Awareness, use and willingness to self-test for HIV: An analysis of cross-sectional population-based surveys in Malawi and Zimbabwe

Background Many southern African countries are nearing the global goal to diagnose 90% of people with HIV by 2020. In 2016, 84% and 86% of people with HIV knew their status in Malawi and Zimbabwe respectively. Despite this progress, gaps remain, particularly among men (≥25 years). We investigated awareness, use and willingness to HIV self-test (HIVST) prior to large scale implementation and explored sociodemographic associations. Methods We pooled responses from two of the first cross-sectional Demographic Health Surveys to include HIVST questions: Malawi and Zimbabwe in 2015-16. Sociodemographic factors and sexual risk behaviours associated with never testing for HIV, and awareness, past use and future willingness to self-test were investigated using univariable and multivariable logistic regression, adjusting for the sample design and limiting analysis to participants with completed questionnaire and a valid HIV result. Analysis of willingness to self-test was restricted to Zimbabwean men as Malawians and women were not asked this question.

Despite low awareness and minimal HIVST experience at that time, willingness to self-test was high, especially in older men with moderate to high sexual risk. These data provide a valuable baseline against which to investigate population-level uptake of HIVST as . As a result, reaching the remaining people with HIV who do not know their status is becoming costly and challenging; with national programmes reporting declining numbers of people with HIV diagnosed through HIV testing services [2,3]. Global and national priorities now include defining sustainable approaches that maintain these high rates of coverage, while reaching individuals and groups still in need of HIV testing, prevention and treatment.
Across southern Africa, men continue to be less well served by HIV programmes than women, being less likely to have ever-tested [4] and more likely to develop advanced HIV disease, reflecting late diagnosis and/or treatment initiation [5]. Men have fewer opportunities for HIV testing compared to women, as well as social-cultural, economic and systemic barriers which reduce access to and uptake of services [6,7].
HIV self-testing (HIVST) is recommended by the World Health Organization (WHO) [2] and is a key intervention for reaching populations who may not test otherwise, particularly among men [8]. Results from multiple evaluations show HIVST has high uptake, can increase the population-coverage of HIV testing and has high safety and acceptable, globally [9,10]. As of July 2019, this recommendation has been taken up globally, with nearly 7 million HIVST kits procured by major donors, and 77 countries reporting that they have an HIVST policy; 38 of which are fully implementing self-testing [11,12].
Both Malawi and Zimbabwe were early adopters of self-testing, with pilot studies starting in 2010-15 [13,14], followed by the development of national policies and initiation of large scale implementation in mid-2015 under the STAR (Self-Test AfRica) Initiative [15].
Since then, multiple evaluations of HIVST in each country have shown community-based and partner-delivered HIVST to be a feasible and effective way to reach first time testers, men, young people, as well as partners of people with HIV [10,[16][17][18][19]. Recent mathematical modelling suggests that, with appropriate targeting toward men in southern Africa among other priority groups, HIVST can also be cost-effective [20,21].
As both countries move toward broader self-testing scale-up, we used Demographic Health Survey (DHS) data from 2015-16 to analyse population-level awareness, use and willingness to self-test prior to large scale implementation [22,23]. These questions were first added as optional additions to the DHS questionnaire in 2015. As such, the objective of this study was to provide a point of comparison with future evaluations post national scale-up, as well as to inform future implementation, of HIVST. Here we assess early implementation of HIVST questions in population-based surveys, and assess associations with awareness, use and future willingness to self-test.

