The Effect of Human Immunodeficiency Virus Prevention and Reproductive Health Text Messages on Human Immunodeficiency Virus Testing Among Young Women in Rural Kenya: A Pilot Study.

BACKGROUND: More than half of human immunodeficiency virus (HIV)-infected individuals in Kenya are unaware of their status, and young women carry a disproportionate burden of incident HIV infections. We sought to determine the effect of an SMS intervention on uptake of HIV testing among female Kenyan college students.
METHODS: We conducted a quasi-experimental study to increase HIV testing among women 18 to 24 years old. Four midlevel training colleges in Central Kenya were allocated to have their study participants receive either weekly SMS on HIV and reproductive health topics or no SMS. Monthly 9-question SMS surveys were sent to all participants for 6 months to collect data on HIV testing, sexual behavior, and HIV risk perception. We used multivariate Cox proportional hazards regression to detect differences in the time to the first HIV test reported by women during the study period.
RESULTS: We enrolled 600 women between September 2013 and March 2014 of whom 300 received weekly SMS and monthly surveys and 300 received only monthly surveys. On average, women were 21 years of age (interquartile range, 20-22), 71.50% had ever had sex and 72.62% had never tested for HIV. A total of 356 women reported testing for HIV within the 6 months of follow-up: 67% from the intervention arm and 51% from the control arm (hazard ratio, 1.57; 95% confidence interval, 1.28-1.92).
CONCLUSIONS: Use of weekly text messages about HIV prevention and reproductive health significantly increased rates of HIV testing among young Kenyan women and would be feasible to implement widely among school populations.

RCT Entities:   Population Interventions Outcomes

An estimated 35 million people globally were living with human immunodeficiency virus (HIV) at the end of 2013, with the majority (70%) residing in Sub-Saharan Africa where women account for 58% of infections.[1,2] HIV testing is a fundamental component of HIV prevention programs and the entry point into HIV care. Yet in sub-Saharan Africa, only 15% of women 15–24 years old are aware of their HIV status, despite having the highest HIV incidence rates and being a priority population for HIV prevention programs.[2,3] In a recent population-based survey, the majority (53%) of HIV-infected Kenyans were not aware of their status and one third of these reported not testing because they did not perceive themselves to be at high risk.[4] Additionally, only half (54%) of young Kenyan women had comprehensive knowledge of HIV/acquired immune deficiency syndrome (AIDS),[5] including routes of transmission and prevention strategies. Using the Health Belief Model,[6] interventions to improve young women's knowledge of HIV, their perceived risk of acquisition of HIV, and finally prompt them to action may increase HIV testing uptake. Mobile phone interventions provide an ideal method to provide such personalized interventions.

Kenya currently has 33.6 million mobile phone subscribers, representing an 82.6% country penetration[7] and subscribers prefer to use text messages (also known as short message service or SMS) rather than calling.[8] Text messages have been used successfully in Kenya to increase antiretroviral (ART) adherence,[9] retain HIV-exposed infants into care,[10] improve childhood vaccination campaigns,[11] retain men in postoperative care after medical male circumcision[12] and provide contraceptive information to young people.[13] Despite successes in mobile-based interventions to improve issues in health, these interventions to date have been implemented on a small scale, and there remains great potential for widespread use and substantial public health impact in Kenya.[14]

Government-run middle level colleges in Kenya typically have a resident community health nurse to provide basic health care services to students, including provision of condoms if needed. However, HIV testing and treatment services are not routinely available and can only be accessed through external health facilities. Knowledge of HIV status is the first key step in linkage to care for those known to be HIV infected. Thus, strategies with expansive coverage are needed to increase HIV testing among young women in Kenya and provide opportunities to engage in discussions about their personal HIV risk, risk behaviors, and improve their autonomy to reduce their HIV risk. We sought to increase HIV testing by improving HIV awareness, enhancing HIV risk perception, and reducing high risk behavior among young women in a rural region of Kenya through automated text messages.

