PAMBAYESIAN: PAtient Managed decision-support using Bayesian networks

Lead Research Organisation: Queen Mary, University of London
Department Name: Sch of Electronic Eng & Computer Science

Abstract

Patients with chronic diseases must take day-to-day decisions about their care and rely on advice from medical staff to do this. However, regular appointments with doctors or nurses are expensive, inconvenient and not necessarily scheduled when really needed. Increasingly, there are low cost and highly portable sensors that can measure a wide range of physiological values. Can such 'wearable' sensors be used to improve the way that chronic conditions are managed? Patients could have more control over their own care if they wished; doctors and nurses could monitor their patients without the expense and inconvenience of visits, except when they are actually needed. Remote monitoring of patients is already in use for some conditions but there are barriers to its wider use: it relies too much on clinical staff to interpret the sensor readings; patients, confused by the information presented, may become more dependent on health professionals, whose work may be increased rather than reduced.

The project seeks to overcome these barriers by addressing two weaknesses of the current systems. First is their lack of intelligence. Intelligent systems that can help medical staff in making decisions already exist and can be used for diagnosis, prognosis and advice on treatments. One especially important form of these systems uses belief or Bayesian networks, which show how the relevant factors are related and allow beliefs, such as the presence of a medical condition, to be updated from the available evidence. However, these intelligent systems do not yet work easily with data coming from sensors. The second weakness is any mismatch between the design of the technical system and the way the people - patients and professional - interact. We will work on these two weaknesses together: patients and medical staff will be involved from the start, enabling us to understand what information is needed by each player and how to use the intelligent reasoning to provide it. The medical work will be centred on three case studies, looking at the management of rheumatoid arthritis, diabetes in pregnancy and atrial fibrillation (irregular heartbeat). These have been chosen both because they are important chronic diseases and because they are investigated by significant research groups in our Medical School, who are partners in the project. This makes them ideal test beds for the technical developments needed to realise our vision and allow patients more autonomy in practice.

To advance the technology, we will design ways to create belief networks for the different intelligent reasoning tasks, derived from an overall model of medical knowledge relevant to the diseases being managed. Then we will investigate how to run the necessary algorithms on the small computers attached to the sensors that gather the data as well as on the systems used by the healthcare team. Finally, we will use the case studies to learn how the technical systems can integrate smoothly into the interactions between patients and health professionals, ensuring that information presented to patients is understandable, useful and reduces demands on the care system while at the same time providing the clinical team with the information they need to ensure that patients are safe.

If successful, our results will be useful not only for the examples of chronic diseases studied on the project but also for managing other chronic medical conditions, when the same techniques can be applied. Although the project will produce prototype systems, several stages of product development and clinical trials will be needed before real systems are available for patients; we will prepare for these and make a first evaluation of the economic benefits of the proposed systems during the project. Also, several technology companies are involved in the project's Advisory Board to help ensure effective commercial exploitation in the long run.

Planned Impact

Technology has the potential both to improve healthcare and reduce its cost, to the benefit of both patients and healthcare professionals. This project aims to develop a new generation of intelligent decision-support systems to work alongside the latest portable sensors so that patients can manage chronic medical conditions with efficient support from healthcare professionals. Achieving these benefits clearly relies on many strands of research (e.g. sensors, communications), but the contribution of this project will allow digital health technology developers to design systems that place less reliance on continuous supervision by health professionals and allow more patient autonomy.

The project's strategy for impact is: (i) early engagement with clinical stakeholders and patients, leading into clinical evaluation through a sequence of pilots and then trials, in parallel with (ii) progressive engineering refinement of the research prototypes developed within the project, in partnership with companies, all supported by (iii) attention to ethical and regulatory requirements and economic viability from an early stage so that these do not become barriers. This strategy will be applied firstly in the context of three medical case studies, in the management of rheumatoid arthritis, diabetes in pregnancy and atrial fibrillation but equally extends to the application of the technical advances to the management of other chronic medical conditions.

