Predictive Health Intelligence – the story so far

Introduction

Predictive Health Intelligence (PHI) was established in 2019. What started as an idea and developed into a successful grant application to the National Institute for Health and Care Research is now an active company; publishing papers, exhibiting at conferences worldwide and benefiting patients on a daily basis who otherwise would have been missed. This report sets out the story of PHI; the activities undertaken, the thinking behind the company and the progress made to date. This is not an academic paper nor an appeal for funds, but a summary of where PHI came from, where it is now and where it wants to be.

Background

The liver is the largest gland and the largest solid organ in the body, weighing some 1.8kg in men and 1.3kg in women. It holds approximately 13% (about one pint or 0.57 litres) of your total blood supply at any given time and has over 500 functions.
Liver disease (which has many causes) places a huge burden on public health. One in nine people in the UK suffers from the condition. The British Liver Trust say that in the UK there are more than 10,000 deaths due to liver disease each year. That’s more than 27 deaths a day. There has been a four-fold increase in death rates over the last 50 years.
There are three key facts about the liver that most are not aware of:

1. Liver disease is treatable: The liver has an amazing ability to heal, it can even grow back whole sections after surgery. However, if liver disease is too advanced, the scarring stops the regeneration process meaning that, in the best case, only limited improvements can be made if a patient presents with advanced disease. The key is finding problems early and getting people on treatments as soon as possible. Lifestyle changes are sometimes all it takes to improve your liver health.
2. Liver disease is usually silent: Liver disease usually doesn’t show symptoms until serious damage has already happened, making early diagnosis and treatment challenging. Patients frequently present with advanced and incurable disease, yet the signs of early liver disease are often hidden in previous routine test results.
3. Liver disease is common: Liver disease is a major cause of illness and premature death. It affects around one in nine people, including many people who don’t consume alcohol. It is estimated to cost the NHS £6bn a year. It is sadly unique among the major disease groups in that the mortality rate has actually increased in the last 50 years – for almost all other disease groups it is falling. While as a society we have become far better at managing and treating heart disease, cancer, diabetes and others, the mortality of liver disease has increased by almost 400%.

These three facts combine to create a perfect storm of liver disease in the population, and set the scene for the idea behind PHI.

Why PHI?

The clinical problem that was being seen and needed solving was quite clear:

• The patient perspective: they often didn’t understand why they’d been found so late in their disease’s progression. This was often distressing.
• The clinicians’ frustrations that they were seeing people so late when treatment options were far more limited, despite there being signals that advanced disease was imminent.

Given this troubling situation, Dr Tim Jobson and colleagues at Somerset NHS Foundation Trust wanted to do something about it. Their proposal was to develop a ‘case-finding’ approach to liver disease; using existing blood test results already in the laboratory systems to attempt to identify patients likely to have, or be at risk of developing, significant liver disease. The principle of ‘case-finding’ was already established in several other disease areas including Hepatitis C, and it is recognised that this approach is much cheaper than ‘population screening’, as no new tests are initially performed. To do this, it was necessary to look for patterns of blood tests in the laboratory computer systems, which are owned and used by the NHS, to identify potential patients. The thinking was that there is a vast amount of intelligence in these systems: multiple blood tests going back many years. The NHS already possesses a wealth of information that is underused and could be of great value to people with an undiagnosed, progressive, and treatable disease. Although new in many respects, this was simply a different way of doing on a computer system what liver doctors routinely do in clinic. There was nothing particularly difficult in terms of computer science, no complex ‘Artificial Intelligence’, and no incomprehensible mathematical algorithms. But there were problems. Healthcare computer systems are not designed to be used in this way. Whilst they can report on all sorts of data, their primary use is to show a clinician all the information on a known patient, already identified, not finding a cohort of patients with particular blood test combinations, or trends over time. The team wanted to make it possible for clinicians in hospitals to easily analyse data to look for ‘missing’ liver patients themselves, not rely on supercomputers and highly skilled analysts. The vision was of a liver health ‘Airbnb’ or ‘Rightmove’ equivalent; easy to use and giving quick results. After one meeting Dr Jobson outlined the problems that were being found with this new case-finding approach to Neil Stevens, an expert in healthcare information systems and the use of data in health. After the usual surprise about how patients with potential liver problems can be overlooked, there followed more surprise, this time from the clinicians as it was explained that this problem might well be realistic to solve. Patient involvement in guiding PHI’s approach.
The involvement of the public and patients has been central to PHI from the very outset. At the inception of the project, a public / patient meeting was held with 20 current patients, ex-patients and members of the public attending. The initial part of this meeting was educational – explaining in clear, lay terminology the problem (late/missed diagnosis), the impact (advanced and often incurable liver disease) and the potential solution (data-driven approach to risk-stratification and case-finding – using historic blood tests and other laboratory data). The overwhelming feeling of the group was that people would welcome this if it stopped them becoming ill and that this would outweigh any concerns about use of data. Some were surprised this sort of ‘double checking’ of results wasn’t already taking place. At the same time, it was important to fully include GPs throughout the process and ensure all safety requirements were met for the use of personal data. The outputs of this discussion were used to shape the application to the National Institute for Health and Care Research. For example, the public and patients felt that data from multiple sources should be combined to provide as holistic as possible care record, and that this record should be used for early identification of liver disease. They felt that if this approach to case-finding is used then waiting times for treatment should be kept to a minimum and thus case-finding strategies should be tailored to the treatment capacity available (something which was implemented once the project reached that stage). Another key recommendation was about the involvement of GPs. The project team have been mindful of this whilst also having to take into consideration the impact of capacity shortfalls in primary care and the impact of COVID-19. Thus, GPs are engaged and kept fully informed, but the project team strives to minimise workload impact on GPs.