MATERIALS AND METHODS

Study Design

This was a quasi-experimental study using text messages to increase HIV testing. There were 2 technical colleges and 2 teachers' training colleges within the study cohort, and students within these colleges are assigned by selection from the central government. A coin toss was used to assign all participants at each college to receive weekly HIV-related text messages (intervention group) or no messages (control group). To minimize differences in student characteristics, 1 technical college and 1 teacher training college were assigned to each group. Because study participants were boarding at their college, there was high potential for participants to discuss the study with one another, which influenced the decision to randomize institutions, rather than individuals and avoid bias from cross-contamination.

Study Population

Participants were enrolled between September 2013 and March 2014 from four 2-year colleges located in Kiambu County in Central Kenya, a predominantly rural area. Eligible participants were women ages 18 to 24 years, were HIV uninfected or unaware of their HIV status, had not tested for HIV in the preceding 12 months, owned a mobile phone and had regular access to electricity to charge their mobile phone, and knew how to send and receive text messages.

Study Procedures

The college administrative bodies granted permission to conduct the study within the college compounds, suggested appropriate times to contact students for study recruitment, and designated areas to conduct study screening and enrollment procedures. Study staff approached potential participants, described the study, and offered enrollment during the same day. At screening and enrollment, demographic and eligibility data were collected via face-to-face standardized quantitative interviews in a private area. Staff verified that each participant could read and respond to SMS and provided training on SMS survey questions and procedures, including instructions to delete the messages after responding as a privacy measure. Participants selected a 4-digit numeric password and a preferred time to receive the survey and messages (for the intervention group) and completed a practice survey on their phone.

Participants at colleges assigned to the intervention arm received weekly messages on HIV and reproductive health-related topics that were developed after interactive discussions with area college students during formative community entry activities. Convenience sampling was used to select students who were invited to participate in informal discussions about concerns that young people, particularly women, may have concerning sexuality and HIV. A total of 4 sessions were held—2 with more than 100 participants (men and women) at 2 colleges and 2 sessions with female participants only with up to 30 participants at 2 colleges. Six categories of topics (with 63 total messages) were developed: pregnancy, contraceptives, sexually transmitted infections, condoms, anal and oral sex, and HIV risk (Table 1). Messages were not static and evolved throughout the study to avoid repetition and maintain participant interest. All messages ended with the statement “Get tested for HIV.” Recipients had the option to send a return SMS requesting additional messages on the same topic or access a menu enabling exploration of the other topics. Each woman could request up to 3 additional messages per week.

TABLE 1

Examples of Weekly HIV Sensitization Messages

All participants received monthly SMS surveys to collect data on HIV testing and sexual behavior. SMS surveys were automated to begin the day after enrollment and continue once a month for 5 consecutive months. The survey comprised of nine questions, including password verification, HIV testing, number of new and recurrent sex partners, sexual frequency and condom use, pregnancy and pregnancy intent, and perceived risk of HIV during the past one month. Survey questions were resent if responses were not one of the designated codes. Participants were not charged for their survey responses and on completion of the survey, they were compensated with 50 KSH (∼US $0.50).

SMS Survey Platform

An automated SMS system was developed for delivery, receipt, and recording of SMS surveys and weekly messages (mSurvey Inc., Nairobi, Kenya), and the database was hosted on cloud-based technology. SMS messages were sent out directly through the mobile provider's network. Study staff accessed the SMS survey operational metrics online in real time to view participants' responses, track survey completion, and weekly message responses.

Statistical Analysis

The study sample size of 600 was designed to have 80% power to detect a 20% increase in reported HIV testing between the participants in the 2 study groups. Descriptive statistics were used to summarize participant characteristics and survey responses. χ2 statistics were used to describe differences in sexual behavior between study arms. Our primary analytic method was an intent-to-treat analysis using Cox proportional hazards regression to compare the time elapsed from enrollment to the first reported HIV test between participants in the 2 study arms. Cumulative probability curves with a log rank test were used to describe the proportion of women testing by arm. Because of our quasi-experimental design that did not individually randomize women, we also conducted a multivariate analysis with adjustment for covariates determined a priori—age, number of sex partners, and condom use with all sex acts—due to their known associations with HIV testing behaviors.[15-17] Additional demographic, behavioral, and medical characteristics were considered as potential confounders and included in final models if they substantially changed the hazard ratio (by ≥10%). In separate statistical models, generalized estimating equations were used to determine associations between HIV testing and sexual behaviour on a per visit basis and whether women who were consistently sexually active reported more instances of testing. Less than 10% of the data were missing from each data variable, and thus we excluded missing data from the analysis, rather than using imputation methods or other methods for missing data. Data were analyzed using STATA version 13.1 (College Station, TX).

