The CIVIC Project: A Sustainable Platform for COVID-19 syndromic-surveillance via Health, Deprivation and Mass Loyalty-Card Datasets
Lead Research Organisation:
University of Nottingham
Department Name: Nottingham University Business School
Abstract
In light of ongoing COVID-19 infections, and approaching second waves, there is urgent need to:
N1. Vastly improve estimation of UK-wide unrecorded cases.
N2. Identify key antecedents of COVID in mass, UK-wide behavioural data, that can power urgently needed early-warning systems at scale; sustainably; and without reliance on self-reporting apps.
N3. Model impact to hidden, vulnerable communities (e.g. food poverty, BAME), to help long-term intervention strategies.
CIVIC is ideally placed to address these needs via unparalleled granularity of access to mass behavioural data; A unique partnership: private-sector data-providers (e.g. Boots, OLIO, Fareshare), academic expertise (Epidemiology, Behavioural Science, AI/Statistics), and public-sector impact partners (ONS, JBC, NHS-X) building an unprecedented platform via 3 interlinked work-packages:
WP1. Partnership with Boots/NHS to generate first-ever, sustainable models of untested COVID-19 cases through interrogation of mass, line-item health/pharmacy transaction data (validated against 111-call-data).
WP2. Identification of behavioural and clinical antecedents of COVID-19 outbreak; processing mass retail loyalty-card/point-of-sale logs via AI/machine-learning techniques, generating near-future forecasts, underpinning early-warning systems.
WP3. Modelling of hidden social/economic impacts to key vulnerable communities, identified in actual behavioural patterns not simple demographic projections.
Each WP has 2 stages. Stage-1 focuses on strictly-anonymized, aggregated data derived from >1.5 billion transactional records, providing crucial deliverables and revolutionizing insights for each of the UK's 32,884 neighbourhoods (LSOAs) within just 4 months. Stage-2 increases fidelity, via individual-level modelling via a ground-breaking "Data Donation" framework.
N1. Vastly improve estimation of UK-wide unrecorded cases.
N2. Identify key antecedents of COVID in mass, UK-wide behavioural data, that can power urgently needed early-warning systems at scale; sustainably; and without reliance on self-reporting apps.
N3. Model impact to hidden, vulnerable communities (e.g. food poverty, BAME), to help long-term intervention strategies.
CIVIC is ideally placed to address these needs via unparalleled granularity of access to mass behavioural data; A unique partnership: private-sector data-providers (e.g. Boots, OLIO, Fareshare), academic expertise (Epidemiology, Behavioural Science, AI/Statistics), and public-sector impact partners (ONS, JBC, NHS-X) building an unprecedented platform via 3 interlinked work-packages:
WP1. Partnership with Boots/NHS to generate first-ever, sustainable models of untested COVID-19 cases through interrogation of mass, line-item health/pharmacy transaction data (validated against 111-call-data).
WP2. Identification of behavioural and clinical antecedents of COVID-19 outbreak; processing mass retail loyalty-card/point-of-sale logs via AI/machine-learning techniques, generating near-future forecasts, underpinning early-warning systems.
WP3. Modelling of hidden social/economic impacts to key vulnerable communities, identified in actual behavioural patterns not simple demographic projections.
Each WP has 2 stages. Stage-1 focuses on strictly-anonymized, aggregated data derived from >1.5 billion transactional records, providing crucial deliverables and revolutionizing insights for each of the UK's 32,884 neighbourhoods (LSOAs) within just 4 months. Stage-2 increases fidelity, via individual-level modelling via a ground-breaking "Data Donation" framework.
Publications
Ban L
(2021)
External validation of a model to predict women most at risk of postpartum venous thromboembolism: Maternity clot risk
in Thrombosis Research
Bonander C
(2022)
A regression discontinuity analysis of the social distancing recommendations for older adults in Sweden during COVID-19.
in European journal of public health
Brewer HR
(2023)
Association Between Purchase of Over-the-Counter Medications and Ovarian Cancer Diagnosis in the Cancer Loyalty Card Study (CLOCS): Observational Case-Control Study.
in JMIR public health and surveillance
Dolan E
(2023)
Assessing the value of integrating national longitudinal shopping data into respiratory disease forecasting models.
