National age-of-consent laws and adolescent HIV testing in sub-Saharan Africa: a propensity-score matched study

Abstract Objective To estimate the association between legal age of consent and coverage of human immunodeficiency virus (HIV) testing among adolescents in countries with high HIV-burden. Methods We analysed data from adolescents aged 15–18 years, who participated in Demographic and Health Surveys or AIDS Indicator Surveys between 2011 and 2016, in 15 sub-Saharan African countries. To improve balance in the distribution of measured individual- and country-level characteristics, we used propensity score matching between adolescents in countries with more versus less restrictive age-of-consent laws (≤ 15 years versus ≥ 16 years). We estimated the percentage of individuals who self-reported that they have done an HIV test in the past 12 months and compared the differences in such testing rates among adolescents exposed to lower versus higher age-of-consent laws. We also investigated effect modifications by sex and age. Findings Legal age of consent below 16 years was associated with an 11.0 percentage points higher coverage of HIV testing (95% confidence interval, CI: 7.2 to 14.8), corresponding to a rate ratio of 1.74 (95% CI: 1.35 to 2.13). HIV testing rate had a stronger association with lower age of consent among females than males. The testing rates differences were 14.0 percentage points (95% CI: 8.6 to 19.4) for females and 6.9 percentage points (95% CI: 1.6 to 12.2) for males (P-value for homogeneity = 0.07). Conclusion This study provides evidence to support the recent World Health Organization’s recommendations that countries should examine current laws and address age-related barriers to uptake of sexual and reproductive health services.


Introduction
In 2017, an estimated 1.8 million adolescents were living with human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS) globally. 1 Of these, 1.5 million (84%) HIV-infected young people live in sub-Saharan Africa, where AIDS is the leading cause of death among people aged 15-19 years. 2 Most HIV-infected adolescents acquired HIV through mother-to-child transmission and were not diagnosed during infancy. Others were infected through sexual contact; injecting drugs; or through HIV transmission in health-care settings (e.g. blood transfusions). 3 In 2017, girls aged 15-19 years in sub-Saharan Africa were nearly three times as likely to be newly infected with HIV than adolescent boys. 1 Adolescents are a group that have been largely left behind in the global AIDS response and AIDS-related deaths declined between 2000 and 2016 for all age groups except adolescents. 1 However, this neglected group has recently come into focus and the global consensus is that scaling up effective HIV/AIDS prevention, treatment and care for adolescents, especially girls, is urgently needed. 3,4 In 2015, the Joint United Nations Programme on HIV and AIDS and United Nations Children's Fund launched the All In To End the Adolescent AIDS Epidemic (All In) campaign, which focuses on the 25 countries that contribute to 86% of all new HIV infections in adolescents. 3 By the year 2020, the campaign aims to reduce new HIV infections among adolescents by at least 75%, reduce AIDS-related deaths by at least 65% and end stigma and discrimination for adolescents living with HIV. Strategies to achieve these targets focus on achieving wider access to effective treatment, holistic prevention programmes that combine behavioural, biomedical and structural prevention strategies and increasing HIV testing among adolescents. Specifically, the campaign aims for 90% of adolescents living with HIV to know their status. 3 The World Health Organization (WHO) recommends HIV testing and counselling, with linkage to prevention, treatment and care, for all adolescents living in generalized epidemic settings, defined as countries where HIV prevalence is consistently over 1% among pregnant women. 5 However, according to recent data for eastern and southern Africa, only an estimated 23% of girls and 16% of boys aged 15-19 years report being tested for HIV and receiving the result in the past 12 months. 2 Well-researched barriers to adolescent HIV testing and counselling include difficulties in accessing testing services, fear of discrimination and family reaction and fear of a positive diagnosis and AIDS-related illness or death. 5 In some countries, adolescents also face legal and policy barriers to HIV testing and counselling, in particular those related to requirements for parent or guardian consent to access HIV testing and counselling services. 6 Restrictive age-of-consent legislation is acknowledged as a potential barrier to HIV testing among adolescents and there is an increasing interest to review and reform age-of-consent laws. 3,5 According to a review of age-of-consent laws in African countries, 19 of the 33 countries had clear national laws supporting independent HIV testing and counselling for people younger than 18 years. 7 Most African countries set the age of consent at 16 or 18 years; however, some countries like South Africa and Uganda permit independent access to HIV testing and counselling services as early as 12 years of age. 5 The intent of policies on age of consent is to protect youth minors; yet, requiring parental consent to access HIV testing and counselling may be a barrier to such services. 8 Objective To estimate the association between legal age of consent and coverage of human immunodeficiency virus (HIV) testing among adolescents in countries with high HIV-burden. Methods We analysed data from adolescents aged 15-18 years, who participated in Demographic and Health Surveys or AIDS Indicator Surveys between 2011 and 2016, in 15 sub-Saharan African countries. To improve balance in the distribution of measured individual-and country-level characteristics, we used propensity score matching between adolescents in countries with more versus less restrictive ageof-consent laws (≤ 15 years versus ≥ 16 years). We estimated the percentage of individuals who self-reported that they have done an HIV test in the past 12 months and compared the differences in such testing rates among adolescents exposed to lower versus higher age-ofconsent laws. We also investigated effect modifications by sex and age. Findings Legal age of consent below 16 years was associated with an 11.0 percentage points higher coverage of HIV testing (95% confidence interval, CI: 7.2 to 14.8), corresponding to a rate ratio of 1.74 (95% CI: 1.35 to 2.13). HIV testing rate had a stronger association with lower age of consent among females than males. The testing rates differences were 14.0 percentage points (95% CI: 8.6 to 19.4) for females and 6.9 percentage points (95% CI: 1.6 to 12.2) for males (P-value for homogeneity = 0.07). Conclusion This study provides evidence to support the recent World Health Organization's recommendations that countries should examine current laws and address age-related barriers to uptake of sexual and reproductive health services.
National age-of-consent laws and adolescent HIV testing Britt McKinnon & Ashley Vandermorris HIV testing and counselling, due to fear of disclosure or violence. 9,11,12 Healthy parental involvement and communication have been shown to have positive effects on youth decision-making around sexual and reproductive health issues, 8 however, there is little evidence to suggest that parental consent fosters positive parental involvement. 7 The review on age-of-consent laws recommended that countries consider setting the minimum legal age of consent for HIV testing and counselling to 12 or 14 years of age, but the authors acknowledge there are little data to support this recommendation. 7 Although several countries in sub-Saharan Africa have lowered the legal age of consent in the past decade, little is known about the impact of these changes. Several stakeholders have called for more research to understand the effects of lowering the legal age to consent to HIV testing and counselling. 7,13 The objective of this study was to estimate the association between legal age of consent and the rates of HIV testing among adolescents, using a quasi-experimental approach and data from 15 sub-Saharan countries.

