Assessing and adjusting for differences between HIV prevalence estimates derived from national population-based surveys and antenatal care surveillance, with applications for Spectrum 2013

Triangulation of Routine Antenatal HIV Prevalence Data and Adjusted HIV Estimates in Mozambique

BACKGROUND

Routine health system data are central to monitoring HIV trends. In Mozambique, the reported number of women receiving antenatal care (ANC) and antiretroviral therapy for prevention of mother-to-child transmission (PMTCT) has exceeded the Spectrum-estimated number of pregnant women since 2017. In some provinces, reported HIV prevalence in pregnant women has declined faster than epidemiologically plausible. We hypothesized that these issues are linked and caused by programmatic overenumeration of HIV-negative pregnant women at ANC.

METHODS

We triangulated program-reported ANC client numbers with survey-based fertility estimates and facility birth data adjusted for the proportion of facility births. We used survey-reported ANC attendance to produce adjusted time series of HIV prevalence in pregnant women, adjusted for hypothesized program double counting. We calibrated the Spectrum HIV estimation models to adjusted HIV prevalence data to produce adjusted adult and pediatric HIV estimates.

RESULTS

ANC client numbers were not consistent with facility birth data or modeled population estimates indicating ANC data quality issues in all provinces. Adjusted provincial ANC HIV prevalence in 2021 was median 45% [interquartile range 35%-52% or 2.3 percentage points (interquartile range 2.5-3.5)] higher than reported HIV prevalence. In 2021, calibrating to adjusted antenatal HIV prevalence lowered PMTCT coverage to less than 100% in most provinces and increased the modeled number of new child infections by 35%. The adjusted results better reconciled adult and pediatric antiretroviral treatment coverage and antenatal HIV prevalence with regional fertility estimates.

CONCLUSIONS

Adjusting HIV prevalence in pregnant women using nationally representative household survey data on ANC attendance produced estimates more consistent with surveillance data. The number of children living with HIV in Mozambique has been substantially underestimated because of biased routine ANC prevalence. Renewed focus on HIV surveillance among pregnant women would improve PMTCT coverage and pediatric HIV estimates.

Mapping age- and sex-specific HIV prevalence in adults in sub-Saharan Africa, 2000-2018

BACKGROUND

Human immunodeficiency virus and acquired immune deficiency syndrome (HIV/AIDS) is still among the leading causes of disease burden and mortality in sub-Saharan Africa (SSA), and the world is not on track to meet targets set for ending the epidemic by the Joint United Nations Programme on HIV/AIDS (UNAIDS) and the United Nations Sustainable Development Goals (SDGs). Precise HIV burden information is critical for effective geographic and epidemiological targeting of prevention and treatment interventions. Age- and sex-specific HIV prevalence estimates are widely available at the national level, and region-wide local estimates were recently published for adults overall. We add further dimensionality to previous analyses by estimating HIV prevalence at local scales, stratified into sex-specific 5-year age groups for adults ages 15-59 years across SSA.

METHODS

We analyzed data from 91 seroprevalence surveys and sentinel surveillance among antenatal care clinic (ANC) attendees using model-based geostatistical methods to produce estimates of HIV prevalence across 43 countries in SSA, from years 2000 to 2018, at a 5 × 5-km resolution and presented among second administrative level (typically districts or counties) units.

RESULTS

We found substantial variation in HIV prevalence across localities, ages, and sexes that have been masked in earlier analyses. Within-country variation in prevalence in 2018 was a median 3.5 times greater across ages and sexes, compared to for all adults combined. We note large within-district prevalence differences between age groups: for men, 50% of districts displayed at least a 14-fold difference between age groups with the highest and lowest prevalence, and at least a 9-fold difference for women. Prevalence trends also varied over time; between 2000 and 2018, 70% of all districts saw a reduction in prevalence greater than five percentage points in at least one sex and age group. Meanwhile, over 30% of all districts saw at least a five percentage point prevalence increase in one or more sex and age group.

