Alcohol consumption is a critical driver of the HIV epidemic worldwide, particularly in sub-Saharan Africa (SSA), where high-risk alcohol use and HIV are prevalent. Unhealthy alcohol use, defined as drinking more than the recommended amount of alcohol [1], is associated with increased sexual risk behavior, increased HIV transmission [2-8], and diminished treatment outcomes among people with HIV, including reduced antiretroviral (ARV) adherence [9-13] and reduced HIV viral load suppression [9,11,13-16]. Thus, reducing alcohol use among those with HIV is a public health priority.

In Uganda, more than half of the population abstains from alcohol use; however, among people who drink alcohol, most of whom are male, the yearly average consumption is 26 L of absolute alcohol, which translates to 474 L of 5.5% of alcohol by volume (typical for beer) [17]. The prevalence of heavy episodic drinking, defined as consuming at least 60 g of pure alcohol on one occasion in the previous 30 days, is 68.8% among men and 32.6% among women who consume alcohol [17]. Among people with HIV in SSA, meta-analyses have found the pooled prevalence of alcohol use disorder, defined as problem drinking that is at risk of becoming severe [18], to be 22.9% to 29.8% [19,20]. Brief interventions to reduce alcohol use have shown efficacy in reducing drinking by 15% to 30% one year after the intervention, as well as good feasibility and cost-effectiveness in primary care settings worldwide [21-28]. However, the usefulness of brief alcohol interventions depends on effective screening for unhealthy alcohol use [23,29-40], which is often underreported [41-44]. For example, we have found substantial underreporting of alcohol use by people in HIV care in Uganda to clinicians at clinic intake visits [43]. Other studies have reported high rates of underreporting of alcohol use when compared with biological measures such as phosphatidylethanol [41,44-46]. Thus, there is a need to develop methods to improve reporting of unhealthy alcohol use as an essential step toward referral to brief alcohol interventions.

The use of tablets for self-completion of clinic intake forms is increasing in many resource-rich settings; for instance, several clinical settings in the United States and Sweden have examined the use of digital screening and found that reporting of risk behaviors was comparable with more traditional screening methods [34,47-57], and the digital screening over a wide range of patient characteristics was acceptable [58-62]. Audio computer-assisted self-interviewing (ACASI) has been associated with increased reporting of stigmatized behaviors such as forced sex, especially in Africa and rural settings [63]. ACASI has also been useful for populations with low literacy, and studies have shown high levels of comfort with its use among men and women as well as in older adults across international settings [64-66]. The rise in mobile device use worldwide makes digital screening technology ubiquitous. In addition, the simplicity of touch screen computers (tablets or smartphones) allows populations with low literacy, such as in Uganda, where 24% of the adult population (aged ≥15 years) have low literacy [67], to self-administer questionnaires that would normally require the assistance of a third party. Feasible, acceptable, and efficient methods for the assessment of sensitive behaviors in settings with low literacy are essential for comprehensive HIV care. Thus, we sought to develop and test a brief (3-5 minutes), digital, easy-to-use self-administered health screener for implementation in a busy, underresourced, low-literacy HIV treatment setting.

We conducted patient and clinician focus group discussions (FGDs) and in-depth interviews (IDIs) to develop and pilot-test a digital self-administered screener, called the Self-Administered Digital Screener for Health (SASH), for potential use in HIV clinic waiting rooms to increase reporting of unhealthy alcohol use in people with HIV in Uganda. We examined the acceptability of the SASH by exploring what participants thought of the health screener and their experience with the SASH, including their ability to complete it on a tablet. We also explored whether the use of the SASH would be feasible in our study setting, specifically asking questions about the practical and logistical issues of implementing the new technology within standard HIV care.

