Formulation and validation of a regional household wealth index for sub-Saharan Africa

A new era in global health assistance requires a focus on efficiently using limited and declining donor funds. This shift requires better evaluation methods to allocate resources effectively. Most evaluations in low- and middle-income countries (LMICs) examine health disparities within countries, but it is also crucial to assess health outcomes at an inter-country level based on national wealth. Cross-country studies support resource reallocation to the neediest nations and help transition programs like HIV responses within countries with better health infrastructure. This paper presents an unsupervised machine learning method, Principal Component Analysis (PCA), applied to household surveys from 15 African countries to create a universal wealth index that allows multiple countries to be compared on a common scale. Our method places households on a regional wealth scale, enabling cross-country comparisons of health indicators. We used a pooled dataset of 136,086 households from 15- Population-based HIV Impact Assessment (PHIA) countries and validated our universal ranking approach against a local wealth indicator adjusted for macroeconomic differences. The results showed coherence between the macroeconomic-adjusted multinational scale and the PCA-created regional scale, supporting the method's usability for regional household rankings. The proposed method relocates households, as citizens of the world, on a regional wealth scale compared to most surveys that rank them by income placements in their local states. The validation results suggest that the direction and magnitude of mobility of households from national to regional scale in both methods were adequately coherent, ensuring the usability of our approach in ranking households regionally. The PCA-created border-agnostic wealth quintiles enable policymakers to optimize their efficiency improvement efforts, which promises superior efficiency gains over the siloed localized efficiency improvements. Our approach, tested on PHIA-participating countries, can be replicated for similar surveys to study utilization patterns and health outcomes globally.

BackgroundThe increase in cell phone ownership in low- and middle-income countries (LMIC) has created an opportunity for low-cost, rapid data collection by calling participants on their cell phones. Cell phones can be mobilized for a myriad of data collection purposes, including surveillance. In LMIC, cell phone–based surveillance has been used to track Ebola, measles, acute flaccid paralysis, and diarrheal disease, as well as noncommunicable diseases. Phone-based surveillance in LMIC is a particularly pertinent, burgeoning approach in the context of the COVID-19 pandemic. Participatory surveillance via cell phone could allow governments to assess burden of disease and complements existing surveillance systems.ObjectiveWe describe the protocol for the LeCellPHIA (Lesotho Cell Phone PHIA) project, a cell phone surveillance system that collects weekly population-based data on influenza-like illness (ILI) in Lesotho by calling a representative sample of a recent face-to-face survey.MethodsWe established a phone-based surveillance system to collect ILI symptoms from approximately 1700 participants who had participated in a recent face-to-face survey in Lesotho, the Population-based HIV Impact Assessment (PHIA) Survey. Of the 15,267 PHIA participants who were over 18 years old, 11,975 (78.44%) consented to future research and provided a valid phone number. We followed the PHIA sample design and included 342 primary sampling units from 10 districts. We randomly selected 5 households from each primary sampling unit that had an eligible participant and sampled 1 person per household. We oversampled the elderly, as they are more likely to be affected by COVID-19. A 3-day Zoom training was conducted in June 2020 to train LeCellPHIA interviewers.ResultsThe surveillance system launched July 1, 2020, beginning with a 2-week enrollment period followed by weekly calls that will continue until September 30, 2022. Of the 11,975 phone numbers that were in the sample frame, 3020 were sampled, and 1778 were enrolled.ConclusionsThe surveillance system will track COVID-19 in a resource-limited setting. The novel approach of a weekly cell phone–based surveillance system can be used to track other health outcomes, and this protocol provides information about how to implement such a system.International Registered Report Identifier (IRRID)DERR1-10.2196/31236

