Abstract
Among 145 individuals surveyed in Tanzania, 42% indicated willingness to test for HIV in response to a confidential, phone-based text message. On average, participants were likely to provide contact information for 1.5 members of their social networks, suggesting high feasibility and moderate acceptability of a novel mHealth HIV testing intervention.
Keywords: HIV testing, social networks, confidentiality, mHealth, sub-Saharan Africa
Summary Statement
Preliminary data suggest that confidential SMS-based HIV testing invitations to social and sexual network contacts are feasible, moderately acceptable, and potentially cost-effective among two high risk populations in Tanzania.
1. Background/Introduction
In order to end the HIV epidemic, persons with HIV (PWH) must become aware of their diagnosis, engage in care, and adhere to effective treatment. Consequently, UNAIDS set the ambitious “95-95-95” targets of diagnosing 95% of all PWH, initiating antiretroviral therapy for 95% those diagnosed, and attaining virologic suppression for 95% of those treated.(1, 2) Although significant progress has been made towards achieving these targets, advances remain uneven across sub-Saharan Africa. Nearly 17% of PWH in Tanzania are unaware of their serostatus,(3) and many hesitate to test because of fear, stigma, and discrimination.(4)
One promising approach for increasing HIV testing rates is using confidential mobile phone-based testing invitations to social and sexual network contacts of persons testing for HIV. Social and sexual networks play a critical role in HIV transmission and testing uptake,(5-7) and mobile phones provide opportunities for efficiently extending testing invitations to networks of HIV testers. Carefully designed Short Message Service (SMS) invitations could efficiently and discretely reach populations that are at the margins of the healthcare system. This approach aligns with WHO recommendations to promote equitable services and optimize impact for populations most likely to have higher numbers of sexual partners.(8)
Text messaging has been used to communicate health information across diverse settings, including promotion of sexually transmitted infection (STI) testing.(9) Leveraging the ubiquity of mobile phones, coupled with the low cost of text-based messaging, SMS interventions have been used to promote HIV testing among gay men in Australia(10), random mobile phone users in South Africa,(11) and young college women(12) and female sex workers(13) in Kenya.
To our knowledge, SMS interventions to promote HIV testing have not, however, probed the social networks of individuals at risk for HIV infection. Given the correlations of HIV risk behaviors within social networks,(14) which result from an individual’s tendency to choose partners similar to themselves (a property called homophily(15)), HIV testing strategies that confidentially penetrate social networks of HIV testers create new opportunities for identifying at-risk individuals and linking them to care. Concurrently, the very low costs of SMS suggest that even a marginally effective intervention could be highly cost-effective. We explored the potential feasibility and acceptability of SMS-based confidential social network referrals for HIV testing (“CONSORT”) to reach the social network contacts of male mountain porters and female bar workers, two groups at high risk for HIV infection in the Kilimanjaro Region of Tanzania.(16)
2. Methods
As part of a study evaluating novel approaches to improve the uptake of HIV testing among high-risk populations in the Kilimanjaro Region of Tanzania (ClinicalTrials.gov Protocol NCT02714140),(17) we systematically recruited and surveyed members of two high risk populations to assess the feasibility and acceptability of CONSORT. Participants in the parent study comprised 773 women employed in randomly selected bars, restaurants, and guesthouses serving alcohol to patrons (“bars” and “female barworkers”, respectively) and 707 male porters supporting climbers of Mount Kilimanjaro (“male porters”) who were sequentially approached as they exited Mount Kilimanjaro National Park. Eligible participants were ages 18 to 49 and invited to come to the study offices for consent and enrollment. Acceptance rates in the parent study were 66.4% for female barworkers and 54.7% for male porters.
In the context of continuous data collection for the parent study, all study participants presenting at the study offices for in-person follow-up surveys between 5 February and 31 March 2020 were asked to complete a survey that assessed their demographic, social network, and HIV risk characteristics. We used a name generator/interpreter survey module to assess the size and composition of participants’ egocentric networks. A ‘name generator’ question asked each participant (“index”) to name family members, friends, loved ones, co-workers, neighbors, or other persons (their “alters”) from whom they have sought advice or help, or with whom they have discussed important matters. ‘Name interpreter’ questions, cycled through up to 5 named alters for each participant and assessed alters’ key characteristics (e.g., gender, relationship, communication frequency). CONSORT feasibility was probed by recording index participants’ knowledge of phone numbers for each alter. Acceptability was described by index participants’ likelihood of sending CONSORT messages to each alter, and their own likelihood of testing for HIV if they received a confidential testing invitation. Likelihood was assessed using a 4-point Likert scale (very likely; somewhat likely; somewhat unlikely; not likely). Preferences for CONSORT versus physical invitation cards were also assessed.