We conducted the study at the Immune Suppression Syndrome (ISS) Clinic of the Mbarara Regional Referral Hospital (MRRH) in Uganda, an 11,000-patient clinic with 45 patients seen per day per clinician. The ISS Clinic uses electronic medical records via the Open Medical Record System [68]. Adherence counseling, which sometimes includes brief advice on alcohol use, is conducted by the HIV counselors in individual and group formats at the clinic. The MRRH is a government referral hospital in the western region of Uganda with a bed capacity of 600. It is also the teaching hospital for Mbarara University of Science and Technology, home to the second-largest medical school in Uganda. The MRRH is located in the semirural city of Mbarara in southwestern Uganda, approximately 250 km from the capital city of Kampala. Although ISS Clinic patients come from throughout Western Uganda, our previous studies, from which the participants were sampled, limited inclusion to those who lived within 60 km or 2-hour driving distance of the ISS Clinic for ease of follow-up. Patient participants for this study were purposefully sampled to meet study criteria from existing research databases. All clinicians attending a weekly clinic meeting were invited to participate; during the meeting, those interested we given additional study information by study staff before providing informed consent.

The study was conducted in 4 phases summarized in Table 1, and as described in the following sections.

In phase 1, we conducted 4 FGDs to elicit input on the content, format, and feasibility of the proposed screener. All FGDs were held in a private space at the clinic. Clinicians were invited to participate after the study team gave a presentation of the study during a weekly ISS Clinic meeting. We recruited patients by sampling participants from our previous study databases [69,70] (all consented to future contact) based on previous unhealthy alcohol use underreporting, sex, and literacy levels, defined as follows. Previous underreporting was defined as not meeting self-report criteria for unhealthy drinking via the Alcohol Use Disorders Identification Test-Consumption (ie, women scoring <3 and men scoring <4) [71] but having a phosphatidylethanol alcohol biomarker level of ≥50 ng/mL. Literacy (literate vs low or no literacy) was defined as the ability to read a prescribed sentence on a card when asked during study interviews. For this phase, we aimed to recruit 20 people who had underreported their alcohol use, including 10 people with low literacy and 10 people who were literate. We sought to include people with low literacy to represent the clinic population with low literacy. The FGD groups were balanced by literacy level and segregated by sex. Eligible participants were invited either via phone or in person during an Mbarara ISS Clinic visit.

We conducted clinician FGDs in English and patient FGDs in Runyankole, as desired by the participants. The discussions were conducted by a research assistant (RA; AK, hereafter referred to as the RA) while another researcher (CN) recorded the sessions, took notes, and kept time. Both are Ugandan with over a decade of HIV and alcohol research experience. For this study, the RA received training in qualitative methodology from an experienced qualitative researcher on the team (SWK) and in Dedoose (SocioCultural Research Consultants, LLC) [72] by the project director (NE). Participants were asked questions about their experience discussing alcohol use at the clinic, the acceptability and feasibility of digital screening for alcohol use, suggestions for implementing digital screening in the clinic, and suggestions for the content of a screening tool. All sessions were audio recorded, transcribed, translated into English, and uploaded to Dedoose [72] for coding by the RA. Following each FGD, the RA wrote summary notes with key observations and reflections on emergent themes, tone of the discussions, and nonverbal communication such as body language and gestures. The study team analyzed the notes using a rapid approach [73].

We based our initial set of questions on information obtained from the phase 1 FGDs (eg, which screening questions to include) and on an existing automated screening tool using interactive voice response in a study that developed and evaluated the acceptability and use of the tool in a primary care clinic waiting room [59]. The study concluded that the use of their electronic screener in the clinic setting was feasible and accepted by both clinicians and patients, with reporting rates comparable with published written questionnaires [59]. We created a tablet-based prototype of the health screener to run on the Android operating system, the most widely used operating system in Africa [74], and used CommCare (Dimagi, Inc) [75] software as it was free and user-friendly; allowed for offline use; and had capabilities for including more than one language, pictures, and audio, which were all important criteria for implementation in a limited-resource setting. On the basis of the phase 1 FGDs, we determined that we needed to include a training module. The training module, conducted on the tablet, included written and recorded instructions on how to use the device, respond to questions, select response options, repeat questions, and end the session. Two sample questions—“What year is it?” and “How old are you?”—were included as practice questions.

