This was a cross-sectional analysis of baseline data from a study of persons living with HIV who were recruited from a large HIV clinic in southwestern Uganda, entitled the Alcohol Drinkers’ Exposure to Preventative Therapy for Tuberculosis (ADEPTT) study (registered under NCT03302299 in the ClinicalTrials.gov registry). The study was a single-arm prospective trial examining the safety, tolerability, and adherence to a 6-month daily isoniazid (INH) regimen among persons living with HIV who were coinfected with latent TB, who recently consumed alcohol (two-thirds of the sample) or abstained from alcohol consumption (one-third of the sample).

We conducted screening for the ADEPTT study using a multistep process. Eligibility criteria at the initial screening step included being an adult (≥18 years old) living with HIV, being fluent in Runyankole (the local language) or English, having been prescribed ART for at least 6 months, living within 2 hours’ driving distance of or 60 km from the study site, having no plans to relocate from the catchment area, and having no history of active TB or taking TB preventive medications. Alcohol use eligibility additionally included self-reporting current (prior 3 months) alcohol use, or alcohol abstinence (no use in the prior year), with a target of enrolling 200 persons in the former and 100 persons in the latter category. Exclusion criteria are currently taking or having taken nevirapine in the prior 2 weeks, or taking anticonvulsant medications (both contraindications to INH). Further eligibility criteria included alanine transaminase (ALT) and aspartate transaminase (AST) levels being <2× their normal upper limits, being cleared of active TB (for those reporting TB symptoms), and not being pregnant. Finally, participants were screened for evidence of latent TB by research assistants trained in administration and interpretation of the tuberculin skin test (TST), latent TB was defined as a positive TST finding with an induration of ≥5 mm read 48-72 hours after injection with purified protein derivative (PPD). Those with positive TST results were invited to participate in the study. Data included in this analysis are baseline data.

All study participants were reimbursed for transport costs following completion of study procedures at each study visit.

The study was approved by the University of California, San Francisco Institutional Review Board (16-19093), the Mbarara University of Science and Technology Research Ethics Committee (11/10-16), and the Ugandan National Council for Science and Technology (HS2183). Participants provided written consent prior to study participation.

Data were collected using an interviewer-administered laptop-based survey using the Computer Assisted Survey Information Collection system. The surveys were conducted in private spaces, for confidentiality reasons, in Runyankole or English, depending on the participants’ preference. Baseline blood draws were conducted for testing for a biomarker of alcohol use—phosphatidylethanol [19]. Phosphatidylethanol is a sensitive and specific marker of prior 3 weeks’ alcohol consumption.

We described sample characteristics using proportions for categorical variables and median (IQR) values for continuous variables. We used unadjusted and adjusted logistic regression models to examine associations with cell phone use (primary outcome), and sending and receiving text messages (secondary outcomes). The multivariable models controlled for the following potential confounders selected a priori: age, sex, literacy, HHI, and organized religious activity. In the multivariable model for frequent sending of text messages, we included education instead of literacy owing to a zero cell (ie, no one in the nonliterate group reported sending text messages frequently). Finally, we conducted exploratory analyses using the 3-level alcohol use variable rather than any alcohol use as the main independent variable of interest.

A total of 1434 people were screened for the ADEPTT study. Main reasons for ineligibility included a history of active TB (n=26) or TB medications (n=42), elevated AST or ALT (n=79), and testing negative on the TST (n=848). Of the screened participants, 308 people were found to be eligible for enrollment. Six people declined enrollment, one was later found to be ineligible and was excluded, and one was missing cell phone use data; 300 participants were included in this analysis.

Of the 300 participants included, approximately half of them (n=146, 48.7%) were male and their median age was 40 (IQR 33-47) years (Table 1). The majority of the participants (n=219, 73.0%) had no more than primary school (6 years) education. Two-thirds (n= 199, 66.6%) of the participants were either married or cohabiting. Approximately half (n=158, 52.7%) of the participants reported attending organized religious services at least once a week. The majority of the participants (n=263, 87.7%) were literate. By design, over two-thirds (n=214, 71.3%) of participants had consumed alcohol in the past 3 months.

The majority (n=267, 89.0%) of participants reported they had access to a cell phone; of them, 262 (98.1%) owned the phone and 26 (9.7%) shared the phone. The majority (n=262, 87.3%) of participants used a cell phone (owned a cell phone and had it on for more than half of the day; Table 1). In unadjusted analysis, any alcohol use was not significantly associated with cell phone use (odds ratio [OR] 0.76, 95% CI 0.34-1.67; P=.49) (Table 2). Adjusted analyses showed lower odds of cell phone use among those who consumed alcohol than among those who did not, although this did not reach statistical significance (adjusted OR [aOR] 0.48, 95% CI 0.18-1.25; P=.13).

We observed associations between literacy and cell phone use (aOR 4.15, 95% CI 1.74-9.89; P<.01) and also between middle or high HHI and cell phone use (aOR 8.03, 95% CI 3.26-19.79; P<.01).

We performed exploratory analyses to assess the impact of level of drinking on cell phone use but did not find an association between level of alcohol use (moderate and heavy vs no use) and cell phone use (moderate use: aOR 0.62, 95% CI 0.19-2.04; heavy use: aOR 0.42, 95% CI 0.15-1.17; P=.25; Table 3).