Descriptive statistics were used to analyze participants’ social networks and the potential feasibility and acceptability of CONSORT. Student’s t-tests and chi-squared tests assessed the statistical significance of differences between male and female participants. To account for the small sample size, confidence intervals around the estimated proportions were calculated using the Wilson method.(18) To characterize the potential impact and cost-effectiveness of CONSORT, sample estimates of acceptability were applied to a hypothetical cohort of 1000 index participants whose characteristics were assumed to resemble those of the participants in this study.
3. Results
Sample characteristics
Participants comprised 113 female barworkers and 32 male porters. Most female bar workers were unmarried, whereas the majority of male porters were married (Table 1, Panel A). The numbers of lifetime and recent sexual partners were significantly higher for men than women. Nearly 4% of female barworkers reported being infected with HIV. Nearly one third of women and half of the men reported testing for HIV on two or fewer occasions in their lifetime.
Table 1.
Characteristics of study participants and their social networks
| Panel A. Index respondents1 | All participants (n=145) |
Female barworkers (n=113) |
Male porters (n=32) |
p-value2 | |
|---|---|---|---|---|---|
| Age | in years | 31 (8.0) | 31 (7.9) | 34 (8.3) | 0.102 |
| Education | Primary school or less | 62 (42.8) | 45 (39.8) | 17 (53.1) | 0.225 |
| Any secondary school | 83 (57.2) | 68 (60.2) | 15 (46.9) | ||
| Marital status | Not married | 79 (54.5) | 68 (60.2) | 11 (34.4) | 0.015 |
| Married | 66 (45.5) | 45 (39.8) | 21 (65.6) | ||
| # of sexual partners (lifetime) | 0-1 | 20 (13.8) | 20 (17.7) | 0 (0.0) | <0.001 |
| 2 | 28 (19.3) | 24 (21.2) | 4 (12.5) | ||
| 3-4 | 45 (31.0) | 40 (35.4) | 5 (15.6) | ||
| 5+ | 52 (35.9) | 29 (25.7) | 23 (71.9) | ||
| # of sexual partners (past 12 months) | 0-1 | 112 (77.2) | 93 (82.3) | 19 (59.4) | 0.003 |
| 2 | 19 (13.1) | 14 (12.4) | 5 (15.6) | ||
| 3+ | 14 (9.7) | 6 (5.3) | 8 (25.0) | ||
| Self-reported HIV serostatus | Negative / not tested | 141 (97.2) | 109 (96.5) | 32 (100.0) | 0.576 |
| Living with HIV | 4 (2.8) | 4 (3.5) | 0 (0.0) | ||
| # of HIV tests (lifetime) | None | 10 (7.1) | 6 (5.5) | 4 (12.5) | 0.084 |
| 1-2 | 40 (28.4) | 28 (25.7) | 12 (37.5) | ||
| 3-4 | 47 (33.3) | 36 (33.0) | 11 (34.4) | ||
| 5+ | 44 (31.2) | 39 (35.8) | 5 (15.6) | ||
| Number of alters3 named by the index | 0 | 2 (1.4) | 1 (0.9) | 1 (3.1) | 0.064 |
| 1 | 20 (13.8) | 17 (15.0) | 3 (9.4) | ||
| 2 | 62 (42.8) | 48 (42.5) | 14 (43.8) | ||
| 3 | 39 (26.9) | 33 (29.2) | 6 (18.8) | ||
| 4 | 9 (6.2) | 6 (5.3) | 3 (9.4) | ||
| 5 | 8 (5.5) | 3 (2.7) | 5 (15.6) | ||
| 6 or more | 5 (3.4) | 5 (4.4) | 0 (0.0) | ||
| Panel B. Alters1,3 | Alters of all participants (n=362) |
Alters of female barworkers (n=267) |
Alters of male porters (n=86) |
p-value2 | |
| Gender | Female | 200 (55.2) | 178 (64.5) | 22 (25.6) | <0.001 |
| Male | 162 (44.8) | 98 (35.5) | 64 (74.4) | ||
| Relationship to the index | Neighbor | 15 (4.1) | 11 (4.0) | 4 (4.7) | 0.123 |
| Friend | 118 (32.6) | 82 (29.7) | 36 (41.9) | ||
| Co-worker | 25 (6.9) | 18 (6.5) | 7 (8.1) | ||
| Family member | 145 (40.1) | 119 (43.1) | 26 (30.2) | ||
| Partner | 55 (15.2) | 44 (15.9) | 11 (12.8) | ||
| Other | 4 (1.1) | 2 (0.7) | 2 (2.3) | ||
| Contact frequency with the index | Weekly or less | 94 (26.0) | 73 (26.4) | 21 (24.4) | 0.779 |
