In 2020, there were an estimated 1·7 million children younger than 15 years living with HIV worldwide, and only about half (54%) were on antiretroviral therapy (ART).¹ Most new paediatric HIV infections are a result of vertical transmission and approximately 90% of children living with HIV live in sub-Saharan Africa.² Children living with HIV who are not on ART experience substantial morbidity and mortality,^3,4 and WHO guidelines have called for immediate treatment initiation at diagnosis for children living with HIV, regardless of age, to improve survival and development.⁵ To identify children living with HIV and to ensure rapid ART initiation, prevention of mother-to-child transmission (PMTCT) services should include infant testing for all HIV-exposed infants, including early infant diagnosis by age 2 months and repeated testing throughout breastfeeding.⁵

Although coverage of paediatric HIV testing through PMTCT services has increased considerably over the past decade, progress has been slow and gaps remain across countries. In 2015, only an estimated 54% of HIV-exposed infants were tested at 2 months across the Start Free, Stay Free, AIDS Free focus countries comprised of the 21 countries (all in sub-Saharan Africa) with the highest burden of pregnant women living with HIV.⁶ By 2020, early infant diagnosis had increased to approximately 63% overall, but remained at less than 50% in five countries.^7,8 Substantial challenges continue to contribute to missed diagnoses among children living with HIV, including retention of women living with HIV on ART throughout breastfeeding, identification of women who acquire HIV during breastfeeding (which poses high vertical transmission risk), and testing of children throughout and after cessation of breastfeeding.^7,9,10

It is crucial to identify children living with HIV who are undiagnosed and ensure that they are initiated on ART because of the high mortality among children living with HIV who are not on treatment.¹¹ Finding older children living with HIV who were missed in PMTCT programmes is challenging.⁷ Provider-initiated testing and counselling has been the primary approach to case finding; however, it has not been widely implemented, in part due to an absence of perceived need to test older children, as well as due to issues related to disclosure and consent.^12,13 Unfortunately, there are few available empirical studies that have estimated the proportions of undiagnosed children living with HIV in countries with a high HIV burden and these data are not part of routine reporting efforts. Information about the size and characteristics of this population is needed to design effective and targeted testing strategies.

We aimed to estimate the prevalence of undiagnosed HIV among children living with HIV, as well as the proportions of children living with HIV on ART and with viral suppression in Eswatini, Lesotho, Malawi, Namibia, Tanzania, Zambia, and Zimbabwe.

Between 2015 and 2017, 42 248 children aged 1–14 years were included in the seven country surveys, of whom 594 were living with HIV. The estimated weighted prevalence of HIV among children aged 1–14 years from the PHIA surveys was 0·9% (probability band 0·7–1·1) overall, ranging from 0·4% in Tanzania to 2·7% in Eswatini (appendix p 2). In all of the countries except Tanzania and Zambia, HIV prevalence was higher among children aged 10–14 years than among children younger than 10 years.

The median age of children living with HIV was 6·8 years (IQR 3·9–10·4) and 39·5% were aged 5–9 years (table 1). 63·0% of children living with HIV across all countries had previously been tested for HIV before the survey. Among children living with HIV who had previously been tested, 80·3% were reported to have had HIV-positive results, 15·9% were reported as negative at most recent test, and 3·8% had indeterminate or no results received. The mothers of most children living with HIV (81·3%) were alive; 72·1% of mothers tested HIV-positive in the survey, 3·7% tested HIV-negative, and 24·2% had missing results or were not available to be tested. Using information reported by mothers about previous testing, we estimated that 10·6% of mothers were diagnosed with HIV before pregnancy, 20·7% were diagnosed during pregnancy (with the HIV-positive child), 41·4% were diagnosed after the child was born, and an additional 18·9% were newly identified in the PHIA survey. Among mothers who reported HIV-positive status at the time of the survey, 88·6% had detectable antiretroviral medication blood concentrations and among all mothers who were HIV-positive in the survey, 49·2% had a viral load of less than 50 copies per mL (table 1).

