Setting
The study was conducted in eight roadside wellness clinics in Kenya run by the North Star Alliance. The North Star Alliance is an organization providing primary and secondary health services to hard-to-reach populations, including sex workers and truck drivers, through its 36 clinics located at major transit hubs across Africa. In 2015, the North Star Alliance provided services to 253,227 client-visits, of which 18% included HIV testing [10]. When a person visits any North Star Alliance clinic, his/her information is entered into the electronic health record system, including a mobile phone number if the client has one and is willing to share it.
Standard of Care
At every client encounter in a North Star Alliance clinic, HIV testing is offered and the test used is a blood-based (finger-prick) provider-administered test. HIV testing by clients at North Star clinics is tracked in the electronic health record system and a few times a year a text message reminder is sent to those clients who do not have a record of having tested for HIV in the previous 3 months. The message for clients in East Africa reads “North Star Alliance East Africa would wish to kindly remind you to visit any of our roadside wellness centres for HIV testing. Your health, our priority.”
Sample, Eligibility and Consent
For this study, we selected a sample of female sex workers registered in the North Star Alliance electronic health record system who: (1) had no indication that they were HIV-positive, (2) resided in Kenya, (3) had a valid mobile phone number listed, (4) had fewer than four HIV tests recorded in the system in the past 12 months (indicating that they were not following the recommendation to test every 3 months for four tests per year [11]), and (5) had not had an HIV test in the past 3 months.
Once the sample of eligible participants was selected and their data cleaned, the North Star Alliance sent the following passive consent text message twice, once in Kiswahili and once in English, a week apart.
North Star Alliance is evaluating our programs for their improvement using client information from our system. The information we use for this evaluation will not be linked to your name and you will not be contacted or have any expenses related to your inclusion. If you have questions about the use of your data, call [phone number of clinic where they had last been seen]. To have your data excluded, reply “NO” to this text.
After each consent message, any clients who contacted us indicating they wanted to opt out of having their data included were removed from the sample prior to randomization.
Randomization and Intervention
The eligible individuals in our samples who did not communicate their desire to opt out of the evaluation were randomized to one of three study arms.
-
1.
Intervention The intervention consisted of a text message informing participants that HIV self-test kits were available at all North Star Alliance clinics in Kenya. The message was sent three times, one week apart, first in Kiswahili, then in English and then again in Kiswahili, and read:
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.
-
2.
Enhanced Standard of Care (Enhanced SOC) Those randomized to the enhanced SOC arm received the SOC message reminding clients to come to a clinic for HIV testing (described above under Standard of Care), sent three times, one week apart, first in Kiswahili, then in English and then again in Kiswahili.
-
3.
Traditional Standard of Care (Traditional SOC) Those randomized to the traditional SOC arm received the SOC message one time sent simultaneously in both Kiswahili and English.
Sample Size and Power
Our primary outcome of interest was the comparison of HIV testing rates over a 2-month period following the initial text message about HIV testing between the intervention and enhanced SOC arms. We calculated sample size assuming the enhanced SOC would achieve 48% testing rate (38–48% were expected to test after a text reminder based on past records in the electronic health record system), and found that in order to detect a risk ratio (RR) of 1.2 (odds ratio [OR] 1.4) at 80% power and 95% confidence level, we would need a sample of about 750 female sex workers in each study arm. Therefore, we set our target sample size to 750 in each of the two study arms of primary interest, the intervention and enhanced SOC, and determined the randomization ratio after selecting the eligible sample in order to achieve this goal, which ended-up being 1:1:0.93 for the intervention, enhanced SOC, and SOC arms respectively.
HIV Testing Procedures
Study participants who presented at any North Star Alliance clinic in the SOC arms (enhanced or traditional) were offered only the standard provider-administered blood-based HIV test, which is offered to all North Star Alliance clinic attendees. Participants from the intervention arm who presented at a North Star Alliance clinic in Kenya were given a brief demonstration of the self-testing kit and then offered a choice among: (1) the standard provider-administered blood-based HIV test; (2) the self-administered oral HIV test for use in the clinic with provider supervision; or (3) the self-administered oral HIV test kit for home use. Study arm was identified by the clinic receptionist by looking-up the client’s mobile phone number on an Excel spreadsheet listing the numbers of those in the intervention arm. The counselor was informed when an intervention client came in so she would be given a demonstration of the self-test kit and then offered the testing choices. Those in all study arms who visited a North Star Alliance clinic outside of Kenya would be offered the SOC test only, as those clinics did not have self-test kits. In addition, if someone not in the intervention arm came to a Kenyan clinic and specifically requested a self-test kit, presumably having heard about it from someone in the intervention arm, they were given the self-test so as not to miss an HIV testing opportunity. The HIV testing procedures were as follows:
Those who accepted the standard provider-administered blood-based test underwent the standard pre- and post-test counseling and testing process.
