Study design
The study was a randomized controlled trial in which access to SRH information, goods, and services using a MPA (MPA-SRH) was compared to the standard of care of access to SRH information, goods, and services (SOC-SRH). A protocol of the trial, which was a part of a broader app development and impact evaluation project, has been published previously[9]. The trial was approved by the Mbarara University of Science and Technology Ethical Review Committee and the Uganda National Council of Science and Technology and was registered at MUREC1/7 No. 07/05–18 on 29th June 2018.
Participants
The trial included students of Kyambogo University (KYU), the second-largest university in Uganda, which is situated in the capital city, Kampala. Eligibility criteria for inclusion into the trial were: (1) age 18 to 30 years, (2) self-reported sexual activity in the last six months, (3) more than 12 months to graduation, (4) access to an internet-enabled Android smartphone, and (5) informed consent. Trial participants were recruited in KYU halls of residence and KYU-affiliated hostels. Data were collected by in-person interviews in students’ rooms.
Intervention
In the intervention group, trial participants were granted access to a MPA to enable their access to SRH information, goods, and services over a period of six months. Details of the development and features of the app are described elsewhere[10]. Briefly, the MPA was developed in partnership with Gershom Technologies Ltd using the Android operating system and archived on the Google Play store. The MPA was designed to link the different goods and services providers: healthcare facilities which provided SRH goods and services; Beyonic Uganda (https://beyonic.com/), a payments company which managed payments and co-payments; SafeBoda (https://safeboda.com/ug/), a transport company which managed the transportation of ordered goods and app users to receive services; and GHE Consulting (https://www.gheuganda.com/) for coordination, management, and oversight.
The MPA included the following features: (1) sign-up and sign-in; (2) a user module for ordering SRH goods (sanitary pads, male condoms, contraceptives, pregnancy tests, and pain killers) and services (HIV voluntary testing and counseling (VTC), STI diagnosis and treatment (D&T), family planning counseling, and general SRH consultation); (3) an SRH information module (menstrual period tracker, frequently-asked questions (FAQs), SRH tips, and a live chat); (4) a payments module to enable provider payments by GHE Consulting, copays by clients, and payments for transportation; (5) a delivery module to enable clients to track shipments, set up pickups for in-facility visits, and set up pick up points for products; and (6) a security module for authentication and password protection. Embedded in the MPA was an advertising interface to test potential sustainability at the scale of in-MPA advertising. While the study provided a major subsidy for SRH goods and services as well as transportation, there was a modest co-pay that was paid by app users through a link between mobile money and the Beyonic payment system. The co-pay was designed to test the utility of client payments in potential future iterations of the MPA after the pilot period.
Procedures
Intervention group participants received a text message with a link to download the MPA, including directions on the set up of a MPA account and instructions on the use of the MPA. Upon download of the MPA, participants could gain access to SRH information through the four in-app portals. Participants could order SRH goods and have them delivered to their rooms or to a designated spot for pick up. Participants could also connect with health service providers to book SRH services and connect with transport providers to organize transportation to and from healthcare facilities. Participants in the SOC-SRH group received no intervention, i.e., accessed SRH information, goods, and services as they did before the onset of the trial.
Outcomes
There were four primary outcomes in the trial all reflecting changes from baseline to end-line (end of six months follow-up period): SRH knowledge score (SRH information), use of contraceptives (SRH goods), use of HIV VTC (SRH services), and use of STI D&T (SRH services). There were two secondary outcomes in the trial, both reflecting changes from baseline to end-line: use of condoms and use of alcohol during the last sexual encounter (both behavioral and attributable to SRH information).
Sample size
The sample size was calculated to detect a change between baseline and end-line in use of modern contraceptive goods. The prevalence of modern contraceptive use among sexually active youth aged 18 to 30 years at the time of study planning was 30.2%. Using the formula for the use of the Z-test for two sample proportions in studies with behavioral components, we calculated a sample size of 435 participants per group in order to obtain 90% power to detect a 15% change in modern contraception use in this population at a two-sided significance level of 5%. The sample size calculation assumed no design effect (d = 1) and was adjusted for a 10% non-response rate to obtain a minimum sample size of 479 per group (a total of 958 participants).
Randomization and masking
After recruitment and informed consent, participants were informed that there was a 50-50 chance that they would be randomly assigned to gain access to the MPA (MPA-SRH) or be a part of the control group (SOC-SRH). The participants’ telephone numbers and unique study identification numbers were given to the app developer, who was responsible for randomization. Participants were randomized 1:1 to MPA-SRH and SOC-SRH using computer-generated random numbers. The research team, including providers at health facilities, transport providers, and payment technicians, and participants were blind to the intervention group, but the app developer was not.
Statistical analysis
Data were double-entered and cross-validated, and all analyses were conducted in Stata version 16 (College Station TX, USA). We compared outcomes between participants randomly assigned to MPA-SRH versus SOC-SRH, ignoring potential changes in access to SRH information, goods, and services in the SOC-SRH group that may have been triggered by the use of the app by participants in the MPA-SRH group.
SRH knowledge scores were summed from correct responses from 17 different attributes and ranged from 0 -17 (each question was awarded 1 point). Contraceptive use, HIV VTC, STI D&T, condom use during last sexual encounter, and alcohol use during the last sexual encounter were binary variables. We estimated the difference in changes from baseline to end-line (with 95% confidence intervals (CIs) and p-values) using differences-in-differences. We used the Wilcoxon rank-sum test for the knowledge score and the Z-test for proportions.
Using quantile regression (knowledge score) and a logit link function (access to modern contraceptive use, HIV voluntary counseling and testing, and STI testing and treatment), we estimated a mean differences-in-differences effect, adjusting for demographic characteristics. Models were estimated using maximum likelihood.