Designing Data Stewardship Models for Artificial Intelligence (AI) R&D

Description

Summary of the work We want an organisation to design and assess the feasibility and merit of data stewardship models that could increase visibility over health data, transparency over its use, and empowerment of patients and the public in decisions about granting access to it for AI purposes. Expected Contract Length The project may span up to 4 months and must be completed by mid August 2023 Latest start date Monday 17 April 2023 Budget Range The budget is up to £150,000 (ex VAT) for the entire contract Why the Work is Being Done The NHS AI Lab in the NHS Transformation Directorate seeks to enable the safe, ethical and effective use of AI-driven technologies in health and care. The Lab launched the AI Ethics Initiative in February 2021 with the aim of supporting research and practical interventions that could strengthen the ethical adoption of AI within the sector. This call for research intends to explore how participatory approaches to data stewardship models could be implemented in practice. A participatory approach could mean involving people in the “design, development and deployment of data governance frameworks”, or with emerging technological advances like Personal Online Datastores, it could mean enabling patients to have greater visibility of data about themselves and making decisions that they would typically entrust to third-parties. This specification was informed by a deliberative public dialogue held with Sciencewise in July 2022 to understand the risks and benefits of particular data stewardship approaches from the perspective of the public and their expectations for how patients and the public should be engaged in decisions about data access for AI purposes in health and care. The project may span up to 4 months and must be completed by mid August 2023. Problem to Be Solved Data stewardship involves making decisions about who gets to access data for research and development purposes. Current approaches to data stewardship lack adequate and appropriate involvement of the public, which may be needed given the particular ethical risks that AI systems pose and the need to increase the transparency and accountability of these systems. The public dialogue we conducted highlighted a need for alternative data stewardship models that can increase visibility over health data, transparency over its use, and empowerment of patients and the public in decisions about granting access to it for AI R&D purposes. We are looking for an organisation to design options for data stewardship models under these objectives and the themes and principles surfaced by the dialogue participants (see accompanying reports in G Drive). The successful bidder will also be required to demonstrate the feasibility and merits of the models that they design, including that these can be reasonably tested through a subsequent pilot stage (The piloting of data stewardship models will be conducted by separate contractors in a subsequent stage to this competition). More details can be found in the accompanying G Drive: https://drive.google.com/drive/folders/15eSSo-zXWGRas5QmKYqf8LgsTu7y2vtj?usp=share_link Who Are the Users The users could vary, but we anticipate that they may include doctors, data scientists, legal and ethical experts from within, as well as independent of, the NHS, patients and lay members of the public. The users will depend on the type of data stewardship model being designed. Early Market Engagement As noted, the specifications for this research call were shaped by the findings from a public dialogue we held to deliberate on different approaches to data stewardship. The findings from the dialogue can be found in the accompanying reports in the G Drive. The designed data stewardship models will need to be responsive to the cross-cutting themes and principles surfaced by the dialogue participants. In their design of data stewardship models, bidders can consider: • how the models deliberated on as part of the public dialogue can be strengthened, including through adaptation and/or elaboration, in response to the limitations and questions that participants identified • how some aspects of the different models deliberated during the public dialogue could be combined or work in tandem (as opposed to operating as mutually exclusive) to create a novel approach that meets the identified cross-cutting themes and principles • how an entirely new data stewardship model can be designed to better meet the public dialogue’s cross-cutting themes and principles. Work Already Done N/A Existing Team The supplier will be overseen by the AI Ethics Initiative team within the NHS AI Lab. Frequency of contact and liaison will be decided with the bidder. The main point of contact will be a Research Manager, with oversight from the Head of AI Research & Ethics. Current Phase Discovery Skills & Experience • Demonstrated experience in GDS discovery research • Demonstrated experience in user research involving health and care • Demonstrated experience in conducting research involving the management and use of data or related areas • Demonstrated experience in conducting feasibility studies involving options for the management and use of data or related areas • Strong quantitative and qualitative user research experience • Strong stakeholder management skills and experience, including experience of engaging with healthcare professionals, patients and the public Nice to Haves • Work collaboratively, sharing knowledge and experience • Provide evidence of leading in project governance and delivery • Provide evidence of building relationships quickly Work Location The work can be performed from any location, but it will be coordinated by staff at the Department of Health and Social Care, 39 Victoria Street, London, SW1H 0EU Working Arrangments Current circumstances mean that