Risk prediction for Women's Health and Rights in Tanzania: novel statistical methodology to target effective interventions

Lead Research Organisation: University of Nottingham
Department Name: Sch of Mathematical Sciences

Abstract

This programme will extend novel advances in mathematical sciences to identify, measure and rectify previously intractable humanitarian abuses related to Rights of Women (SDG5, 3.1, 5.3). The innovations established will feed directly into government supported health and education interventions in Tanzania, East Africa - a country where women continue to suffer from pernicious inequality, leading to horrendous and sustained humanitarian abuses: widespread Female Genital Mutilation (FGM), Forced Marriage and continued unacceptable rates of Perinatal Mortality.

To address these issues a new mathematical framework is required, motivated by a single key issue underpinning SDG5, and distinct from most others. Dreadful as they are, the challenges facing other SDGs, whether they be floods, disease or even extreme poverty, are visible: they can be observed, modelled, monitored. The challenge of Women's Rights abuses stand in contrast: here data is obscured, censored and hidden from sight, often intentionally so. What data does exist is partial, unrepresentative and multi-viewed, arriving piecemeal from disparate sources. As a consequence, solutions based on mathematical modelling are often passed over wholesale. Only aggregate, region level figures tend to be known. Vulnerability and risk across individuals are left unmodelled; interventions fail and abuses are perpetuated.

This destructive pathway is symptomatic of many rights issues of girls and women, and calls for a statistical approach designed specifically to handle the highly sparse, noisy and unbalanced data common to problems of this nature. Technical work to address this issue will be undertaken in three phases. First, a methodology for probabilistic data assembly will be developed to address hidden and obfuscated data challenges. This is followed by key extensions to Object Oriented and Topological Data Analysis that handle multi-view and temporally unaligned datasets. A final stage sees integration of developments into full predictive models, and the completion of the framework.

Through partnership with in-country academics, government, private-sector partners and NGOS, and application of novel data sources (digital health data; drone/earth observation imagery; mobile-money; cell network data; crowd-sourced event reporting), resulting models will be used in two key intervention streams during project lifetime: 1. Perinatal mortality modelling with partners d-tree and the Ministry of Health (SDG 3.1); 2. FGM/Forced Marriage modelling (SDG 5.3) for target educational interventions with partners onebillion, Hope for Girls and Women and the Tanzania Development Trust. While these interventions provide initial focus for mathematical sciences developments, we expect the framework generated to have application beyond this geographical extent, and to a wide range of Sustainable Development Goals.

Planned Impact

The project will provide impact to benefit direct partner organizations, citizens, community leaders and workers, NGOs and other organisations.

In particular, the project will provide direct and ongoing societal impact aimed at reducing perinatal mortality (SDG 3.1, Intervention Package 1) and reducing incidences of FGM/Forced Marriage (SDG 5.3, Intervention Package 2). Direct beneficiaries will be vulnerable women and girls, our partner organisations (D-Tree, onebillion, Hope for Girls and Women, Tanzania Development Trust), community leaders and workers, and other NGOs. Resulting in otherwise unobtainable vulnerability and health models, the project will deliver learnt economic and societal information; this much needed knowledge will benefit and enable community leaders, NGOs and other organisations beyond our partners in the project to drive evidence based policy change.

Our partners will benefit from increased economic efficiency from implementation of the methodology and models. Longer term, economic opportunities exist for these and other data products being developed as a service with regular updates though continued collaboration with partners outside the timeframe of the project.

Underpinning this societal and economic impact, advances in knowledge in statistical methodology will be delivered (WP1-3) and through dissemination routes the impact will be driven beyond the partners in this project: to further government, commercial, parastatal and other interested organisations (including the World Bank, DFID). In addition the project will be used to drive increased awareness of the prevalence, challenges and issues both to the general public (via public outreach activities) and provide benefit for further investigations in contexts outside the scope of this grant within related multidisciplinary fields where similar problems of hidden, censored, noisy and sparsely observed data arise (e.g. food poverty - University of Nottingham Food Beacon, issues of slavery - University of Nottingham Rights Lab).