The population-based HIV impact assessment (population-based HIV impact assessments) surveys are among the first to estimate national adult HIV incidence, subnational prevalence of viral load suppression, and pediatric HIV prevalence. We summarize the survey methods implemented in Zimbabwe, Malawi, and Zambia, as well as response rates and quality metrics. Each cross-sectional, household-based survey used a 2-stage cluster design. Survey preparations included sample design, questionnaire development, tablet programming for informed consent and data collection, community mobilization, establishing a network of satellite laboratories, and fieldworker training. Interviewers collected demographic, behavioral, and clinical information using tablets. Blood was collected for home-based HIV testing and counseling (HBTC) and point-of-care CD4+ T-cell enumeration with results immediately returned. HIV-positive blood samples underwent laboratory-based confirmatory testing, HIV incidence testing, RNA polymerase chain reaction (viral load), DNA polymerase chain reaction (early infant diagnosis), and serum antiretroviral drug detection. Data were weighted for survey design, and chi square automatic interaction detection-based methods were used to adjust for nonresponse. Each survey recruited a nationally representative, household-based sample of children and adults over a 6-10-month period in 2015 and 2016. Most (84%-90%) of the 12,000-14,000 eligible households in each country participated in the survey, with 77%-81% of eligible adults completing an interview and providing blood for HIV testing. Among eligible children, 59%-73% completed HIV testing. Across the 3 surveys, 97.8% of interview data were complete and had no errors. Conducting a national population-based HIV impact assessment with immediate return of HIV and other point-of-care test results was feasible, and data quality was high.

Background The wealth index is a commonly-used measure of socio-economic position (SEP) in low- and middle-income countries. The approach arose from Demographic and Health Surveys (DHS) where few other options for SEP-measurement are available. Despite many unanswered questions the wealth index approach, and the methods of wealth index construction used by the DHS, have been widely-adopted by the epidemiological community. This thesis explores the appropriateness of the wealth index as a measure of SEP, using data from the Malawi Integrated Household Survey 2004/5 (IHS2) and Brazil DHS 1996. Main findings 1) The wealth index and consumption expenditure Some proponents of the wealth index claim it to be a reliable and rational proxy for consumption expenditure; a systematic review of the literature demonstrated this to be an unreasonable assumption. Analyses of IHS2 data showed that the agreement of the wealth index with consumption expenditure is largely unaffected by alternative equivalence scales for adjusting consumption expenditure for household size and composition, or by the range of items included in the consumption expenditure aggregate. 2) Are the methods of wealth index construction used by the DHS the most appropriate? The DHS use principal components analysis (PCA) to weight the indicators in a wealth index; issues in the use of PCA were considered and alternative weighting methods explored. When nominal or ordered categorical indicators are used, alternatives to PCA are considered preferable. A single wealth index is often constructed for urban and rural areas together; generating separate indices for each area had little effect on the final index. Agreement with consumption expenditure was lower in rural areas; various approaches to wealth index construction did not alter this. Expanding the range and number of indicators used by DHS wealth indices did not increase agreement with consumption expenditure, but it did reduce 2 the observed urban-rural differences and enhance the ability of the wealth index to differentiate between rural households. 3) What socio-economic processes contribute to the wealth index hierarchy? A key issue for any measure of SEP is its conceptual clarity; analyses of the IHS2 data demonstrate that the socio-economic processes leading to a wealth index hierarchy remain largely unknown, although both household- and community-level factors play a role. 4) Alternatives to the wealth index Potential alternatives to the wealth index were explored, and the consequences of using the wealth index versus these alternatives were assessed for different purposes. The uncertainty about the socio-economic processes being captured by the wealth index implies that in all situations, the wealth index should only be used after careful consideration of available alternative SEP indicators. Conclusions Using the wealth index in the DHS has allowed the quantification and comparison of health inequalities in low- and middle-income countries on an unprecedented scale, thereby playing a vital role in advocacy of health equity. Some alterations to the methodology of wealth index construction are recommended for future studies using the wealth index, but the use of the wealth index in primary data collection is questionable given the uncertainty surrounding the socio-economic processes it is capturing.