| Daily or almost daily | 268 (74.0) | 203 (73.6) | 65 (75.6) | ||
| Index knows a phone number of the alter | No | 15 (4.1) | 11 (4.0) | 4 (4.7) | |
| Yes | 347 (95.9) | 265 (96.0) | 82 (95.3) | ||
| Index ever talked with the alter about HIV | No | 141 (39.0) | 107 (38.8) | 34 (39.5) | 0.900 |
| Yes | 221 (61.0) | 169 (61.2) | 52 (60.5) |
For age, numbers represent means (with standard deviations in parentheses); all other variables are categorical in nature and numbers represent counts (with percentages in parentheses).
p-values describe the statistical significance of differences between female barworkers and male porters; significance was evaluated using a t-test (for age) and Fisher’s exact tests for all other variables.
Alters were defined as family members, friends, loved ones, co-workers, neighbors, or other persons (“alters”) from whom index participants have sought advice or help, or with whom they have discussed important matters.
Social network characteristics
Participants reported an average of 2.8 individuals with whom they discussed important matters in the past year (Table 1, Panel B). Additional information was collected for up to 5 of these “alters” (N=362). Male index clients were more likely to list male alters (74%); female index clients were more likely to list female alters (64%; p<0.001). Over 40% were family members, 33% friends, 15% partners, 7% colleagues. Contact frequency with alters was high (74% daily or almost daily). Index clients reported having spoken to 61% of alters about HIV.
Feasibility and acceptability of CONSORT
Index participants reported knowing a phone number for 96% of alters (Table 2). On average, each participant was likely to invite 2.1 contacts (84% of alters) for HIV testing using physical invitation cards, and 1.5 contacts (60% of alters) using CONSORT. When given the choice between the two, two-thirds preferred invitation cards, whereas 25% preferred CONSORT SMS. When asked "If you received an SMS suggesting that you test for HIV but you do not know who suggested that you test, how likely is it that you would get tested for HIV?", 42% of index clients indicated that they would be somewhat or very likely to test for HIV.
Table 2.
Feasibility and potential acceptability of CONSORT
| Feasibility and acceptability among alters1 | All participants (n=362) | Female barworkers (n=267) | Male porters (n=86) | p-value2 | ||||
|---|---|---|---|---|---|---|---|---|
| n (%) | 95% CI | n (%) | 95% CI | n (%) | 95% CI | |||
| Index knows phone number of the alter | No | 15 (4.1) | [2.5;6.7] | 11 (4.0) | [2.2;7.0] | 4 (4.7) | [1.8;11.4] | 0.761 |
| Yes | 347 (95.9) | [93.3;97.5] | 265 (96.0) | [93.0;97.8] | 82 (95.3) | [88.6;98.2] | ||
| Index would personally give the alter an HIV testing invitation card | Very likely | 235 (64.9) | [59.9;69.7] | 170 (61.6) | [55.7;67.1] | 65 (75.6) | [65.5;83.4] | 0.001 |
| Somewhat likely | 68 (18.8) | [15.1;23.1] | 59 (21.4) | [17.0;26.6] | 9 (10.5) | [5.6;18.7] | ||
| Somewhat unlikely | 38 (10.5) | [7.7;14.1] | 35 (12.7) | [9.3;17.1] | 3 (3.5) | [1.2;9.8] | ||
| Not likely | 21 (5.8) | [3.8;8.7] | 12 (4.3) | [2.5;7.4] | 9 (10.5) | [5.6;18.7] | ||
| Index would send a confidential SMS message with an HIV testing invitation to the alter | Very likely | 144 (41.4) | [36.3;46.6] | 106 (40.0) | [34.3;46.0] | 38 (45.8) | [35.5;56.4] | 0.332 |
| Somewhat likely | 66 (19.0) | [15.2;23.4] | 55 (20.8) | [16.3;26.0] | 11 (13.3) | [7.6;22.2] | ||
