Public involvement and engagement in health data science: A national project on reproducible machine learning
Roles and responsibilities
We are looking to recruit 8-12 people to be members of a steering group which will provide support to Health Data Research UK – the national institute for health data science. We would like to recruit members of the public who have an interest in helping to improve how doctors and patients can trust clinical decisions suggested by computers via machine learning.
This steering group of public panel members will be reimbursed for their time and lend their individual perspectives to key discussions about different aspects of the possible use of machine learning to better improve future healthcare decisions.
Our ambition is to draw on the principles of co-production on how we work together, building high quality relationships and valuing your thoughts and perspectives as a key part of the research process. We know that public contributors can help make research better and more useful.
Machine learning offers the potential to help doctors prevent and treat a wide range of human diseases, where computers try to automatically learn new clinical insights from vast amounts of healthcare data. One example is understanding how physical activity and sleep relate to cardiovascular disease, by measuring these behaviours from wearable ‘fitbit-like’ devices. Another example is by using helathcare records to prioritise patients who have additional needs by better predicting their risk of being readmitted to hospital.
However, there are some challenges. How do we know machine learning is reliable? We sometimes talk about reproducibility in machine learning, where the results of many scientific studies are difficult, or even impossible, to recreate. This calls into question whether any clinical decisions supported by machine learning are reliable, and above all trustworthy. At Health Data Research UK, we are creating tools and training material to help researchers responsibly use machine learning methods in clinical care.
In particular, we are keen to ask for your input as we believe public involvement and engagement in research offers an important angle to improve the acceptability, relevance and robustness of health data science research. You and the other public contributors will have the opportunity to contribute as partners in our research process. You can be involved in all stages of the research process, influencing the aims of the research, the design of the study and its analysis and dissemination. In particular we would value your input on questions such as peoples’ preference for computer ‘algorithms,’ what they mean, how we create them and how we understand them, so they can be as accurate as possible.
• Attend and contribute to regular online meetings. Options include once a month, once every 2 months, or once every 3 months. We will record them to help us keep an accurate record. We want to work with you to identify the contribution you can make to this area and maybe to work together to write this up.
• Review emailed documents concerning the design of the research in advance of group meetings.
• Advise on research priorities, their study design, what the findings mean and their possible impact on clinical practice. g
• You may be asked to provide advice by phone or email between meetings depending upon your availability and expertise.
• Interest in and commitment to improving health and social care
• Be able to maintain confidentiality
• Have the time to attend meetings and read documents beforehand
• Be willing to learn about machine learning with the research team.
• Knowledge of experience of health data science e.g. ischaemic heart disease, ischaemic stroke, and peripheral arterial disease
• An understanding of the NHS
• Have access to a computer and e-mail
We will ensure the provision of ongoing support that will be tailored to your individual needs. This will include:
• Advice and assistance from a dedicated point of contact
• Written guidance
Members of this steering group will be paid £20 per hour as recommended by the National Institute for Health Research.
We are looking to recruit 8-12 members by 30th November 2020.
Nuffield Department of Population Health, University of Oxford