Use of data

Secure use of data is key to the success of PHI and is a high-risk area if it is not properly managed. One example is the National Care Data system which ran from 2013-2016 and was subsequently closed down – mainly because of concerns about the use of data. One particular public concern was the sharing of patients’ confidential data with private companies. Such an approach is often seen to be damaging and PHI set the project up in such a way that only NHS clinicians would have access to confidential data, and only for patients looked after in their own care system.
Data is key to enabling the care team to find those who are most at risk from liver disease in the future. It is necessary to look at as many blood tests as possible and identify those which show early signs that the person may have problems in the future.
It works like this: Somerset NHS Foundation Trust (SFT) stores a copy of the results of all blood tests taken in Somerset and processed in the central laboratory as part of the patient records that it keeps. This ensures that anyone treating patients in the Somerset health community has access to results which may be critical for diagnosis and monitoring. SFT and the wider health community (e.g. GPs) remain the controllers for all the healthcare data used in this project, and this data is not passed to Predictive Health Intelligence Ltd.
GPs have responsibility for their patients and are legal controllers of some of their data, and so it is important that GPs provide consent for this project to use the results of any blood tests that were taken following requests by their practices. After extensive communications and work with the Local Medical Committee in Somerset over 85% of GP practices have agreed to work with the team and share data. The team also had to work closely with the Data Protection and Information Governance Teams at the Trust and the Caldicott Guardian (a clinician who is accountable for the safe use of patient data).
To prioritise confidentiality, the only pieces of information used by the case-finding database are the individual’s NHS number, date of birth and blood test results which are relevant in the diagnosis of chronic liver disease. No other personal health information is used. Before any data is shared, the specialist data team at SFT removes all other information. This further improves the security of patient data.
The project follows strict NHS guidelines on confidentiality and information governance. This includes a high level of data security which ensures that only authorised members of the project team can access data. The team has also obtained appropriate approvals throughout this project prior to working with any patient data.

Developing a pilot

Once the necessary consents had been obtained – a Data Sharing Agreement with GP surgeries and a Data Protection Impact Assessment with the Trust – it was important to test the concept to see if it could work.
The team decided to recall a small number of patients identified by the system – now named hepatoSIGHTTM – who might be at risk of developing Primary Biliary Cholangitis (PBC), a chronic disease that can, little by little, destroy some of the tubes linking the liver to the gut (bile ducts) but which is manageable with medication if caught early.
The letters to be sent out to people identified by the system were crucial, as it was important to get the balance right between making people aware of the risks, so they responded, but not causing undue alarm. To help this the clinical team supported by the Communications Lead held a workshop in Taunton with about 15 invited members of the public. They were given sample letters to comment on , and also asked their views on other publicity and communication that would help. The letters were amended in the light of their comments.
A detailed communication plan was also developed to ensure that everyone who needed to know about the recall was informed, and measures were put in place to manage any misunderstandings causing concern or generating negative publicity.

