Spatial distribution and characteristics of HIV clusters in Ethiopia

Spatial mapping and predictors of ever-tested for HIV in adolescent girls and young women in Ethiopia

Background

Adolescent girls and young women (AGYW) are expected to be healthy in life. However, the unique health challenges faced by AGYW include unsafe sex practices and substance abuse. Only 46.3% of AGYW in Africa are aware of their HIV status, and difficulties are underlined in HIV testing among adolescents and young people. To demarcate the areas with low and high HIV testing, this study aimed to map predictors of ever-tested for HIV among adolescent girls and young women in Ethiopia.

Methods

Secondary data analysis was conducted using the dataset from the 2016 Ethiopia Demographic and Health Survey (EHDS). We conducted spatial autocorrelation and Moran's I statistics to investigate the regional variance of HIV being ever-tested in AGYW. In addition, spatial regression analyses such as ordinary least squares (OLS) regression and geographically weighted regression (GWR) were carried out to determine the predictors of being ever-tested for HIV among AGYW.

Results

Addis Ababa, some parts of Amhara, Dire Dawa, Gambela, and Tigray were the primary regions and city administrations for being ever-tested for HIV among AGYW. A lesser proportion of AGYW being ever-tested for HIV was found in Somalia, Afar, Benshangul Gumuz, and southern nations. Spatial regression analyses identified an age range of 15-19 years, being Muslim, having no formal education, having no knowledge about HIV, and experiencing severe stigma as predictors of being ever-tested for HIV among AGYW.

Conclusion

The proportion of AGYW being ever-tested for HIV was high in Addis Ababa, some parts of Amhara, Dire Dawa, Gambela, and Tigray. Spatial regression analyses identified that AGYW aged 15-19 years, having no formal education, having no knowledge about HIV, and experiencing severe community stigma as predictors negatively affecting the proportion of being ever-tested for HIV, while being Muslim was a predictor that positively affected the proportion of being ever-tested for HIV. The governments and other stakeholders should focus on increasing HIV testing among these special groups of the population.

Patterns of basic pneumococcal conjugated vaccine coverage in Ethiopia from 2015 to 2018; further analysis of Ethiopian DHS (2016-2019)

Background

Geographic variation is crucial in spotting performance gaps in immunization programs, including the Pneumococcal Conjugated Vaccine (PCV). This will help speed up targeted vaccination and disease elimination programs in resource-limited countries. Thus, this study aimed to investigate the geographic variation and determinants of PCV vaccination coverage among children aged under five years old in Ethiopia.

Methods

This analysis was carried out based on the 2016 and 2019 nationally representative Ethiopia Demographic and Health Survey (EDHS). We included two surveys of 10,640 children aged 12-23 months. The spatial analysis also covered 645 and 305 clusters with geographical information for both 2016 and 2019, respectively. We explored the spatial distribution, global spatial autocorrelation, spatial interpolation, and Stats Can windows of children with PCV-3 vaccination. P-values were generated using 999 Monte Carlo simulations to identify statistically significant clusters. To understand the coverage of PCV-3 in all areas of the country, we employed the ordinary Kriging interpolation method to estimate the coverage in unsampled areas. We also used hierarchical multivariate logistic regression to identify the factors associated with the utilization of the PCV vaccine (full vaccination).

Results

Except for Addis Ababa, children in all regions have lower odds of receiving all three PCV vaccines compared to the Tigray region. Residence, sex of a child, mother's literacy status, household wealth index, and place of delivery were significant factors associated with receiving the third dose of PCV. Spatial analysis also showed the Somali and Afar regions had the lowest coverage, while the Addis Ababa and Tigray regions had higher coverage in both surveys.

Conclusion

Even though the coverage of the full PCV vaccine improved from 2016 to 2019, variation was observed among regions and between rural and urban areas. The wealth index and educational status of mothers were the most important determinants of PCV vaccine utilization. Hence, the mass campaign might boost coverage in nomadic and semi-nomadic regions and rural areas. Similarly, programs that narrow the gap due to low socioeconomic differences should be formulated and implemented to increase uptake and general coverage.

Use of unsupervised machine learning to characterise HIV predictors in sub-Saharan Africa

INTRODUCTION

Significant regional variations in the HIV epidemic hurt effective common interventions in sub-Saharan Africa. It is crucial to analyze HIV positivity distributions within clusters and assess the homogeneity of countries. We aim at identifying clusters of countries based on socio-behavioural predictors of HIV for screening.

METHOD

We used an agglomerative hierarchical, unsupervised machine learning, approach for clustering to analyse data for 146,733 male and 155,622 female respondents from 13 sub-Saharan African countries with 20 and 26 features, respectively, using Population-based HIV Impact Assessment (PHIA) data from the survey years 2015-2019. We employed agglomerative hierarchical clustering and optimal silhouette index criterion to identify clusters of countries based on the similarity of socio-behavioural characteristics. We analyse the distribution of HIV positivity with socio-behavioural predictors of HIV within each cluster.

