Secondary Logo

Share this article on:

Pediatric HIV Treatment Gaps in 7 East and Southern African Countries: Examination of Modeled, Survey, and Routine Program Data

Saito, Suzue, PhD, MIA, MA*,†; Chung, Hannah, MPH*; Mahy, Mary, ScD, MHSc; Radin, Anna K., DrPH, MPH§; Jonnalagadda, Sasi, PhD; Hakim, Avi, MA, MPH; Awor, Anna C., PhD, MPH, Mstat; Mwila, Annie, MD; Gonese, Elizabeth, PhD, MPH; Wadonda-Kabondo, Nellie, PhD; Rwehumbiza, Patrick, MD, MPH; Ao, Trong, ScD; Kim, Evelyn J., PhD; Frederix, Koen, MD, MPH*; Nuwagaba-Birbomboha, Harriet, MD, PhD*,†; Musuka, Godfrey, MSc (Med), MPhil, DVM*; Mugurungi, Owen, MD, Msc; Mushii, Jeremiah, MSc#; Mnisi, Zandile, MSc**; Munthali, Gloria, MD††; Jahn, Andreas, MD, PhD‡‡; Kirungi, Wilford L., MBChB, MSc§§; Sivile, Suilanji, MBChB, MMED║║; Abrams, Elaine J., MD*,†

JAIDS Journal of Acquired Immune Deficiency Syndromes: August 15, 2018 - Volume 78 - Issue - p S134–S141
doi: 10.1097/QAI.0000000000001739
Supplement Article

Background: Remarkable success in the prevention and treatment of pediatric HIV infection has been achieved in the past decade. Large differences remain between the estimated number of children living with HIV (CLHIV) and those identified through national HIV programs. We evaluated the number of CLHIV and those on treatment in Lesotho, Malawi, Swaziland, Tanzania, Uganda, Zambia, and Zimbabwe.

Methods: We assessed the total number of CLHIV, CLHIV on antiretroviral treatment (ART), and national and regional ART coverage gaps using 3 data sources: (1) Joint United Nations Programme on HIV/AIDS model-based estimates and national program data used as input values in the models, (2) population-based HIV impact surveys (PHIA), and (3) program data from the President's Emergency Plan for AIDS Relief (PEPFAR)–supported clinics.

Results: Across the 7 countries, HIV prevalence among children aged 0–14 years ranged from 0.4% (Uncertainty Bounds (UB) 0.2%–0.6%) to 2.8% (UB: 2.2%–3.4%) according to the PHIA surveys, resulting in estimates of 520,000 (UB: 460,000–580,000) CLHIV in 2016–2017 in the 7 countries. This compared with Spectrum estimates of pediatric HIV prevalence ranging from 0.5% (UB: 0.5%–0.6%) to 3.5% (UB: 3.0%–4.0%) representing 480,000 (UB: 390,000–550,000) CLHIV. CLHIV not on treatment according to the PEPFAR, PHIA, and Spectrum for the countries stood at 48% (UB: 25%–60%), 49% (UB: 37%–50%), and 38% (UB: 24%–47%), respectively. Of 78 regions examined across 7 countries, 33% of regions (PHIA data) or 41% of regions (PEPFAR data) had met the ART coverage target of 81%.

Conclusions: There are substantial gaps in the coverage of HIV treatment in CLHIV in the 7 countries studied according to all sources. There is continued need to identify, engage, and treat infants and children. Important inconsistencies in estimates across the 3 sources warrant in-depth investigation.

*ICAP at Columbia University, New York, NY;

Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, NY;

Joint United Nations Programme on HIV/AIDS, Geneva, Switzerland;

§U.S. Department of State, Office of the U.S. Global AIDS Coordinator and Health Diplomacy, Washington, DC;

Centers for Disease Control and Prevention, Atlanta, GA;

Zimbabwe Ministry of Health and Child Care, AIDS and TB Programme, Harare, Zimbabwe;

#Tanzania Ministry of Health, National AIDS Control Program, Dar es Salaam, Tanzania;

**Swaziland Ministry of Health, Health Research Department, Mbabane, Swaziland;

††Directorate of Clinical Care and Diagnostic Services, Zambia Ministry of Health, Lusaka, Zambia;

‡‡International Training and Education Center for Health, Lilongwe, Malawi;

§§Uganda Ministry of Health, Kampala Uganda; and

║║ Infectious Diseases Unit, Department of Internal Medicine, University Teaching Hospitals - Adult Hospital.

Correspondence to: Suzue Saito, PhD, MIA, MA, Mailman School of Public Health, ICAP at Columbia University, 722 West 168th Street, 13th Floor, New York, NY 10032 (e-mail: ss1117@columbia.edu).

Supported by the President's Emergency Plan for AIDS Relief through the Centers for Disease Control and Prevention under the terms of #U2GGH001226.

The authors have no conflicts of interest to disclose.