| Somewhat unlikely | 71 (20.4) | [16.5;24.9] | 56 (21.1) | [16.6;26.4] | 15 (18.1) | [11.3;27.7] | ||
| Not likely | 67 (19.3) | [15.5;23.7] | 48 (18.1) | [13.9;23.2] | 19 (22.9) | [15.2;33.0] | ||
| Index's preference for personally giving an invitation card vs. sending a confidential SMS HIV testing invitation to alter's mobile phone | Invitation card | 235 (67.7) | [62.6;72.4] | 171 (64.8) | [58.8;70.3] | 64 (77.1) | [67.0;84.8] | 0.131 |
| Confidential SMS | 87 (25.1) | [20.8;29.9] | 73 (27.7) | [22.6;33.3] | 14 (16.9) | [10.3;26.3] | ||
| Neither | 15 (4.3) | [2.6;7.0] | 13 (4.9) | [2.9;8.2] | 2 (2.4) | [0.7;8.4] | ||
| No preference | 10 (2.9) | [1.6;5.2] | 7 (2.7) | [1.3;5.4] | 3 (3.6) | [1.2;10.1] | ||
| Acceptability among index particpants1 | All participants (n=145) | Female barworkers (n=113) | Male porters (n=32) | p-value2 | ||||
| n (%) | 95% CI | n (%) | 95% CI | n (%) | 95% CI | |||
| Index client: Would test after receiving a CONSORT message | Very likely | 31 (21.53) | [15.6;28.9] | 24 (21.24) | [14.7;29.7] | 7 (22.58) | [11.4;39.8] | 0.991 |
| Somewhat likely | 30 (20.83) | [15.0;28.2] | 23 (20.35) | [14.0;28.7] | 7 (22.58) | [11.4;39.8] | ||
| Somewhat unlikely | 25 (17.36) | [12.0;24.4] | 20 (17.70) | [11.8;25.8] | 5 (16.13) | [7.1;32.6] | ||
| Not likely | 58 (40.28) | [32.6;48.4] | 46 (40.71) | [32.1;49.9] | 12 (38.71) | [23.7;56.2] | ||
Abbreviations: CONSORT–confidential social network referral for HIV testing; SMS–short message system message; CI–confidence interval
Numbers represent counts (with percentages in parentheses).
p-values describe the statistical significance of differences between female barworkers and male porters; significance was evaluated using Fisher’s exact tests. Confidence intervals around the estimated proportions were calculated using a method by Wilson (1927).
Potential impact and cost-effectiveness of CONSORT
Our results allow for preliminary estimates of the potential impact and cost-effectiveness of CONSORT. Assuming 1000 persons with characteristics similar to our study participants were offered the opportunity to facilitate confidential SMS-based HIV testing invitations, 1,507 confidential SMS-based testing invitations could be sent to contacts within their social networks. Among SMS recipients, 638 would be very likely or somewhat likely to test for HIV. At a cost of Tsh1000 (~$0.44) for 1000 SMS, the SMS cost per CONSORT-attributable HIV test would be approximately 0.1 cent. Given plausible assumptions (93% previously tested for HIV; 0.1% HIV incidence among prior testers; 3% HIV prevalence among first-time testers), these 638 CONSORT-attributable HIV tests would be expected to result in 1.93 new HIV diagnoses; the SMS cost per new HIV diagnosis would be approximately 34 cents.
4. Discussion
Among 145 individuals surveyed, we observed high feasibility and moderate acceptability of confidential SMS messaging to reach social network referrals for HIV testing. On average, each participant was very likely or somewhat likely to invite 2.1 network contacts using physical invitation cards, and 1.5 contacts using CONSORT SMS. More than 40 percent of participants indicated a willingness to test for HIV if they received a CONSORT invitation. Extrapolations of these estimates to a hypothetical cohort of 1000 index clients suggest a yield of more than 600 CONSORT-attributable HIV tests and 2 newly diagnosed HIV infections at an SMS cost of less than 1 US Dollar.
