Investigating health impacts of natural resource extraction projects in Burkina Faso, Ghana, Mozambique, and Tanzania: Protocol for a mixed methods study

Background: Natural resource extraction projects offer both opportunities and risks for sustainable development and health in host communities. Often, however, the health of the community suffers. Health impact assessment (HIA) can mitigate the risks and promote the benefits of development but is not routinely done in the developing regions that could benefit the most. Objective: Our study aims to investigate health and health determinants in regions affected by extractive industries in Burkina Faso, Ghana, Mozambique, and Tanzania. The evidence generated in our study will inform a policy dialogue on how HIA can be promoted as a regulatory approach as part of the larger research initiative called the HIA4SD (Health impact assessment for sustainable development) project. Methods: The study is a concurrent triangulation, mixed methods, multi-stage, multi-focus project that specifically addresses the topics of governance and policy, social determinants of health, health economics, health systems, maternal and child health, morbidity and mortality, and environmental determinants, as well as the associated health outcomes in natural resource extraction project settings across four countries. To investigate each of these health topics, the project will (1) use existing population-level databases to quantify incidence of disease and other health outcomes and determinants over time using time series analysis; (2) conduct two quantitative surveys on mortality and cost of disease in producer regions; and (3) collect primary qualitative data using focus groups and key informant interviews describing community perceptions of the impacts of extraction projects on health and partnership arrangements between the projects and local and national governance. Differences in health outcomes and health determinants between districts with and without an extraction project will be analyzed using matched geographical analyses in quasi-Poisson regression models and binomial regression models. Costs to the health system and to the households from diseases found to be associated with projects in each country will be estimated retrospectively. Results: Fieldwork for the study began in February 2019 and concluded in February 2020. At the time of submission, qualitative data collection had been completed in all four study countries. In Burkina Faso, 36 focus group discussions and 74 key informant interviews were conducted in three sites. In Ghana, 34 focus group discussions and 64 key informant interviews were conducted in three sites. In Mozambique, 75 focus group discussions and 103 key informant interviews were conducted in four sites. In Tanzania, 36 focus group discussions and 84 key informant interviews were conducted in three sites. Quantitative data extraction and collection is ongoing in all four study countries. Ethical approval for the study was received in all four study countries prior to beginning the fieldwork. Data analyses are underway and results are expected to be published in 2020 and 2021. Conclusions: Disentangling the complex interactions of resource extraction projects with their host communities requires an integrative approach drawing on many methodologies under the HIA umbrella. By using complementary data sources to address the question of population health in project areas from several angles, bias and missing data will be reduced, generating high-quality evidence to aid countries in moving toward sustainable development.

The study is designed as a concurrent triangulation mixed methods design [24], with simultaneous collection of (1) quantitative data used to measure resource extraction project effects on population health by describing incidence of disease and other health determinants and outcomes over time, as well as (2) qualitative data describing community perceptions of the impacts of projects on health. In this design, the data will be analyzed first separately and then integrated, with the advantage that each data source can complement and strengthen the findings of the other data sources, as well as drive further research questions. Based on the initial study aims, seven major research topics of interest were identified: governance and policy, social determinants of health, health economics, health systems, maternal and child health, morbidity and mortality, and environmental determinants and associated health outcomes (see Figure 1). Data will be collected on every research topic by the local team within each country and then shared across all project teams, with the aim of making comparisons both within and between each country. Study design. The headings are the research topics relevant to population health in natural resource extraction project (NREP) areas identified in conjunction with local partners. The specific research topics are investigated with both qualitative (white) and quantitative (gray) research methods. AIS: AIDS Indicator Survey; DHIS2: District Health Information System 2; DHS: Demographic and Health Survey; GIS: geographic information system; MICS: Multi-Indicator Cluster Survey; MIS: Malaria Indicator Survey; SARA: Service Availability and Readiness Assessment; SPA: Service