Ongoing Approach to Public and Patient Involvement (PPI)

As well as the initial meeting held to inform the grant application to NIHR, the teams involved in developing the product and using it for both research and case-finding for treatment have engaged with the public and patients to help shape the programme of work. Throughout, it has been a priority to follow a best practice approach to PPI to ensure maximum benefit from these activities. The PHI team has a key member of the team (an experienced communications professional) who takes responsibility for managing this aspect of the work including:

• Developing and shaping the PPI plans with public contributors
• Setting and refining overall PPI strategy as the project progresses
• Liaising and planning with the research team regarding PPI activities throughout the project lifecycle
• Acting as a single point of contact for all public contributors involved in the project
• Recruiting public contributors and working with them to develop support and training needs
• Providing appropriate induction and training to public contributors as well as ongoing mentoring and support
• Ensuring that involvement is aligned to UK Standards for Public Involvement
• Communicating to public contributors the project status, providing feedback on PPI activities and their impact
• Planning and organising the logistics (including access and support needs) for all project-related meetings
• Synthesising results and conclusions of PPI activities and feeding back to researchers – only informally
• Developing and implementing appropriate evaluation, monitoring and reporting of PPI
• Writing up PPI sections with public contributors for project reports

This is key as the whole concept behind the Somerset Liver Improvement Programme and the subsequent development of the hepatoSIGHT search engine was driven by patients who were identified late with advanced liver diseases, yet questioned repeatedly in clinic and other healthcare encounters why their results had not been acted on sooner. These negative patient and clinician experiences drove the initial conceptual development of the project.
Post covid pandemic, a more formal Patient and Public Reference Group was established to monitor the progress of the project and ensure regular challenge and questioning of the decisions of the Project Team. Terms of Reference were drafted and agreed, with input from the Trust, Healthwatch and the NHS PPI engagement teams. People were recruited from existing contacts and through the above groups, with articles placed in the newsletters of the ICB, SFT and HealthWatch, and articles placed in the local press to encourage interest.

This group now meets regularly with a formal agenda and with minutes. Each meeting has a significant educational and briefing component with key members of the research team providing background information and project updates. The group meets every 2-3 months and influences key tactical aspects of the company as well as long-term strategy. This group has, for example, been influential in promoting the focus on deprivation in the case-finding approach.

Key principles for our PPI process include:

• INCLUSIVE OPPORTUNITIES – the team advertises widely and supports attendance through expense reimbursement, and accessible venues.
• WORKING TOGETHER – the team values all contributions and builds productive relationships.
• SUPPORT AND LEARNING – the team promotes learning and understanding at all stages, through the meetings, written materials, blogs, podcasts and the project website.
• GOVERNANCE – the team involves the public in research management, regulation, leadership and decision-making – as evidenced by formal records of meetings and project updates.
• COMMUNICATIONS – the teams use extensive communication materials. Plain language is used – with multiple channels including written press, local radio, blogs, podcasts, websites, press releases and social media.
• IMPACT – identifying and sharing the difference that public involvement makes to research.
• TRAINING – It is important to strike a balance between informing PPRG members and offering training and support while retaining their ability to challenge and question. Too little training leaves them ill-informed; too much could reduce their effectiveness in challenging.

The first part of any meeting is dedicated to a training session from the PHI Medical Director. This often covers the types and prevalence of liver diseases and the latest research into liver disease. The PPRG members ask questions and guide where future training might go. The team recently carried out a detailed Training Needs Analysis with the group to ask whether the training was useful and what additional training they would like. There was very positive feedback about the training sessions, with the group feeling informed and involved, and a strong desire for it to continue regularly. In addition requests were made for new training subjects such as social media use and good practice in PPI, and these have been added to the programme.
As well as general updates and information sharing, for specific research projects the approach to data access and management has been discussed and reviewed with the group. Specific feedback is documented in formal minutes of the meetings. Key recent discussions have been around the use of data for identifying patients with hepatitis D and also for population-level risk stratification. The group has been highly supportive of the approach – recognising the validity of data used for these two purposes.
To ensure that the discussions are genuine and not simply supportive, the group is given a range of options and asked to prioritise one over the others. For example, the group was strongly of the opinion that deprivation was a key data set (from a given list of three or four) for the team to identify and analyse. This was therefore prioritised over other options and has led to the paper about Hepatitis and Deprivation stats – see section on publications below.
There has been minimal negative feedback to date. One example was of delay in calling back a patient to book them in for an appointment – and this led to tightening up of the recall process. In general however feedback from patients is positive, and reinforces the importance of our case-finding approach.