in Nature communications
Description | The UKRI CIVIC programme team (University of Nottingham, Imperial College London, University Bristol), along with our stakeholders, has made real-world impact to future plans for attending to UK disease response. This is thanks in no small part to the genuine advances that have already been made during this exciting UKRI project: * Development of a unprecedented data platform housing >2.1billion real-time transactions, focusing on over-the-counter medication sales across the UK (Boots/Tesco/Olio), but allowing linkage with: - EPD (Pharmacy Data) - Supply Data (Alliance) - Wellbeing Data (Coop) - Line-item healthcare transaction data (Boots PLC) - Individually Donated Line-item transaction data (Tesco) - Disease related public events (UK Government) - Mobility Data (Google) - Demographic Data (ONS) - Food insecurity (OLIO) - Geographical/Geospatial Data (Google, OS, ONS) * Thanks to this data platform and the work of its partners (NHS, Boots, Co-op) the CIVIC programme has now made significant breakthroughs in passive syndromic surveillance for the UK. CIVIC's AI models are now not only achieving 81% accuracy in predicting respiratory disease deaths across all UK local but doings so: -- At Local Authority Levels -- Forecasting over three weeks in advance -- When case data for the disease does not exist. -- Even in the face of rapid behavioural changes of the UK population * Due to the success of model forecasting given the integration of sales data, initial work has commenced to operationalize outputs with the NHS-X Early Warning Team. NHS-X are now keenly seeking to extend and implement this work to take models to practice, if we are granted continuation in follow on proposals. * The advances made on CIVIC, combining self-mediation loyalty card data with machine learning models, are proving the genuine possibility of operationalizing an "early warning" systems for fine-grained UK syndromic surveillance that: - Passively use retail behavioural data. - Do not require external apps / encumbrance on the general public - Function even in the complete absence of new testing procedures (e.g. so would apply to new COVID variants / forms of respiratory disease). Beyond disease surveillance this the CIVIC programme has shown further impacts via its modelling of UK vulnerability, specifically in nutrition and food insecurity. Please see: - Work demonstrating pre- and post- COVID impacts on food insecurity: https://foodinsecuritymap.uk (user:guest/pass:guest) - National mapping of nutritional and food vulnerability, to better understand COVID long term impacts: http://www.foodinsecuritymap.co.uk/ (user:olio, pass: r9a5b2) Thanks to UKRI's support, and the success of this work, we aim to seamlessly move our stakeholders (Boots/NHS/Co-op/ONS/JBC) towards a next-stage operationalization plan, but also to make possible a new opportunity to improve the system coverage/fidelity even further through introduction of a new partner seeking to contribute to the platform for purposes of social good (Co-operative Ltd), to increase UK surveillance for future disease risks. |
Exploitation Route | We are deeply committed to taking the outputs of CIVIC forward via: * Integration into NHS Early Warning Systems (ongoing) * Through the CIVIC program's direct contribution to the establishment of the UK Data Donation Advisory Council (www.datadonation.co.uk). * Further integration of its outputs into policy/evidence collection (as per CIVIC's contribution to "Sharing data for public good: How can privacy be protected?" UK Parliament Evidence" - see https://committees.parliament.uk/writtenevidence/43526/html/) * Full extension and operationalization of CIVIC's "Epi-consumer" methodology, linking individual cohorts to healthcare transactions, to improve disease understanding/prediction. Specifically in follow on work: - Operationalize models within NHS into daily business practice - Advance underlying AI (addressing key issues of feature drift and model explainability) - Further integrate individual data linkage to increase model robustness/responsiveness (via ZOE app, BioBank CSS, and the National Cohorts) - Add further data streams via (a) new partners (Co-op/Alliance) and (b) Data donation programmes (Our Future Health) - Explore sustainable business models to ensure the longevity of the system with stakeholders |
Sectors | Digital/Communication/Information Technologies (including Software) Environment Healthcare Government Democracy and Justice Retail |
URL | https://www.nlab.org.uk/project/civic |
Description | Please see Key Impacts/Findings Research fish entry for full details (this has been combined with "narrative impact" for the programme) |
First Year Of Impact | 2021 |
Sector | Communities and Social Services/Policy,Healthcare,Government, Democracy and Justice,Retail |
Impact Types | Societal Policy & public services |