Among children living with HIV, 61·0% were diagnosed and 39·0% were undiagnosed (table 1). 87·6% of diagnosed children living with HIV were reported to have been tested for HIV previously compared with only 24·5% of undiagnosed children living with HIV (p<0·0001). We estimated that 14·6% of mothers of diagnosed children were identified as having HIV before pregnancy, 30·2% were diagnosed during pregnancy, and only 2·1% were newly diagnosed in the PHIA survey (table 1). Among undiagnosed children living with HIV, 4·2% of mothers were diagnosed before pregnancy, 6·0% were diagnosed during pregnancy, and 44·9% were newly identified in the PHIA survey (p<0·0001). Slightly more than half (57·3%) of HIV-positive mothers of diagnosed children living with HIV had a viral load of less than 50 copies per mL, whereas 36·3% of HIV-positive mothers of undiagnosed children living with HIV had viral suppression (p=0·027). Characteristics of the samples from each country are shown in the appendix (pp 3–10).

Among diagnosed children living with HIV, 88·4% had detectable antiretroviral medication concentrations in their blood, 5·0% were reported to be on ART but had no detectable antiretroviral medication concentrations, and 6·6% were not on ART according to antiretroviral medication detection and caregiver reports (table 2). Namibia had the highest proportion of diagnosed children living with HIV with detectable antiretroviral medication concentrations (96·0%) and Zambia had the lowest (80·2%). Among diagnosed children living with HIV, 48·3% had a viral load of less than 400 copies per mL; the proportion ranged from 27·1% in children aged 1–4 years to 58·2% in those aged 10–14 years. Among all children living with HIV, including those without previous diagnosis, 54·7% had detectable antiretroviral medication concentrations, 3·0% were reported to be on ART but had no detectable antiretroviral medication concentrations, and 42·3% were not on ART according to antiretroviral medication detection and caregiver reports. Only one-third (32·6%) of all children living with HIV had a viral load of less than 400 copies per mL (table 2). Namibia had the highest proportion of all children living with HIV with detectable antiretroviral medication concentrations (81·0%) and Eswatini had the highest proportion with a viral load of less than 400 copies per mL (59·7%). Zambia had the lowest proportion of all children living with HIV with detectable antiretroviral medication concentrations (43·1%) and Tanzania had the lowest proportion of all children living with HIV with a viral load of less than 400 copies per mL (17·3%).

Across the seven countries, we estimated that—at the time of the PHIA surveys—there were 425 000 children aged 1–14 years living with HIV (probability band 365 000–485 000) and of these, 259 000 (216 000–303 000; 61·0%) were diagnosed and 166 000 (128 000–204 000; 39·0%) were undiagnosed (table 3). Around half (50·9%) of children living with HIV aged 1–4 years, 35·2% aged 5–9 years, and 34·4% aged 10–14 years were undiagnosed (figure). Zambia (51·5%) and Tanzania (50·0%) had the highest proportions of undiagnosed children living with HIV, while Namibia (15·6%) and Eswatini (16·6%) had the lowest. In Lesotho, Namibia, Tanzania, and Zimbabwe, children living with HIV aged 1–4 years were the age group with the highest proportion who were undiagnosed (table 3). In Malawi and Zambia, 10–14 years was the age group with the highest proportion of undiagnosed children living with HIV (figure).

Examination of agreement between HIV status of children as reported by parents or guardians and detection of antiretroviral medications in the child’s blood (as a proxy for diagnosis) showed only fair to good concordance (κ 0·61; 95% CI 0·55–0·68; table 4). Among children with detectable antiretroviral medication concentrations, 20·4% were reported by caregivers to be HIV-negative or have unknown HIV status and 15·2% of children who were reported by caregivers to have been diagnosed with HIV did not have detectable antiretroviral medication concentrations (table 4).

Using data from PHIA surveys conducted across seven countries from 2015 to 2017, we estimate that 39% of children living with HIV had not been diagnosed at the time of the survey, amounting to an estimated 166 000 undiagnosed children living with HIV who were not yet on treatment. Among children living with HIV who were previously diagnosed, most had detectable antiretroviral medication concentrations but only about half had a suppressed viral load. In the total population of children living with HIV, including those who were previously undiagnosed, just more than half had antiretroviral medications detected in their blood and a third had a suppressed viral load. Overall, these findings highlight substantial gaps in paediatric HIV diagnosis services and underscore the urgent need to accelerate testing to identify undiagnosed children living with HIV and to ensure they initiate treatment. For the children with known HIV-positive diagnosis (most of whom were on ART), these findings are further evidence of suboptimal treatment outcomes among children living with HIV.