RESULTS

Two principal components were obtained, with the first describing 62.3% and 70.1% and the second explaining 18.3% and 20.6% variance of the total socio-behavioural variation in females and males, respectively. Two clusters per sex were identified, and the most predictor features in both sexes were: relationship with family head, enrolled in school, circumcision status for males, delayed pregnancy, work for payment in last 12 months, Urban area indicator, known HIV status and delayed pregnancy. The HIV positivity distribution with these variables was significant within each cluster.

CONCLUSIONS /FINDINGS

The findings provide a potential use of unsupervised machine learning approaches for substantially identifying clustered countries based on the underlying socio-behavioural characteristics.

Novel Location-Based Survey Using Cognitive Interviews to Assess Geographic Networks and Hotspots of Sex and Drug Use: Implementation and Validation Study

BACKGROUND

The Ending the HIV Epidemic initiative in the United States relies on HIV hotspots to identify where to geographically target new resources, expertise, and technology. However, interventions targeted at places with high HIV transmission and infection risk, not just places with high HIV incidence, may be more effective at reducing HIV incidence and achieving health equity.

OBJECTIVE

We described the implementation and validation of a web-based activity space survey on HIV risk behaviors. The survey was intended to collect geographic information that will be used to map risk behavior hotspots as well as the geography of sexual networks in Los Angeles County.

METHODS

The survey design team developed a series of geospatial questions that follow a 3-level structure that becomes more geographically precise as participants move through the levels. The survey was validated through 9 cognitive interviews and iteratively updated based on participant feedback until the saturation of topics and technical issues was reached.

RESULTS

In total, 4 themes were identified through the cognitive interviews: functionality of geospatial questions, representation and accessibility, privacy, and length and understanding of the survey. The ease of use for the geospatial questions was critical as many participants were not familiar with mapping software. The inclusion of well-known places, landmarks, and road networks was critical for ease of use. The addition of a Google Maps interface, which was familiar to many participants, aided in collecting accurate and precise location information. The geospatial questions increased the length of the survey and warranted the inclusion of features to simplify it and speed it up. Using nicknames to refer to previously entered geographic locations limited the number of geospatial questions that appeared in the survey and reduced the time taken to complete it. The long-standing relationship between participants and the research team improved comfort to disclose sensitive geographic information related to drug use and sex. Participants in the cognitive interviews highlighted how trust and inclusive and validating language in the survey alleviated concerns related to privacy and representation.

CONCLUSIONS

This study provides promising results regarding the feasibility of using a web-based mapping survey to collect sensitive location information relevant to ending the HIV epidemic. Data collection at several geographic levels will allow for insights into spatial recall of behaviors as well as future sensitivity analysis of the spatial scale of hotspots and network characteristics. This design also promotes the privacy and comfort of participants who provide location information for sensitive topics. Key considerations for implementing this type of survey include trust from participants, community partners, or research teams to overcome concerns related to privacy and comfort. The implementation of similar surveys should consider local characteristics and knowledge when crafting the geospatial components.

Spatial variation and attributable risk factors of anaemia among young children in Uganda: Evidence from a nationally representative survey

Anaemia continues to be a burden especially in developing countries that not only affects the physical growth and cognitive development of children but also increases their risk to death. Over the past decade, the prevalence of anaemia among Ugandan children has been unacceptably high. Despite this, spatial variation and attributable risk factors of anaemia are not well explored at national level. The study utilized the 2016 Uganda Demographic and Health Survey (UDHS) data with a weighted sample of 3805 children aged 6-59 months. Spatial analysis was performed using ArcGIS version 10.7 and SaTScan version 9.6. This was followed by a multilevel mixed-effects generalized linear model for the analysis of the risk factors. Estimates for population attributable risks (PAR) and fractions (PAF) were also provided using STATA version 17. In the results, intra-cluster correlation coefficient (ICC) indicates that 18% of the total variability of anaemia was due to communities within the different regions. Moran's index further confirmed this clustering (Global Moran's index = 0.17; p-value<0.001). The main hot spot areas of anaemia were Acholi, Teso, Busoga, West Nile, Lango and Karamoja sub-regions. Anaemia prevalence was highest among boy-child, the poor, mothers with no education as well as children who had fever. Results also showed that if all children were born to mothers with higher education or were staying in rich household, the prevalence would be reduced by 14% and 8% respectively. Also having no fever reduces anaemia by 8%. In conclusion, anaemia among young children is significantly clustered in the country with disparities noted across communities within different sub-regions. Policies targeting poverty alleviation, climate change or environment adaptation, food security as well interventions on malaria prevention will help to bridge a gap in the sub regional inequalities of anaemia prevalence.