Sexual and reproductive health (SRH) surveys around the world, especially in low- and middle-income countries, have been and continue to be the primary sources of data about individual-, community-, and population-level sexual and reproductive health. Beginning with the Knowledge, Attitudes, and Practices surveys of the late 1950s, SRH surveys have been crucial tools for informing public health programming, healthcare delivery, public policy, and more. Additionally, major demographic and health modeling and estimation efforts rely on SRH survey data, as have thousands of research studies. For more than half a century, surveys have met major SRH information needs, especially in low- and middle-income countries. And even as the world has achieved impressive information technology advances, increasing by orders of magnitude the depth and breadth of data collected and analyzed, the necessity and importance of surveys have not waned. As of 2021, four major internationally comparable SRH survey platforms are operating in low- and middle-income countries—the Demographic and Health Surveys Program (DHS), Multiple-Indicator Cluster Survey (MICS), Population-Based HIV Impact Assessment (PHIA), and Performance Monitoring for Action (PMA). Among these platforms, DHS collects the widest range of data on population, health, and nutrition, followed by MICS. PHIA collects the most HIV-related data. And PMA’s family planning data are collected with the most frequency. These population-based household surveys are rich data sources, collecting data to measure a wide range of SRH indicators—from contraceptive prevalence to HIV prevalence, from cervical cancer screening rates to skilled birth delivery rates, from age at menarche to age at first sex, and more. As with other surveys, SRH surveys are imperfect; selection bias, recall bias, social desirability bias, interviewer bias, and misclassification bias and error can represent major concerns. Furthermore, thorny issues persist across the decades, including perpetual historic, measurement, and methodological concerns. To provide a few examples with regard to history, because the major survey programs have historically been led by donors and multilateral organizations based in the Global North, survey content and implementation have been closely connected with donor priorities, which may not align with local priorities. Regarding measurement, maternal mortality data are highly valued and best collected through complete vital registration systems, but many low- and middle-income countries do not have complete systems and therefore rely on estimates collected through household surveys and censuses. And regarding methods, because most surveys offer only a snapshot in time, with the primary purpose of monitoring key indicators using a representative sample, most analyses of survey data can only show correlation and association rather than causation. Opportunities abound for ongoing innovation to address potential biases and persistent thorny issues. Finally, the SHR field has been and continues to be a global leader for survey development and implementation. If past is prelude, SRH surveys will be invaluable sources of knowledge for decades to come.

Interspecific associations detected in phytosociological data sets sampled in local areas can reflect locally specific combinations of environmental factors and may thus differ from the interspecific associations existing on a regional scale. As a result, vegetation units derived from numerical classifications of local data sets can accurately reflect local environmental gradients, but their boundaries or spectra of diagnostic species must be frequently adjusted when transferred to the regional scale. Local vegetation classifications can be useful for some purposes, but regional classifications are superior, as they facilitate communication among the researchers from different areas. We demonstrated changes in interspecific associations between regional and local scale, using a data set of 14 589 releves of herbaceous vegetation of the Czech Republic, and 16 local subsets of this national data set. We focused on sociological species groups, derived statistically in the national data set. Changes in coherence of these groups when applied to the local data sets were described on the basis of statistical association between the releves containing some species of these groups and the species belonging vs. not belonging to these groups. The results were summarized using the principal components analysis (PCA). In addition, releve data sets were compared with respect to presence/absence of sociological species groups, using the principal coordinate analysis (PCoA). The results of PCA and PCoA were compared by Procrustean analysis. Local data sets differed from the national data set to different extent. The national data set was more remote to the local data sets if the analysis focused on the coherence of species group rather than on presence/absence. The species groups from the national data set retained most of their coherence in low-altitude hilly landscapes with thermophilous flora, i.e., the most diverse landscape type of the Czech Republic. On the other hand, many species groups from the national data set could not be recognized in mountainous areas or flat lowlands. These results suggest that interspecific associations existing on regional scale are best reproduced in those local areas which have a high habitat heterogeneity or which have a central position along the major gradients existing on regional scale.