CONCLUSIONS

As the HIV epidemic persists and evolves in SSA, geographic and demographic shifts in prevention and treatment efforts are necessary. These estimates offer epidemiologically informative detail to better guide more targeted interventions, vital for combating HIV in SSA.

Challenges in estimating HIV prevalence trends and geographical variation in HIV prevalence using antenatal data: Insights from mathematical modelling

BACKGROUND

HIV prevalence data among pregnant women have been critical to estimating HIV trends and geographical patterns of HIV in many African countries. Although antenatal HIV prevalence data are known to be biased representations of HIV prevalence in the general population, mathematical models have made various adjustments to control for known sources of bias, including the effect of HIV on fertility, the age profile of pregnant women and sexual experience.

METHODS AND FINDINGS

We assessed whether assumptions about antenatal bias affect conclusions about trends and geographical variation in HIV prevalence, using simulated datasets generated by an agent-based model of HIV and fertility in South Africa. Results suggest that even when controlling for age and other previously-considered sources of bias, antenatal bias in South Africa has not been constant over time, and trends in bias differ substantially by age. Differences in the average duration of infection explain much of this variation. We propose an HIV duration-adjusted measure of antenatal bias that is more stable, which yields higher estimates of HIV incidence in recent years and at older ages. Simpler measures of antenatal bias, which are not age-adjusted, yield estimates of HIV prevalence and incidence that are too high in the early stages of the HIV epidemic, and that are less precise. Antenatal bias in South Africa is substantially greater in urban areas than in rural areas.

CONCLUSIONS

Age-standardized approaches to defining antenatal bias are likely to improve precision in model-based estimates, and further recency adjustments increase estimates of HIV incidence in recent years and at older ages. Incompletely adjusting for changing antenatal bias may explain why previous model estimates overstated the early HIV burden in South Africa. New assays to estimate the fraction of HIV-positive pregnant women who are recently infected could play an important role in better estimating antenatal bias.

Analysis of HIV/AIDS Epidemic and Socioeconomic Factors in Sub-Saharan Africa

Sub-Saharan Africa has been the epicenter of the outbreak since the spread of acquired immunodeficiency syndrome (AIDS) began to be prevalent. This article proposes several regression models to investigate the relationships between the HIV/AIDS epidemic and socioeconomic factors (the gross domestic product per capita, and population density) in ten countries of Sub-Saharan Africa, for 2011–2016. The maximum likelihood method was used to estimate the unknown parameters of these models along with the Newton–Raphson procedure and Fisher scoring algorithm. Comparing these regression models, there exist significant spatiotemporal non-stationarity and auto-correlations between the HIV/AIDS epidemic and two socioeconomic factors. Based on the empirical results, we suggest that the geographically and temporally weighted Poisson autoregressive (GTWPAR) model is more suitable than other models, and has the better fitting results.

Mapping HIV prevalence in sub-Saharan Africa between 2000 and 2017

HIV/AIDS is a leading cause of disease burden in sub-Saharan Africa. Existing evidence has demonstrated that there is substantial local variation in the prevalence of HIV; however, subnational variation has not been investigated at a high spatial resolution across the continent. Here we explore within-country variation at a 5 × 5-km resolution in sub-Saharan Africa by estimating the prevalence of HIV among adults (aged 15-49 years) and the corresponding number of people living with HIV from 2000 to 2017. Our analysis reveals substantial within-country variation in the prevalence of HIV throughout sub-Saharan Africa and local differences in both the direction and rate of change in HIV prevalence between 2000 and 2017, highlighting the degree to which important local differences are masked when examining trends at the country level. These fine-scale estimates of HIV prevalence across space and time provide an important tool for precisely targeting the interventions that are necessary to bringing HIV infections under control in sub-Saharan Africa.