The findings and conclusions in this paper are those of the authors and do not necessarily represent the official position of the funding agencies.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's Web site (www.jaids.com).

Back to Top | Article Outline

INTRODUCTION

Remarkable success in the prevention and treatment of HIV in children has been achieved in the past decade. Expanded access to antiretroviral treatment (ART) for pregnant and breast-feeding women for the prevention of mother-to-child HIV transmission has reduced new infections by 60% from 420,000 (Uncertainty Bounds (UB): 260,000 – 620,000) in 2000 to 180,000 (UB: 110,000–260,000) in 2017.1 In parallel, scale-up of early infant diagnosis (EID) and pediatric testing and treatment has led to a dramatic reduction in pediatric AIDS-related deaths from 270,000 (UB: 150,000–400,000) in 2000 to 110,000 (UB: 62,000–160,000) in 2017.1 The Accelerating Children's HIV/AIDS Treatment (ACT) Initiative (2014–2016) funded by the President's Emergency Plan for AIDS Relief (PEPFAR) and the Children's Investment Fund Foundation substantially contributed to the recent increase in diagnosis and treatment coverage among high-burden countries1,2 by committing dedicated resources to accelerate identification, linkage, and treatment of children living with HIV (CLHIV).3 Notwithstanding these achievements, large differences remain between the estimated number of CLHIV and those that are identified and treated through the national HIV programs, including PEPFAR. The persistence of gaps in some countries have given rise to calls to confirm both the model-based estimates of CLHIV and reported numbers of CLHIV identified and started on HIV treatment with direct measurements using population-based surveys.

The population-based HIV impact assessment (PHIA) project is assessing the status of the HIV epidemic and the impact of the HIV treatment scale-up in 14 high-burden countries.4 Most PHIA surveys include measurement of HIV prevalence among the pediatric population aged 0–14 years. Total number of CLHIV at national and subnational levels are estimated using survey weights developed with population projections generated by the national statistical authorities from each country or from the United Nations Population Division for the year of the survey implementation. The PHIA surveys also measure ART coverage for children using both parent/guardian report and detectable antiretroviral drugs (ARVs) in blood specimens.

To assess the model-based CLHIV estimates published by UNAIDS and the total number of CLHIV on ART reported from PEPFAR and national programs, we conducted a study using data from calendar years 2016–2017 with 2 objectives: (1) to compare pediatric HIV prevalence and total number of CLHIV using (a) Joint United Nations Programme on HIV/AIDS (UNAIDS) model-based estimates and (b) PHIA-based estimates; (2) to compare pediatric treatment coverage gaps using PEPFAR, PHIA, and UNAIDS data. We examined data for 5 Southern African countries, namely, Lesotho, Malawi, Swaziland, Zambia, Zimbabwe, and 2 East African countries, namely, Uganda and Tanzania.

Back to Top | Article Outline

METHODS

Data Sources

We examined 3 data sources: (1) UNAIDS model-based estimates and national program data used as input values in the models, (2) PHIA surveys, and (3) program data from PEPFAR-supported nonmilitary clinics. UNAIDS data are publicly accessible through the UNAIDS Web site.5 Survey protocols for the Lesotho, Malawi, Swaziland, Tanzania, Uganda, Zambia, and Zimbabwe PHIAs were approved by the Centers for Disease Control and Prevention Institutional Review Board (IRB), the Columbia University Medical Center IRB, and the local IRB in each country (Ministry of Health Research and Ethics Committee of Lesotho, National Health Science Research Committee of Malawi, Swaziland National Health Research Review Board, National Institute of Medical Research of Tanzania, Uganda National Council for Science and Technology and Uganda Virus Research Institute, the Tropical Disease Research Center in Zambia, and the Medical Research Council of Zimbabwe). Program data from PEPFAR-supported clinics are routinely collected as part of service provision, are aggregated at the clinic level, and do not contain any personally identifiable information. They are also made available publicly through the PEPFAR website.6

Back to Top | Article Outline

UNAIDS Model-Based Estimates From Spectrum Software

Spectrum files were produced by country teams (comprised of national and international experts) in each of the 7 countries. Results from these files were compiled for this analysis and are also available on the UNAIDS Web site for the 7 countries. A detailed description of the Spectrum model is available in the Spectrum/AIM manual and in past publications.7–9 In brief, the model uses country-specific data from HIV surveillance (both sentinel surveillance and routine testing data) and national surveys, as well as numbers on ART (for children, men, and women separately) and pregnant women receiving either prophylaxis or ART compiled by ministries of health to produce estimates of and determine trends in HIV prevalence and incidence. These are combined with global/regional epidemiological patterns (rates of disease progression, mortality, and mother-to-child transmission probabilities) to produce estimates of key indicators.7 Spectrum version 5.63 was used for these calculations (Avenir Health, Glastonbury, CT). We used national-level estimated pediatric HIV prevalence, total number of CLHIV, and their associated Uncertainty Bounds for 2016 (objective 1) and the number of children on ART compiled by ministries of health for 2016 (objective 2) for this study. Subnational estimates from Spectrum were not used. Of note, adult prevalence from the PHIA surveys in Zimbabwe, Malawi, and Zambia were inputs in the 2016 estimates for these countries. However, they are not directly used in the child prevalence calculation.

