In recent years, a new approach for building artificial intelligence (AI) systems has appeared. The idea is to build a single AI system using a huge amount of information creating what is known as a "foundation model". This foundation model learns from the data some of the basic rules but is not necessarily useful in itself. Instead, it is a building block that can be taken by someone to then further train and construct an AI to perform a task. A single foundation model can therefore give rise to many AI applications. In this project, we will explore whether foundation models could be produced for the development of clinical risk prediction models. Clinical risk prediction models take in information about a person's health and then produce output which tells us how likely they might experience some health outcome of interest in the future. For instance, a risk prediction model might use a patient's current health conditions, blood pressure measurements and body mass index to predict their future risk of having a stroke. Risk models are able to do this by learning from historical data about how often people with similar health conditions experienced the same type of health outcome in the past. When building risk prediction models it is typical that we first find data on a relevant group of patients, analyse their data and then create the prediction model. However, if we change the data (e.g. to include cholesterol measurements) or we want to predict something different (e.g. the risk of heart failure instead of stroke), we will often have to build an entirely new prediction model from scratch. If this is done manually, it would take a large amount of time and resource, and even if it were possible, medical regulators do not have time to evaluate all these possible creations. This is where foundation models could be useful since they could be used to rapidly produce risk prediction models based on any combination of input and output data that is desired. This is important for patients with two or more chronic health conditions (or multiple long-term conditions). They are often under-served because most medical guidance and practice is often based upon how to treat one condition at a time. It is not always clear to doctors how to treat multiple conditions but using AI to analyse historical medical records, they could identify what has worked best in the past and use this as guidance as to how to best treat patients in the future. The challenge is that one moment doctors could be faced with a patient with type 2 diabetes, high blood pressure and osteoarthritis who is looking for long-term chronic pain management and in the next they might be concerned about the risk of stroke of someone with dementia, obesity and high cholesterol. Do we need to build separate prediction models for both? What if another patients arrives with a different set of conditions, do we need a model for them as well? Our research proposes to build an automated system, based upon a general clinical risk prediction foundation model, which would allow doctors to build their own customised prediction model on-the-fly. Given an individual's conditions, the system would automatically construct an appropriate prediction model for that person, giving doctors the individualised guidance for that patient. We will investigate how to build such systems safely and whether the performance is at least as good as models which are built manually. The ultimate aim is to use AI to empower doctors with the ability to make better decisions based on the use of health data.

Thanks to the development of drugs known as antiretrovirals, people living with HIV (PLHIV) can live long, healthy lives. Unfortunately, PLHIV need to take antiretrovirals for the rest of their lives. Antiretrovirals are generally safe but can cause serious side effects in some people, particularly with long-term use. Common side effects are discovered in clinical trials. If a drug causes side effects that are too severe or too common it will fail the trial. It isn't possible to test enough people in a clinical trial to discover less common side effects. These are found by monitoring people taking the drug in the real world. It is also essential that the safety of a drug is monitored in people of all ethnicities because some side effects are more common in people belonging to a particular ethnic group. Our work focuses on the 3.5% of the Ugandan population - 1.5 million people - who live with HIV. At the moment we have very little information about how many PLHIV suffer side effects due to antiretrovirals. The importance of encouraging and enabling healthcare professionals to report drug-related side effects is recognised by the Ugandan government. However, systems for monitoring drug-related side effects have only recently been developed in Uganda and the number of reports is very low. Only 400 reports on side effects due to antiretrovirals were made during the 12-months from October 2018 to September 2019. We urgently need to improve reporting of drug-related side effects due to recent changes in the treatment offered to PLHIV in Uganda. In 2018, Uganda began a programme to rapidly roll-out antiretroviral combinations including dolutegravir (DTG), the new drug recommended by the World Health Organisation (WHO), to PLHIV. Uganda is also rolling-out Isoniazid Preventive Therapy (IPT) to prevent active tuberculosis - the main cause of death in PLHIV. Although DTG has some important advantages over other antiretrovirals, we know that in some people it can cause liver damage, high blood sugar, anxiety, insomnia or depression. In addition, the risk of side effects is likely to be higher when DTG and IPT are taken together. We aim to test whether reporting via a mobile application is effective at increasing reporting of antiretroviral-related side effects by healthcare professionals. If successful, our project will also improve our understanding of which side effects are