Updates to the Spectrum/Estimations and Projections Package model for estimating trends and current values for key HIV indicators

BACKGROUND

The Spectrum model is used by national programs and UNAIDS to prepare annual estimates of the status of the HIV epidemic in 160 countries. The model and assumptions are updated regularly under the guidance of the UNAIDS Reference Group on Estimates, Modelling and Projections in response to new data, studies, and program needs. This study describes the most recent updates for the 2016 round of estimates.

METHODS

Meetings of the UNAIDS Reference Group include individuals with extensive knowledge of HIV programs, research, statistics, and public policy. The Reference Group also collaborates with other institutions (such as the United Nations Population Division and the US Census Bureau) and projects (such as the ALPHA Network and IeDEA Consortium) to ensure that latest methods and data are used in the preparation of the annual estimates. In the past year new methods and data have been introduced for pediatric estimates, incidence fitting, and ART mortality (described elsewhere in this supplement).

RESULTS

The 2016 version of Spectrum includes a number of other enhancements, including updated demographic data from the United Nations Population Division, program options for treatment (treat all), and programs to prevent mother-to-child transmission (option B+), improved methods to aggregate uncertainty to regional and global levels, several options for generating incidence trends, adjustments to the Estimations and Projections Package model to better incorporate aging effects, adjustments to account for the changing bias in prevalence from antenatal clinic surveillance, and an option to fit incidence among all adults 15+ in addition to 15-49.

CONCLUSION

The new methods and data implemented in the 2016 version of Spectrum allow national programs more flexibility in describing their programs and improve the estimates of key indicators and their uncertainty.

Empirical validation of the UNAIDS Spectrum model for subnational HIV estimates: case-study of children and adults in Manicaland, Zimbabwe

BACKGROUND

More cost-effective HIV control may be achieved by targeting geographical areas with high infection rates. The AIDS Impact model of Spectrum - used routinely to produce national HIV estimates - could provide the required subnational estimates but is rarely validated with empirical data, even at a national level.

DESIGN

The validity of the Spectrum model estimates were compared with empirical estimates.

METHODS

Antenatal surveillance and population survey data from a population HIV cohort study in Manicaland, East Zimbabwe, were input into Spectrum 5.441 to create a simulation representative of the cohort population. Model and empirical estimates were compared for key demographic and epidemiological outcomes. Alternative scenarios for data availability were examined and sensitivity analyses were conducted for model assumptions considered important for subnational estimates.

RESULTS

Spectrum estimates generally agreed with observed data but HIV incidence estimates were higher than empirical estimates, whereas estimates of early age all-cause adult mortality were lower. Child HIV prevalence estimates matched well with the survey prevalence among children. Estimated paternal orphanhood was lower than empirical estimates. Including observations from earlier in the epidemic did not improve the HIV incidence model fit. Migration had little effect on observed discrepancies - possibly because the model ignores differences in HIV prevalence between migrants and residents.

CONCLUSION

The Spectrum model, using subnational surveillance and population data, provided reasonable subnational estimates although some discrepancies were noted. Differences in HIV prevalence between migrants and residents may need to be captured in the model if applied to subnational epidemics.

The effects of HIV on fertility by infection duration: evidence from African population cohorts before antiretroviral treatment availability

OBJECTIVES

To estimate the relationship between HIV natural history and fertility by duration of infection in east and southern Africa before the availability of antiretroviral therapy and assess potential biases in estimates of age-specific subfertility when using retrospective birth histories in cross-sectional studies.

DESIGN

Pooled analysis of prospective population-based HIV cohort studies in Masaka (Uganda), Kisesa (Tanzania) and Manicaland (Zimbabwe).

METHODS

Women aged 15-49 years who had ever tested for HIV were included. Analyses were censored at antiretroviral treatment roll-out. Fertility rate ratios were calculated to see the relationship of duration of HIV infection on fertility, adjusting for background characteristics. Survivorship and misclassification biases on age-specific subfertility estimates from cross-sectional surveys were estimated by reclassifying person-time from the cohort data to simulate cross-sectional surveys and comparing fertility rate ratios with true cohort results.