Methods
We obtained data from the 2015-16 Malawi and Zimbabwe DHS surveys [22,23], which provide data from a representative sample of men (15-54 years) and women of reproductive age (15-49 years) living in Malawi and Zimbabwe, with linked laboratory HIV results. We limited our analysis to participants who had completed intervieweradministered questionnaires, provided blood specimens for HIV testing, and had a valid result from this HIV test.
Our main outcomes of interest were self-reported: ever testing for HIV; awareness of and use of HIVST and willingness to self-test in the future. Willingness to self-test was only asked in Zimbabwe, and only included in the male questionnaire.
Independent variables used in the analysis included country (i.e. Malawi or Zimbabwe), sex (i.e. male or female), household wealth (i.e. measured by standard quintiles), age (i.e. measured five-year age bands from 15-45 years and 45 + years) education (i.e. measured by secondary education or lower), literacy (i.e. measured by the ability to read a full sentence), employment (i.e. measured by actively working), marital status (i.e. measured by married or co-habituating), and HIV status reported during the survey (i.e. HIV-positive or HIV-negative). A three-category HIV risk variable was defined from reported sexual activity in the past four weeks, and the following high-risk exposures in the previous 12 months: multiple (i.e. ≥2) partners; any paid sex (asked to men); having received gifts, cash or other compensation in exchange for sex (asked to women); and having a sexually transmitted infection (STI). Individuals with one or more of these risk variables were classified as "high-risk". The remaining respondents reporting no other risk exposures were classified as "moderate risk" if sexually active in the past 4-weeks and "low-risk" if reporting no sexually activity in the past four weeks.

Data analysis
Analysis used Stata version 11 (College Station, Texas). Standard country-specific sampling and cluster weights provided by DHS were set using the survey (svy) commands.
Participants with missing data for outcomes or independent variables were excluded from the analysis. The choice of independent covariates was informed by literature on factors influencing testing for HIV and adaptation of the simplified hierarchical framework for HIV testing (including self-testing) among men in sub-Saharan Africa (See Fig. 1). Independent variables were all categorical, and so were reported as proportions for baseline and other tables. We also pre-specified a stratified analysis by testing history to explore differences in awareness and use of, and willingness to, self-test for HIV.
Associations between independent and outcome variables were investigated using adjusted chi-squared tests for sampling weights and univariable odds ratios (OR) with 95% confidence intervals (CI). Before multivariable analysis, confounding and collinearity between independent variables was explored by investigating associations between variables for all with significant associations with any given outcomes. were and Potential effect modification [24] was explored using stratified analyses by sex, HIV status, previous HIV testing history and HIV risk category.
Variables for inclusion in multivariable models were selected by the putative causal framework (Fig. 1), and by investigating effect modification and collinearity. Univariable and multivariable analyses used logistic regression. P-values across age, wealth and HIV sexual risk were calculated using the Wald Test. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [25]. (Supplementary data S1 includes the checklist.) Men reporting more sexual risk behaviours also reported greater willingness to self-test than men reporting less risk (87.6% vs 78.8%; p<0.001). Most men willing to self-test had tested in the past 12 months (88.5%). However, a high proportion of men who had not tested for HIV in the previous two or more years were also willing to self-test (86.4%).

Baseline characteristics
(Supplementary table S1 includes for baseline characteristics of Zimbabwean men on willingness to self-test, 2015-16).

Ever tested for HIV
Factors associated with ever having tested for HIV were: being Malawian, female, older, urban residents, married or cohabitating, employed, having secondary or higher-level education, being literate, having an HIV-positive test in the survey and having either moderate or reporting more sexual risk behaviours (

Awareness and use of self-testing
Few respondents (287, 1.2%) had self-tested previously, although 12.6% were aware of self-testing. In univariable analyses, previous self-testing was associated with age (more likely to have self-tested if 20 years of age and older), urban residence, increasing affluence, better education, literate and with high HIV risk. On multivariable analysis, across wealth quintiles, being wealthier was associated with previous self-testing (p<0.001) with wealthiest men having the greatest odds of past self-testing (aOR for wealthiest vs. least wealthy=3.59; 95%CI:1.79-7.18, p<0.001). Older men, 30-34 years old, also had greater odds of past self-testing compared to younger men (age [15][16][17][18][19] (aOR= 2.89; 95%CI: 1.47-5.68, p<0.002). (A complete analysis of ever self-testing is shown in supplementary table S2.) Table 3 provides outcomes from univariable and multivariable analysis for awareness of HIV self-testing. On univariable analysis, the following variables were significantly associated with being aware of HIVST: being from Zimbabwe, male, older (≥20 years), urban, wealthy, married, actively employed, more educated, literate, moderate or highrisk, and having tested previously.
On multivariable analysis, the following variables were significantly associated with being aware of HIVST: being male (male vs female: aOR=1.55;95%CI:1-37-1. 75 from a rural setting were negatively associated with awareness of self-testing (see Table   3).