Back to Top | Article Outline

PHIA Surveys

The PHIA surveys are nationally representative, cross-sectional, household-based surveys with a 2-stage cluster sampling design. PHIA surveys measure national adult HIV incidence, adult and pediatric HIV prevalence, and progress toward the UNAIDS 90-90-90 treatment targets (90% of people living with HIV know their status, 90% of those who know their status are receiving ART, and 90% of those on treatment are virally suppressed).10 HIV status for children between 18 months and 14 years was determined in a randomly selected subset of households by in-home counseling and HIV rapid testing according to each country's HIV national testing guidelines. Children younger than 18 months who tested positive with the rapid test in the home were confirmed with a DNA polymerase chain reaction test on dried blood spot specimens from blood collected during the household visit. For current use of ARVs, in Malawi, Zambia, Swaziland, and Uganda, current ARV use is determined either via the detection of ARVs in blood for nevirapine, efavirenz, atazanavir, and lopinavir or parent- or guardian-reported current ARV use. The list of ARVs was determined based on the prevailing national guidelines and included at least 3 of the above drugs per country. At the time of this publication, only parent or guardian report of current ARV use was available for this analysis in Zimbabwe, Lesotho, and Tanzania.11,12

Estimates from PHIA were adjusted for the 2-stage cluster sample design and nonresponse using the χ2 automatic interaction detector.13 In addition, poststratification procedures were applied to adjust for noncoverage to calibrate weighted sample counts to match the population control totals from projected 2016 national population counts by gender and 5-year age groups published by the national statistical authorities in Zimbabwe (ZIMSTAT),14 Malawi (National Bureau of Statistics),15 and Zambia (Central Statistical Offices),16 and from projected 2017 national population counts for Swaziland (Central Statistical Office)17 and Uganda (Bureau of Statistics).18 For Lesotho, the 2016 census data supplied the population control totals (Bureau of Statistics)19 and for Tanzania population control totals from the UN population division were used.20 For this study, we used national and regional HIV prevalence and ART coverage estimates and the estimated number of CLHIV and CLHIV on ART using weighted numbers reflecting the poststratification calibration described above.

Back to Top | Article Outline

PEPFAR Program Data

We used PEPFAR Fact View analytic data sets, which are extracted at the end of each quarterly reporting period from the PEPFAR Data for Accountability, Transparency Impact Monitoring (DATIM) database, through which implementing partners submit data.21 Specifically, we used data on the numbers of CLHIV currently on ART at the end of the reporting quarter (indicator: TX_CURR) at PEPFAR-supported nonmilitary clinics, aggregated at district level from the 7 countries for calendar years 2015–2016.22 These data are compiled from routinely reported program data collected through electronic or paper-based records in clinics supported by PEPFAR providing pediatric HIV services and reported to the Office of the Global AIDS Coordinator via DATIM. Fact View data sets do not include data from clinics that are not supported by PEPFAR. For this study, we aggregated district-level numbers of CLHIV on ART for ages <1, 1–4, and 5–14 years and the sum of the disaggregations in PEPFAR-supported sites to national and regional levels.

Back to Top | Article Outline

ANALYSIS

Comparing PHIA and Spectrum Pediatric National HIV Prevalence and CLHIV Estimates

We compared national 2016 pediatric HIV prevalence and estimates of the number of CLHIV from PHIA surveys in Lesotho, Malawi, Swaziland, Tanzania, Uganda, Zambia, and Zimbabwe collected between October 2015 and July 2017 with those from Spectrum. Because PHIA and Spectrum point estimates and Uncertainty Bounds are calculated using different methods, we did not attempt a statistical comparison of the estimates but instead interpreted the estimates and described programmatically important differences and similarities.

Back to Top | Article Outline

Comparing Gaps in ART Coverage From PEPFAR, PHIA, and Spectrum

We first calculated ART coverage among children aged 0–14 years by taking the ratio of the number of CLHIV currently on ART from PEPFAR to the total CLHIV estimates from PHIA at regional and national levels. We merged the PEPFAR ART data with the PHIA CLHIV data by region. The numbers of CLHIV on ART from the July–September quarter of 2016 were used for all disaggregated age-group data (ie, <1, 1–4, 5–14). If data were not available for the July–September quarter of 2016, we took data that were most complete from the closest preceding quarter between 2015 January–March quarter and 2016 July-September quarter. Supplemental Digital Content, Table S1, http://links.lww.com/QAI/B184 summarizes the distribution of data used by country and the amount of missing data we had.