RESULTS

HIV-negative and HIV-positive women contributed 15 440 births and 86 320 person-years; and 1236 births and 11 240 000 person-years, respectively, to the final dataset. Adjusting for age, study site and calendar year, each additional year since HIV seroconversion was associated with a 0.02 (95% confidence interval 0.01-0.03) relative decrease in fertility for HIV-positive women. Survivorship and misclassification biases in simulated retrospective birth histories resulted in modest underestimates of subfertility by 2-5% for age groups 20-39 years.

CONCLUSION

Longer duration of infection is associated with greater relative fertility reduction for HIV-positive women. This should be considered when creating estimates for HIV prevalence among pregnant women and prevention of mother-to-child transmission need over the course of the HIV epidemic and antiretroviral treatment scale up.

Statistical models for incorporating data from routine HIV testing of pregnant women at antenatal clinics into HIV/AIDS epidemic estimates

OBJECTIVE

HIV prevalence data collected from routine HIV testing of pregnant women at antenatal clinics (ANC-RT) are potentially available from all facilities that offer testing services to pregnant women and can be used to improve estimates of national and subnational HIV prevalence trends. We develop methods to incorporate these new data source into the Joint United Nations Programme on AIDS Estimation and Projection Package in Spectrum 2017.

METHODS

We develop a new statistical model for incorporating ANC-RT HIV prevalence data, aggregated either to the health facility level (site-level) or regionally (census-level), to estimate HIV prevalence alongside existing sources of HIV prevalence data from ANC unlinked anonymous testing (ANC-UAT) and household-based national population surveys. Synthetic data are generated to understand how the availability of ANC-RT data affects the accuracy of various parameter estimates.

RESULTS

We estimate HIV prevalence and additional parameters using both ANC-RT and other existing data. Fitting HIV prevalence using synthetic data generally gives precise estimates of the underlying trend and other parameters. More years of ANC-RT data should improve prevalence estimates. More ANC-RT sites and continuation with existing ANC-UAT sites may improve the estimate of calibration between ANC-UAT and ANC-RT sites.

CONCLUSION

We have proposed methods to incorporate ANC-RT data into Spectrum to obtain more precise estimates of prevalence and other measures of the epidemic. Many assumptions about the accuracy, consistency, and representativeness of ANC-RT prevalence underlie the use of these data for monitoring HIV epidemic trends and should be tested as more data become available from national ANC-RT programs.

Measuring the Impact of Antiretroviral Therapy Roll-Out on Population Level Fertility in Three African Countries

Background

UNAIDS official estimates of national HIV prevalence are based on trends observed in antenatal clinic surveillance, after adjustment for the reduced fertility of HIV positive women. Uptake of ART may impact on the fertility of HIV positive women, implying a need to re-estimate the adjustment factors used in these calculations. We analyse the effect of antiretroviral therapy (ART) provision on population-level fertility in Southern and East Africa, comparing trends in HIV infected women against the secular trends observed in uninfected women.

Methods

We used fertility data from four community-based demographic and HIV surveillance sites: Kisesa (Tanzania), Masaka and Rakai (Uganda) and uMkhanyakude (South Africa). All births to women aged 15–44 years old were included in the analysis, classified by mother’s age and HIV status at time of birth, and ART availability in the community. Calendar time period of data availability relative to ART Introduction varied across the sites, from 5 years prior to ART roll-out, to 9 years after. Calendar time was classified according to ART availability, grouped into pre ART, ART introduction (available in at least one health facility serving study site) and ART available (available in all designated health facilities serving study site). We used Poisson regression to calculate age adjusted fertility rate ratios over time by HIV status, and investigated the interaction between ART period and HIV status to ascertain whether trends over time were different for HIV positive and negative women.