Title | Food Vulnerability Mapping |
Description | An output of the CIVIC platform, first ever UK wide food vulnerability models (acting at >30,000 MSOAs across the UK). https://foodinsecuritymap.co.uk (user:olio, pass: r9a5b2) Follow on pre- post- COVID impact work on vulnerable communities with Havering County Council: https://foodinsecuritymap.uk (pass:guest/user:guest) |
Type Of Material | Data analysis technique |
Year Produced | 2021 |
Provided To Others? | Yes |
Impact | * Use by Havering County Council to direct post-COVID resource management, in partnership with OLIO * Engagement from Greater London Authority to replace current insecurity surveying methodologies * Investment to expand the platform from Guy & St Thomas Foundation For more details please see: https://foodinsecuritymap.uk (pass:guest/user:guest) |
URL | https://foodinsecuritymap.uk |
Title | The CIVIC Data Platform |
Description | An unparalleled data-source for the analysis of "epi-consumption": the relationship between consumption/medication sales/disease incidence, for hidden COVID incidence estimation, hospitalization forecasting and variable importance analysis. This database platform links together data from: * EPD (Pharmacy Data) * Supply Data (Alliance) * Wellbeing Data (Coop) * Line-item transaction data (Boots PLC) * Individually Donated Line-item transaction data (Tesco) * Disease related public events (UK Government) * Mobility Data (Google) * Demographic Data (ONS) * Food insecurity (OLIO) * Geographical/Geospatial Data (Google, OS, ONS) The platfom maintains ~2.1billion real-time transactions, focusing on over-the-counter medication sales across the UK (Boots/Tesco/Olio). Due to this plaform the CIVIC programme has now made significant breakthroughs in passive syndromic surveillance for the UK, with AI models using the data platform now not only achieving 81% accuracy in predicting respiratory disease deaths across all UK local but: * Forecasting over three weeks in advance * Doing so, crucially, even when case data for the disease does not exist. Due to commercial sensitivity this data platform is only accessible through application. For more details please see: https://www.nlab.org.uk/project/civic https://www.datadonation.org |
Type Of Material | Database/Collection of data |
Year Produced | 2021 |
Provided To Others? | No |
Impact | Due to this dataset, advances have been made in: * Vulnerability Modelling * Food Insecurity Analysis (Havering Council, Greater London Health Authority) * Respiratory Health analysis * COVID forecasting * Inputs into NHS Early Warning Systems (ongoing) It is the aim of the CIVIC program team to continue to consolidate this data platform and work with project stakeholders (via follow on applications) to extend this invaluable future resource for external access. |
URL | https://www.nlab.org.uk/project/civic |
Description | ALSPAC |
Organisation | University of Bristol |
Department | Avon Longitudinal Study of Parents and Children (ALSPAC) |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | Support for "data linkage infrastructure" to support modelling of transactional datasets with the ALSPAC cohort participants, with specific focus on estimation and forecasting of COVID, Respiratory disease prevalence and insecurity/vulnerability in local communities. |
Collaborator Contribution | Ethics review; Data Provision (COVID survey data from ALSPAC participants); Technical Support; Staff time; Meetings attendance; |
Impact | * Outputs and outcomes as per the whole CIVIC project report (all partnerships on the program are intrinsic to deliverables). * Multidisciplinary collaboration (Computer Science, Maths, Epidemiology, Consumer Behaviour) |
Start Year | 2021 |
Description | Boots UK Ltd |
Organisation | Boots UK |
Country | United Kingdom |
Sector | Private |
PI Contribution | Processing and modelling of transactional datasets, providing key insights/features/variables to the CIVIC platform key to estimation and forecasting of COVID, Respiratory disease prevalence and insecurity/vulnerability in local communities. |
Collaborator Contribution | Data Provision (anonymized line-item point-of-sale data covering all UK Boots stores, ~1.5 billion logs); Technical Support; Staff time; Meetings attendance; Advisory Board membership; (n.b. the estimated contribution of Boots in-kind data support is unprecedented here, and as such it is extremely difficult to put a figure on the overall data-access, on top of the £120k estimate for general support) |
Impact | * Outputs and outcomes as per the whole CIVIC project report (all partnerships on the program are intrinsic to deliverables). * Multidisciplinary collaboration (Computer Science, Maths, Epidemiology, Consumer Behaviour) |
Start Year | 2021 |
Description | Cooperative Ltd |
Organisation | The Co-operative Group Ltd |