Spatial distribution and determinants of HIV prevalence among adults in urban Ethiopia: Findings from the Ethiopia Population-based HIV Impact Assessment Survey (2017–2018)

The design and evaluation of national HIV programs often rely on aggregated national data, which may obscure localized HIV epidemics. In Ethiopia, even though the national adult HIV prevalence has decreased, little information is available about local areas and subpopulations. To inform HIV prevention efforts for specific populations, we identified geographic locations and drivers of HIV transmission. We used data from adults aged 15–64 years who participated in the Ethiopian Population-based HIV Impact Assessment survey (October 2017–April 2018). Location-related information for the survey clusters was obtained from the 2007 Ethiopia population census. Spatial autocorrelation of HIV prevalence data were analyzed via a Global Moran’s I test. Geographically weighted regression analysis was used to show the relationship of covariates. The finding indicated that uncircumcised men in certain hotspot towns and divorced or widowed individuals in hotspot woredas/towns might have contributed to the average increase in HIV prevalence in the hotspot areas. Hotspot analysis findings indicated that, localized, context-specific intervention efforts tailored to at-risk populations, such as divorced or widowed women or uncircumcised men, could decrease HIV transmission and prevalence in urban Ethiopia.

Spatial codistribution of HIV, tuberculosis and malaria in Ethiopia

Background

HIV, tuberculosis (TB) and malaria are the three most important infectious diseases in Ethiopia, and sub-Saharan Africa. Understanding the spatial codistribution of these diseases is critical for designing geographically targeted and integrated disease control programmes. This study investigated the spatial overlap and drivers of HIV, TB and malaria prevalence in Ethiopia.

Methods

HIV, TB and malaria data were obtained from different nationwide prevalence surveys, and geospatial covariates were obtained from publicly available sources. A Bayesian model-based geostatistical framework was applied to each survey leveraging the strength of high-resolution spatial covariates to predict continuous disease-specific prevalence surfaces and their codistribution.

Results

The national prevalence was 1.54% (95% CI 1.40 to 1.70) for HIV, 0.39% (95% CI 0.34 to 0.45) for TB and 1.1% (95%CI 0.95 to 1.32) for malaria. Substantial subnational variation was predicted with the highest HIV prevalence estimated in Gambela (4.52%), Addis Ababa (3.52%) and Dire Dawa (2.67%) regions. TB prevalence was highest in Dire Dawa (0.96%) and Gambela (0.88%), while malaria was highest in Gambela (6.1%) and Benishangul-Gumuz (3.8%). Spatial overlap of their prevalence was observed in some parts of the country, mainly Gambela region. Spatial distribution of the diseases was significantly associated with healthcare access, demographic, and climatic factors.

Conclusions

The national distribution of HIV, TB and malaria was highly focal in Ethiopia, with substantial variation at subnational and local levels. Spatial distribution of the diseases was significantly associated with healthcare access, demographic and climatic factors. Spatial overlap of HIV, TB and malaria prevalence was observed in some parts of the country. Integrated control programmes for these diseases should be targeted to these areas with high levels of co-endemicity.

Characterizing and mapping the spatial variability of HIV risk among adolescent girls and young women: A cross-county analysis of population-based surveys in Eswatini, Haiti, and Mozambique

Background

To stem the HIV epidemic among adolescent girls and young women (AGYW), prevention programs must target services towards those most at risk for HIV. This paper investigates approaches to estimate HIV risk and map the spatial heterogeneity of at-risk populations in three countries: Eswatini, Haiti and Mozambique.

Methods

We analyzed HIV biomarker and risk factor data from recent population-based household surveys. We characterized risk using three approaches: complementary log-log regression, latent class analysis (LCA), and presence of at least one risk factor. We calculated the proportion and 95 percent confidence intervals of HIV-negative AGYW at risk across the three methods and employed Chi-square tests to investigate associations between risk classification and HIV status. Using geolocated survey data at enumeration clusters and high-resolution satellite imagery, we applied algorithms to predict the number and proportion of at-risk AGYW at hyperlocal levels.

Results

The any-risk approach yielded the highest proportion of at-risk and HIV-negative AGYW across five-year age bands: 26%-49% in Eswatini, 52%-67% in Haiti, and 32%-84% in Mozambique. Using LCA, between 8%-16% of AGYW in Eswatini, 37%-62% in Haiti, and 56%-80% in Mozambique belonged to a high vulnerability profile. In Haiti and Mozambique, the regression-based profile yielded the lowest estimate of at-risk AGYW. In general, AGYW characterized as “at risk” across the three methods had significantly higher odds of HIV infection. Hyperlocal maps indicated high levels of spatial heterogeneity in HIV risk prevalence and population density of at-risk AGYW within countries.

Conclusion

Characterizing risk among AGYW can help HIV prevention programs better understand the differential effect of multiple risk factors, facilitate early identification of high-risk AGYW, and design tailored interventions. Hyperlocal mapping of these at-risk populations can help program planners target prevention interventions to geographic areas with populations at greatest risk for HIV to achieve maximal impact on HIV incidence reduction.