We modified the SASH prototype through an iterative process that comprised 2 rounds of demonstrations, participant hands-on use, and interviews, followed by a rapid analysis and modification of the prototype’s content and appearance customization. For this phase, we selected 8 previous FGD participants, including people from each of the categories of respondents (clinicians, people with low or no literacy, and those underreporting unhealthy drinking), who emerged from the FGDs as opinion leaders to participate in the demonstration of the prototype. In each session, the RA demonstrated the SASH prototype, trained the participants on its use, and asked participants to use it in her presence. The RA solicited feedback on the wording and comprehensibility of the content in the local language, Runyankole; the layout of the screen (buttons, colors, and icons); the sound or accents of the voice recordings; and the participants’ preferences for a stylus or finger to touch the screen. The RA also requested feedback on the adequacy and effectiveness of the training module, which she summarized in written documents.

Following each individual demonstration and testing session, the RA wrote summary notes highlighting the user experience, content areas for discussion because of lack of clarity or other discrepancies, suggestions, questions, and comments raised by the participants. On the basis of a rapid analysis of the summary notes [73], we modified the prototype and conducted a second round of demonstrations with feedback solicitation 6 to 8 weeks later with the same participants.

In the final phase, we pilot-tested the SASH followed by IDIs with 20 new participants recruited from our previous studies and who met the criteria for unhealthy drinking, including 10 (50%) who had previously underreported their drinking and 10 (50%) with low or no literacy, as described previously.

The RA demonstrated the use of the SASH while the participants observed. Next, the participants proceeded to the training module, in English or Runyankole, on the tablet. Participants were allowed to use the training module repeatedly and with the RA’s (AK) assistance as needed. The RA noted where difficulties occurred during the training module and screening questions. After completion of the SASH, participants were asked to share their experiences using the SASH in a 30- to 60-minute IDI using a semistructured interview guide. All interviews were audio recorded, transcribed, translated into English, and uploaded to Dedoose by the RA for coding and analysis. Following each IDI, the RA wrote summary notes that captured the general tone of the interview and reflected on emergent themes, nonverbal communication observed, and points for discussion with the larger study team.

Two Ugandan staff members (AK and CK) and the US-based project director (NE) initially reviewed all transcripts and written summaries from all phases of the study for completeness, language, and translation accuracy. All the documents were uploaded to Dedoose for data management and coding. The 3 individuals, all experienced in qualitative analysis, worked together to code the data concurrently using both inductive and deductive methods. A predetermined set of codes was informed by the domains of inquiry explored in the FGD and IDI guides. In addition, the analysis team open-coded transcripts and summaries to identify emerging themes. Inductive codes were defined and agreed upon by the team. Next, to ensure reliability in coding, the team triple-coded the first 4 transcripts, met as a team to ensure consistency in coding, and then completed the coding. The team worked on coded data to identify themes using thematic analysis [76,77]. We used content analysis [76,77] to interpret the individual interview data, including the systematic assignment of the predetermined codes.

The study was approved by the University of California, San Francisco Institutional Review Board (15-6933); the Mbarara University of Science and Technology Research and Ethics Committee (02/08-15); and the Ugandan National Council for Science and Technology (HS 1977). All participants provided written informed consent before study participation in their preferred language (English or Runyankole). Transportation and refreshments were provided as incentives.

A total of 45 people (n=5, 11% clinicians and n=40, 89% patients) participated in the study. The 5 clinicians included 1 (20%) medical officer, 1 (20%) clinical officer, 1 (20%) nurse, 1 (20%) counselor, and 1 (20%) peer educator, from whom we did not collect data on age. We attempted phone contact with 114 participants from previous National Institutes of Health–funded alcohol research studies conducted by the research team at the ISS Clinic. We reached 66 people by phone, of whom 26 (39%) declined participation and 40 (61%) enrolled in the study. Participants declined mainly because of time constraints. The 40 patient participants had a median age of 38 (IQR 32-44) years, 65% (26/40) were male, 43% (17/40) were not literate, and 70% (28/40) had previously underreported their drinking (Table 2).

