For this study, we used methods similar to those we used in the study among female sex workers in Kenya . Below we briefly describe the main design characteristics of the study.
Setting and population
This study was conducted in the North Star Alliance clinic system, which aims to bring primary and secondary health services to hard-to-reach groups across Africa, including truckers and sex workers. In 2015, the North Star Alliance reported 253,227 client-visits at their 36 roadside wellness clinics across Africa, which includes the eight clinics in Kenya participating in this study. Eighteen percent of these visits included HIV testing . Information on clinic clients is entered into the electronic health record system (EHRS), including their mobile phone number if the client has one and is willing to provide it.
Current HIV testing standard of care
Clients presenting at North Star Alliance clinics are offered a blood-based (finger-prick), provider-administered rapid HIV test (standard HIV test). HIV testing is tracked in the EHRS and text message reminder is sent to those clients who have not tested in the past three months. The message reads “North Star Alliance East Africa would wish to kindly remind you to visit any of our roadside wellness centers for HIV testing. Your health, our priority.”
Sample and eligibility criteria
Study participants were selected from the EHRS. Eligibility criteria included: (a) no indication of being HIV-positive, (b) Kenya resident, (c) valid mobile phone number listed, (d) had fewer than four HIV tests in the past year (suggesting that they were not testing every three months as recommended for those at high risk ), (e) no indication of an HIV test in the past three months, (f) had not participated in our previous study on self-administered HIV testing [3, 9, 17], (g) were male, and (h) worked as truckers, including drivers and assistants (turn boys).
Once the sample of eligible participants was selected, the North Star Alliance sent a passive consent text message twice, a week apart, first Kiswahili and then in English that informed clients that their medical record data would be used for program evaluation and provided instructions for how clients could opt out of being included if they chose to do so. Clients who indicated they wanted to opt out of the evaluation were removed from the sample prior to randomization. Requests to opt out after randomization were considered study drop-outs and were excluded from the data analysis.
Randomization and intervention
The eligible individuals who remained in the sample after the consent process were randomized to one of three study groups.
Intervention, in which a text message was sent three times, one week apart, alternating in English or Kiswahili that stated:
“You can now self-test at home or in the clinic for HIV using a new test kit available from all North Star Alliance clinics in Kenya. Your health, our priority.”
Enhanced SOC, in which the SOC text previously described was sent three times, alternating in English or Kiswahili.
SOC in which the SOC text was sent one time in both Kiswahili and English concurrently.
Sample size and power
We powered our study for to compare the HIV testing rates over the two months following the first text message between the intervention and enhanced SOC groups. We determined the required sample size for these two groups using data from the EHRS, which suggested that the HIV testing rates after sending the SOC text message among truckers was ≤68%. We used the maximum rate of 68% in order to account for any small increase that might occur when sending the text three times instead of only once in the enhanced SOC. In order to detect a 20% increase in the proportion testing in the intervention group over the anticipated 68% in the enhanced SOC group (RR = 1.2, OR = 1.4) at 80% power and 95% confidence level, we required a sample size 750 truckers in each of the two groups. Therefore, we set our target sample size to 750 truckers in the intervention and 750 in the enhanced SOC. If we had > 1500 eligible clients, we planned to randomize the extra in the sample to the SOC group. Thus the probability of being randomized into each of the three study groups was determined by the number of eligible participants in order to ensure 750 in the intervention and 750 in the enhanced SOC groups. It turned out that we had 2262 eligible clients participating, giving us a randomization ratio of 1:1:1.02 for the intervention, enhanced SOC and SOC groups respectively.
Masking & HIV testing procedures (program)
Participants were not informed about the specific research question or that they would be randomized to different HIV testing programs in order to avoid bias. Participants in both SOC groups (enhanced or standard) were offered only the standard HIV test when seeking clinic services. Those in the intervention group who sought services in one of the clinics in Kenya were given a demonstration of the self-testing kit and offered choices among (1) the SOC HIV test, (2) the self-administered oral HIV test for use in the clinic with provider supervision, or (3) a self-administered oral HIV test kit to take and use outside of the clinic (e.g. home use).
Those who accepted the standard HIV test underwent the standard pre- and post-testing counselling and testing process. Those who chose the supervised self-test in the clinic received pretest counseling and then were given the OraQuick HIV self-test kit  to use while a counselor sat in the room to answer questions and provide guidance if needed, followed by posttest counseling. Those who took a self-test kit home were given pre-test counseling in the clinic and then instructed to use their test kit within three days and to call or send a text message if they had any questions while testing and after using the test to receive a call-back for post-test counseling and referrals. Participants who did not contact the clinic staff by day three were called repeatedly by clinic staff until they were reached.
The study was approved by the City University of New York (CUNY) Institutional Review Board, the Kenya Medical Research Institute Ethics Committee (KEMRI), and the University of KwaZulu-Natal Biomedical Research Ethics Committee (BREC).
We described the sample overall and by group to ensure that the randomization worked as expected. The statistical significance of any differences by group was assessed using the chi-square test for categorical variables and the Kruskal Wallis test for numeric variables. We compared the proportion who tested for HIV during the two-month follow-up period between clients in the intervention and those in the enhanced SOC group (primary comparison) as well as between those in the enhanced SOC and those in the SOC group (secondary comparison) to estimate the impact of the text message content (i.e. about self-testing kits or about HIV testing in general) and number (one versus three) on HIV testing, respectively. We also looked at differences in clinic contact for any reason (i.e. the client came to a North Star Alliance clinic during the follow-up period for any services, not just HIV testing) between the groups to determine if the text message influenced how many truckers came to the clinics even if some of them did not test. We used logistic regression for these comparisons and all models were based on intent-to-treat; even if someone did not receive the text messages we sent, they were included in the arm to which they were randomized.
EHRSs are not a perfect data source and we found that there were 5 clients without an indication of HIV testing in the EHRS who were documented as having self-tested in the administrative records kept by the clinics to track the use of self-test kits for inventory purposes and to follow-up with those who took self-test kits for home-use. We considered the EHRS data as our primary data source, since we did not have administrative data on non-self-testers, and first analyzed the data as reported in the EHRS. However, as a sensitivity analysis we also ran the models coding those five participants as having tested, as documented in the administrative records, to determine if our results changed substantially.