A key finding from our analysis is the uneven progress in HIV diagnosis and ART coverage among children at the time of the surveys across these seven countries, which reflects the severity of their HIV epidemics, as well as the performance of their PMTCT and adult ART programmes. Although our data do not provide direct information about specific gaps in PMTCT services and paediatric testing efforts, they do provide some indications. In the years before the surveys, there remained serious challenges with infant diagnosis in some of these countries, particularly Tanzania, as evidenced by high proportions of undiagnosed children living with HIV in the age group of 1–4 years. In addition, across the seven countries, two-thirds of undiagnosed children living with HIV had reportedly never been previously tested for HIV, and many mothers of children living with HIV were newly identified as HIV-positive through the surveys. If these women accurately reported knowledge of their HIV status, this indicates missed HIV testing in mothers, including identification of incident HIV infections during pregnancy and breastfeeding, which could account for unidentified children living with HIV. These data also suggest that provider-initiated HIV testing and counselling efforts had not been effectively implemented at the time of the surveys, given the high proportions of older children who remained undiagnosed, including more than 100 000 children living with HIV in the age group of 5–14 years. It is important to note that our data are from 2015 to 2017 and progress has been made since the time of the PHIA surveys. In Malawi, Namibia, Tanzania, and Zambia, for instance, early infant diagnosis coverage increased from less than 50% in 2015 to between 65% (Zambia) and 99% (Namibia) by 2020.⁶ However, despite improvements in early infant diagnosis, overall progress has been slow with only an estimated 54% of children living with HIV globally receiving ART in 2020.⁶

As previously reported by Saito and colleagues,¹⁷ the PHIA data have been pivotal in validating existing UNAIDS estimates of ART coverage and further quantifying the number of children not on treatment. Our data provide further insights into the gaps in ART coverage and show that these are primarily driven by missed diagnosis. Based on detectable antiretroviral medication concentrations, ART coverage among previously diagnosed children living with HIV in most countries was high, but less than half of diagnosed children living with HIV had a suppressed viral load. These findings are further evidence of poor treatment outcomes previously identified among children living with HIV in sub-Saharan Africa, which are driven by late ART initiation, high pretreatment viral loads, suboptimal ART regimens, and pretreatment drug resistance.^18–20 ART coverage in children living with HIV varied by country and was consistent with our findings on previously undiagnosed children living with HIV. Eswatini and Lesotho, which had the lowest proportions of undiagnosed children, also had the highest ART coverage among all children living with HIV, whereas Zambia and Tanzania, where about half of all children living with HIV were undiagnosed, had the lowest ART coverage. These data are clear evidence that to address the paediatric treatment gap, we must do more to identify children living with HIV.

Despite improved coverage of PMTCT services, including ART for pregnant women and testing at age 6–8 weeks for HIV-exposed infants,⁷ challenges remain in retaining women living with HIV on ART and ensuring early and continued infant testing throughout breastfeeding. Interventions that can improve retention of mothers and infants include integrated post-partum ART services for mothers within the maternal–child health platform, provision of additional training of health-care workers, and involvement of peer and other lay health workers.^21,22 Efforts should also include pre-exposure prophylaxis (PrEP) for pregnant and breastfeeding women to prevent new HIV infections,²³ along with retesting for previously HIV-negative mothers during breastfeeding to identify new maternal infections.¹⁰ Many countries have introduced retesting guidelines for pregnant and breastfeeding women,²⁴ which should be coupled with testing for infants with mothers who have been identified as HIV-positive. Our finding that one-quarter of undiagnosed children living with HIV had been previously tested and that most were HIV-negative underscores the need to ensure all HIV-exposed infants receive a final status at the end of breastfeeding, which remains challenging in many countries.⁷ Infant tracking systems, integrated maternal infant care throughout breastfeeding, and point of care testing could be considered to improve final status determination.^22,25,26