Study design and measures
This is a cross-national study to estimate the association between legal age of consent and coverage of HIV testing and counselling among adolescents aged 15-18 years. We used a cut-off of 18 years because individuals older than this age were eligible to consent to HIV testing and counselling across all countries. We used a quasi-experimental propensity score matching approach to achieve balance in the distribution of measured covariates between countries with more and less restrictive consent legislation. We employed this approach expecting that factors such as national HIV prevalence would confound the relationship between age-of-consent policies and uptake of HIV testing among adolescents, as countries with higher HIV prevalence tend to have higher coverage of HIV testing and counselling. These countries may have also lowered the legal age of consent in response to the country's HIV epidemic. 6 By matching the groups being compared for variables that might predict the likelihood of exposure to less restrictive age-of-consent laws, the propensity score approach can improve control of confounding factors and reduce bias. 14 We used nationally representative household survey data from Demographic and Health Surveys (DHS) and Aids Indicator Surveys. 15 We included Sub-Saharan African countries targeted in the All In campaign 3 with at least one nationally-representative survey, conducted between 2011 and 2016, that asked participants whether they had received an HIV test within the past 12 months. Adolescents aged 15-18 years and who had not previously given birth and were not currently pregnant, were eligible for the study. We set these inclusion criteria as HIV testing is frequently linked to antenatal care services regardless of consent laws. 16 We linked individual-level data on self-reported HIV testing and other covariates to information on national age-of-consent laws for independent HIV testing and counselling. We obtained information about such laws from a review conducted in 2011 and 2012. 7 We also reviewed national policy documents to ensure each survey data set were matching the correct legislation in place at the time of the survey.

Statistical analysis
We used 1-1 nearest neighbour propensity score matching without replacement and tested different caliper sizes to ensure that potential confounders were balanced between adolescents in the two groups, that is, age-of-consent policy ≤ 15 years versus ≥ 16 years. We identified potential country-and individual level confounders a priori. The four country-level variables were: (i) HIV prevalence: (ii) adolescent fertility rate; (iii) health expenditures per capita; and (iv) comprehensive HIV knowledge among youth, and the seven individual-level variables were: (i) age; (ii) sex; (iii) marital status; (iv) education; (v) rural residence; (vi) previous sexual activity; and (vii) assets-based

Original sample Matched sample
Notes: We used a caliper of 0.01 for the matched samples. Countries included in the study are listed in Table 1.