While other studies have explored the use of SMS messaging to promote HIV testing in sub-Saharan Africa,(11, 12, 19, 20) to our knowledge the feasibility and acceptability of using confidential SMS messaging to recruit social network referrals have not been explored. Prior work has shown that SMS messaging increased HIV testing among female sex workers(13) and college students(12) in Kenya and among a random sample of mobile phone users in South Africa.(11) Although each of these studies had limitations, taken together, they suggest that SMS messaging interventions can increase the uptake of HIV testing.
In addition to its low cost, we note other potential advantages of the CONSORT approach. First, chain referral methods, such as respondent-driven sampling, are a proven means of accessing hard-to-reach populations.(21) Reaching “populations that are left behind” is a priority of the global health sector strategy on STIs.(8) Additionally, the privacy and confidentiality afforded by mHealth technologies may overcome the stigma associated with personal invitations to test for HIV. Finally, because infrastructure and SMS messaging costs are low, CONSORT could be readily scaled within resource-limited settings and are extensible beyond HIV-messaging.
We note limitations. First, the sample may not be representative of the population of barworkers and porters, as enrollment and retention in the parent study was subject to self-selection. Second, our estimates may be subject to hypothetical bias and social desirability bias and may not be representative of other populations at increased risk of HIV infection. As such, these results are considered preliminary, but sufficient to prompt further feasibility and acceptability assessments of CONSORT.
In summary, we describe interest in CONSORT among mountain porters and bar workers in the Kilimanjaro Region of Tanzania. Given its low cost, capacity to reach hidden populations, and potential for widespread scalability, further study is warranted. Specifically, additional research should identify desirable healthcare provider- and client-side characteristics of CONSORT, including acceptable and effective content of SMS messages, evaluate feasibility and acceptability among clients presenting for testing, and provide preliminary estimates of its efficacy.
Supplementary Material
ACKNOWLEDGMENTS
The authors are grateful to the study participants and to the study research assistants for input on study procedures and study implementation.
The authors thank the staff of the Kilimanjaro Clinical Research Institute, the University of South Carolina’s Arnold School of Public Health, especially the Department of Health Services Policy & Management and the Center for Health Care Quality, the Duke Global Health Institute and Duke University’s Center for Health Policy and Inequalities Research, for administrative support, and the Kilimanjaro Regional Administration and the Moshi District Council Administration for their support of the study’s development and implementation.
FUNDING
Research reported in this publication was supported by the National Institute of Mental Health of the National Institutes of Health in the United States under Award Numbers R01MH106388 and R21MH96631 and by the Duke University Center for AIDS Research (CFAR), an NIH funded program (P30AI064518). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Footnotes
DECLARATIONS
The study protocol was approved by the Institutional Review Boards at Duke University and the University of South Carolina in the United States, the Ethics Review Committee at Kilimanjaro Christian Medical University College, and the National Institute for Medical Research.
COMPETING INTEREST
No potential competing interest was reported by the authors.
Contributor Information
Jan Ostermann, Department of Health Services Policy & Management, University of South Carolina, USA South Carolina Smart State Center for Healthcare Quality, University of South Carolina, USA Duke Global Health Institute, Duke University, USA.
Bernard Njau, Kilimanjaro Christian Medical Centre, Moshi, Tanzania.
Martha Masaki, Kilimanjaro Christian Medical Centre, Moshi, Tanzania
Tara Mtuy, Kilimanjaro Christian Medical Centre, Moshi, Tanzania Department of Global Health and Development, London School of Hygiene and Tropical Medicine, London, UK.
Dafrosa Itemba, Tanzania Women's Research Foundation, Moshi, Tanzania
Amy Hobbie, Center for Health Policy & Inequalities Research, Duke University, Durham, North Carolina, USA.
Valerie Yelverton, University of South Carolina
Spencer Moore, Wageningen University, The Netherlands.
Thespina Yamanis, School of International Service, American University, Washington DC, USA.
Nathan M Thielman, Duke University, UNITED STATES.
References
- 1.UNAIDS 2021;Pages. Accessed at https://www.unaids.org/en/topics/2025_target_setting. Accessed June 22, 2021 2021.