The first version of the SASH prototype was developed by the study team (phase 2). We considered 5 topics (pain, smoking, alcohol use, physical activity, and depressed mood) used in a previous waiting room screener [59] plus ARV adherence. The chosen topics for inclusion in the SASH were agreed upon by the study team and were based on what we considered to be the most relevant in our study setting from our previous experience in the Mbarara ISS Clinic. It was also important for the study team to screen for behaviors that would allow for feasible interventions in the future in this setting. We included questions about ARV adherence in the SASH as it was frequently discussed during the FGDs and we felt that including general health questions along with questions on sensitive behaviors would help normalize the SASH as a general clinic tool, making it more acceptable to patients in this setting where sensitive behaviors such as alcohol use are often stigmatized. We used validated questions on adherence [78], depression [79], and alcohol use [71].

The final content of the SASH included a training module with 2 questions plus 13 questions that covered sex, general health status, adherence to ARVs and other medicines and herbs, alcohol use, sexual health, and depression, as well as a final question asking participants if they wanted to talk to a clinician about their health issues (Table 3). The responses to this last question, as with others in the SASH, were not followed by clinic referrals as the scope of the study was limited to testing the SASH. The SASH included color-coded multiple-choice questions and response options (Figure 1), with an additional option for participants to play the corresponding audio clip by pressing the audio symbol shown. The entire SASH was available in English or Runyankole, the local language.

The SASH was further refined through an iterative process with the patients (phase 3). Following the demonstrations of the SASH, patients provided input on the content, order of questions, and mode of delivery. We determined that we needed to allow for multiple training sessions after all participants requested additional training time and repeated the training module several times. Thus, we separated the training module from the actual screening questions to allow participants to repeat the training module until they felt comfortable with and confident about the technology. The inclusion of multiple-choice as opposed to open-ended response options, particularly with questions on the “number of days drinking,” was also guided by feedback from patients who felt that it was difficult to recall these measures with accuracy. Both clinicians and patients in the FGDs felt that it was best to place perceived difficult questions, such as those about depression, at the end. Some suggestions were outside the scope of a screening tool (eg, a female patient suggested that the digital screener serve as a medication adherence intervention).

We developed and pilot-tested a touch screen, digital health screener with the potential to increase reporting of unhealthy alcohol use in people with HIV in Uganda, a low-resource setting with varying levels of literacy and reasons for underreporting alcohol use. The resulting SASH is a health screener with 15 screening questions illustrated with colorful images and a voice option that reads the questions and response options to the participants. We found that the SASH was acceptable to clinicians and patients who consumed unhealthy levels of alcohol, including those with a history of underreporting their alcohol use, and usable by patients with a range of literacy levels. Although our scope did not include feasibility studies, we explored patient and clinician perceptions of the feasibility of implementing a digital screener in their clinic setting. Their responses focused on practical concerns such as where and when to use the tablet within the clinic; charging the tablet; cost; time taken to complete the screener, including training; and the potential additional workload for clinic staff who would have to train patients to use the tablet and screener. Clinicians and patients in our study shared a desire for the SASH to be used to improve patient care through data sharing with clinicians to mitigate communication barriers between patients and clinicians and save clinicians’ time.