We then calculated the gap in ART coverage among children aged 0–14 years at national and regional levels by merging the PHIA and PEPFAR data and using the formula as below:

For the regional level ART coverage gap analysis, we grouped regions into 2 categories: (1) regions with equal to or greater than 81% ART coverage (ie, less than 19% treatment gap), a figure derived from multiplying the first and second 90 in the UNAIDS 90-90-90 treatment targets10; and (2) regions with less than 81% ART coverage (ie, equal to or greater than 19% treatment gap). Uncertainty Bounds for the treatment coverage gaps calculated using the PEPFAR:PHIA ratio approach were determined by the following formulas:

We compared these ART coverage gaps calculated taking the ratio of PEPFAR program data to PHIA CLHIV estimates with ART coverage gaps generated from self-reported current ARV use or detectable ARVs in blood data from PHIA and with ART coverage gaps reported by national programs used as inputs into Spectrum. Note that PEPFAR is a large funding source for pediatric care and treatment but not the only source. Inputs into the Spectrum model should reflect the national coverage most accurately as national governments report all data on CLHIV on ART in the country irrespective of funding source. In countries where PEPFAR coverage of pediatric ART is high, PEPFAR estimates and Spectrum estimates are expected to be similar. In countries where PEPFAR focuses on specific regions, Spectrum estimates are expected to be higher. Supplemental Digital Content, Table S2, http://links.lww.com/QAI/B184 summarizes the scope of PEPFAR support for sites providing pediatric ART and also summarizes pediatric patients by country. Jackknife estimates of variance were used to calculate Uncertainty Bounds around PHIA treatment coverage gap estimates. Spectrum software (Avenir Health, Glastonbury, CT) generated Uncertainty Bounds were used for Spectrum treatment coverage gap estimates.

All analyses were conducted in SAS 9.4 (SAS Institute, Cary, NC).

Back to Top | Article Outline

RESULTS

Comparing PHIA and Spectrum National HIV Prevalence and CLHIV Estimates

Table 1 summarizes the HIV prevalence among children aged 0–14 years and the total number of CLHIV from PHIA and Spectrum. The HIV prevalence ranged from 0.4% (UB: 0.2%–0.6%) in Tanzania using PHIA data and 0.5% (UB: 0.5%–0.6%) in Uganda using Spectrum data to 2.8% (UB: 2.2%–3.4%) in Swaziland using PHIA data and 3.5% (UB: 3.0%–4.0%) using Spectrum data. In PHIA, Malawi had the highest estimated number of CLHIV at 119,510 (UB: 89,028 – 149,974) followed by Tanzania at 103,785 (UB: 61,296 – 146,274). In Spectrum, Tanzania was the only country with an estimated CLHIV of more than 100,000; CLHIV in Malawi was estimated at 74,000 (56,000 – 86,000) in Spectrum. National-level estimates of the number of CLHIV were comparable across all countries, except in Malawi where Uncertainty Bounds between PHIA and Spectrum did not overlap (Table 1).

TABLE 1

TABLE 1

Back to Top | Article Outline

Comparing Gaps in ART Coverage From PEPFAR, PHIA, and Spectrum

We first compared the number of CLHIV on ART 0–14 from PEPFAR-supported clinics, PHIA, and Spectrum (Table 2). The number of CLHIV on ART 0–14 ranged from approximately 10,000 in Lesotho and Swaziland to 70,000 in Malawi based on PHIA data. Using PEPFAR data, the number ranged from approximately 7000 in Swaziland to 53,000 in Malawi and Uganda. Using Spectrum data, the number ranged from 7600 in Lesotho to 66,000 in Zimbabwe. The numbers of CLHIV on ART across the 3 sources were roughly comparable, demonstrated by the PHIA Uncertainty Bounds containing both Spectrum and PEPFAR numbers in all countries, except in Malawi. In Malawi, the Uncertainty Bounds of PHIA included the PEPFAR estimate, but the number of CLHIV on ART from Spectrum was lower than the lower bound of the PHIA Uncertainty Bounds.

TABLE 2

TABLE 2

The point estimates of CLHIV on ART from PEPFAR, although within the Uncertainty Bounds of PHIA, were higher than the PHIA point estimates in Lesotho, Tanzania, Uganda, and Zambia and lower in Malawi, Swaziland, and Zimbabwe. The point estimates of CLHIV on ART from Spectrum were higher than PHIA and PEPFAR point estimates in Swaziland, Tanzania, Uganda, Zambia, and Zimbabwe and lower in Lesotho and Malawi.

Comparing PHIA and PEPFAR age-disaggregated number of CLHIV on ART for those younger than 5 years and 5–14 year olds, PEPFAR numbers fell within PHIA Uncertainty Bounds for both, except in the 5–14 age group in Malawi where the PEPFAR data appeared to underrepresent the number of CLHIV on ART.