Progress in the HIV response in Uganda: findings from two sequential population-based HIV impact assessment surveys, 2017-21.

Initiatives to monitor progress in health interventions like sanitation are increasingly focused on disparities in access. We explored three methodological challenges to monitoring changes in sanitation coverage across socio-economic and demographic determinants: (i) confounding by wealth indices including water and sanitation assets, (ii) use of individual urban and rural settings versus national wealth indices and (iii) child-level versus household-level analyses. Sanitation coverage by wealth for children and households across settings was estimated from recent Demographic and Health Surveys in six low-income countries. Household assignment to wealth quintiles was based on principal components analyses of assets. Concordance in household quintile assignment and estimated distribution of improved sanitation was assessed using two wealth indices differing by inclusion or exclusion of water and sanitation assets and independently derived for each setting. Improved sanitation was estimated using under five children and households. Wealth indices estimated with water, and sanitation assets are highly correlated with indices excluding them but can overstate disparities in sanitation access. Independently, derived setting wealth indices highly correlate with setting estimates of coverage using a single national index. Sanitation coverage and quintile disparities were consistently lower in household-level estimates. Standard asset indices provide a reasonably robust measure of disparities in improved sanitation, although overestimation is possible. Separate setting wealth quintiles reveal important disparities in urban areas, but analysis of setting quintiles using a national index is sufficient. Estimates and disparities in household-level coverage of improved sanitation can underestimate coverage for children under five.

BackgroundHIV/AIDS remains the leading cause of death in sub-Saharan Africa. Due to multiple constraints experienced by households that seem to be disproportionally affected, families generally seek assistance from the community and external economic support. Previous researchers studied socioeconomic and gender inequality in HIV/AIDS prevalence in sub-Saharan African countries. However, very few researchers have paid attention to the external economic support for HIV/AIDS affected households in Tanzania. This study investigates the difference in economic support among households affected or not affected by the HIV/AIDS epidemic in Tanzania.MethodsData used stemmed from the Tanzania HIV Impact Survey 2016–2017 (THIS) of the Population-based HIV Impact Assessment (PHIA) project, collected between 2016 and 2017 in Tanzania. The study population were the heads of households (adults) with age greater than 15. The dependent variable for the study was economic support. This consisted of both material and non-material assistance obtained from outside the household. Socio-demographic (economics) characteristics constituted the predictors of the study. Descriptive statistics and econometric modelling were used to analyse determinants associated with external economic support. Oaxaca-Blinder decomposition method was also performed to investigate the difference in economic support depending on households’ serological status in Tanzania.ResultsA total of 12,008 households were included. Almost 11% of the household heads indicated that their households received economic support. HIV/AIDS affected 7% of households. The mean age of the household heads was 45 years (SD ± 15) with a range of 16–80. The majority of household heads were men (72%). Being a household head affected by HIV/AIDS increases the probability to receive external economic support (p < 0.05). The difference in external economic support between the two groups (HIV/AIDS and no- HIV/AIDS households) was - 0.032 (p < 0.01). This gap was observed to favour households affected by HIV/AIDS. Almost 72% (− 0.023/− 0.032) of this difference was explained by characteristics such as the wealth index (p < 0.01), residence area (urban) (p < 0.01), marital status (widowed (p < 0.05) and divorced or separated) (p < 0.1) and age (p < 0.01).ConclusionThe difference in economic support across households affected or not affected by HIV/AIDS was explained by wealth index, residence area, marital status and age. These findings represent important implications for health policy regarding future economic support strategies for HIV/AIDS-affected households.