Early in the project, we will set-up clinical focus groups to ensure that technology solutions fit the actual problems. Initial engagement with patients and health care professionals will be around the case studies with patient advisory groups formed from existing PPI groups. Later, we will also engage with wider patient groups such as Patient Opinion, UK eHealth Association, Healthwatch and NHS Choices to explore the next stage of the translation of our research into practice. Agile and startup companies have an important role in introducing this disruptive technology, and a group of UK medical technology companies (BeMoreDigital, Mediwise, Rescon, SMART Medical, uMotif) as well as IBM UK and Hasiba Medical GmbH, have committed to advising the project (minimum 2 Advisory Group meetings per year) with a view to exploiting the technology.

We will expand this with advice from our innovation team on IP, licensing and spin-outs. The project includes an initial study of the potential health economic impact in the case studies. To address the traditional 'regulatory barrier' we have made safety a core research objective and will progressively engage key regulators (e.g. the Information Governance Alliance) and the Caldicott Guardian and Digital Health lead for Barts Health, with whom initial contracts have already been made.

In addition to the medical beneficiaries, the project promises major impact in the data sciences with advances in methods for building and running Bayesian network (BN) models that combine expert judgment and data. Although many researchers and system developers already use BNs for decision-support and probabilistic analysis, more widespread use is limited by constraints on the size and complexity of models. The project will deliver a framework for integrating models at different levels of granularity, providing a practical way to build more complex models efficiently. Since current state-of-the-art BN technology has extremely limited support for the modelling concepts proposed the research will be of benefit to those applying BNs in other applications. The project will deliver open source code so that academic researchers and users can use the new techniques in their own work without restriction.
 
Description We have developed a systematic way to elicit clinicians' knowledge using the idea of 'caremaps' (in the areas of rheumatoid arthritis, gestational diabetes, and chronic heart disease) to transform these into the required Bayesian network models to be used for intelligent decision support.
Exploitation Route Medical researchers will be able to use the methods to develop caremaps and Bayesian networks for other chronic conditions, while a number of medical companies are already investigating ways to exploit the technology.
Sectors Healthcare

URL https://pambayesian.org/
 
Description Collaborative R&D
Amount £31,000 (GBP)
Funding ID 24111 
Organisation Innovate UK 
Sector Public
Country United Kingdom
Start 03/2019 
End 08/2019
 
Description 10th Annual CAS Conference Talk 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Schools
Results and Impact A Talk for teachers on what machine learning /AI is about and practical ways to include it in the curriculum
Year(s) Of Engagement Activity 2018
URL https://teachinglondoncomputing.org/machine-learning/
 
Description CAS London Annual Conference 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Schools
Results and Impact A workshop to explain AI/Machine Learning and how it can be practically done in the curriculum.
Year(s) Of Engagement Activity 2018
URL https://teachinglondoncomputing.org/machine-learning/
 
Description Invited Keynote Speech 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Public/other audiences
Results and Impact Keynote speaker at the International Symposium on Advanced Electrical and Communication Technologies (ISAECT 18), 21 to 23 November 2018, Rabat, Morocco.
Year(s) Of Engagement Activity 2018
URL http://www.isaect.org/akram-alomainy/
 
Description Invited Lecture to the Bournemouth Skeptics Society "Fallacies of Probability and Risk" 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Public/other audiences
Results and Impact This was attended by about 80 members of the general public (and the talk jas been viewed by many more online). There was a lively debate that followed and multiple follow-up interactions in which people said that their views had changed as a result.
Year(s) Of Engagement Activity 2018
URL http://bournemouth.skepticsinthepub.org/Event.aspx/16558/Fallacies-of-Probability-and-Risk
 