Regulatory

The Kenyatta National Hospital Ethics Review Committee approved the study protocol, and all participants provided written informed consent. This study is registered with clinicaltrials.gov (NCT02527135).

RESULTS

Participant Characteristics

A total of 1039 women were screened for the study: 600 were enrolled into the study, 421 were ineligible due to having tested for HIV within the previous 12 months, and 10 were eligible but did not enroll (Fig. 1). Half (300) were from the colleges allocated to control arm, and 300 were from colleges allocated to intervention arm. The median age of participants was 21 years (interquartile range, 20–22; Table 2) with similar age distribution between the 2 groups. Most women were unmarried (93%), had engaged in vaginal and/or anal sex at least once in their lifetime (64% in control arm and 79% in intervention arm), had never tested for HIV (72.62%), and approximately half were in their first year of college (55.59%). The majority of the participants had been sexually active in the past month (62.69% in intervention arm and 57.71% in control arm), and nearly half of these women reported having new sexual partners (40% in intervention arm and 45.35% in control arm) and using condoms for all sexual encounters in the previous month (Table 2).

FIGURE 1

Consort diagram of study design.

TABLE 2

Baseline Characteristics of Study Participants Collected by Interview and SMS Survey

Monthly Survey and Weekly Message Response Rates

There was no substantial difference in survey response rates between the 2 arms. Of 3600 monthly surveys sent out over 6 months, 3272 received a response (90.89%), and 3136 (87.11%) were fully completed. Four hundred thirty-four (72.33%) participants completed all the survey questions sent to them, whereas 5 (<1%) participants did not respond to any questions throughout the study period. Of the 166 participants without complete survey data, 114 (68.67%) completed 3 or more surveys, whereas 17 (10.24%) responded to at least 1 question from 3 or more surveys (Fig. 2).

FIGURE 2

SMS survey completion rates by study month.

Of 7200 weekly messages sent out to intervention arm participants, 7078 (98.31%) were delivered to the participants' phones. Of these, 63.07% of participants reengaged once for more information, 59.06% reengaged twice, and 55.51% reengaged 3 times.

HIV Testing

A total of 356 women reported testing at least once for HIV in the 6 months of study follow-up: 201 (67%) in the intervention arm and 155 (51%) in the control arm (log rank, P < 0.0001), indicating a 57% increase in reported HIV testing by the intervention arm compared with the control arm (95% confidence interval [95% CI], 28–92%) (Table 3). Adjusting for age, condom use, and number of sex partners or ever having a sex partner did not substantially change the point estimate (adjusted harzard ratio (HR), 1.54; 95% CI; 1.25–1.90). The median time to first HIV test was 12 weeks for women in the intervention arm and 20 weeks for women in the control arm (Fig. 3).

TABLE 3

Comparison of HIV Testing Incidence by Study Arm and Other Demographic and Behavioral Factors

FIGURE 3

Cumulative probability curve of HIV testing by study arm.

Overall, there were 884 monthly SMS surveys in which participants reported HIV testing during the past month, 542 reports by intervention arm participants and 342 reports by control arm participants. Of the 356 participants who reported testing for HIV at least once during study follow-up, 106 (17.66%) women reported testing only once during study follow-up, 102 (17%) tested twice, and 148 (24.67%) tested 3 or more times.

Sexual Behaviour and HIV Testing

Among 3228 answered survey questions about sexual behavior, sexual partners were reported 2043 times (63%). Women reporting a current sex partner were 1.42 times as likely to report testing for HIV during the previous month (95% CI, 1.15–1.76) compared with those without a partner. However, there was no association between HIV testing at each visit and having a new sex partner, condom use, or HIV risk perception during the previous month (Table 4).

TABLE 4

Association of HIV Testing With Longitudinal Sexual Behavior and HIV Risk Perception

DISCUSSION

In this quasi-experimental study, women receiving weekly text messages tested for HIV in significantly higher numbers than women who did not receive weekly messages. Approximately half of the participants receiving intervention messages tested within 12 weeks of the intervention, a rate that is almost twice as fast as those participants not receiving intervention messages.