Research
National age-of-consent laws and adolescent HIV testing Britt McKinnon & Ashley Vandermorris household wealth. We selected a caliper size of 0.01 after comparing changes in mean standardized bias and sample size (comparison available from the corresponding author). Histograms showing the estimated propensity scores by legal age of consent for HIV testing and counselling for the original and matched samples are presented in Fig. 1.
Using the prop ensity-scorematched sample, we assessed the association between age-of-consent polices and adolescent HIV testing with logistic regression models. To facilitate interpretation and reporting of associations on the absolute probability scale, we calculated rate ratios (RR) for the proportion of people tested for HIV in the past 12 months and rate differences from average marginal probabilities estimated from the regression coefficients. 17 In this primary analysis we included all adolescents, since more adolescents in sub-Saharan Africa have been infected through mother-to-child transmission than through sexual contact. Furthermore, the WHO's recommendations state that all adolescents should receive HIV testing and counselling in generalized epidemic settings. 5 We conducted a second analysis restricted only to those who reported ever having had sex.
We also investigated whether the association between age-of-consent policies and HIV testing rates differed by sex or age (15-16 years versus17-18 years of age), by including interaction terms between these variables and the age-ofconsent policy. A χ 2 test for formal statistical comparison of homogeneity of the estimated RRs and rate differences was used to assess whether the associations differed by sex or age. 18 Given the low statistical power of homogeneity tests, we considered more liberal P-value cutoffs of 0.10 to be suggestive of potential effect modification. 19 We accounted all analyses for clustering at the country level and conducted the analyses using Stata version 14.1 (StataCorp LCC, College Station, United States of America).

Results
The full sample included 62 628 adolescents, of which 39 339 were females and 23 289 were males. The legal age of consent for independent HIV testing and counselling was ≤ 15 years in six of the countries and ≥ 16 years in the remaining nine countries (Table 1). Table 1 also presents estimated coverage of HIV testing in the past 12 months for female and male adolescents. Table 2 presents characteristics of the sample stratified by age of consent. The HIV prevalence did not differ between countries with more or less restrictive age-of-consent policies (7.1% in countries with ≥ 16 years as age consent and 6.9% in countries with ≤ 15 years). The two groups were similar on other characteristics, except for level of education and rural residence. In countries with ≥ 16 years as the age of consent, 59% of the adolescents had some secondary schooling (95% CI: 48 to 69%) versus 36% (95% CI: 28 to 44%) in countries with < 16 years as the age of consent.
In the propensity score-matched sample, lower age of consent (≤ 15 years versus ≥ 16 years) was associated with a 11.0-percentage point higher coverage of HIV testing in the past 12 months (95% CI: 7.2 to 14.8; Table 3). This difference corresponds to a RR of 1.74 (95% CI: 1.35 to 2.13). There were also differences in the coverage of HIV testing by sex and age. Females were more likely to have been tested for HIV than males (rate difference: 8.6 percentage points; 95% CI: 4.2 to 13.1). As expected, HIV testing increased with age, coverage was 9.6 percentage points higher (95% CI: 7.7 to 11.6) among adolescents aged 18 than 15 years. In the analysis restricted to sexually active adolescents, the rate difference between lower age of consent versus higher was similar to the full sample (11.8 percentage points; 95% CI: 7.0 to 16.6), while the RR was slightly smaller (1.52; 95% CI: 1.21 to 1.83).
Examining variables that could potential modify the association between legal age of consent and HIV testing rate showed that the increased coverage seen in lower legal age-of-consent group is greater among females than males (rate difference:14.0 for females and 6.9 for males; P-value for homogeneity: 0.07). Among adolescents who reported being sexually active, the association between age of consent and testing rate by sex was slightly attenuated (P-value for homogeneity: 0.24). There was no evidence that associations between age of consent and HIV testing varied by age, either in the full or sexually active sample (P-value for homogeneity: 0.83 and 0.66, respectively; Table 4).