Provision Assessment. The health impacts of resource extraction projects vary based on the baseline characteristics and environment of the host community [25]. Therefore, we partnered with four different countries with a history of natural resource extraction and low health-related SDG index values: Burkina Faso, Ghana, Mozambique, and Tanzania (see Figure 2). Within those countries, three to four large mining sites were chosen for qualitative and quantitative field work. The mines were chosen based on type, size, length of operation, and type of ownership. Quantitative data on disease and health outcomes is available country-wide through District Health Information System 2 (DHIS2) and Demographic and Health Survey (DHS) databases (see Table 1) [26]. Location of study countries in sub-Saharan Africa. Primary quantitative outcomes and data availability from the District Health Information System 2 (DHIS2), the Demographic and Health Survey (DHS), and other national-level databases. aSDG: Sustainable Development Goal. In Burkina Faso, primary data collection was conducted in three gold mining sites: (1) Houndé Mine in Houndé district (population in 2006: 77,000); (2) Yaramoko Gold Mine in Bagassi district (population in 2006: 33,000); and (3) Bissa Mine in Kongoussi district (population in 2006: 71,000). The three mines have been operational since 2017, 2016, and 2013, respectively. While the Houndé and Bissa mines are open-pit mines, operations in Bagassi are underground. Houndé is a small city located approximately 5 km from the mine with formal and informal settlements reaching closer. In Bagassi, villages are scattered around the mining premises. Sabcé is the closest town from the Bissa mine. During the field visit, artisanal mining sites were observed around all three mines. In Ghana, primary data collection was conducted in three gold and manganese mining sites: (1) Edikan Gold Mine (Perseus Mining) in the Upper Denkyira West district (population in 2019: 71,425); (2) Tarkwa (Anglo Gold Ashanti, Gold Fields, and Ghana Manganese Company) in the Tarkwa-Nsuaem Municipal district (population in 2019: 117,550); and (3) Newmont Ahafo Mine (Newmont Goldcorp) in the Asutifi North Municipal (population in 2010: 52,259). The three mines have been operational since 2011, 1961, and 2003, respectively. All three are open-pit mines. Edikan Gold Mine has four nearby communities—Ayanfuri, Fobinso, Nkonya, and Abenabena—located in both the Upper Denkyira West and Amenfi West districts. The sites near Tarkwa include the small communities of Akoon, Tarkwabanso, Wangarakrom and Badukrom, and Iduaprim. Four communities are near the Newmont Ahafo Mine: Tutuka, Kenyase 1, Kenyase 2, and Ntrotroso. In Mozambique, primary data collection was conducted in three types of mining sites—ruby, titanium, and coal—involving communities of four administrative districts and four mining companies: (1) Montepuez Ruby Mining (MRM) in Montepuez district (population in 2017: 261,235); (2) Kenmare Moma Mining in Larde district (population in 2017: 85,971) and Moma district (population in 2017: 310,706); (3) Vale Mozambique; and (4) International Coal Ventures (ICVL), the latter two in Moatize district (population in 2017: 343,546). All four are open-pit mines that have been operating since 2007 (Kenmare Moma Mining) and 2011 (MRM, Vale, and ICVL). MRM is surrounded by four small villages—N´sewe, N´thorro, M´pene, and Namanhumbir—whose main economic activities are agriculture and unregulated artisanal mining. Near Kenmare Moma Mining, Moma and Larde are the main and small emerging coastal towns located approximately 80 km and 20 km from the Kenmare mining company, respectively. The company activities affect communities from Moma district (ie, Pilivili locality) and Larde district (ie, Topuito locality). The eight affected villages at Pilivili locality (ie, Moma district) are the closest, located 5 km away from the mining company. There are 11 neighborhoods and villages affected by Kenmare Moma Mining, of which seven are located within the mining concession area and four are less than 100 meters from the mining pit: Tipane, Mutiticoma, Izoua, and Topuito. Agriculture and fishing are the main economic activities; however, commerce is an activity emerging mainly after mining implementation in 2005. Moatize is a small town 20 km from the city of Tete. It is an industrial complex composed of at least six large-scale coal mining companies—Vale Mozambique, ICVL, Nkodezi Coal Company, Minas de Revubue, Minas de Moatize Riversdale Mozambique Limitada, and Jindal Mozambique Minerals (JSPL)—some being subsidiaries of Vale Mozambique. At the time of site visit, very active commerce activities were observed along the main road. Apart from the town of Moatize, there are 12 small affected communities surrounding mining companies, four belonging to Moatize-sede locality—Catete, Mphandue, Matambanhama, and Ntchenga—and eight to Benga locality—Cancope, Capanga Gulo, Campanga lowane, Chitambo, Chitondo, Nyambalualu, Kangale, and Benga-sede. The last five communities are along the Zambeze River, the main local source of water. Agriculture and fishing are the main economic activities. In Tanzania, three gold mining sites were chosen for fieldwork: Geita Gold Mine (Geita district), Buzwagi Gold Mine (Kahama district), and