We next summarized the national treatment coverage gap using the proportion of CLHIV not on ART from PEPFAR, PHIA, and Spectrum (Table 3 and Fig. 1). Across the 7 countries, the overall treatment gap was 49% (UB: 43%–55%), 48% (UB: 25%–60%), and 38% (UB: 24%–47%) using PHIA, PEPFAR, and Spectrum data, respectively. Uncertainty Bounds for the treatment gap overlapped across the 3 sources, except in Lesotho where the Spectrum estimates suggested a significantly higher treatment gap compared with PHIA and PEPFAR estimates. Averaging across the 3 sources, Tanzania had the highest treatment gap at 57% (UB: 36%–70%), whereas Swaziland had the lowest treatment gap at 30% (UB: 17%–40%).

TABLE 3

TABLE 3

FIGURE 1

FIGURE 1

We further disaggregated the treatment gap by younger (0–4 years) and older (5–14 years) age groups using PHIA and PEPFAR data. Pooling the 7 countries together, although the Uncertainty Bounds overlapped, the treatment gap was higher in young children at 60% compared with older children at 44% using PHIA data (Table 3). In PEPFAR data, no age trend was observed; the treatment gap among young children was 49% compared with 48% in older children. There was also substantial variation by country. Using PHIA data, younger children had a larger treatment gap than older children in all countries except in Malawi where the trend was reversed and consistent with the trend observed in PEPFAR data. Using PEPFAR data, younger children had a smaller treatment gap than older children in all countries except Tanzania and Uganda where the trend was reversed and consistent with the trend in PHIA data.

We also examined treatment coverage gaps at the regional level using PHIA and PEPFAR data. We assessed whether there are any geographical patterns to the overall gap in coverage found in the national analysis. Table 4 summarizes the distribution of regions by ART coverage for all children and disaggregated by younger (0–4 years) and older (5–14 years) age groups. Overall, 67% and 59% of all regions had lower than 81% ART coverage according to PHIA and PEPFAR data, respectively. Using PHIA data, more regions did not meet the 81% ART coverage target for younger children compared to older children (55% vs. 47%). Using PEPFAR data, on the other hand, there was no age trend as roughly the same number of regions were not meeting the 81% ART coverage target for younger and older children (57% vs. 56%).

TABLE 4

TABLE 4

Back to Top | Article Outline

DISCUSSION

Our study analyzing data from PHIA, PEPFAR, and Spectrum demonstrated that despite increased efforts to diagnose and treat CLHIV, notable treatment gaps among CLHIV remain in the 7 countries examined. With a few exceptions, findings from the PHIA surveys largely confirmed the model-based estimates of the number of children younger than 15 years living with HIV infection as well as program data on CLHIV on ART, both of which are used to set program targets and resource allocation. Still, low levels of precision, particularly for age- and region-disaggregated data, will continue to pose challenges for program planners and implementers.

The overall treatment gap across the 7 countries based on PHIA and PEPFAR was consistent at nearly 50% with Spectrum showing a lower gap at 38%. The Spectrum-based treatment coverage gap was consistently lower compared with PHIA- and PEPFAR-based treatment coverage gaps in all countries except in Lesotho where the Spectrum-based treatment coverage gap was significantly higher than the PHIA and PEPFAR estimates and in Tanzania where Spectrum estimates were nearly identical to the PEPFAR estimate. Furthermore, these results also appear to track the performance toward the 90-90-90 targets in adults aged 15 years and older showing that Swaziland is close to achieving the targets with others lagging behind in varying degrees.23,24

Furthermore, the disaggregated analysis by age group showed inconsistent results between PHIA and PEPFAR with PHIA showing a larger gap in younger children compared with older children in all but 1 country and PEPFAR showing no age trend. The larger gap in younger children observed in PHIA is consistent with the continued challenge in implementing effective and efficient EID programs in many settings. Furthermore, the preferred antiretroviral agent for this age-group, lopinavir/ritonavir, until recently was only available as a liquid formulation and likely stymied program implementation and clinical management.25–30 In Uganda, specifically, the smaller coverage gap in older children may also be a result of early expansion of “Treat All” from children younger than 3 years to all children under 15 years.31,32

In Malawi, on the other hand, both PHIA and PEPFAR analyses demonstrated a lower ART coverage gap among children younger than 5 years. It is not unlikely that the early rollout of Option B+ resulted in general improvement in the pediatric cascade, with new mothers identified as HIV positive and engaged in care, facilitating identification and engagement of their HIV-exposed infants.33 Furthermore, Malawi continued to endorse the use of a nevirapine-based pediatric fixed-dose combination regimen for children younger than 3 years, rather than the World Health Organization–preferred lopinavir/ritonavir-based ART regimen.34 This may have facilitated a more rapid and sustained increase in access to treatment without the operational complexity that comes with providing a poorly palatable liquid formulation with cold chain requirements.

