The Longitudinal Research Platform
Agincourt Health and socio-Demographic Surveillance System (HDSS)
Established in 1992, the Agincourt Health and socio-Demographic Surveillance System (HDSS) covers a population of some 116,000 people residing in 21,000 households in 31 villages in Bushbuckridge sub-district, Mpumalanga province. The HDSS database contains data from the comprehensive coverage of demographic events within this geographically defined population over some 30 years, coupled with health and socio-economic information. Variables measured routinely include pregnancy outcomes, deaths, in- and out-migrations, household membership relationships, resident status, nationality, marital status, and education. A verbal autopsy, to establish probable cause of death, is conducted for every death within a year of the event. Information is obtained from the most knowledgeable household respondent available, while the verbal autopsy is conducted with the closest caregiver of the deceased. Data quality checks include duplicate surveying of a random sample and rigorous consistency and plausibility checks embedded in the data management system. Data are entered into a longitudinal relational database system that contains information on all individuals ever recorded as residing in the study area.
The multi-year longitudinal database contributes substantial scientific output, including trend data, in its own right. The HDSS is a robust data resource, serving as a powerful and versatile sampling frame for observational and intervention studies, as well as policy evaluations. Project data can be linked to the HDSS extending analytic possibilities.
Agincourt is a founding node of the DSI / SAMRC South African Population Research Infrastructure Network (SAPRIN) – a national research infrastructure funded by the National Department of Science and Innovation (DSI) and hosted by the South African Medical Research Council (SAMRC) SAPRIN: South African Population Research Infrastructure Network.
Clinic and Hospital Link
The “Agincourt HDSS-Clinic-Hospital link” is an ongoing longitudinal data platform that captures patient information from primary healthcare facilities in the Agincourt study area and two nearby district hospitals, and links them in real-time to the Agincourt population database. Linkage of these data from 8 primary care facilities (since 2014) and 2 hospitals (since 2020) to the health and demographic surveillance system (HDSS) enables monitoring of the burden and impact of various illnesses, coverage, and selective use of healthcare services. The ensuing analytic data provide researchers and local / district health managers with much-improved statistics on the utilisation of services, integrated chronic care, and the burden/impact of key conditions, including co-morbidities, on decentralised levels of the healthcare system.
Linkage is done using both deterministic and probabilistic record linkage approaches. When a new patient presents at the clinic, a data clerk obtains written informed consent to collect their clinical data and then links these to the Agincourt population. A data clerk collects identifiers that are used to search the HDSS database. Linkage is first attempted using a national ID number; where there is no match a combination of mobile phone number, first name, and date of birth is used. If still no match, an algorithm based on the Fellegi-Sunter probabilistic record linkage model is used. Potential matches are reviewed by the data clerk with the patient to resolve any uncertainty about his/her identity. The data clerk then extracts clinical information from clinic files and logs all follow-up visits. The privacy of patients is protected via rigorous data logging, storage, and encryption procedures. The linked data is been used in various studies and for the evaluation of trials.
HDSS-based mortality surveillance in Africa and South Asia
- PI(s): Stephen Tollman, Kathleen Kahn
- Co-investigators: Chodziwadziwa Kabudula, Kobus Herbst
- Programme Manager: Tshegofatso Seabi
- Funder: Bill and Melinda Gates Foundation, USA
- Collaborating Institutions: SAPRIN, African Health Research Institute, DIMAMO Population Health Research Centre, and Wits-VIDA Soweto CHAMPS (SA); Manhica Health Research Centre (Mozambique); African Population and Health Research Centre, Kaloleni/Rabai, Siaya-Karemo, Manyatta (Kenya); Iganga-Mayuge (Uganda); Magu (Tanzania); Nyanja Health Research Institute (Malawi); Kersa (Ethiopia); Navronga (Ghana); Nanora (Burkina Faso); Matlab, Dhaka, Chakaria (Bangladesh); Vadu (India)
Following the declaration of COVID-19 as a pandemic by the World Health Organization, there have been high levels of reported deaths, at least in countries with functioning civil registration and vital statistics (CRVS). These largely under-represent the true mortality due to COVID-19. A fundamental question, then, is what is the impact of COVID-19 on mortality and the scale of excess deaths, and the population sub-groups most affected, particularly in low- and middle-income settings? Constructing a true representation of COVID-19 deaths can be useful for social policies and future pandemic preparedness planning. The goal of this initiative is to characterise all-cause mortality rates and trends, by age and sex, across a range of rural and urban sub-Saharan African and South Asian settings under continuous health and demographic surveillance.