While new innovations and testing interventions are needed to identify older children living with HIV, expansion of effective existing approaches is also urgently needed.²⁷ By our estimates, around half of undiagnosed children living with HIV had mothers who were diagnosed with HIV, most of whom were on treatment, which is evidence of missed index testing opportunities and low prioritisation of case finding in children. Provider-initiated HIV testing and counselling in health-care settings, particularly malnutrition units, tuberculosis clinics, and inpatient wards, has shown high yield and often identifies children with acute health needs.^13,28 Index case finding and family testing, which identify children living with HIV through their HIV-positive parents, siblings, and other household members, are also effective strategies that should be offered through both facility or home-based testing.²⁹ An evaluation of index case finding in Malawi in 2014–15 found that 65% of adults living with HIV reported a child or young person in their household who had not been tested, and, in 10% of households with untested children, at least one new child living with HIV was identified.³⁰ Major initiatives, including PEPFAR’s accelerating children’s HIV/AIDS treatment initiative (2014–16), might have improved paediatric case finding; however, their impact has not been well characterised²⁹ and ART coverage among children remains low.⁶

To our knowledge, our analysis is unique in examining concordance between parent or guardian reports of HIV-positive status and detection of antiretroviral medications in the child’s blood (as a marker of known HIV-positive status). We found only fair to good agreement between the report of a child’s HIV-positive status and detectable antiretroviral medications, partly because 15% of children previously diagnosed with HIV were not on ART, which reflects poor treatment coverage even among known children living with HIV. In addition, we found that about 20% of children with detectable antiretroviral medication concentrations were reported as HIV-negative or to have unknown status, which could have resulted from respondents not knowing the child’s HIV status but might also reflect hesitation among parents and guardians to report HIV-positive status of children, even for children on ART. These data indicate that surveys relying on parental reporting of children’s HIV status (without testing) might underestimate the true proportions of children living with HIV, and suggest that caregivers might experience stigma associated with a child’s HIV diagnosis.^31,32

We believe these are the first national estimates of undiagnosed children living with HIV based on population-level data from sub-Saharan Africa and as such these findings provide new and important information. Routinely reported country indicators do not include data on undiagnosed children living with HIV. Our estimates are therefore an important contribution to efforts to estimate the magnitude of missed HIV diagnosis and to increase paediatric ART coverage. The rigour and high participation rates of the PHIA surveys are strengths of this study, as are the biological data collected, including testing of mothers and children and detection of antiretroviral medication blood concentrations which we used to identify children with known HIV-positive status. Our study has some limitations. We were not able to include infants younger than 12 months in our analysis due to missing antiretroviral medication drug concentrations from blood samples in most countries; therefore our conclusions are limited to children aged 1–14 years. There might also be some misclassification of HIV diagnosis status. It is possible that adults reporting children’s HIV status were unaware of or refused to report HIV-positive status and, if these children were missing detectable antiretroviral medication concentrations due to non-adherence or not having started ART, we might have overestimated the number of undiagnosed children. Another limitation is the potential for misreporting of mother–child biological relationships (women might have been misidentified as biological mothers), which might be responsible for a small number of serodiscordant mother–child pairs. In addition, our estimates of the timing of the mother’s HIV diagnosis were imprecise as we did not have access to children’s dates of birth and could therefore only roughly estimate the timing of the mother’s HIV diagnosis in a window around the child’s birth. Finally, the PHIAs were conducted from 2015 to 2017, providing evidence about PMTCT and HIV testing coverage for as long ago as the early 2000s for older children. There has been substantial progress in expanding access and coverage of PMTCT and paediatric testing initiatives over the past 15 years, including changes to guidelines to start all pregnant women living with HIV on lifelong ART in all of the countries included in this study, point of care testing, and provider-initiated testing for older children. Our findings describe challenges to PMTCT programmes in the past; however, the data on the youngest children in these surveys confirm that gaps have persisted in many countries, contributing to missed diagnosis and low treatment coverage among children living with HIV.

In conclusion, across seven countries with a high HIV burden, we estimate that in 2015–17, 39% or 166 000 children aged 1–14 years living with HIV had not been diagnosed and were not on treatment. Furthermore, due to missed diagnosis, only about half of children living with HIV in these countries were on ART and only a third had viral suppression. To close the paediatric treatment gap, urgent efforts are needed to accelerate HIV testing in these settings.