Bulyanhulu Gold Mine (Mslala district). Geita and Buzwagi are open-pit mines, while Bulyanhulu is an underground mine (based on observation). At the time point of data collection (ie, March to May 2019), the Geita Gold Mine was fully operational and the Buzwagi and Bulyanhulu mines were both in reduced production status, meaning that they were processing already extracted material and no longer extracting new raw material. The Geita Gold Mine, about 70 km south of Lake Victoria, is located between several villages and Geita, a main city of the district. The Buzwagi mine is surrounded by several villages and is 6 km away from Kahama, the capital of the district. The vibrant villages of Kakola, Bushing’we, and Kakola Namba 9 are next to the Bulyanhulu mine. During the field visit, artisanal mining sites were observed around all three mines. The main quantitative component of the study will be a retrospective observational longitudinal study using routine health data extracted from DHIS2 and other national-level databases. Specifically, this quantitative part of the study seeks to answer the following questions: (1) Which districts have been directly, indirectly, or not impacted by projects? (2) What differences in health-related SDG indicators and other health indicators (eg, health systems) can be observed in districts impacted by resource extraction projects compared to nonimpacted districts, including maternal and child health–related indicators? (3) What are the costs and benefits to the health system of project implementation? (4) How do projects impact on environmental determinants of health? and (5) What are the strengths and limitations of the national, routine, health information systems and other data sources and repositories at the national level to monitor how projects impact on health-related SDG target indicators and other health indicators? The specific research topics belonging to this work package are effects on health systems, morbidity and mortality, environmental determinants of health, maternal and child health, and health economics. The research in this work package uses existing DHIS2 surveillance data routinely collected by the governments of the study countries, along with supplementary national datasets—DHS, Service Availability and Readiness Assessment (SARA), Malaria Indicator Survey (MIS), Multi-Indicator Cluster Survey (MICS), AIDS Indicator Survey (AIS), and Service Provision Assessment (SPA) (see Table 2)—and remote sensing data (eg, geographical positioning, weather, and pollution data). Sources of existing national-level data used in the HIA4SD (Health impact assessment for sustainable development) study. As part of the project, information on natural resource extraction projects (ie, type, size, and associated projects) and their exact geographic position will be extracted from government and other public record databases in the four study countries and mapped using ArcGIS software (Esri), a geographic information system (GIS). The data exist in diverse sources in different countries (ie, Ministries of Land and Natural Resources, Ministries of Water and Sanitation, Ministries of Science and Environment, Ministry of Energy, Forest Commissions, and local government records) as well as international monitoring organizations (ie, the International Finance Corporation [IFC] and the World Bank). Satellite imagery from the Landsat database will be used to estimate the impact of mining on settlement growth. The extracted satellite scenes will cover a time range spanning the years prior and after the opening of selected large-scale mining projects. With the aid of ancillary ground-truth information from historic Google Earth imagery, the size of settlements will be determined annually using machine learning algorithms. The annual growth of settlements will then be compared between mining and comparison sites. To analyze the differences in health outcomes between districts with and without extractive industries, a matched geographical analysis will be performed within the framework of the HIA4SD project to provide a stratified sampling framework. In a first stage, natural resource extraction projects will be mapped in each country together with their key attributes (ie, size and number of workers, length and age, type, and geographical footprint of the project). Each country will be spatially stratified into four levels for sampling purposes: (1) areas in direct proximity to a project (highly impacted areas), (2) areas within a 20-30 km buffer zone of the directly impacted regions (impacted areas), (3) regions bordering project areas that contain access roads or other economic or physical links to the project (potentially impacted areas), and (4) regions greater than 30 km away from a project that do not contain access roads or other economic or physical links to the project (nonimpacted areas). Comparison study sites will be selected from the nonimpacted areas and matched with important baseline characteristics (ie, community socioeconomic activities, vegetation, altitude, and ecological zone) to the highly impacted areas. All facilities, including public and private, that fit within perimeter boundaries and are registered in the DHIS2 database will be selected. In a second stage, highly impacted districts will be matched to two or