Those who chose the self-test for supervised use in the clinic received the standard pre-test counseling and then were given an OraQuick HIV self-test kit [12] which has an insert with written (English and Kiswahili) and pictorial instructions included as part of the kit. An HTC counselor sat in a private room with the study participant while she used the HIV test (supervised self-administration) in order to answer any questions that arose during the test administration and offer correction if needed. Upon the availability of the HIV test results 20 min later, the client was given the option to view the results in private or with the counselor. After viewing the HIV test results, the client received the standard post-test counseling and any needed referrals. If the client chose to view the test results in private, she was encouraged to disclose the test results during post-test counseling, but the final decision whether or not to disclose was the client’s. If she did not disclose the results, the counselor was to provide the post-test counseling information for both scenarios (HIV-positive and HIV-negative test result), including information about accessing HIV care in the case of a positive test result.
Those who chose to take a self-test kit for use outside of the clinic (i.e., home use) were given pre-test counseling in the clinic and then instructed to use their test within 3 days and to call or send a text message after using the test to receive a call-back for post-test counseling and any necessary referrals. Participants who failed to contact the clinic staff within 3 days after obtaining a test kit were called to inquire about the use of the test kit and provided counseling and referrals if needed. Clients were also told that they could call or send a text at any time while self-testing should they have any questions or concerns. As with in-clinic self-testing, clients were encouraged to disclose their test results during post-test counseling, but whether or not they did so was the client’s choice and if she did not disclose, the counselor was to provide information about both HIV test outcome scenarios.
Data Collection Methods
For this study, we relied on data from two sources: (1) the North Star Alliance electronic health record system which documented HIV testing and which test was used (provider-administered test or self-test), and (2) administrative data collected at the clinics in a password-protected Excel spreadsheet for tracking the number of self-test kits used to order resupplies when needed and for tracking time since a client took a self-test kit for home use to ascertain when to contact the client if they failed to call for post-test counseling.
The study procedures were approved by the City University of New York Institutional Review Board, the Kenya Medical Research Institute Ethics Committee, and the University of KwaZulu-Natal Biomedical Research Ethics Committee.
Data Analysis
We described the sample in terms of the basic demographic characteristics (those available in the health record system) overall and by study arm and assessed the statistical significance of any differences by study arm using a Chi square test for categorical variables and a Kruskal–Wallis test for numeric variables. We then conducted logistic regression analysis to compare HIV testing during the 2-month follow-up period among clients in the intervention arm versus those in the enhanced SOC arm (primary comparison) as well as among those in the enhanced SOC versus those in the traditional SOC arms (secondary comparison) to look at the impact of the content of the text message (i.e., about self-testing kits or HIV testing in general) and of the number of text messages (3 vs. 1) on HIV testing, respectively. We also used logistic regression to look at differences in clinic contact for any reason (i.e., for HIV testing or some other service) between the groups to see if the text message brought more clients to the clinic even if some chose not to test. Finally, we looked at whether the differences in HIV testing by study arm were modified by HIV testing history (whether the client had an HIV test at a North Star Alliance clinic in the past year) and, for those in the intervention arm, we describe the HIV testing choices made.
We found some discrepancies between the electronic health record data and the clinic administrative records that were kept on self-testers in the number of people who self-tested for HIV in the intervention arm. Specifically, there were 38 female sex workers whose data in the electronic health record system did not indicate an HIV test but the clinics listed them as having self-tested. This might occur for a number of reasons, such as the counselor forgot to enter the data in the online health record system, or entered it but the internet connection was disrupted while the data were being sent to the server, or the data were entered after we downloaded the data for these analyses. The data were downloaded from the health record system 2 months after completing follow-up. Because of this discrepancy, we first analyzed the data including these 38 female sex workers as not having tested (as indicated in the electronic health records) because we did not have similar administrative data on HIV testing from the clinics for those in the SOC arms. Differential data cleaning could bias the results and incorrectly elevate the association between the intervention and HIV testing and we felt it best to err on the conservative side knowing that we might have bias toward the null. However, we also ran the analysis recoding those 38 female sex workers as having tested, as indicated in the administrative clinic records, to see if it changed our results substantively; while the strength of the effect increased as would be expected, the conclusions regarding the significance of the associations remained unchanged in all comparisons. In addition, after the study began, two participants in the intervention arm disclosed that they were HIV-positive to clinic staff when offered HIV testing. There was no indication in the electronic heath record system of their HIV-positive status. We included these two individuals in the intervention arm for analysis to maintain the randomization, despite the fact that it would bias our findings toward the null.