In Lesotho, Swaziland, Zimbabwe, and Zambia, the age trend in the treatment coverage gap was inconsistent between PHIA and PEPFAR—in PHIA, the treatment gap was higher in the younger age group, whereas using PEPFAR data, the younger age group had higher ART coverage. One potential explanation is that program data may not fully account for death or loss-to-follow-up after ART initiation given that infants and young children on ART are at disproportionately higher risk for mortality compared with older children.35–37 Furthermore, the aging of CLHIV on treatment may not be properly reflected in the routine reporting. These issues may cause overestimation of younger children and underestimation of older children on ART in program data. This limitation in program data has been noted in previous studies.38,39

Our analysis confirmed the treatment gap and the need for further intensified efforts to identify and treat CLHIV. Although the national level gap stood at close to 50% when pooling the 7 countries together, the regional level analysis showed 67% of regions using PHIA data or 59% of regions using PEPFAR data had less than 81% ART coverage, suggesting that the national treatment gap is not distributed evenly across all regions of the country but possibly concentrated in specific regions. Regional differences in availability and access to prevention of mother-to-child HIV transmission, EID, and pediatric ART services likely contribute to this variation in the treatment gap and should be addressed to close the treatment gap.

Our analysis had several limitations. First, the treatment coverage estimated from PEPFAR program data may underestimate the total number of children on ART nationally in some countries and thereby overestimate treatment gaps, as PEPFAR does not support all clinics offering ART in the country. This may be a particular problem in countries, such as Lesotho and Zimbabwe, where national governments or other bilaterals support a substantial part of the pediatric treatment scale-up independent of PEPFAR support (see Table S2, Supplemental Digital Content, http://links.lww.com/QAI/B184). However, we did not observe any clear pattern of overestimation of the treatment gap by PEPFAR when comparing estimates with PHIA and Spectrum. For example in both Lesotho and Zimbabwe, PEPFAR-supported clinics accounted for ~75% of the CLHIV on ART. Yet in Lesotho PEPFAR data showed lower treatment gap compared to Spectrum while in Zimbabwe PEPFAR data showed higher treatment gap compared to Spectrum. In Zimbabwe, the 25% of the CLHIV on ART may not account for the reduction in the treatment gap by more than half from 48% in PEPFAR to 22% in Spectrum. These differences are likely caused by a combination of many factors, including limited PEPFAR program coverage and also data quality issues in PEPFAR and/or national program data.

Similarly, as noted in the Methods section, because of incomplete data and changing reporting requirements, we could not always map the PEPFAR quarterly program data back to the PHIA data collection period. We therefore took the most complete data available in the quarter closest to the PHIA data collection period. The PEPFAR coverage, therefore, particularly at regional level, may over- or underrepresent true coverage.

In PHIA surveys, we identified a total of only 591 HIV-positive children across the 7 countries. National age disaggregated estimates and subnational estimates are therefore relatively imprecise evidenced by the large Uncertainty Bounds. Adjusting the self-reported ART coverage estimate with results of the biomarker test for ARVs in blood has, on average, increased estimated ART coverage by 6% among children (unpublished analysis of PHIA data from Malawi, Zambia, Swaziland, and Uganda). The estimates for Zimbabwe, Lesotho, and Tanzania may therefore be underestimated because biomarker data were not yet available at the time of this publication. For infants under 18 months, PHIA conducted DNA polymerase chain reaction test on those who were screened positive by 1 rapid test in the field irrespective of the status of the mother. We cannot rule out the possibility that there were infants who were missed, among those infants exposed to HIV who tested negative on the rapid test. The PHIA-based number of CLHIV therefore may underestimate those younger than 18 months.40 PHIA-based number of CLHIV reflects the age–sex distribution at the national level, published in population projections. To the extent that there is regional-level variability in these distributions, our estimates may be an over- or undercount. In addition, response rates for children averaged at 80% (range across country, 62%–98%). χ2 automatic interaction detector nonresponse adjustments adjusted only for observable differences between respondent and nonrespondent. If there were any unobserved differences, our estimates will be biased.

The models used for the UNAIDS estimates are modified every year in an effort to improve model accuracy. As a result, the comparability to the PHIA results may change from year to year. The ministries of health–compiled ART data entered into Spectrum are likely to suffer from quality challenges, which could lead to overcounts in some countries and undercounts in other countries. Finally, as PHIA and Spectrum Uncertainty Bounds were calculated using different methods, a statistical comparison of the 2 estimates was beyond the scope of this article.

