Know your audience: Predictors of success for a patient-centered texting app to augment linkage to HIV care in rural Uganda

Background: Despite investments in infrastructure and evidence for high acceptability, few mHealth interventions have been implemented in sub-Saharan Africa. Objective: We sought to (1) identify predictors of uptake of an mHealth application for a low-literacy population of people living with HIV (PLWH) in rural Uganda and (2) evaluate the efficacy of various short message service (SMS) text message formats to optimize the balance between confidentiality and accessibility. Methods: The trial evaluated the efficacy of a SMS text messaging app to notify PLWH of their laboratory results and request return to care for those with abnormal test results. Participants with a normal laboratory result received a single SMS text message indicating results were normal. Participants with an abnormal test result were randomized to 1 of 3 message formats designed to evaluate trade-offs between clarity and privacy: (1) an SMS text message that stated results were abnormal and requested return to clinic ("direct"), (2) the same message protected by a 4-digit PIN code ("PIN"), and (3) the message "ABCDEFG" explained at enrollment to indicate abnormal results ("coded"). Outcomes of interest were (1) self-reported receipt of the SMS text message, (2) accurate identification of the message, and (3) return to care within 7 days (for abnormal results) or on the date of the scheduled appointment (for normal results). We fit regression models for each outcome with the following explanatory variables: sociodemographic characteristics, CD4 count result, ability to read a complete sentence, ability to access a test message on enrollment, and format of SMS text message. Results: Seventy-two percent (234/385) of participants successfully receiving a message, 87.6% (219/250) correctly identified the message format, and 60.8% (234/385) returned to clinic at the requested time. Among participants with abnormal tests results (138/385, 35.8%), the strongest predictors of reported message receipt were the ability to read a complete sentence and a demonstrated ability to access a test message on enrollment. Participants with an abnormal result who could read a complete sentence were also more likely to accurately identify the message format (AOR 4.54, 95% CI 1.42-14.47, P=.01) and return to clinic appropriately (AOR 3.81, 95% CI 1.61-9.03, P=.002). Those who were sent a PIN-protected message were less likely to identify the message (AOR 0.11, 95% CI 0.03-0.44, P=.002) or return within 7 days (AOR 0.26, 95% CI 0.10-0.66, P=.005). Gender, age, and socioeconomic characteristics did not predict any outcomes and there were no differences in outcomes between those receiving direct or coded messages. Conclusions: Confirmed literacy at the time of enrollment was a robust predictor of SMS text message receipt, identification, and appropriate response for PLWH in rural Uganda. PIN-protected messages reduced odds of clinic return, but coded messages were as effective as direct messages and might augment privacy. Trial Registration: Clinicaltrials.gov NCT 01579214; https://clinicaltrials.gov/ct2/show/NCT01579214 (Archived by WebCite at http://www.webcitation.org/6Ww8R4sKq).