This a multinational initiative bringing together 17 sites/centres from Africa and South Asia. Cho Kabudula and Stephen Tollman of the MRC/Wits-Agincourt Research Unit, with Kobus Herbst of the South African Population Research Infrastructure Network (SAPRIN) and Beth Tippett-Barr of Nyanja Health Research Institute, have co-hosted three face-to-face workshops to date: in March and November 2022, and July 2023. The goal of these workshops has been to strengthen capacity in data management, analysis, scientific writing and dissemination, and so better understand the impact of COVID-19 on excess mortality in African and South Asian settings.
Multimorbidity in Africa: Digital Innovation, visualisation and application
- PI(s): Scott Hazelhurst (PI of record), Stephen Tollman, Michele Ramsay, Catherine Kyobutungi
- Project Manager: Kerry Glover
- Funder: NIH, USA
- Collaborating Institutions: IBM Research Africa, DS-I Africa – Data Science for Health Discovery and Innovation in Africa and SAPRIN (SA); African Population and Health Research Center (Kenya); Vanderbilt University Medical Center (USA).
- Website: https://www.madiva.africa/
The MADIVA Research Hub is dedicated to creating data science methods and solutions to address multimorbidity challenges in Africa, where co-occurring diseases contribute significantly to the health burden. Our primary research sites, located in rural Bushbuckridge, South Africa, and urban Nairobi, Kenya, possess extensive longitudinal data collected through health and demographic surveillance systems, alongside emerging clinical health records and genomic data.
Our focus involves developing and implementing data science techniques to connect diverse datasets, constructing dashboards for stakeholders, and employing novel machine learning methods to stratify populations based on risk profiles for different diseases. This includes utilising polygenic risk scores.
DMAC (Data Management and Analysis Core):
Led by Cho Kabudula, the DMAC plays a crucial role in the research hub by managing and analysing complex data related to multimorbidity. This core provides technical expertise, infrastructure, and tools to integrate, analyse, and visualise diverse data types. By centralising data and enabling advanced analyses, DMAC contributes to understanding the burden of multimorbidity and informs healthcare improvements in sub-Saharan Africa.
TCDPC (Training, Capacity Development and Pilot Core):
The TCDPC is vital for pilot projects and training within the research hub. It offers direction and support for new researchers and capacity development initiatives. The core emphasises early career development through pilot projects, while also addressing training needs. By fostering inclusive approaches and mentorship, TCDPC aims to bridge the gap between existing healthcare models and the evolving nature of healthcare in sub-Saharan Africa.
- Project 1:
Project 1 is led by Xavier Gómez-Olivé. Utilising data from diverse African settings, work focuses on understanding and responding to the challenges of evolving multimorbidity. By integrating data from Nairobi and rural South Africa, the project aims to create an effective health system response by linking and visualising various datasets. This data-driven approach has the potential to significantly impact the management of multimorbidity across sub-Saharan Africa.
- Project 2:
Project 2 aims to enhance existing models for predicting health outcomes related to diseases and multimorbidity. Through rich longitudinal data and genetic information from African communities, the project will develop predictive models using both established risk algorithms and advanced machine learning techniques. The goal is to identify high-risk clusters for targeted interventions and provide risk estimates for both individuals and populations, aiding clinical decisions and public health strategies.
Minimally Invasive Tissue Sampling (MITS)
Integrating minimally invasive tissue sampling into established community-based mortality surveillance in rural South Africa: contributing to strengthening verbal autopsy
- Co-PIs: Kathleen Kahn, Ryan Wagner
- Project Manager: Lucky Mondlane
- Funder: The Bill & Melinda Gates Foundation through the MITS Alliance, USA
- Collaborating Institutions: The MITS Alliance, and The Ohio State University (USA); Wits Department of Pathology (SA)
Understanding who dies from what can assist researchers and policymakers in strengthening health systems and designing interventions that cater to the evolving disease burden and changing population dynamics. However, determining cause of death (CoD) in rural South Africa where clinical autopsy services are limited, upwards of at least half of deaths occur outside of health facilities and rapid and complex epidemiological transitions are underway, can be difficult. While considerable efforts have been made to develop a standardized VA tool and robust methods for assigning the most likely CoD, certain conditions with nonspecific signs and symptoms make determining a CoD with the VA method difficult. As such, strengthening the current VA method by including a minimally invasive tissue sampling component will likely allow for more accurate individual-level cause of death determination and, undertaken strategically, can further strengthen the VA method.
As a new member of the Minimally Invasive Tissue Sampling (MITS) Alliance and with funding received for the next 3 years, we aim to:
- develop and pilot methods for including minimally invasive tissue sampling (MITS) as part of routine population-based mortality surveillance in conjunction with the VA method;
- compare the CoD ascertained through the MITS procedure with the standard VA approach in four groups of deaths (neonates, children, maternal and adults), and
- broaden current scope of focus of MITS work in South Africa by ascertaining population-based CoDs in adults, piloting innovative technology and allowing for rural and urban comparative work. In undertaking this work, we aim to contribute meaningfully to the MITS surveillance alliance and its partners.