Willingness to self-test among Zimbabwean men
On crude analysis, older aged Zimbabwean men who were working, those with moderate to high-risk, and men who had previously tested men reported greater willingness to selftest (Table 4).
On univariable analysis, the following were significantly associated with overall willingness to self-test (unstratified): being older (aged 25 The relationship between willingness to test and socio-economic variables (wealth and working) and HIV status, substantially differed according to both high and low-sexual risk ( While employment in high-risk men was associated with less willingness to self-test (working vs not working: aOR: 0.57, 95%CI: 0.34-0.95, p=0.030), employed low-risk men were significantly more willing than unemployed low risk men to self-test (aOR 1.41; 95%CI: 1.13-1.77, p=0.003). The association with previous testing and willingness to test was also more pronounced factor for low risk men (previously tested vs never tested: aOR 1.48; 95%CI: 1.18-1.85, p<0.001) than high risk men (previously tested vs never tested: aOR 1.20; 95%CI: 0.76-1.90, p=0.435), while associations with age (p=0.106) and wealth (p=0.102) were less pronounced than for high risk men (described above).
We additionally conducted an a priori-defined stratified analysis to investigate whether willingness to self-test varied by past HIV-testing behaviour (i.e. previously tested or not) and HIV testing history (supplementary table S3). Patterns of willingness to self-test were similar for the 2,437/7,372 (33.1%) men who had never previously tested as for those with at least one past HIV test, with greater willingness in older men. Those not previously tested who tested HIV-positive in the survey were however more likely to be willing to self-test (aOR=1.83, 95%CI: 1.03-3.24, p=0.039), which was not significant among evertested men.