three nonimpacted comparison districts within each country. Districts will be matched based on important baseline characteristics (ie, population, urbanization and aggregate night satellite brightness, square area, number of health care facilities, and disease rates) during the year before project implementation in the highly impacted district. In order to assess the quality and completeness of the DHIS2 data, a comparison of DHIS2 and other data repositories with health statistics being collected at the local level under other work packages will be done. Next, potential positive and negative associations between health outcomes and health determinants (independent variables) and the existence of natural resource extraction projects (dependent variable) will be analyzed in the fourth working step by means of quasi-Poisson regression models and binomial regression models. The time span to be analyzed will be determined by the development history of the projects of interest in combination with the availability of data, which will vary between datasets. For the regression model, setting-specific cluster effects (eg, urban, rural, and type of project) will be taken into account. In order to maximize statistical power, pooled cross-country analysis and meta-analysis will be employed in addition to country-specific evaluations. Finally, based on the previous working steps, strengths and limitations of the national routine health information systems and other datasets at the national level to monitor impacts on health-related SDG target indicators and other health indicators of extraction projects will be determined and systematically reported [27]. Costs to the health system and to the households arising from prespecified disease conditions in each country will be estimated based on a retrospective approach. Health system costs will be collected from published information and on-site in selected health facilities in impacted districts through key informant interviews. Cost-generating components, such as required medical resources related to the corresponding illness, will be identified and a monetary value will be attributed to them. Costs incurred by the households, which are associated with the health care received, will be obtained through exit interviews in health facilities. The combination of excess cases associated with the presence of resource extraction projects (ie, data generated in the quantitative part of the study) and corresponding costs in each study country will allow the estimation of the cost incurred to the health system and households. A cost-of-illness analysis will be employed for estimating the costs incurred because of specific diseases or conditions (eg, HIV incidence rates, accidents, and chronic respiratory diseases) that have been identified in the first work package as being significantly increased due to the presence of extractive industry projects. The comparison of incidence data from impacted districts with incidence data from matching comparison districts (ie, districts with similar characteristics as impacted districts but without the presence of extractive industries) will allow for calculating the number of excess cases (eg, per 100,000 inhabitants) over the duration of 1 year. Probabilistic sensitivity analysis will be employed to allow for uncertainty around the cost estimates. Economic benefits of resource extraction projects will be measured monetarily in terms of direct and indirect financial contributions from the projects to the health sector. Financing of health infrastructure is considered to be a direct financial contribution, while the share of the health budget in the incremental tax revenues is considered to be an indirect contribution. We will also measure any other contributions. The key informant interviews, focus group discussions, and in-depth interviews will be recorded using digital voice recorders for subsequent transcription. The transcripts will then be imported into software for qualitative data analysis—NVivo (QSR International)—to code the transcripts for thematic content and framework analysis based on the COREQ (COnsolidated criteria for REporting Qualitative research) criteria [28,29]. The information obtained from different sources will be used for systematically assessing the different topics of the PhD projects (ie, partnership arrangements and the perception of health care services availability and accessibility), with a particular angle on maternal and child health, sexual and reproductive health of adolescent girls, and key social determinants linked to the perceived health impacts. Using triangulation methodologies, the most striking quantitative and qualitative results will be compared within and across the research topics and countries. Synergies and discrepancies in national data sources and local perspectives will be identified and used to develop new research questions and tools. In addition, the qualitative results will be used to explore whether or not the national-level data adequately capture the full range of health impacts from resource extraction projects as reported by the local community.

N/A

• farnham, andrea
• cossa,
• dietler, dominik
• engebretsen, rebecca
• leuenberger, andrea
• lyatuu, isaac
• nimako, belinda afriyie
• zabré, hyacinthe r.
• brugger, fritz
• winkler, mirko s.