Male circumcision (MC) offers men lifelong partial protection from heterosexually acquired HIV infection. The impact of MC on HIV incidence has not been quantified in nationally representative samples. Data from the population-based HIV impact assessments were used to compare HIV incidence by MC status in countries implementing voluntary medical MC (VMMC) programs. Data were pooled from population-based HIV impact assessments conducted in Eswatini, Lesotho, Malawi, Namibia, Tanzania, Uganda, Zambia, and Zimbabwe from 2015 to 2017. Incidence was measured using a recent infection testing algorithm and analyzed by self-reported MC status distinguishing between medical and nonmedical MC. Country, marital status, urban setting, sexual risk behaviors, and mean population HIV viral load among women as an indicator of treatment scale-up were included in a random-effects logistic regression model using pooled survey weights. Analyses were age stratified (15-34 and 35-59 years). Annualized incidence rates and 95% confidence intervals (CIs) and incidence differences were calculated between medically circumcised and uncircumcised men. Men 15-34 years reporting medical MC had lower HIV incidence than uncircumcised men [0.04% (95% CI: 0.00% to 0.10%) versus 0.34% (95% CI: 0.10% to 0.57%), respectively; P value = 0.01]; whereas among men 35-59 years, there was no significant incidence difference [1.36% (95% CI: 0.32% to 2.39%) versus 0.55% (95% CI: 0.14% to 0.67%), respectively; P value = 0.14]. Medical MC was associated with lower HIV incidence in men aged 15-34 years in nationally representative surveys in Africa. These findings are consistent with the expected ongoing VMMC program impact and highlight the importance of VMMC for the HIV response in Africa.

For over 25 years, new programs to attempt to stem the HIV epidemic have been developed in Africa by country governments as well as external donors. These programs and activities have built and operated facilities, trained clinicians, financed drugs and commodities, supported and helped finance government health planning and operations, and contributed in other ways. Who has benefited from this massive mobilization? While some single country and narrowly focused studies have been done, the issue of equity of HIV programs for vulnerable populations has not been examined in a large set of countries. Using Population-based HIV Impact Assessment (PHIA) data, we examine equity of the HIV programs in 13 African countries to determine if vulnerable groups (such as those with low wealth, rural populations, young adults, and females) have achieved comparable levels of access to HIV program services. In contrast, we also compare the equity of the HIV response to rural and low-wealth populations with the equity of corresponding domestic health systems using Demographic and Health Survey data. This study found that in over half of the countries, the HIV response indicators were equitable for vulnerable population segments including the low-wealth population (in seven countries) and rural population segment (in nine countries). In no country was the domestic health system equitable for these two groups. However, HIV programming does show some clear patterns of inequity for low-wealth and rural populations in some countries. For gender and young adults, the HIV response indicators show that in all 13 countries men and young adults are consistently underserved relative to their counterparts.

The global HIV response has had notable effects on the trajectory of the epidemic. Dramatic declines in morbidity and mortality have been noted in association with the scale-up of effective...

1. Methods that assess patterns of phylogenetic relatedness, as well as character distribution and evolution, allow one to infer the ecological processes involved in community assembly. Assuming niche conservatism, assemblages should shift from phylogenetic clustering to evenness with decreasing geographic scale because the relative importance of mechanisms that shape assemblages is hypothesized to be scale-dependent. Whereas habitat filtering is more likely to act at regional scales because of increased habitat heterogeneity that allows sorting of ecologically similar species in contrasting environments, competition is more likely to act at local scales because low habitat heterogeneity provides few opportunities for niche partitioning. 2. We used species lists to assess assemblage composition, data on ecologically-relevant traits, and a molecular phylogeny, to examine the phylogenetic structure of antbird (Thamnophilidae) assemblages at three different geographical scales: regional (ecoregions), intermediate (100-ha plots) and local (mixed-flocks). In addition, we used patterns of phylogenetic beta diversity and beta diversity to separate the factors that structure antbird assemblages at regional scales. 3. Contrary to previous findings, we found a shift from phylogenetic evenness to clustering with decreasing geographical scale. We argue that this does not reject the hypothesis that habitat filtering is the predominant force in regional community assembly, because analyses of trait evolution and structure indicated a lack of niche conservatism in antbirds. 4. In some cases, phylogenetic evenness at regional scales can be an effect of historical biogeographic processes instead of niche-based processes. However, regional patterns of beta diversity and phylogenetic beta diversity suggested that phylogenetic structure in our study cannot be explained by the history of speciation and dispersal of antbirds, further supporting the habitat-filtering hypothesis. 5. Our analyses suggested that competitive interactions might not play an important role locally, which would provide a plausible explanation for the high alpha diversity of antbirds in Amazonia. 6. Finally, we emphasize the importance of including trait information in studies of phylogenetic community structure to adequately assess the mechanisms that determine species co-existence.