Discussion
The main findings from this analysis of 2015-16 survey data captured immediately before HIVST implementation in Malawi and Zimbabwe were that awareness of and life-time use of self-testing were low, with only 12.6% of respondents aware of self-testing and only 1.2% having ever self-tested for HIV. Willingness to self-test was high, although this question was asked only of male Zimbabweans, with 84.5% respondents reporting themselves willing, including 30.4% of all previously untested men. Self-testing appeared to appeal most strongly to midlife-older men, with the highest willingness in those aged 35-39, as well as those with high or moderate sexual HIV risk and in rural settings, where having never previously tested for HIV was more common than in urban settings. Factors independently associated with greater awareness of HIVST included men, urban residence, and literacy: with many of these same factors also associated with having tested for HIV at least once in this analysis of 2015-16 data. Poorer and unemployed individuals were less likely to be aware of self-testing.
The importance of high willingness to self-test among older Zimbabwean men, including those with higher risks, should not be underestimated. This challenges perceptions that men may not want to or are afraid test for HIV and underscores the importance of providing more opportunities and HIV testing options which are acceptable to men. As reported in a recent analysis among never tested men in sub-Saharan Africa, nearly all those offered HIV testing in the survey accepted and learned their results [4].
Despite significant gains and scale-up of HIV testing in both Malawi and Zimbabwe, men continue to be missed [1,2]. According to recent first-90 estimates in sub-Saharan Africa, men with HIV aged ≥ 25 years are much less likely to know their HIV-positive status than younger men and women [26]. As the median age of all people with HIV continues to increase [27], identifying and scaling-up strategies which appeal to older age groups, especially men, will be needed. Greater efforts are needed to roll-out evidence-based HIVST approaches for reaching men, such as secondary distribution from a female partners at antenatal care, or through networks of other high risk sexual, drug injecting or social contacts, including those with HIV [9,16,28]. Considering the high willingness to self-test in high-risk men in rural areas, additional community outreach strategies may still be needed. HIVST in workplaces and through faith-based organizations should also be considered, as early programmatic data suggests it may be particularly useful for reaching older men [29]. However, more focused programmatic efforts and communication strategies for workplace HIVST may be needed, as in contrast to low-risk men, high-risk men who were working were less willing to self-test. Further evaluation is needed to understand the utility of HIVST through formal and informal workplace programmes and how well they can reach high-risk men. It will be important to assess differences in HIVST awareness, use and willingness among older and higher risk men in future surveys.
Since these surveys, HIVST alongside conventional testing has been rapidly scaled-up, notably so for Malawi and Zimbabwe. Between May 2015 and July 2017, the STAR Initiative alone distributed 172,830 and 265,091 kits in Malawi and Zimbabwe respectively [10].
Following WHO guidelines and WHO-prequalification of four HIVST products, and multiple large scale implementation studies [2,30], volumes continue to increase annually, with latest estimates suggesting that between 2017 and 2020, with existing donor support, both countries will have procured at least 4 million self-tests [12].
High willingness to self-test in Malawi and Zimbabwe has also been underscored by observed high uptake in community-based HIVST interventions, with uptake by 45-75% reported by endline surveys between 2016 and 2019 in three population-level cluster randomized trials in rural communities [18,19,31]. In 2017, a survey following community-based HIVST distribution in rural Zimbabwe with or without supply-side financial incentives for post-test linkage showed that 81.7% of residents were aware of self-testing and 55.8% had self-tested [18]. Two trials of community distribution of HIVST in rural Malawi showed high uptake of HIVST, with significantly increases in ever-testing for HIV in men and adolescents [19,31]. Even in the standard of care arms, 31.5% of participants in the 2016-17 trial and 32.3% in the 2018-19 trial, respectively, were aware of HIVST [18,19,31].
These are substantial increases compared to the low awareness and use of self-testing in 2015-16 DHS surveys, and highlight the broader impact on awareness from large implementation science studies, such as STAR. In 2015-16, HIVST was limited to small pilot studies in each country, as national and international policies were still in development and there were no nationally registered or WHO-prequalified products available [32].
As HIVST continues to expand globally, monitoring overall HIVST awareness, use and willingness will contribute to better understanding of the reach and impact of HIVST.
Ideally, the extent to which social determinants such as urban residence, literacy and affluence dictate awareness of HIVST will diminish with more comprehensive distribution strategies such as those through community outreach, health facilities, by sexual partners and in other venues such as workplace and private sector pharmacies. Population-based surveys, like DHS surveys, will then provide an important source of information for countries implementing HIVST, as well as those planning to add HIVST as part of existing HIV testing services. Together with routine programmatic data and special studies, DHS surveys that have included questions on HIVST can then provide a meaningful baseline and point of comparison for future analyses and important insights for future implementation.
Although Malawi and Zimbabwe have scaled-up HIV testing, gaps remain. Efforts to reach the shrinking number of people with HIV who do not know their status is becoming more challenging and also less cost-effective unless targeted toward specific populations and settings with lower knowledge of status among people with HIV [3]. Maintaining the high testing coverage and knowledge of status achieved will not be inexpensive and HIVST is likely to play a role. Furthermore, HIVST also addresses patient costs of accessing services and equity concerns which also need to be considered, especially as programmes get closer to the national goals.
Programmes will need to carefully evaluate how they both maintain essential HIV testing services in facilities, while also deploy highly-focused and effective outreach, with limited resources. Strategies such as offering HIVST through highly focused channels among priority populations, or through time-limited and geographically focused community outreach (such as every 5 years), may be more cost-effective and affordable as more people with HIV learn their status and new infections decline [20].