Over two-thirds (n=200, 66.9%) of the total participants reported rarely or never receiving text messages and the majority (n=269, 89.7%) reported rarely or never sending text messages (Table 1). Of those who had access to a cell phone (n=267), only one-third (99/267, 37.1%) reported frequently receiving text messages, while only (31/267, 11.6%) reported frequently sending text messages.

In adjusted analyses, any alcohol use was not significantly associated with frequent sending (aOR 0.82, 95% CI 0.28-2.37; P=.71) or receiving (aOR 1.31, 95% CI 0.70-2.47; P=.40) text messages (Tables 4 and 5).

Age was significantly associated with frequent sending of text messages (aOR 0.93, 95% CI 0.88-0.98; P<.01) but not with frequent receiving of text messages (aOR 0.98, 95% CI 0.95-1.00; P=.09) (Tables 4 and 5).

We performed exploratory analyses to assess the impact of level of drinking on frequent sending and receiving of text messages and found an association between level of alcohol use and frequent receipt of text messages (aORs for moderate and heavy alcohol use, respectively: 0.76, 95% CI 0.35-1.65 and 1.84, 95% CI 0.93-3.66; P=.03), but not with frequent sending of text messages (aORs for moderate and heavy alcohol use, respectively: 0.64, 95% CI 0.17-2.31 and 0.96, 95% CI 0.30-3.05; P=.73; Table 6).

We evaluated cell phone ownership and use (frequency of sending and receiving text messages) in a cohort of persons living with HIV in HIV care, who were coinfected with latent TB in Uganda. We also evaluated the association between alcohol use and cell phone use. The results showed a high prevalence of both cell phone ownership and use in this population. This is consistent with many other studies and surveys of cell phone access and use in Uganda [3,29] and in Africa [30]. This high prevalence of cell phone ownership may be largely attributed to the ease of use and relatively low cost of cell phones [11,31]. Persons living with HIV have been shown to find cell phones convenient as an mHealth tool [32] and have reported that it helps reduce transportation costs and still provides access to much needed health information [3]. We found high levels of cell phone use, indicating the potential utility of cell phone ownership in providing care and treatment for persons living with HIV, and as a tool that can be leveraged for mHealth and interventions.

In this analysis, alcohol use was not significantly associated with cell phone use. This finding held up in exploratory analyses examining the level of drinking. Our findings were different from a study conducted in Uganda among persons living with HIV in a fishing community that found that cell phone use was associated with alcohol use before sexual intercourse, although this was only among young females aged 14-24 years [26]. Other studies, conducted in high-resource settings, in the general population (ie, not focused on persons living with HIV) and among young persons, have found positive associations between alcohol use and cell phone use [33,34]. The differences observed may be due to the fact that in these other studies, the population was much younger than that in our study. In our study, the median age was 40 years while many of these studies were among younger persons. Younger persons may use cell phones for social reasons, including making plans for drinking and entertainment [35] with peers, while older persons may use cell phones more for economic reasons, as they have to provide for families and have more responsibilities; hence, we observed a lack of association with alcohol and instead observed an association with HHI and literacy. These findings suggest the need to carefully evaluate the target population for cell phone–based interventions, and that these interventions should take into consideration the level of literacy and economic status of the population.

We also found that participants rarely sent or received text messages (approximately 30%). This was different from other studies, especially those conducted in middle- and high-resource countries where sending and receiving of text messages was high (60%) [36], but similar to a study among youths with HIV in Uganda (27%) [5]. A suggested explanation for the low text message use is the loss of novelty of sending or receiving text messages on a mobile phone in Africa [5] and therefore decreased use of the function. A great majority of the participants were literate; literacy was associated with frequent receiving of texts, and none of the persons who frequently sent text messages were nonliterate. While we could not include literacy in the model because no nonliterate person reported sending text messages; there was a strong effect of education with sending of text messages, which is consistent with prior reports [37,38]. We did not observe an association between any alcohol use and frequent sending and receiving of text messages; however, the exploratory analysis suggests that level of alcohol use may be associated with odds of frequent receipt of text messages. While cell phone use was high, using text messages alone as an mHealth intervention may not be effective, especially with interventions that require responses.

This study was limited in the fact that the sample was selected for the main study enrollment purposes (ie, having latent TB); this may limit the generalizability of the study. However, we did not observe any major differences between patients enrolled and those not enrolled when we analyzed major factors associated with one of the selection criteria (PPD+ skin test) [39]. Another limitation is that this study may have been underpowered to detect associations, as it was not designed to evaluate the relationships being examined. In a post hoc power calculation, we found the study would have approximately 80% power to detect an odds ratio of 0.34 or lower for the relationship between any alcohol use and any cell phone use, assuming 89% cell phone use among those without alcohol use as observed in our sample. The study was, therefore, likely underpowered to detect the magnitude of the association observed in this cohort.

The strength of this study was that alcohol use was measured both by self-report and objectively by use of a biomarker to confirm actual alcohol use.

Many persons living with HIV own cell phones that are turned on for more than half of the day. There is, however, very little use of text messaging among persons living with HIV in this low-income setting.

There is hope that mHealth interventions in this population can be carried out using cell phones owing to their popularity; however, the interventions may need to employ methods that do not rely exclusively on the sending and receiving of text messages. Future mixed methods studies with large sample sizes should be conducted to further evaluate preferred alternatives to text messages such as interactive voice response and the relationship between alcohol use and cell phone use.