we will work with individuals by phone or video conferencing. This provides more flexibility, particularly when also communicating with staff outside the department. There are currently no Covid-19 restrictions. It is possible that we will expect face to face meetings where appropriate. Security Clearance Baseline Personnel Security Standard (BPSS) is the lowest level of security control available for personnel. Additional T&Cs No additional terms and conditions will be used No. of Suppliers to Evaluate 4 Proposal Criteria • Approach and methodology • How the approach would meet our requirements • Estimated timeframes for the work • Identified risks and dependencies and offered approaches to manage them • Value for money • Team experience Cultural Fit Criteria • Take responsibility for their work • Be transparent • Share knowledge and expertise • Have a no-blame culture and encourage people to learn from their mistakes Payment Approach Fixed price Assessment Method • Case study • Work history • Reference • Presentation Evaluation Weighting Technical competence 70% Cultural fit 10% Price 20% Questions from Suppliers 1. Is there an expectation to test these models with patients? The piloting of the designed data stewardship models will be conducted by separate contractors in a subsequent stage to this competition.However, we expect that you will use quantitative and/or qualitative research to involve users (including patients and the public) to inform the design of the data stewardship models. For example, such research activities can be used to address the limitations and questions that the public dialogue participants have raised about data stewardship approaches. The extent of these activities will be framed by the available timeline and budget for this project. 2. Please could you clarify the meaning of the phrase “management and use of data or related areas”. Specifically could you provide examples of what you mean by related areas and by management do you mean governance? These requirements relate to broad areas of skills and experiences in conducting research and feasibility studies in relation to data governance (collection, maintenance, and access of data), data architecture, data integration, and data lifecycle management. 3. By AI in this context are you referring specifically to AI as medical device or more broadly? The context for this research call involves decisions about granting access to health data for AI purposes. These purposes can be broad, and not only relating to the development of medical devices. For example, they could relate to forecasting and the use of natural language processing. 4. Could you please provide clarity on the rates for the opportunity? It is not clear what clarity is being requested here, however please refer to the 'Budget range' and 'Payment approach' within the published opportunity which will provide you with the necessary information at this stage. 5. Your brief identifies needs for deep technical knowledge & the ability to suggest, shape and progress data-specific ideas. Your questions are more typically ‘user engagement / market research’ themes. Our interpretation is that you need BOTH deep data-technical capabilities AND user/market research competence. Please confirm: have we interpreted this correctly, and what is your preference on mix between these different capabilities (80:20 / 50:50 etc)? Your interpretation is correct. As per the ‘Essential skills and experience’ section, we will be evaluating experiences in user research and research involving the management and use of data. Please note that there are no weightings attached to each of the individual criteria at this first stage. This will be introduced for those invited to the second stage. In your considerations for technical capabilities, please note that this is a discovery phase and there will be a subsequent phase for trialling the designed models to be conducted by separate contractors. 6. When you use the term feasibility study, what are you looking to see that differs to a discovery? This project will involve discovery phase activities, including feasibility studies as described in the ‘Service Description’ document. This will involve the successful bidder demonstrating the feasibility and merits of the models of data stewardship that they design in two ways: as part of the broader data access system; and that these can be reasonably tested through a subsequent pilot stage. 7. The opportunity says that, “The successful bidder will…be required to demonstrate the feasibility and merits of the models that they design”. Please confirm that you expect this to be achieved during the design phase through a Proof-of-Concept model? This competition relates to a discovery phase, and all requested activities are to be completed as part of this phase. 8. The opportunity says that, “The successful bidder will…be required to demonstrate the feasibility and merits of the models that they design, including that these can be reasonably tested through a subsequent pilot stage”. Will the subsequent pilot stage be awarded through a future competitive process, or will the successful bidder for the design stage be offered a contract for the pilot stage? Following the completion of this project, the piloting of the designed data stewardship models will be awarded through a separate and subsequent competition. 9. Could you confirm if the work required is any of the following expectation points below ?Expectations :• Manage Validations and DQ requirements or issues?• Creating guideline documentation regards to the proposed data ?• Privacy and Data security planning?• IG and Data Governance regulations ?• Management of metadata, overall Data management?• Requirement of data analytics and provide solution to potential problems? Please refer to the ‘Skills and experience’ section for guidance on what is expected for this work.