Limitations
This study has many strengths, namely its large sample size and that it is one of the first to provide an assessment of HIVST awareness, use, and willingness to self-test in two early adopter African countries prior to wide-scale implementation. As such, it provides insight into the progress and changes made since HIVST has been rolled-out, serving as an example for countries monitoring HIVST implementation and scale-up. However, as a cross-sectional survey using self-reported information there may be reporting bias due to social desirability [33]. Previous studies have highlighted challenges with collecting selfreported data, particularly related to sexual risk behaviours and HIV testing history [34,35]. Thus, it is possible that there may be differences between what people reported and their actual behaviour. Given that HIVST was relatively new during the surveys, it is possible that willingness may also change as more people have experiences self-testing. It will be important to assess following broader implementation and scale-up. Second, given that we included two of the first countries to include questions on HIVST, there were discrepancies in implementation such that not all those surveyed were asked about selftesting and willingness to self-test could not be assessed in Malawians or in Zimbabwean women. Willingness to self-test may be similar or different among women and among Malawians, and it will be important to ensure that they are included in future surveys.

Conclusions
Even in 2019, the percentage of people who had never tested for HIV remained above target for Malawian and Zimbabwean men aged ≥ 25 years [26]. Reaching these men will be critical to achieving 2030 goals and maintaining low HIV incidence. Despite very low awareness and previous use of HIVST among 2015-16 DHS respondents, willingness to self-test was high, especially among older Zimbabwean men with high sexual risk.
Reaching these groups is a priority for HIV testing, prevention and care services as we move towards HIV elimination. Social determinants -notably urban residence, paid employment, literacy and wealth -had a pronounced impact on awareness of HIVST in 2015-16, a time that preceded programmatic implementation.
These data provide a valuable baseline against which to investigate population-level HIVST uptake and equity as programmes scale-up. Countries conducting population-based surveys, especially those where HIVST is being used or soon to be introduced, should consider including questions to assess knowledge, awareness and willingness to self-test aiming to provide baseline data and to better understand the potential impact of HIVST overtime and across and within countries. Individuals with any "high-risk" exposures were classified as "high-risk", with the remaining respondents classified as "low risk" if reporting no sexually activity in the past four weeks, and as "moderate risk" otherwise. § P-value based on cluster-adjusted chi-squared test This excludes 37 people who did not report on sexual activity and risk behaviours and are missing from the "HIV risk category". **Represents variables which were not included in the multivariable analysis due to identified collinearity. *** HIV risk is defined in this analysis includes reported sexual activity in the past four partners, any paid sex (asked to men), having received gifts, cash or other compensation in exchange for sex (asked to women), and having a sexually transmitted infection (STI).
Individuals with any "high-risk" exposures were classified as "high-risk", with the remaining respondents classified as "low risk" if reporting no sexually activity in the past four weeks, and as "moderate risk" otherwise. § P-value based on Wald Test. P-values for variables with more than two categories are shown. partners, any paid sex (asked to men), having received gifts, cash or other compensation in exchange for sex (asked to women), and having a sexually transmitted infection (STI).
Individuals with any "high-risk" exposures were classified as "high-risk", with the remaining respondents classified as "low risk" if reporting no sexually activity in the past four weeks, and as "moderate risk" otherwise. § P-value based on Wald Test. P-values for variables with more than two categories are shown. men did not respond and 1 not provide information on sexual activity (HIV risk). Sexual activity was not reported by 1 respondent and could not be used in HIV risk variable.
These variables have a total sample size of 7 371. **Represents variables which were not included in the multivariable analysis due to identified collinearity. *** HIV risk is defined in this analysis includes reported sexual activity in the past four weeks, and the following high-risk exposures in the previous 12 months: multiple (i.e. 2) partners, any paid sex (asked to men), having received gifts, cash or other compensation in exchange for sex (asked to women), and having a sexually transmitted infection (STI).
Individuals with any "high-risk" exposures were classified as "high-risk", with the remaining respondents classified as "low risk" if reporting no sexually activity in the past four weeks, and as "moderate risk" otherwise. § P-value based on Wald Test. P-values for variables with more than two categories are shown. Figure 1 Mechanisms affecting HIV testing uptake in adults (aged 15+) in southern Africa,