most common in Ugandan PLHIV and how many people they affect. These are essential first steps in our work to make sure that every PLHIV is treated with the right antiretrovirals at the right dose in the future. The mobile application that we will test is called Med Safety(R). Med Safety(R) was developed by a European drug safety project and adapted for Uganda's National Drug Authority (NDA) by the UK's Medicines and Healthcare products Regulatory Agency (MHRA) but isn't yet widely used. We will recruit 3820 healthcare professionals from 382 HIV treatment centres to: 1) investigate factors that affect the success of rolling out Med Safety(R) among healthcare professionals and how healthcare providers feel about using the application; 2) discover whether using Med Safety(R) leads to more reports of drug-related side effects than the traditional web- and paper-based forms; and 3) whether using Med Safety(R) saves money for healthcare providers. We will also train researchers in drug safety. This project will show whether Med Safety is effective at improving the reporting of drug-related side effects by healthcare professionals. Our learning from deploying the Med Safety(R) application across a population that encompasses large, developed cities and isolated rural areas will be invaluable for wider global efforts in drug safety monitoring. Our strong links with National and International agencies including the NDA, MHRA and WHO will help to ensure that our work improves the safety of PLHIV.

Regenerative medicine aims to replace or regenerate human cells, tissues or organs to restore or establish normal function. Any newly developed products much be subject to stringent pre-clinical safety tests before use in humans. Currently, we mainly use animal models for pre-clinical safety testing, but not only does this bring ethical concerns, but animals are biologically different to humans, so sometimes the testing can provide poor indication of the product is actually safe in humans. We urgently need more representative models which contain human cells, to see how our bodies will react to the new products. Our project is focused on the design of new, improved models for the predictive safety testing of new regenerative medicine products. The models we will create are called "organ-on-a-chip" models, as they create a miniature biologically correct human organ on a small device reminiscent of a computer chip. We can input the new regenerative medicine products into the organ of choice built onto the chip, to see how that organ will respond. We will work one of the largest manufacturers of organ-on-a chip models, a company called Emulate Inc. They have already created organ-on-a chip models of lung and liver which we will establish in our laboratories, before we create brand new models of the different musculoskeletal tissues prone to injury, so we can test new regenerative medicine products developed for any of these tissues. Working with Emulate is important as they have developed a commercially available platform within which we can develop the models, making it easier and faster to translate our new models to commercial products, available to all companies, clinicians and researchers trying to create new drugs or regenerative medicine products.

While it has been recognised for some time that small particles from vehicle exhausts and other traffic related pollutants cause a range of health effects, policy within the UK and Europe has not directly targeted these. Emissions from vehicles and ambient air itself are regulated in terms of total particles, with no specific targeting of one component or another. While this is clearly prudent in that it potentially drives reductions in all types of particles, it is also inefficient as it is likely that some particles (or particle components) are more toxic than others. This project seeks to elucidate the more toxic components of the pollution mix in London, with particular emphasis on traffic generated particles. If successful this will inform a more focussed and more efficient policy process for regulating vehicle emissions and ambient air quality. As well as regulating vehicle emissions and ambient air, policy makers - particularly at local and regional government level-can influence air pollution impacts through traffic management and wider planning decisions. Here the project will provide better information on spatial and temporal exposures and their relation to adverse impacts of air quality. This dynamic exposure information will be a major step forward in assessing the scope for more focussed traffic and infrastructure planning and management in London, with possible applications elsewhere in the UK. Even though there is a substantial literature confirming the impact of traffic pollution on health there are still substantial gaps. There is very strong evidence that exposure to traffic pollution causes asthma exacerbations in children and reasonably strong evidence that it may cause other health effects including the onset of childhood asthma, non asthma respiratory symptoms, impaired lung function, total and cardiovascular mortality and cardiovascular morbidity. In this project we will undertake a number of new investigations to examine the relationship between chronic exposure to traffic pollution and health. These will include studies of mothers in pregnancy right through to senior citizens. These innovative studies will include health outcomes rarely if ever available for investigation of air pollution effects (e.g. primary care data, child cardiovascular risk factors). The use of exposure metrics on a fine spatial scale that are in routine use for policy in London will enable exposure response relationships to be used for quantifying policy options in terms of health impact. Further this will enable us to evaluate the health impact of trends in exposure to traffic related air pollution, most specifically the Low Emission Zone for London (LEZ).