Direct impact upon the skills and capacity of researchers (2 UK RFs, 1 PhD student) and African mathematicians will be delivered as an integral part of the project. The former will build research capacity and future research leaders in this interdisciplinary field within the UK. The latter will importantly facilitate the in-country applications and refinement of the proposed solutions in addition to providing capacity for the ongoing use, maintenance and extensions of the proposed approach in the region.

Publications

10 25 50
 
Description We have developed a method to estimate deprivation where a set of comparative judgments are used to indicate whether one area is more/less/equally deprived as another. These judgments can be used to estimate a relative ordering of the areas by deprivation, taking into account spatial relationships between areas and accounting for uncertainty. We have demonstrated that the method works well when only partial data are available, which is important for saving time and resources on organising such surveys in the field. The key application involves estimating deprivation in Dar es Salaam, Tanzania. Such information will be used as an important variable in the study of perinatal maternal health and the risk of FGM (UN Sustainable Development Goals SDG 5, SDG 3.1 and SDG 5.3). We have also developed methods for estimating smooth paths for certain types of structured data, like networks between rural or city locations, and this will be important for modelling changes in risk over time. We have also developed a joint and individual analysis of image, phone and network data.
Exploitation Route We are still developing the methods, and are working with our partners.
Sectors Education,Healthcare

 
Description Scheme 6: Research Workshop Grant. (awarded to RG Seymour and KE Severn)
Amount £4,410 (GBP)
Organisation London Mathematical Society 
Sector Academic/University
Country United Kingdom
Start 12/2021 
End 03/2022
 
Title Comparative Judgment Data from Dar es Salaam 
Description The R package contains the Comparative Judgement data set collected in Dar es Salaam for assessing relative deprivation 
Type Of Material Database/Collection of data 
Year Produced 2021 
Provided To Others? Yes  
Impact The dataset has been used as the main motivating application to develop Bayesian comparative judgment methods in spatial settings. The method requires much less data, and so there is a saving in cost and time in collecting future datasets. 
URL https://cran.r-project.org/package=BSBT
 
Description D-Tree International 
Organisation D-Tree International
Country United States 
Sector Charity/Non Profit 
PI Contribution We are working with D-tree on the data generation part of the project, including perinatal mortality analysis, survey analysis incorporating comparative judgement and other covariate data. This work addresses sustainable development goal SDG 5, 3.1 through our work on perinatal mortality analysis.
Collaborator Contribution D-tree have worked with us on the data generation part of the project, including perinatal mortality data transfer and analysis.
Impact 1. Smith G, Mansilla R, Goulding J. (2020). Model Class Reliance for Random Forests. 34th Conference on Neural Information Processing Systems (NeurIPS 2020). 2. Seymour RG, Sirl D, Preston SP, Dryden IL, Ellis M, Perrat B, ... Goulding J. (2020). The Bayesian Spatial Bradley--Terry Model: Urban Deprivation Modeling in Tanzania. arXiv. 3. Seymour RG, Briant J. (2020). BSBT: The Bayesian Spatial Bradley--Terry Model. R package version 1.0.0. Vignette. The collaboration is multidisciplinary involving mathematical sciences, computer science and health sciences.
Start Year 2019
 