Results

Age-adjusted fertility rates declined significantly over time for HIV negative women in all four studies. However HIV positives either had no change in fertility (Masaka, Rakai) or experienced a significant increase over the same period (Kisesa, uMkhanyakude). HIV positive fertility was significantly lower than negative in both the pre ART period (age adjusted fertility rate ratio (FRR) range 0.51 95%CI 0.42–0.61 to 0.73 95%CI 0.64–0.83) and when ART was widely available (FRR range 0.57 95%CI 0.52–0.62 to 0.83 95%CI 0.78–0.87), but the difference has narrowed. The interaction terms describing the difference in trends between HIV positives and negatives are generally significant.

Conclusions

Differences in fertility between HIV positive and HIV negative women are narrowing over time as ART becomes more widely available in these communities. Routine adjustment of ANC data for estimating national HIV prevalence will need to allow for the impact of treatment.

Do HIV prevalence trends in antenatal clinic surveillance represent trends in the general population in the antiretroviral therapy era? The case of Manicaland, East Zimbabwe

OBJECTIVE

National estimates of HIV trends in generalized epidemics rely on HIV prevalence data from antenatal clinic (ANC) surveillance. We investigate whether HIV prevalence trends in ANC data reflect trends in men and women in the general population during the scale-up of antiretroviral therapy (ART) in Manicaland, Zimbabwe.

METHODS

Trends in HIV prevalence in local ANC attendees and adults aged 15-49 years in towns, agricultural estates, and villages were compared using five rounds of parallel ANC (N = 1200) and general-population surveys (N = 10 000) and multivariable log-linear regression. Changes in the age pattern of HIV prevalence and the age distribution of ANC attendees were compared with those in the general population. Age-specific pregnancy prevalence rates were compared by HIV infection and ART status.

RESULTS

Cumulatively, from 1998-2000 to 2009-2011, HIV prevalence fell by 60.0% (95% confidence interval, 51.1-67.3%) in ANC surveillance data and by 34.3% (30.8-37.7%) in the general population. Most of the difference arose following the introduction of ART (2006-2011). The estates and villages reflected this overall pattern but HIV prevalence in the towns was lower at local ANCs than in the general population, largely because of attendance by pregnant women from outlying (lower prevalence) areas. The ageing of people living with HIV in the general population (52.4% aged >35 years, 2009-2011) was under-represented in the ANC data (12.6%) because of lower fertility in older and HIV-infected women.

CONCLUSION

After the introduction of ART in Manicaland, HIV prevalence declined more steeply in ANC surveillance data than in the general population. Models used for HIV estimates must reflect this change in bias.

Improvements in prevalence trend fitting and incidence estimation in EPP 2013

OBJECTIVE

Describe modifications to the latest version of the Joint United Nations Programme on AIDS (UNAIDS) Estimation and Projection Package component of Spectrum (EPP 2013) to improve prevalence fitting and incidence trend estimation in national epidemics and global estimates of HIV burden.

METHODS

Key changes made under the guidance of the UNAIDS Reference Group on Estimates, Modelling and Projections include: availability of a range of incidence calculation models and guidance for selecting a model; a shift to reporting the Bayesian median instead of the maximum likelihood estimate; procedures for comparison and validation against reported HIV and AIDS data; incorporation of national surveys as an integral part of the fitting and calibration procedure, allowing survey trends to inform the fit; improved antenatal clinic calibration procedures in countries without surveys; adjustment of national antiretroviral therapy reports used in the fitting to include only those aged 15-49 years; better estimates of mortality among people who inject drugs; and enhancements to speed fitting.

RESULTS

The revised models in EPP 2013 allow closer fits to observed prevalence trend data and reflect improving understanding of HIV epidemics and associated data.

CONCLUSION

Spectrum and EPP continue to adapt to make better use of the existing data sources, incorporate new sources of information in their fitting and validation procedures, and correct for quantifiable biases in inputs as they are identified and understood. These adaptations provide countries with better calibrated estimates of incidence and prevalence, which increase epidemic understanding and provide a solid base for program and policy planning.