AIMS AND OBJECTIVES: We aim to reduce the cost, burden, and time to conduct randomised controlled trials (RCT) to improve the care of sick and preterm babies and develop medicines, diagnostics and devices that meet their needs. RCT are the most reliable and fair way to test treatments because every patient has an equal chance of receiving the best approach. Data collection is a major cost of RCT hence we will use data recorded as part of clinical care that is already available in a high-quality repository we established, the UK National Neonatal Research Database (NNRD). We will also use digital technologies and innovative study designs. The idea for NNRD-RCT is simple but the technical processes to flow data securely, develop and implement robust, transparent operational procedures, ensure trust in processes, and deliver solutions at scale, are complex, and our principal objectives. NEED: Neonatal care is an area of great national and global need. Newborn health sets trajectories for life; e.g. a baby born preterm, or over or underweight, is at up to 8 times greater risk of developing a chronic disease like diabetes or high blood pressure in adult life, that will decrease healthy life expectancy by as much as 15-20 years. Though improving their health has life-long benefits, newborn babies are disadvantaged in accessing biomedical innovations because of real or perceived difficulties in conducting RCT involving them. There has only ever been a single medicine specifically developed for a newborn condition, and over 90% of medicines used to treat babies have inadequate data on safety, dose, and efficacy because they have only been evaluated in older patients. This is dangerous because the way in which medicines work in babies is often very different from other age groups. Another difficulty is that reliably resolving healthcare uncertainties often requires the participation of large numbers of infants; e.g. to identify whether a treatment reduces severe retinopathy of prematurity, the major cause of childhood blindness, by 25%, would require a study involving about 14,000 infants which would be costly, difficult, and take a long time. Therefore many RCT, though needed, are never done or are too small or poor quality. The consequence is that there are many uncertainties in even routine aspects of care such as the best type of nutrition for preterm babies, and the best way to reduce the risk of infection. WHAT WE WILL DO: We will create standard operating processes to flow data from the NNRD to RCT master-files and electronic forms and digital tools e.g. for automated reminders and staff training. We will obtain stakeholder perspectives to build understanding and trust in NNRD-RCT and develop multi-media communications and transparent, well-governanced processes for commercial RCT. We will involve stakeholders in developing two exemplar NNRD-RCT (to resolve a long-standing, priority-ranked uncertainty in clinical care and to evaluate a neonatal medicine). We will also develop statistical and design solutions that maximise NNRD-RCT efficiency. Outputs will include resources to help clinical investigators, and guidance for industry researchers. WHY A PARTNERSHIP IS NEEDED: A partnership is needed to bring together expertise in scalable technical solutions that meet regulatory standards, develop NNRD-RCT design options, secure stakeholder involvement and engagement, and create impactful communications. Our partners are Bliss (national charity for sick and preterm babies), Health Data Research UK (national institute for health data science), Neonatal Society (clinical research society), Strategic Intelligence Alliance for Healthcare (SME supporting the NHS and industry), OpenClinica (provider of clinical trial cloud technologies) and UK Medicines and Healthcare products Regulatory Agency. The resources developed, know-how and knowledge, will be available to all clinical trials units and researchers around the world.