Description Hope for Girls and Women, Tanzania 
Organisation Hope for Girls and Women
Country Tanzania, United Republic of 
Sector Charity/Non Profit 
PI Contribution We are working with Hope for Girls and Women to statistical models to analyse the extent, prevalence, and underlying drivers of the issue of Female Genital Mutilation (FGM) in rural Tanzania in order to ground interventions. At the current time we are guiding data collection and carrying out preliminary modelling and analysis. This work addresses the sustainable development goal SDG 5, 5.3 through FGM/Forced Marriage modelling.
Collaborator Contribution "Hope for Girls and Women" are an NGO setup by Rhobi Samwelly, a survivor of Female Genital Mutilation (FGM), in order to try and combat this illegal but ongoing social abuse. Hope run a safe house in Tanzania in their continued efforts protect girls from this potential fatal issue. They have also established a "network" of 87 digital champions across remote villages in Northern Tanzania (Mara region), local women in the villages who are able to act as a contact point for those girls in imminent danger. Hope have already lent much domain knowledge to our efforts to understand this crime and are assisting with the necessary data collection processes. Obtaining information about the extent, prevalence, and underlying drivers of this issue in order to ground interventions requires an extensive effort, given 1. the remote nature of the villages; 2. the training and coordination required with the digital champions (who are not IT literate); 3. The digital resources required (e.g. surveying devices); and 4. the challenges of travel/housing digital champions in this process in a region with incredibly poor access and infrastructure.
Impact Several events and activities have been organized, raising awareness of our collaboration, including FGM film screenings and a blog: Introducing the Digital Champions. The collaboration is multidisciplinary involving mathematical sciences, computer science and social sciences.
Start Year 2019
 
Description Tanzania Development Trust 
Organisation Tanzania Development Trust
Country Tanzania, United Republic of 
Sector Charity/Non Profit 
PI Contribution We work closely with the Tanzania Development Trust on the collaboration with Hope For Girls and Women, Tanzania on FGM prevention. This work addresses the sustainable development goal SDG 5, 5.3 through FGM/Forced Marriage modelling.
Collaborator Contribution Collaborating closely with ourselves and Hope For Girls and Women, Tanzania on FGM prevention.
Impact A number of events have been organized to highlight the collaborations in the project with Hope for Girls and Women and the Tanzania Development Trust, including FGM film screenings and a blog: Introducing the Digital Champions. The collaboration is multidisciplinary involving mathematical sciences, computer science and social sciences.
Start Year 2019
 
Title BSBT: The Bayesian Spatial Bradley--Terry Model. R package version 1.2.0 
Description An implementation of the Bayesian Spatial Bradley-Terry (BSBT) model. It can be used to investigate data sets where judges compared different spatial areas. It constructs a network to describe how the areas are connected, and then places a correlated prior distribution on the quality parameter for each area, based on the network. The package includes MCMC algorithms to estimate the quality parameters. 
Type Of Technology Software 
Year Produced 2021 
Open Source License? Yes  
Impact The software is used in Seymour et. al. (2020) to carry out a Bayesian analysis of deprivation in subwards of Dar es Salaam, Tanzania using comparative judgment data. The comparative judgment dataset from Tanzania is also provided in the package. 
URL https://cran.r-project.org/web/packages/BSBT
 
Description Blog: Introducing the Digital Champions 
Form Of Engagement Activity Engagement focused website, blog or social media channel
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Public/other audiences
Results and Impact Members of the research team wrote a blog about the project, which is hosted on the School website.
Year(s) Of Engagement Activity 2020
URL https://www.nottingham.ac.uk/mathematics/news/introducing-the-digital-champions.aspx
 
Description FGM film screenings 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Postgraduate students
Results and Impact Members of the research team (Madeleine Ellis and Katie Severn) have hosted film screenings online of the feature film 'In the name of your daughter' which highlights the harmful and illegal practice of FGM in Tanzania and the work of our partner Hope for Girls and Women.
Year(s) Of Engagement Activity 2020,2021
 
Description Girls in Maths, School Visit, Southwell 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Schools
Results and Impact Our team member Katie Severn organised and led a girls in maths event at a local school (Southwell Minster School) in January 2020 to showcase the many diverse range of maths and demonstrate how mathematical research can affect everyone's life. Our project was discussed as part of the presentation.
Year(s) Of Engagement Activity 2020
URL https://www.minster.notts.sch.uk/news/?pid=3&nid=1&storyid=101
 