Description Invited Talk to the Portsmouth Skeptics Society 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Public/other audiences
Results and Impact 70 people attended this evening talk, but the event was later publicised on twitter and on other websites and the presentations has been viewed widely internationally. This was my second talk at one the UK's skeptics societies (previously did Brighton) and have since been invited to do one at the Bournemouth skeptics in May 2018.
Year(s) Of Engagement Activity 2017
URL https://www.flickr.com/photos/revupreview/sets/72157688176562264/
 
Description Invited seminar at Royal London Hospital: "AI for healthcare relies on smart data rather than big data" 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact This was an invited presentation where the audience were primarily clinicians and clinical researchers. However, the online presentation (see link) has been seen by many others.
Year(s) Of Engagement Activity 2018
URL http://probabilityandlaw.blogspot.com/2018/11/ai-for-healthcare-requires-smart-data.html
 
Description Invited seminar to ERC: "Smart data not big data: Improving critical decision-making with Bayesian networks" 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Policymakers/politicians
Results and Impact 100 policy makers and scientists from the European Research Council attended and the video recording (see url) has been seen by many others. Follow-up discussions suggest the presentation is influencing the way people think about 'big data'.
Year(s) Of Engagement Activity 2018
URL https://webcast.ec.europa.eu/erc-scientific-seminars-prof-norman-fenton-queen-mary-university-of-lon...
 
Description NESTA Public Dialogue on AI 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Public/other audiences
Results and Impact Curzon gave a talk / unplugged demonstration on "What is an algorithm for learning" at a day long deliberative workshop Public Dialogue on Artificial Intelligence run by NESTA/Involve in London, 12 June 2018. He also acted as an expert during the day.
Year(s) Of Engagement Activity 2018
URL https://www.involve.org.uk/our-work/our-projects/practice/artificial-intelligence-what-do-public-rea...
 
Description Ocado Technologies Code for Life Launch 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Media (as a channel to the public)
Results and Impact Gave an invited talk and panel session on the need for a "A Fun Inspiring Computing Education for All" and how AI can play a role at the press launch by Ocado Technologies Code for Life AI:MMO, at CodeNode, London, 22 May 2018.
Year(s) Of Engagement Activity 2018
URL https://www.essentialretail.com/news/ocado-unveils-ai-game-for-schools
 
Description PPI Group Work on Personas for Rheumatoid Arthritis 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Patients, carers and/or patient groups
Results and Impact Project members attended a series of PPI group meetings and held a focus group specifically on the project around personas at the last one of the year. All PPI group members were sent copies of personas developed around rheumatoid arthritis to comment on. The focus group involved discussion around the extent to which the personas captured life with the disease, changes to be made to them and new personas that were needed. This led to a series of changes to the personas and two new personas to be researched and written.
Year(s) Of Engagement Activity 2018
 
Description Presentation of RS-EHR research at Knowledge Management and Data Mining Panel at HEALTHINF 2018 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Postgraduate students
Results and Impact Presentation of research to a panel of academic and clinical research fellows at the HEALTHINF 2018 conference presenting a privacy-protecting framework for generating Realistic Synthetic Electronic Health Records (RS-EHR) using clinical expert input and publicly available incidence and treatment statistics
Year(s) Of Engagement Activity 2018
 
Description Royal Society CPD Session for teachers 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Schools
Results and Impact This was a talk about my research related to Machine Learning given as part of a CPD event for teachers run by the Education Outreach team of the Royal Society in February 2018. The day focused on evidence informed education and dissemination of scientific policy into the classroom. After the talk attending teachers split into groups and worked with facilitators to see how they can embed this input into their teaching to support the curriculum.
Year(s) Of Engagement Activity 2018
URL https://teachinglondoncomputing.org/machinelearning/
 
Description Second Annual PAMBAYESIAN Project Workshop 
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 Workshop for researchers, clinicians, and medical companies associated with the PAMBAYESIAN project (see link below)
Year(s) Of Engagement Activity 2019
URL https://pambayesian.org/2019/02/11/687/