This study is the first examination of pediatric data that pools data from modeled, survey, and routine program data across multiple countries in east and southern Africa. There is still more work to be done in the post-ACT era in the 7 countries described in this article. Too many children are still not being identified and treated in the many communities across Sub-Saharan Africa. Although the ACT initiative successfully enhanced case finding, linkage to care, and treatment coverage, there is continued need to maintain and, in some situations, escalate coverage of routine programs, including provider-initiated testing and counseling for in- and outpatients, TB patients, and patients in nutritional clinics, as well as expanding comprehensive EID and strengthening the linkage between diagnosis and treatment.41,42 Furthermore, accelerating efforts to prevent new infections among pregnant and breast-feeding women, and ensuring all HIV-positive women are successfully engaged in care and receive ART is the most effective way to prevent new pediatric infections.

Back to Top | Article Outline

REFERENCES

1. UNAIDS. HIV estimates with uncertainty bounds, 1990–2016. 2017. Available at: aidsinfo.unaids.org.
2. PEPFAR. Accelerating Children's HIV/AIDS treatment (ACT) initiative. 2014. Available at: https://www.pepfar.gov/partnerships/ppp/234538.htm. Accessed March 16, 2018.
3. Siberry G. What have we learned from the ACT initiative—identifying and treatment children with HIV infection. 2017.
4. ICAP. PHIA project. 2018. Available at: http://phia.icap.columbia.edu/. Accessed March 23, 2018.
5. UNAIDS. National HIV estimates file. 2018. Available at: http://www.unaids.org/en/dataanalysis/datatools/spectrum-epp. Accessed January 15, 2018.
6. PEPFAR. Program results achieved through PEPFAR support. Available at: https://data.pepfar.net/. Accessed February 15, 2018.
7. Stover J, Brown T, Puckett R, et al. Updates to the spectrum/estimations and projections package model for estimating trends and current values for key HIV indicators. AIDS. 2017;31(suppl 1):S5–S11.
8. Mahy M, Penazzato M, Ciaranello A, et al. Improving estimates of children living with HIV from the spectrum AIDS impact model. AIDS. 2017;31(Suppl 1):S13–S22.
9. Bollinger L, DeCormier Plosky W, Pretorious C, et al. Spectrum manual. Spectrum system of policy models. Available at: http://www.avenirhealth.org/software-spectrum.php. Accessed March 23, 2018.
10. UNAIDS. 90–90–90-An ambitious treatment target to help end the AIDS epidemic. 2014. Available at: http://www.unaids.org/en/resources/documents/2017/90-90-90. Accessed March 19, 2018.
11. Saito S, Duong YT, Metz M, et al. Returning HIV-1 viral load results to participant-selected health facilities in national population-based HIV impact assessment (PHIA) household surveys in three sub-Saharan African Countries, 2015 to 2016. J Int AIDS Soc. 2017;20(suppl 7):19–25.
12. Brown K, Williams DB, Kinchen S, et al. Status of HIV epidemic control among adolescent girls and young women aged 15–24 years—seven african countries, 2015-2017. MMWR Morb Mortal Wkly Rep. 2018;67:29–32.
13. SI-CHAID [Computer Program]. Belmont, MA: Statistical Innovations.
14. ZIMSTAT. Population Projections Thematic Report. 2017. Available at: http://www.zimstat.co.zw/sites/default/files/img/publications/Census/population_projection.pdf. Accessed March 20, 2018.
15. Malawi National Bureau of Statistics. Malawi population projections. 2008. Available at: http://www.nsomalawi.mw/images/stories/data_on_line/demography/census_2008/Main%20Report/ThematicReports/Population%20Projections%20Malawi.pdf. Accessed March 20, 2018.
16. Kingdom of Swaziland Central Statistical Office. 2010 census of population and housing: population and demographic projections 2011–2035. 2012. Available at: http://www.zamstats.gov.zm/index.php/publications/category/55-census-projections?download=822:zambia-census-projection-2011-2035. Accessed March 21, 2018.
17. Kingdom of Swaziland Central Statistical Office. Swaziland population projections 2007–2030. 2007. Available at: http://swazistats.org.sz/images/reports/Swaziland%20Population%20Projections%202007_2030%20Report.pdf. Accessed March 1, 2018.
18. Uganda Bureau of Statistics. National population projections 2015–2020. 2015. Available at: http://www.ubos.org/onlinefiles/uploads/ubos/census_projections/National%20by%20single%20year%202015-2020.xlsx. Accessed March 20, 2018.
19. Lesotho Bureau of Statistics. 2016 census summary key findings. 2017. Available at: http://www.bos.gov.ls/2016%20Summary%20Key%20Findings.pdf. Accessed March 20, 2018.
20. UN DESA/Population Division. World population prospects 2017. 2018. Available at: https://esa.un.org/unpd/wpp/Download/Standard/Population/. Accessed March 21, 2018.