Description LGBTQ+ STEM seminar. Comparative judgement for poverty estimation in Tanzania: Letting the citizens decide 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Other audiences
Results and Impact The work on this application on citizen science in Tanzania was presented by team member Rowland Seymour to a very broad group in the LGBTQ+ STEM seminar at Nottingham. The event on February 26th 2020 was part of LGBT History Month to celebrate the research of LGBTQ+ people working in STEM subjects across the University of Nottingham, and was advertised widely.
Year(s) Of Engagement Activity 2020
URL https://www.nottingham.ac.uk/computerscience/events/lgbtq-seminar.aspx
 
Description Maths Civic Showcase Stall as part of International Women's Day in the Market Square, Nottingham 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Public/other audiences
Results and Impact One of our team (Katie Severn) ran a maths civic showcase stall in market square for international women's day, and spoke about our grant.
Year(s) Of Engagement Activity 2020
URL https://exchange.nottingham.ac.uk/blog/international-womens-day-civic-showcase-can-you-join-us-on-su...
 
Description Meeting at the Department for International Development (British Embassy, Tanzania) January 30th, 2020 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Third sector organisations
Results and Impact Meeting at the Department for International Development (British Embassy, Tanzania) discussing our data analysis project work in Tanzania
Year(s) Of Engagement Activity 2020
 
Description Meeting with D-Tree International at the Ministry of Health, Zanzibar, January 31st, 2020 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Third sector organisations
Results and Impact Discussions with D-Tree International about the data analysis of perinatal maternity data collected in Tanzania
Year(s) Of Engagement Activity 2020
 
Description Meeting with the Department for Roads (Zanzibar) on 1st February, 2020. 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Policymakers/politicians
Results and Impact Meeting with Department for Roads (Zanzibar) to discuss data analysis of satellite images and other data
Year(s) Of Engagement Activity 2020
 
Description Open Day presentation for Natural Sciences 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Public/other audiences
Results and Impact The presentation was on "Understanding deprivation in developing countries" Iiven t was a 10-minute "taster lecture" given by Dr David Sirl to ~40 Natural Sciences UCAS visit day prospective students on 25 Feb 2021.
Year(s) Of Engagement Activity 2021
 
Description School talk - Derby College 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Schools
Results and Impact Katie Severn gave presentations on Teams (10/11/20 and 12/11/20) to demonstrate to computer science A-level students at Derby College how the learning on their course is relevant in the real world. The impact was to consolidate the students' learning of the course material and also showcase potential future careers.
Year(s) Of Engagement Activity 2020
 
Description Using citizen knowledge to model urban deprivation, U21 Early Career Researcher Workshop 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Other audiences
Results and Impact The workshop was a Universitas 21 Early Career Workshop on "Modern Slavery, Forced Labour and Human Trafficking: Research Roadmaps to 2030". The workshop marked the UN's International Day for the Abolition of Slavery that took place on 2 December, and team member Rowland Seymour gave a presentation and took part in discussions. Further discussions after the event have taken place, including with the Rights Lab at the University of Nottingham and an NGO, on modelling the perception risk of forced marriage in different regions and ranking legal issues tackling modern slavery.
Year(s) Of Engagement Activity 2020
URL https://universitas21.com/news-and-events/news/u21-early-career-researchers-mark-international-day-a...
 
Description WeAreTechWomen: Techwomen100 Individual Shortlist Katie Severn 
Form Of Engagement Activity Engagement focused website, blog or social media channel
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Industry/Business
Results and Impact A summary of the grant activities was described by Katie Severn on the WeAretechWomen website, and Katie proceeded to be an Individual Winner attending the online awards ceremony.
Year(s) Of Engagement Activity 2020
URL https://wearetechwomen.com/katie-severn-university-of-nottingham/