned that patients tend to have a limited understanding of the impact of alcohol use on HIV and TB drug interactions and on health outcomes such as sexual dysfunction and liver and kidney function. They observed that this knowledge potentially could motivate reduced consumption. Additionally, both in-depth interview and AA focus group participants tended to advocate for abstinence rather than reduction in use of alcohol. Finally, respondents suggested adding behavioral skills to engage participants in (1) religious and spiritual events, (2) learning to refuse alcohol offered by peers, and (3) the willpower to stop drinking alcohol.

Finally, it was suggested that the CBI will be more acceptable if it contains illustrations, especially to aid nonliterate users. Twenty focus group participants suggested that adding more images of major CBI points in addition to the oral presentation by the virtual counselor could improve comprehension and effectiveness.

All participants noted several features of the CBI that would make it feasible for administration in a clinic setting. They reported that intervention duration (about 20 minutes) and amount of information provided were appropriate for use in medical settings. Most in-depth interview participants (8/xx) found the CBI easy to operate. Some in-depth interview participants (2/10) reported that if they were required to use a cellphone instead of the tablet for the intervention, this would be a barrier as some people did not have an Android phone and were not familiar with its operations.

AA focus group participants noted an advantage to the CBI was its capacity for standardized administration, which may not necessarily occur with a person-delivered intervention.

HCP focus group participants also identified challenges to intervention implementation. Currently with no alcohol treatment services in their clinics, a CBI could result in increased provider workload if providers needed to be physically present to assist while patients are using the software program.

In addition, they noted that patients have limited time in the clinics and often limited reading literacy.

Finally, other challenges highlighted were availability of internet access in clinics, confidential space for counseling during outpatient hours, and troubleshooting technical problems. However, they also mentioned that such patients may be handled by appointment in the afternoon time in the clinic.

Computerized interventions delivered by a virtual counselor can reduce unhealthy alcohol use among people with HIV in the United States [36] and with adaptation may show promise in India among patients with HIV or TB, where unhealthy alcohol use is also prevalent. Using 10 in-depth interviews and 3 focus groups, we analyzed the components of a US-developed CBI. Our study findings suggest that HIV and TB patients found Peedy the Parrot to be acceptable as a virtual counselor. Using a virtual counselor rather than a person was seen as an advantage to foster disclosure of alcohol use and provide confidentiality, a demonstrated advantage of CBIs [45]. Virtual counselors are accepted as counselors if they are sufficiently anthropomorphic [46] and empathic, which Peedy was programmed to be. AA and HCP focus group participants specifically suggested that a human virtual counselor (versus animal/avian counselor) could improve acceptability and efficacy of the intervention. This observation reflects an active question in the human-computer interaction literature as to the relative significance of anthropomorphism and realism of virtual agents in fostering connection between users and content [47]. A related feature of virtual counselor evaluation was its optimal gender. In this study, all participants of the AA group believed that a male virtual counselor could facilitate greater disclosure and comfort because unhealthy alcohol use is predominantly reported among men in India. By contrast, the mostly female HCP focus group participants had a preference for the female gender for the virtual counselor.

Such differing evaluations between the HCP and AA focus groups align with findings of numerous studies showing that people tend to prefer congruence with their virtual counselors [48,49]. Our findings underscore the importance of providing choice in a virtual counselor to increase engagement, interest, and comfort in interacting with intervention content and the importance of examining choice in improving alcohol use outcomes [50,51].

The CBI content was rated highly, showing the generalizability of evidence-based, core behavior change components; however, additional key content was identified for inclusion. All study participants suggested adding drinking triggers (influence of surrounding and locality, type of occupation, avoidance of responsibilities) and adding motivators (peer counseling and willpower). Study participants also recommended that the CBI define a standard drink of alcohol for the local setting. As there is much variation in beverage-specific drink size and types of alcoholic drinks across regions in India, local standardization for alcohol interventions is imperative for this CBI [52].

Importantly, there were areas of discrepancy between the HIV/TB patient, HCP, and AA focus groups. The focus groups wanted more information provided on the well-established link between harmful effects of alcohol on HIV and TB health outcomes, whereas HIV/TB patients rejected the link between alcohol and poor HIV/TB health outcomes.

In cessation versus reduction of alcohol use, HIV/TB patients and AA focus groups argued for immediate cessation for all patients, although it is well established that gradual cessation is the medically recommended approach if there are concerns about alcohol withdrawal [1]. Such discrepancies reflect the larger challenge to present credible and persuasive but often unwelcome health information. Many CBI nuances such as visual design and personalization, unrelated to the quality of the health information, are being explored to influence credibility judgments that subsequently influence acceptance of health information [53,54]. This suggests that further refinement of our CBI may be needed to incorporate this emerging evidence base.

While our study found that CBI delivery in government HIV and TB clinical settings is feasible, most respondents believed that it should be delivered in the presence of a counselor or therapist and not as a stand-alone therapy. This qualitative finding corroborates a recent meta-analysis [30] showing that human-supported interventions were more effective than fully automated interventions in reducing mean weekly alcohol consumption among adults with unhealthy alcohol use. In virtual counselor–based care for mental health, both clinicians and participants preferred digital technology as a complement to, rather than a replacement of, face‐to‐face treatments [30,54].

CBIs seem to offer advantages over person-delivered care; however, future investigation will explore these advantages. High fidelity, for example, is presumed; however, it can be attenuated by user inattention to or misinterpretation of intervention content [55]. Incorporating standardized evaluation of human-computer interaction measures will be essential to ensuring fidelity. A second area of future investigation will be to routinely identify the active mechanisms of behavior change. So far, CBIs have varied in their incorporation of evidence-based theory, which limits ability to identify effective components of interventions [56]. Finally, as the accuracy of transdermal alcohol sensors develops, CBIs can be paired via mobile phone to provide needed on-demand delivery of alcohol treatment [57].

A limitation of the study is that currently the CBI is available only in the English language. This necessitated interpretation of the virtual counselor dialog by the authors in the in-depth interview and focus groups. This may have fostered the perception of many respondents that the CBI (apart from technical concerns) could not be effectively delivered without the physical presence of a live counselor. Another limitation of the CBI is that provisions must be made for limited reading and technical literacy. In the hospital setting, this would mean that a minority of patients would attend CBI on an appointment basis for assistance from staff. This could potentially limit administration, especially during busy clinic hours. It may be that future iterations of this CBI include more visually based intervention material that is even simpler to operate.

In summary, CBI appears to be acceptable to HCPs and people with HIV and TB and adaptable to an Indian clinical setting. This promising approach to alcohol counseling based on the IMB model is uniform, structured, organized, and provides intervention fidelity. If effective, CBI would result in not only increased access to evidence-based AUD interventions [58], which are currently not available in Indian HIV/TB clinical settings, but also to improved clinical outcomes and quality of life among people with HIV and TB with unhealthy alcohol use.