Training, literacy, and privacy of information were the primary concerns regarding the use of a digital screener in this low-resource and low-literacy setting. We found that including a training module preceding the screening questions was crucial. The fear of “failing” the screening or mishandling the device made participants particularly anxious, and all (20/20, 100%) requested retraining with the RA during testing because of this. It was unclear whether these concerns about breaking the device while handling it were rooted in patients’ general concerns about the punitive consequences of making mistakes or mere fear of using new technology. Nonetheless, a notable finding in all phases of this study was that additional training beyond the basic instructions and demonstration of the prototype, including repeated practice sessions, was critical for many clinicians and patients. Training was emphasized as essential regardless of the participants’ previous experience with smartphones, electronic devices, or other technology. The need for additional training, as described previously, was similarly noted in another study of new technologies in a clinic setting in SSA [80], as well as in previously mentioned studies that tested ACASI in settings with older adults and patients with low literacy [64-66].

Our results are consistent with previous studies that demonstrated the feasibility and acceptability of digital and web-based previsit screening for substance use [59,81-87]. Although most of these studies were conducted in high-income countries, the results of those conducted in SSA are promising as well [88-92]. A meta-analysis has shown that digital screening followed by brief alcohol interventions is effective in reducing weekly drinking [93]; however, few studies have focused on whether digital screening increases entry into interventions. By contrast, this study aimed to develop a screening tool as a critical precursor for intervention in a setting where alcohol use is both prevalent and stigmatized [43].

We found evidence of tensions between health care providers and patients that affected their ability to discuss sensitive health topics, including sexual health and substance use, during routine clinic visits. This is consistent with a qualitative study with patients and providers in a rural primary care clinic in the United States that found patient-provider relationships to be critical in the feasibility of substance use screening and that patients’ preference was for self-administered, tablet-based screeners [94]. Clinicians in our study expected patients to underreport their drinking but also expressed frustration when patients were not forthcoming about their drinking on consultation. Similarly, patients frequently described their interactions with providers as hostile, with punitive consequences for behaviors considered to be unacceptable, including unhealthy alcohol use. Therefore, avoidance of punitive consequences was a reason for underreporting in addition to maintaining dignity in the presence of health care providers. The preservation of social status for people with HIV, already burdened by the stigma of HIV, was noted as a reason for socially desirable reporting in this study. Both clinicians and patients felt that clinicians did not have enough time for patients during clinic visits and, as such, deeper explorations of mental health and behavioral issues in the context of current HIV care were limited. These findings suggest that digital screening methods may help mitigate the barriers imposed by face-to-face questioning on sensitive topics. More private and anonymous screening methods may prove particularly useful in settings where sensitive behaviors are stigmatized and pressures within patient-provider relationships prohibit accurate self-reporting.

Our success in developing and pilot-testing the SASH with 20 people with HIV who drank alcohol and had varied levels of literacy in the Mbarara ISS Clinic could potentially be replicated and scaled up in similar settings. Although this screener primarily focused on alcohol use, waiting room screening tools have the potential to efficiently and sensitively screen for key health issues such as mental health and medication adherence and ultimately lead to improved referral and treatment for several health and psychosocial issues.

This study had some limitations. The time required to train patients to use a new digital device may have diminished the focus on its actual use as a screener and may have negatively influenced perceptions. We intended for the SASH to be brief (3-5 minutes), but it took participants a relatively long time (approximately 1 hour or more) to complete the 15-question screener with a training module. Furthermore, our translation of “digital screener” more generally as “technology” in Runyankole may have affected patients’ understanding of the larger context of the SASH as a screener and limited their input on content development as well as their discussion of its specific use as a screener in the clinic. Therefore, feedback from some patients focused on the use of the SASH as a learning tool, medical resource, or digital suggestion box to which complaints about their care could be lodged rather than on its use solely as a screener. It was beyond the scope of this study to provide the SASH results to clinicians and determine their impact on patient care or clinical practice. Finally, given our limited scope, we could not fully explore the feasibility of implementing the SASH in this clinic setting.

Our findings suggest that a digital health screener has the potential to improve reporting of unhealthy alcohol use in clinical care for referral to alcohol interventions in HIV clinics in low-resource settings. Further studies are needed to determine the efficacy of the SASH in improving self-reporting and further develop means of implementation.