Country | United Kingdom |
Sector | Private |
PI Contribution | * Processing and modelling of transactional datasets, providing key consumer behaviour insights/features/variables to the CIVIC platform at granular levels, key to estimation and forecasting of COVID, Respiratory disease prevalence and insecurity/vulnerability in local communities. |
Collaborator Contribution | * Data Provision (anonymized line-item point-of-sale data covering all English Co-operative stores); Technical Support; Staff time; Meetings attendance; (n.b. the estimated contribution of Co-op in-kind data support is unprecedented, and as such it is extremely difficult to put a figure on the overall data-access, on top of the estimate for general support) |
Impact | * Outputs and outcomes as per the whole CIVIC project report (all partnerships on the program are intrinsic to deliverables). * Multidisciplinary collaboration (Computer Science, Maths, Epidemiology, Consumer Behaviour) |
Start Year | 2021 |
Description | NHS-X |
Organisation | NHS Digital |
Country | United Kingdom |
Sector | Public |
PI Contribution | Processing and modelling of NHS-X datasets, providing key insights/features/variables to the CIVIC platform necessary to estimation and forecasting of COVID, Respiratory disease prevalence and insecurity/vulnerability in local communities. |
Collaborator Contribution | COVID-19 data liaison; Technical Support/ Schema Support for 111-Call Data access (NHS-Digital); LSOA-level incidence Data; Project discussions; Project Researcher Internship. |
Impact | * Outputs and outcomes as per the whole CIVIC project report (all partnerships on the program are intrinsic to deliverables). * Multidisciplinary collaboration (Computer Science, Maths, Epidemiology, Consumer Behaviour) |
Start Year | 2021 |
Title | https://github.com/nhsx/commercial-data-healthcare-predictions |
Description | Machine learning software that predicts weekly incidence of respiratory disease at and localised death forecasts at local authority resolutions across the UK, using demographic, geographic, ONS and importantly over-the-counter medication sales data. Transferred from the CIVIC project as part of the NHS-X internship program. |
Type Of Technology | Software |
Year Produced | 2022 |
Impact | Ongoing integration into the NHS-England Early Warning Team's software system, mediated by Johnathan Pearson (NHS-X) |
URL | https://github.com/nhsx/commercial-data-healthcare-predictions |
Description | CIVIC project - Model Class Reliance for Demonstrating Variable Importance, NHS Blog |
Form Of Engagement Activity | Engagement focused website, blog or social media channel |
Part Of Official Scheme? | No |
Geographic Reach | National |
Primary Audience | Professional Practitioners |
Results and Impact | A blog written by the CIVIC team in partnership with Johnathan Pearson (NHS-X). Primarily acting an internal facing mechanism within the NHS, this information was available to all NHS departments. Partly due to this exposure the CIVIC project has gained traction with NHS Early Warning systems. |
Year(s) Of Engagement Activity | 2021 |
URL | https://nhsx.github.io/AnalyticsUnit/MCR.html |
Description | Cancer Loyalty Card Study (CLOCS) Annual Meeting 2021, "Lessons from CIVIC project for shopping data donated to investigate ovarian cancer" |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | National |
Primary Audience | Third sector organisations |
Results and Impact | Cancer Loyalty Card Study (CLOCS) Annual Meeting 2021, "Lessons from CIVIC project for shopping data donated to investigate ovarian cancer", given by Goulding J., Dolan. E. on 9th December 2022. This led directly to collaborations with ZOE, Biobank CSS and Our Future Health. |
Year(s) Of Engagement Activity | 2021 |
URL | https://www.clocsproject.org.uk/homepage |
Description | Invited Talk, The Health Foundation (Analysts Seminars), 20th Jan 2022 |
Form Of Engagement Activity | A talk or presentation |
Part Of Official Scheme? | No |
Geographic Reach | National |
Primary Audience | Third sector organisations |
Results and Impact | CIVIC Talk (Goulding, Skatova, 9th March 2022) to analysts at one of the UK's leading health charities, to disseminate knowledge about CIVIC and its approach to disease understanding/surveillance. This has led to ongoing connections, third-sector dissemination and stronger links to Our Future Health (Convened by Adam Steventon, Head of Analytics) |
Year(s) Of Engagement Activity | 2022 |
URL | https://www.health.org.uk/?gclid=Cj0KCQjwuMuRBhCJARIsAHXdnqN3-vZ3tCK5QO6gTarvGLRjspIKDhpFjwgm90VNwEI... |
Description | Invited talk "Etail-connect" conference, 21st Sept 2021, Twickenham London |
Form Of Engagement Activity | A talk or presentation |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Industry/Business |
Results and Impact | Talk (Prof A. Smith, Co-I) on the use of consumer data for social good, and corporate responsibility, focussing on the CIVIC project. The conference has is attended by all UK major brands, |
Year(s) Of Engagement Activity | 2021 |
URL | https://etailconnectautumn.wbresearch.com/speakers/2021 |
Description | Invited talk "Walgreen Boots Alliance" internal research conference, UK, 17 Nov 2021 |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | Talk on the "CIVIC Programme: Diagnosing Disease via Shopping Data", to Boots UK and selected international Walgreen Boots Alliance members. This engagement solidified the CIVIC program with our stakeholder partners, and led to the commitments for follow on operationalization of the platform (convened by Wesley Jones, Boots UK Healthcare Partnership Manager) |
Year(s) Of Engagement Activity | 2021 |
Description | Invited talk, NHSX Analytics Unit, 16st Sept 2021 |
Form Of Engagement Activity | A formal working group, expert panel or dialogue |
Part Of Official Scheme? | No |
Geographic Reach | National |
Primary Audience | Third sector organisations |
Results and Impact | Here we presented progress on the project ot NHS, with the title "Machine Learning analysis of shopping data to investigate respiratory deaths in the UK population, and Explaining the importance of variables (features) using Model Class Reliance (MCR)". This was focussed on NHS-X but attracted a larger audience from NHS-England, and has led directly to ongoing support of the project, and its ongoing operationalization into the NHS Early Warning System (Convened by J. Pearson, NHS-X) |
Year(s) Of Engagement Activity | 2021 |
Description | Keynote at Future Food Symposium, 8th Sept 2021 |
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 | Introductory Talk about CIVIC and the relationship to food vulnerability (Dr John Harvey, Co-I) as the future food symposium, Sept 2021 (Coventry University), to a wide audience of academics, third-sector organizations and local council attendees. |
Year(s) Of Engagement Activity | 2021 |
URL | https://www.coventry.ac.uk/research/about-us/research-events/2020/future-food-symposium/ |
Description | Ovacome Webinar, "Using shopping data to explore diagnosis and donating shopping data", 21 Nov 2021 |
Form Of Engagement Activity | A talk or presentation |
Part Of Official Scheme? | No |
Geographic Reach | National |
Primary Audience | Third sector organisations |
Results and Impact | Ovacome Webinar (Goulding, Dolan), with a talk about CIVIC to women suffering Ovarian Cancer on the potential of using shopping data to explore diagnosis and donating shopping data. This has had a significant impact, gain addition third-sector support for the CIVIC approach, and directly led to engagement with Qualtrics to support studies (and specifically 2000 people recruited for the CIVIC project via their platform) |
Year(s) Of Engagement Activity | 2021 |
URL | https://www.ovacome.org.uk/webinar-recordings |
Description | Social Statistics Seminar, Univeristy of Manchester, "CIVIC and food vulnerability", 12th Jan 2020 |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | Regional |
Primary Audience | Postgraduate students |
Results and Impact | Talk on the 12th January John Harvey (University of Nottingham), at University of Manchester, disseminating knoweldge of the program("CIVIC - Should we measure food insecurity to understand disease impacts? Insights from a novel machine learning approach with the world's largest food sharing network") |
Year(s) Of Engagement Activity | 2021 |
URL | https://events.manchester.ac.uk/event/event:o17k-kguj3nzq-ct0fqp/social-statistics-seminars-202021 |
Description | Talk, CIVIC and Data Donation, Our Future Health, 9th March 2022 |
Form Of Engagement Activity | A talk or presentation |
Part Of Official Scheme? | No |
Geographic Reach | National |
Primary Audience | Professional Practitioners |
Results and Impact | Invited Talk and subsequent Debate/Discussions with the management team, and extended analysts group, at "Our Future Health", leading to ongoing collaborations for extension and operationalization of CIVIC and data donation platforms, as part of Our Future Health's expansion and recruitment goals (w/ Director, Andrew Roddem). |
Year(s) Of Engagement Activity | 2022 |
URL | https://ourfuturehealth.org.uk/ |
Description | Talk, Cooperative PLC, August 31st 2021 |
Form Of Engagement Activity | A talk or presentation |
Part Of Official Scheme? | No |
Geographic Reach | National |
Primary Audience | Professional Practitioners |
Results and Impact | Talk on the "CIVIC Programme: Diagnosing Disease via Shopping Data", to Cooperative UK members. This engagement solidified the CIVIC project with these extended data partners, expanded the program, and led to the commitments for follow on operationalization of the platform (CIVIC-2). Convened by Edward Powell, Co-op. |
Year(s) Of Engagement Activity | 2021 |