- 2.United Nations General Assembly. Political Declaration on HIV and AIDS: Ending Inequalities and Getting on Track to End AIDS by 2030. 2021. [Google Scholar]
- 3.UNAIDS. UNAIDS Data 2020. Geneva: UNAIDS; 2020. [Google Scholar]
- 4.Jin H, Restar A, Beyrer C. Overview of the epidemiological conditions of HIV among key populations in Africa. J Int AIDS Soc. 2021;24 Suppl 3:e25716. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Klovdahl AS, Potterat JJ, Woodhouse DE, et al. Social networks and infectious disease: the Colorado Springs Study. Soc Sci Med. 1994;38(1):79–88. [DOI] [PubMed] [Google Scholar]
- 6.Huang ZJ, He N, Nehl EJ, et al. Social network and other correlates of HIV testing: findings from male sex workers and other MSM in Shanghai, China. AIDS Behav. 2012;16(4):858–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Halkitis PN, Kupprat SA, McCree DH, et al. Evaluation of the relative effectiveness of three HIV testing strategies targeting African American men who have sex with men (MSM) in New York City. Ann Behav Med. 2011;42(3):361–9. [DOI] [PubMed] [Google Scholar]
- 8.World Health Organization. Global health sector strategy on Sexually Transmitted Infections, 2016-2021. 2016. [Google Scholar]
- 9.Unutzer A, Dombrowski JC, Katz DA, et al. Evaluation of Text Message Reminders to Encourage Retesting for Chlamydia and Gonorrhea Among Female Patients at the Municipal Sexually Transmitted Disease Clinic in Seattle, Washington. Sex Transm Dis. 2020;47(7):458–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Bourne C, Knight V, Guy R, et al. Short message service reminder intervention doubles sexually transmitted infection/HIV re-testing rates among men who have sex with men. Sex Transm Infect. 2011;87(3):229–31. [DOI] [PubMed] [Google Scholar]
- 11.de Tolly K, Skinner D, Nembaware V, et al. Investigation into the use of short message services to expand uptake of human immunodeficiency virus testing, and whether content and dosage have impact. Telemed J E Health. 2012;18(1):18–23. [DOI] [PubMed] [Google Scholar]
- 12.Njuguna N, Ngure K, Mugo N, et al. 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. Sex Transm Dis. 2016;43(6):353–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Kelvin EA, George G, Mwai E, et al. A Randomized Controlled Trial to Increase HIV Testing Demand Among Female Sex Workers in Kenya Through Announcing the Availability of HIV Self-testing Via Text Message. AIDS Behav. 2019;23(1):116–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Mulawa M, Yamanis TJ, Hill LM, et al. Evidence of social network influence on multiple HIV risk behaviors and normative beliefs among young Tanzanian men. Soc Sci Med. 2016;153:35–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Kenyon C, Colebunders R. Birds of a feather: homophily and sexual network structure in sub-Saharan Africa. Int J STD AIDS. 2013;24(3):211–5. [DOI] [PubMed] [Google Scholar]
- 16.Ostermann J, Njau B, Mtuy T, et al. One size does not fit all: HIV testing preferences differ among high-risk groups in Northern Tanzania. AIDS Care. 2015;27(5):595–603. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Ostermann J, Njau B, Hobbie A, et al. Using discrete choice experiments to design interventions for heterogeneous preferences: protocol for a pragmatic randomised controlled trial of a preference-informed, heterogeneity-focused, HIV testing offer for high-risk populations. BMJ Open. 2020;10(11):e039313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wilson EB. Probable inference, the law of succession, and statistical inference. J Am Stat Ass. 1927;22:209–12. [Google Scholar]
- 19.Govender K, Beckett S, Masebo W, et al. Effects of a Short Message Service (SMS) Intervention on Reduction of HIV Risk Behaviours and Improving HIV Testing Rates Among Populations located near Roadside Wellness Clinics: A Cluster Randomised Controlled Trial in South Africa, Zimbabwe and Mozambique. AIDS Behav. 2019. [DOI] [PubMed] [Google Scholar]
- 20.Mugo PM, Wahome EW, Gichuru EN, et al. Effect of Text Message, Phone Call, and In-Person Appointment Reminders on Uptake of Repeat HIV Testing among Outpatients Screened for Acute HIV Infection in Kenya: A Randomized Controlled Trial. PLoS One. 2016;11(4):e0153612. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Heckathorn D Respondent-driven sampling: a new approach to the study of hidden populations. Soc Prob. 1997;44(2):174–99. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.