Human immunodeficiency virus testing and counseling is the gateway to HIV prevention and care yet there remains a wide HIV testing gap, especially for young women. In this study, 71.50% of participants had ever had sex in their lifetime, a rate that was higher than the country average of 66.10%.[4] Of all sexually experienced women in this study, only 30.61% had ever tested for HIV at study enrollment. Human immunodeficiency virus testing remains critical for the identification of new infections, linkage to HIV care and as a widely accepted prevention intervention. Human immunodeficiency virus testing and counseling is a free service in Kenya and is widely available and yet many young women engage in condomless sex without testing and with limited perception of risk of HIV, pregnancy and other consequences. Kenyan women predominantly report having their first HIV test at antenatal care clinics and more than half of all HIV infected women learning their status during pregnancy.[4] Interventions targeting young women at the onset of sexual activity have reduced numbers of sex partnerships, improved condom use, and increased HIV testing.[18] In our study, women who reported a sex partner were more likely to test for HIV, highlighting women's awareness of HIV risk and the importance of increasing young women's access to HIV prevention and testing interventions early in their sexual and reproductive lives when they are often most vulnerable.

Given the high response rates elicited within our cohort, our study demonstrated a high degree of feasibility for programs to send SMS messages and conduct health behavior surveys via SMS particularly among young college women. The Kenyan school curriculum includes HIV/AIDS education with teachers being the primary means to disseminate information for the students. A recent review of the implementation of the HIV/AIDS education policy revealed various shortcomings, including an absence of a plan to implement the behavior change component to prevent HIV acquisition among young people.[19] This SMS study used a suite of tailored messages to provide targeted information to college women and encourage them to test for HIV. A personalized SMS school-based education intervention program is a feasible option to increase awareness of HIV risk, improve HIV testing, and reduce risky behavior among young women before sexual debut or early in their sexual lives. In this study, one quarter of women reported testing for HIV 3 or more times, more frequent than the recommended national guidelines. Thus, routine SMS messages may also have a role to play in increasing repeat HIV testing with future studies potentially requiring follow-up for less than 6 months for optimal testing frequency among young women.

In our data set, we were unable to account for underlying differences in the student body at each college or the frequency of additional testing interventions at the colleges. However, adjusting for baseline characteristics in analyses did not alter the results. Data were not collected about other ongoing testing interventions, such as mobile testing campaigns, within the colleges where the study was conducted but all 4 colleges were government-affiliated institutions with similar curricula and practices. In these settings, the content and timing of government or school-sponsored HIV testing campaigns is standardized and would be expected to influence testing rates among students similarly across colleges. There were several instances of server outages and variations in mobile network availability causing delays in message delivery to recipients. These were transient and relatively few, however, with negligible impact on the study. In addition, there may be some inaccuracy in participant's reports about their testing and sexual behavior although the use of SMS provides an anonymous means of self-expression and dialogue and is expected to minimize social desirability bias.[20] Our study intervention was a multipronged approach involving text messages to improve HIV awareness, improve risk perception, and prompt participants to test for HIV. As such, it is difficult to determine the exact mechanism by which the intervention was successful in achieving the primary study aim of improved HIV testing. Future text message studies of HIV testing should consider exploring the exact mechanism of action involved in attaining study objectives. Finally, our study investigated the effect of HIV awareness text messages on HIV testing only within the 6 months of study follow-up. We did not evaluate poststudy effects of the text messages on HIV testing when participants were no longer receiving messages. These implications should be considered by future SMS intervention studies.

Strategies targeting young women for HIV prevention and reproductive and sexual health messaging, especially those that reach women before their first pregnancy, are imperative. SMS are a low-cost health intervention that can be leveraged to increase the frequency of HIV testing among a geographically diverse population. The SMS strategy that delivers messages to young women attending college without disrupting their normal routines has the potential to spark conversation between women within their social networks and could potentially have an effect beyond the individual.[21] Kenya has a high rate of mobile phone use, and SMS is near ubiquitous among young people. The use of SMS messaging strategies in Kenya to increase HIV testing among college students is scalable, and the current network capabilities offer a large opportunity to reach intended targets.