21. PEPFAR. PEPFAR fact view analytic datasets [database online]. Available at: http://www.datim.org. Accessed February 24, 2018.
22. PEPFAR. Interagency collaborative for program improvement (ICPI) fact view data sets. PEPFAR, trans. ICPI_FactView_PSNU_20171222_v2_2 ed2018. 2018.
23. Tanzania HIV impact survey (this) 2016–2017. Summary sheet: preliminary findings. 2017. Available at: http://phia.icap.columbia.edu/wp-content/uploads/2017/11/Tanzania_SummarySheet_A4.English.v19.pdf. Accessed April 27, 2018.
24. Swaziland HIV incidence measurement survey 2: a population-based HIV impact assessment (SHIMS) 2016–2017. Summary sheet: preliminary findings. 2017. Available at: http://phia.icap.columbia.edu/wp-content/uploads/2017/11/Swaziland_new.v8.pdf. Accessed April 27, 2018.
25. Haas AD, Tenthani L, Msukwa MT, et al. Retention in care during the first 3 years of antiretroviral therapy for women in Malawi's option B+ programme: an observational cohort study. Lancet HIV. 2016;3:e175–182.
26. CDC. Impact of an innovative approach to prevent mother-to-child transmission of HIV–Malawi, July 2011-September 2012. MMWR Morb Mortal Wkly Rep. 2013;62:148–151.
27. Essajee S, Bhairavabhotla R, Penazzato M, et al. Scale-up of early infant HIV diagnosis and improving access to pediatric HIV care in global plan countries: past and future perspectives. J Acquir Immune Defic Syndr. 2017;75(suppl 1):S51–S58.
28. Penazzato M, Amzel A, Abrams EJ, et al. Pediatric treatment scale-up: the unfinished agenda of the global plan. J Acquir Immune Defic Syndr. 2017;75(suppl 1):S59–S65.
29. UNICEF. For every child, end AIDS: seventh stocktaking report, 2016. Available at: https://www.unicef.org/publications/index_93427.html. Accessed April 14, 2018.
30. Penazzato M, Gnanashanmugam D, Rojo P, et al. Optimizing research to speed up availability of pediatric antiretroviral drugs and formulations. Clin Infect Dis. 2017;64:1597–1603.
31. Barlow-Mosha L, Musiime V, Davies MA, et al. Universal antiretroviral therapy for HIV-infected children: a review of the benefits and risks to consider during implementation. J Int AIDS Soc. 2017;20:21552.
32. Elyanu P, Magongo EN, Asire B, et al. Impact of Implementing “Test and Treat” Policy on Pediatric ART Enrolments and Coverage in Uganda. Presented at the International AIDS Society (IAS) Conference; Vancouver, Canada; 2015.
33. Kim MH, Ahmed S, Hosseinipour MC, et al. Brief report: impact of option B+ on the infant PMTCT cascade in lilongwe, Malawi. J Acquir Immune Defic Syndr. 2015;70:99–103.
34. World Health Organization. Antiretroviral therapy for HIV infection in infants and children: recommendations for a public health approach: 2010 revision. 2010. Available at: http://www.who.int/hiv/pub/paediatric/infants2010/en/. Accessed April 13, 2018.
35. McNairy ML, Lamb MR, Carter RJ, et al. Retention of HIV-infected children on antiretroviral treatment in HIV care and treatment programs in Kenya, Mozambique, Rwanda, and Tanzania. J Acquir Immune Defic Syndr. 2013;62:e70–e81.
36. Teasdale CA, Yang J, Thome B, et al. Outcomes among children enrolled in HIV care in Mozambique 2009–2013. Pediatr Infect Dis J. 2016;35:1117–1125.
37. Teasdale CA, Sogaula N, Yuengling KA, et al. High risk of loss to follow-up among South African children on ART during transfer, a retrospective cohort analysis with community tracing. J Int AIDS Soc. 2017;20:21748.
38. Braitstein P, Katshcke A, Shen C, et al. Retention of HIV-infected and HIV-exposed children in a comprehensive HIV clinical care programme in Western Kenya. Trop Med Int Health. 2010;15:833–841.
39. Makombe SD, Hochgesang M, Jahn A, et al. Assessing the quality of data aggregated by antiretroviral treatment clinics in Malawi. Bull World Health Organ. 2008;86:310–314.
40. Rutherford GW, McFarland W, Spindler H, et al. Public health triangulation: approach and application to synthesizing data to understand national and local HIV epidemics. BMC public health. 2010;10:447.
41. Kiyaga C, Urick B, Fong Y, et al. Where have all the children gone? High HIV prevalence in infants attending nutrition and inpatient entry points. J Int AIDS Soc. 2018;21:1–7.
42. Simms V, Dauya E, Dakshina S, et al. Community burden of undiagnosed HIV infection among adolescents in Zimbabwe following primary healthcare-based provider-initiated HIV testing and counselling: a cross-sectional survey. PLoS Med. 2017;14:e1002360.
Keywords:

pediatric HIV; strategic information; data triangulation; pediatric antiretroviral therapy

Supplemental Digital Content

Back to Top | Article Outline
Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved.