Partners: Massachusetts Institute of Technology, USA
The goal of this proposal is to develop, implement, and evaluate machine learning algorithms for trustworthy clinical AI. The proposed technology aims to support a wide range of healthcare applications, including diagnostics, treatment personalization, and prediction of treatment outcomes. These applications utilize diverse sources of data, including clinical notes, images, and test results. Rather than designing a disease/application-specific AI algorithm, we are interested in developing a capacity that can be readily embedded into existing AI models to improve their transparency and ability to incorporate human feedback as well as strengthen their data privacy guarantees. The Jameel Clinic team has deployed AI tools in many clinical areas. These tools would be our testing ground for measuring the effectiveness and flexibility of the proposed platform.
The COVID-19 pandemic has once again exposed the lack of coordination and cooperation across the research and development efforts of the infectious disease community. A key indicator of this failing is the lack of international, interoperable data platforms to provide rapid insight into disease pathogenesis and treatment. Access to data is an important tool for accelerating evidence, product development, and scientific innovation. Enhanced technology, infrastructure and systems for data platforms will increase Findability, Accessibility, Interoperability, and Reuse (FAIR), so that science can advance in a rapid, robust and innovative way, saving lives in the affected communities. Recognising the importance and complexity of the task at hand, this project brings together the vast experience of IDDO, Vivli, and other key stakeholders to deliver a broad, effective, and sustainable data platform(s) for infectious diseases. In this project, IDDO will: - Enhance its platform for improved findability, discovery, and persistence which ensures that data are available in the long-term; - Optimise and accelerate its data curation capacity; - Streamline and accelerate its data access workflow; - Develop a business plan to secure the sustainability and success of the platform
OpenSAFELY is a new secure analytics platform for electronic health records in the NHS, created to deliver urgent results during the global COVID-19 emergency. It is now successfully delivering analyses across more than 24 million patients’ full pseudonymised primary care NHS records, with more to follow shortly. All our analytic software is open for security review, scientific review, and re-use. OpenSAFELY uses a new model for enhanced security and timely access to data: we don’t transport large volumes of potentially disclosive pseudonymised patient data outside of the secure environments managed by the electronic health record software company; instead, trusted analysts can run large scale computation across live pseudonymised patient records inside the data centre of the electronic health records software company. This pragmatic and secure approach has allowed us to deliver our first analyses in just five weeks from project start.
Partners: Global Health Centre, Graduate Institute Of International and Development Studies
The Lancet Global Burden of Disease Study 2017: a fragile world indicates that progress toward achieving the SDGs is highly uneven, within and between countries. There is great concern that health targets will not be achieved by 2030 unless there is an extraordinary global and domestic effort. While research has shown accelerated declines in child mortality, some countries, especially in sub-Saharan Africa continue to fall behind. Decades of neglect and chronic under-investment have had detrimental effects on the health and well-being of adolescents aged 10–24 years. While technology races ahead the response by international agencies and governments is slow to address issues of governance and regulation. Rapid digital transformation must urgently be used to improve the health of children and youth through reaching out to individuals, strengthening public health and improving service coverage, particularly at community level thus accelerating progress towards UHC. Ethical guidelines, institutional responsibilities, norms and standards need to be put in place as individual data - especially in health - become one of the most valuable resources on the planet. There are great expectations that digital health and AI could bring transformative benefits to health, but their use- especially in resource-poor settings - remains nascent and uncoordinated.
Pharmaceutical companies understand the need for modeling stochasticity, by which we can improve the reliability and effectiveness of drugs across the whole population. Our 2020 review in Trends in Pharmacological Sciences, one of the highest impact journals in pharmacology, argued that deterministic models that ignore stochasticity bias clinical and pre-clinical trial analyses. The quantitative systems pharmacology (QSP) groups within drug companies are the core units that model and predict patient outcomes, dramatically reducing the research and development timeline by predicting which drugs will fail before clinical trials begin. These groups have recently adopted stochastic dynamical models in order to better achieve their goals. Discussing with CJ Musante, Global Head of Pharmacology at Pfizer, the main reason that stochastic models are not more widely used is due to a lack of software that addresses the needs of QSP modelers. Stochastic models are more predictive of patient outcomes, but computationally expensive. Moreover, turnaround times needed by QSP teams are too short to make use of stochastic models with current software. This proposal addresses this problem, proposing to build software tooling which automates the construction and solution of stochastic models in a way that is orders of magnitude faster than what exists today.
Partners: Massachusetts Institute of Technology, USA
Checklists are simple tools that are often used to promote safety in clinical applications, because they can easily be integrated into a clinical workflow deployment without the need for extensive training or additional technology. Currently, machine learning in health focuses on large, complex deep neural networks. In comparison with deep neural networks, checklists are far easier to use, understand, and scrutinize. In practice, checklists are difficult to develop, and the vast majority of real-world checklists are hand-crafted by panels of experts. Given the widespread integration of electronic health data over the past decade, our goal is to learn checklists from data for clinical tasks. The main application of our method, in partnership with Dr. Leo Celi at Beth Israel Deaconess Medical Center (BIDMC) in Boston, is to predict mortality in patients eligible for Continuous Renal Replacement Therapy (CRRT). A generated checklist for CRRT is clinically desirable in the intensive care unit as a mechanism to trigger a multidisciplinary discussion with the patient’s family, to consider the value of the treatment weighed against the physical burden and cost. Thus we include operational considerations such as number of items in the checklist, the false negative/positive rate tradeoff, and fairness across protected groups.
Partners: Vivli (Centre for Global Clinical Research Data)
Data repositories (such as IDDO and Vivli) have played a significant role in making research data available for secondary use. However, access to data needs to be further enhanced for accelerated innovations on infectious diseases by addressing the friction in data contribution, discoverability, access, and reuse. There is a strong alignment between Vivli the Wellcome’s visions of improving the discovery and accessibility of Individual Participant Data related to COVID-19 in the short term, and for other infectious diseases in the long term. With this grant we seek direct support for discoverability and accessibility of data related to COVID-19, and to directly support Vivli to become an effective node in the “FAIR Data Network”. The grant will focus on the following primary objectives: - Recognition and credit for data contributions, - Accelerated access to Individual Participant Data (IPD) hosted by Vivli, - Discoverability of IDDO studies via the Vivli platform, and - Discoverability of IPD hosted by Vivli and its platform partners (including IDDO) based on rich metadata).
Partners: The Graduate Institute of International and Development Studies
I-DAIR is a Geneva-based global platform to enable inclusive, impactful, and responsible research into digital health and Artificial Intelligence (AI) for health. I-DAIR’s mission is the transformation of personal and public health through collaborative research and development of digital technologies. This proposal for a grant of GBP 2 million over two years (2021- 2022) to fund the key capabilities necessary for I-DAIR to fulfill its primary functions as an enabler of digital health and as a convener of key stakeholders for digital health. The grant will complement a CHF 7 million (GBP 5.4 million) investment by Fondation Botnar, which has allowed I-DAIR to start building a Project Team and begin a two year incubation phase, with launch scheduled for 2022. By enabling key strategic outcomes, Wellcome’s contribution will allow the Project Team to make the case for I-DAIR at launch as a global catalyst for digital health and artificial intelligence (AI) research and development (R&D), and for democratising the R&D landscape by bringing more attention to networks, needs and opportunities in low-and middle-income countries.
Make a clear case for and facilitate UK involvement in the joint action with UK and EU stakeholders. - Act as a secretariat to support the UK partner in the joint action. - Prepare contingency plans for maintaining an influence in the absence of UK representation in the joint action. - Gather a body of evidence of NHS and UK expertise on data processing in the health sector to contribute to the European Code of Conduct - Maintain regular updates and briefings on the scope and objectives of the GPDR Code of Conduct, to maximise UK preparedness. - Follow progress of Europe’s data strategy, the European health data space and the development of a legislative proposal on Artificial Intelligence, using the same UK stakeholder group to feed in expertise where possible and provide Intelligence and updates on possible implications for the health sector.
- Execute UK assigned tasks and responsibilities within the Joint action To a high standard, Establishing the UK as a valued and trusted partner. - Gather a body of evidence of NHS and UK expertise on data processing in the health sector To contribute To the development of the European health data Space. - Regularly convene and consult a wide range of UK stakeholders To ensure that UK health data interests across the sector are fully understood and represented. - Maintain regular updates and briefings on the scope and objectives of the development of the European health data Space, To maximise UK preparedness. - identify and pursue opportunities To progress UK interests in the health sector, and on digital policy development more broadly.