Embracing different approaches to estimating HIV incidence, prevalence and mortality

Background:

Joint United Nations Programme on HIV/AIDS (UNAIDS) and Murray et al. have both produced sets of estimates for worldwide HIV incidence, prevalence and mortality. Understanding differences in these estimates can strengthen the interpretation of each.

Methods:

We describe differences in the two sets of estimates. Where possible, we have drawn on additional published data to which estimates can be compared.

Findings:

UNAIDS estimates that there were 6 million more people living with HIV (PLHIV) in 2013 (35 million) compared with the Murray et al. estimates (29 million). Murray et al. estimate that new infections and AIDS deaths have declined more gradually than does UNAIDS. Just under one third of the difference in PLHIV is in Africa, where Murray et al. have relied more on estimates of adult mortality trends than on data on survival times. Another third of the difference is in North America, Europe, Central Asia and Australasia. Here Murray et al. estimates of new infections are substantially lower than the number of new HIV/AIDS diagnoses reported by countries, whereas published UNAIDS estimate tend to be greater. The remaining differences are in Latin America and Asia where the data upon which the UNAIDS methods currently rely are more sparse, whereas the mortality data leveraged by Murray et al. may be stronger. In this region, however, anomalies appear to exist between the both sets of estimates and other data.

Interpretation:

Both estimates indicate that approximately 30 million PLHIV and that antiretroviral therapy has driven large reductions in mortality. Both estimates are useful but show instructive discrepancies with additional data sources. We find little evidence to suggest that either set of estimates can be considered systematically more accurate. Further work should seek to build estimates on as wide a base of data as possible.

Incorporating incidence information within the UNAIDS Estimation and Projection Package framework: a study based on simulated incidence assay data

OBJECTIVES

The objective of this study is to extend the UNAIDS incidence estimation model, the UNAIDS Estimation and Projection Package (EPP), so that it can incorporate data from incidence assays.

METHODS

We propose combining the likelihood of the incidence assay data with the likelihood of other data, in a manner that is consistent with the biomarker-based incidence estimator using incidence assay data. Two calibrating parameters specify the performance of the incidence assays: the false recent rate and the mean duration of recent infection. We then use synthetic data, based on prevalence data obtained from antenatal clinic surveillances, and in some cases household surveys, from 24 countries, to examine the impact of including incidence assay data, under circumstances wherein the incidence assay data imply the same or a different incidence rate as that inferred from the prevalence data alone, and wherein incorrect calibrating parameters for the incidence assay data are used.

RESULTS

Using incidence assay data, in addition to prevalence data, can improve estimate by narrowing uncertainty intervals in derived HIV incidence estimates, and by providing information on levels or trends in incidence that were not apparent in the prevalence data alone. However, the effect is relatively modest if the sample size of the incidence assay survey is small and results can be biased if the calibrating parameters for the incidence assay data are not known accurately.

CONCLUSION

Incorporating information from incidence assays in the manner proposed has the potential to improve estimates. Further work will examine in more detail the circumstances under which the contribution of incidence assay data would be most valuable.

Learning lessons from operational research in infectious diseases: can the same model be used for noncommunicable diseases in developing countries?

About three-quarters of global deaths from noncommunicable diseases (NCDs) occur in developing countries. Nearly a third of these deaths occur before the age of 60 years. These deaths are projected to increase, fueled by such factors as urbanization, nutrition transition, lifestyle changes, and aging. Despite this burden, there is a paucity of research on NCDs, due to the higher priority given to infectious disease research. Less than 10% of research on cardiovascular diseases comes from developing countries. This paper assesses what lessons from operational research on infectious diseases could be applied to NCDs. The lessons are drawn from the priority setting for research, integration of research into programs and routine service delivery, the use of routine data, rapid-assessment survey methods, modeling, chemoprophylaxis, and the translational process of findings into policy and practice. With the lines between infectious diseases and NCDs becoming blurred, it is justifiable to integrate the programs for the two disease groups wherever possible, eg, screening for diabetes in tuberculosis. Applying these lessons will require increased political will, research capacity, ownership, use of local expertise, and research funding.