Discussion
This study shows that a lower legal age of consent to independent HIV testing and counselling is associated with an increase in HIV testing rate among adolescents in high-HIV burden countries. To date, recommendations to legally lower the age of consent for HIV/AIDS services, such as those in the WHO guidance document 5 have been based on an awareness of the ethical importance of supporting the notion of adolescents' evolving capacities. 23 Such a notion has been delineated in the International Convention on the Rights of the Child. 24 The recommendations were also based on the recognition of the pragmatic and psychosocial barriers to service uptake that age-of-consent laws may introduce. The results of this study provide evidence to support these recommendations.
Age-of-consent laws can be a barrier to adolescents accessing relevant HIV/AIDS services. 10 This study provides new evidence that a lower legal age of consent could address this barrier, especially in countries with more restrictive legislation, and increase HIV testing and counselling uptake. HIV testing and counselling uptake among adolescents, in turn, has been associated with lower incidence of HIV infection over time. 25 If, as the findings of this study suggest, lower age-of-consent laws can be an upstream facilitator for improving access to HIV testing and counselling, such laws may also potentially indirectly act to reduce HIV incidence rates.
The results presented here indicate a stronger association between lower age-of-consent laws and HIV testing rate among females than males. Such a finding suggests that lowering the legal age of consent could have a greater effect on adolescent girls, who are more affected by the HIV epidemic. Adolescent girls are at greater risk, both since they have a higher biological susceptibility to HIV infection and because of sociodemographic characteristics and sociocultural beliefs and practices that deter them from accessing sexual and reproductive health services. 26 With little evidence on effective HIV prevention interventions to protect young women, 27 this study provides a potentially important finding that is well aligned with current efforts to promote gender equality, as articulated by sustainable development goal 5. 28 This study has several strengths. We used comparable high-quality data from the well-established DHS program for 15 of the18 high-HIV burden countries in sub-Saharan Africa. We relied on a comprehensive review of age-of-consent laws conducted by WHO. We employed propensity score matching to minimize potential confounding by individualand country-level factors, including national HIV prevalence and knowledge of HIV. Changes in HIV testing uptake in young people have been attributed to several factors, including access to testing facilities, perceived attitudes of staff towards young people and perceptions around confidentiality. 29 We therefore matched factors likely to be influenced by other policy and programme efforts that affect access and barriers to HIV testing and counselling among adolescents, and that may coincide with ageof-consent laws. The study also has limitations. The propensity score analysis does not account for potential unmeasured confounding factors. Therefore, there may be factors related to a country's sociopolitical acceptance of adolescent sexuality and autonomy, which could influence both the adoption of lower age-of-consent laws and uptake of HIV testing and counselling for reasons other than age-related legal barriers to care. For example, in 2014, the East and Southern Africa Regional Office of the United Nations Population Fund, in collaboration with the Africa Regional Office of the International Planned Parenthood Federation, launched a multi-year initiative to scale up adolescent and youth-friendly health services in the region. 30 Recent assessment of the implementation of such services revealesd significant variation across and within countries, however data on the specific implementation successes  or gaps by country are not yet available. However, youth-friendly health services could have so far an undetermined relationship with age-of-consent laws and adolescent HIV testing and counselling uptake. Potential mechanisms for such a relationship could be that these services enhanced quality of care, improved provider knowledge of the legal rights of adolescents, or increased demand for services. Another limitation is the reliance on self-reports of HIV testing, which may be affected by recall and/ or reporting bias. 31 HIV testing may be either over-reported, if being tested is regarded as the responsible thing to do, or underreported, if getting tested for HIV is perceived as admitting to a socially undesirable behaviour (e.g. premarital sex). 29 Evidence on the likelihood and possible magnitude of these two potential reporting bias scenarios is limited. Although plausible, we have no reason to believe the tendency to under-or over-report HIV testing in a confidential survey would differ systematically according to age-of-consent laws. Furthermore, we note that the 15 countries included in the analysis had five different legal ages of consent, varying between 12 and 18 years. From an analytical standpoint, we were not able to examine differences among these different legal ages and elected to compare laws stipulating ≤ 15 years versus ≥ 16 years. We based this choice on WHO recommendations to consider setting the minimum legal age-of-consent for HIV testing and counselling to 12 or 14 years, as opposed to most countries that set a minimum legal age at 16 or 18. 7 The results in this study should thus be interpreted as a comparison between countries with lower versus higher legal ages of consent, and not as a comparison between specific legal ages.
This study provides evidence that lowering the legal age of consent to HIV testing and counselling may be a potential mechanism for increasing coverages of HIV testing among adolescents in high-HIV burden countries. The findings support WHO's recommendations that countries should examine current laws and address age-related barriers to uptake of sexual and reproductive health services. Additional research would help elucidate any negative or unintended consequences of lowering the legal age of consent for independent HIV testing and counselling.