BackgroundSmartphones have rapidly become an important marker of wealth in low- and middle-income countries, but international household surveys do not regularly gather data on smartphone ownership and these data are rarely used to calculate wealth indices.MethodsWe developed a cross-sectional survey module delivered to 3028 households in rural northwest Burkina Faso to measure the effects of this absence. Wealth indices were calculated using both principal components analysis (PCA) and polychoric PCA for a base model using only ownership of any cell phone, and a full model using data on smartphone ownership, the number of cell phones, and the purchase of mobile data. Four outcomes (household expenditure, education level, and prevalence of frailty and diabetes) were used to evaluate changes in the composition of wealth index quintiles using ordinary least squares and logistic regressions and Wald tests.ResultsHouseholds that own smartphones have higher monthly expenditures and own a greater quantity and quality of household assets. Expenditure and education levels are significantly higher at the fifth (richest) socioeconomic status (SES) quintile of full model wealth indices as compared to base models. Similarly, diabetes prevalence is significantly higher at the fifth SES quintile using PCA wealth index full models, but this is not observed for frailty prevalence, which is more prevalent among lower SES households. These effects are not present when using polychoric PCA, suggesting that this method provides additional robustness to missing asset data to measure underlying latent SES by proxy.ConclusionsThe lack of smartphone data can skew PCA-based wealth index performance in a low-income context for the top of the socioeconomic spectrum. While some PCA variants may be robust to the omission of smartphone ownership, eliciting smartphone ownership data in household surveys is likely to substantially improve the validity and utility of wealth estimates.

The regional spatial scale is a vital linkage for the informed extrapolation of results from local to continental scales to address broad-scale environmental problems. Among-region variation in ecosystem state is commonly accounted for by using a regionalization framework, such as an ecoregion classification. Rarely have alternative regionalization frameworks been tested for variables measuring ecosystem state, nor have the underlying relationships with the variables that are used to define them been assessed. In this study, we asked two questions: (1) How much among-region variation is there for ecosystems and does it differ by regionalization framework? (2) What are the likely causes of this among-region variation? We present a case study using a large data set of lake water chemistry, uni- and multi-scaled hydrogeomorphic and anthropogenic driver variables that likely influence lake chemistry at the subcontinental scale, and seven existing regionalization frameworks. We used multilevel models to quantify and explain within- and among-region variation in lake water chemistry. Our models account for local driver variables of ecosystem variation within regions, differences in regional mean ecosystem state (i.e., random intercepts in multilevel models), and differences in relationships between local drivers and ecosystem state by region (i.e., random slopes in multilevel models). Using one of the best performing regionalization frameworks (Ecological Drainage Units), we found that for lake phosphorus and alkalinity: (1) a majority of all the variation in water chemistry among the studied lakes occurred among regions, (2) very few regional-scale landscape driver variables were required to explain among-region variation in lake water chemistry, (3) a much higher proportion of the total variation among lakes was explained at the regional scale than at the local scale, and (4) some relationships between local-scale driver variables and lake water chemistry varied by region. Our results demonstrate the importance of considering the regional spatial scale for broad-scale research and ecosystem management and conservation. Our quantitative approach can be easily applied to other response variables, ecosystem types, geographic areas, and spatial extents to inform ecosystem responses to global environmental stressors.