Conferences attended.

The team has attended various scientific conferences over the last two years to make connections with interested parties and raise awareness of the technology and the potential use to commissioners, researchers and others. At many of these, abstracts have been presented.

• British Society of Gastroenterology (BSG) – November 2021 – Abstracts presented.
• British Society of Gastroenterology (BSG) – June 2022.
• British Association for the Study of the Liver (BASL) – September 2022 – Abstract presented, see below.
• American Association for the Study of the Liver Diseases (AASLD) – November 2022 – Abstract presented, see below.
• The Healthcare Show – May 2023 – Exhibition stand.
• Digestive Diseases Week (DDW) – May 2023 – Abstract presented, see below.
• NHS Confederation – June 2023 – Exhibition stand.
• British Society of Gastroenterology (BSG) – June 2023.
• European Association for the Study of the Liver (EASL) – June 2023 – Exhibition stand and abstract presented, see below.
• British Association for the Study of the Liver (BASL) – September 2023 – Abstracts presented.
• American Association for the Study of Liver Diseases (AASLD) – November 2023 – Exhibition stand and abstract presented, see below.
• Arab Health – January 2024.

Abstracts published

The team has published a number of abstracts, and presented at conferences, explaining the project. These are summarised in Appendix A

Communications

Throughout the project’s life, there has been interest from both specialist and general media. The team has used the media to raise the profile of the approach and to reassure members of the public who receive a recall letter.
This has included:

• BBC Somerset – Clinical Director interviewed live on three occasions
• Somerset County Gazette – four news stories published
• Surfing the NASH Tsunami podcasts – Medical Director featured twice
• Fatty Liver Alliance podcast with Mike Betel – multiple interviews
• Data Malarkey podcast with Sam Knowles
• MedTech Innovation News – featured multiple times
• Maddyness
• HealthTech Newspaper
• DOTMed
• Inside Precision Medicine
• Digital Health
• European Pharmaceutical Manufacturer

Impact of project

Overall, and looking back at the work of PHI and partners over the last three years, there are five areas of impact worth noting.
Firstly, it is clear that the project has made a difference for individual patients. As the figures above show, the recalls have identified people with PBC, advanced fatty liver disease and hepatitis who would not have been found otherwise. Over 300 people have now been contacted and offered a review of their liver health. Many are now under care that would otherwise not have been provided to them.
Second, the impact has been greater than just the individual. PHI has shown how feasible and easy the case-finding approach is by taking complex and usually difficult to analyse data, shaping it and making it easy to access to present to front line clinicians. This application has significant potential across geographies and disease areas. To be blunt, over time it is possible to change the way patients are found and treated.
Thirdly, PHI has used the data analytical tool to specifically review whether or not the current ‘standard of care’ for liver disease discriminates against people in deprived areas, and after analysis concluded that it does not do so. This new approach to case-finding offers the opportunity for sharper targeting of those areas of deprivation as necessary.
Fourth, this tool has also been seen to be highly effective at finding patients who are appropriate to be entered into clinical trials. This could potentially speed up the recruitment process dramatically, saving money and developing drugs in a far shorter time scale than at present.
Finally, an independent health economics report has looked at the approach and concluded that far from simply being good value for money, the use of hepatoSIGHT will save money each time it is used.

Appendix A – Summary of research abstracts

• PREVALENCE OF UNDIAGNOSED PRIMARY BILIARY CHOLANGITIS AND THE POTENTIAL UTILITY OF A ‘CASE-FINDING’ APPROACH 1 Katharine Hutchison*, 1 Angus Kitchin, 1 Amin Abdulgader, 2 Sophie Stretch, 1 Timothy Jobson. (1 Somerset NHS Foundation Trust, Taunton, UK; 2 Bristol University Medical School, Bristol, United Kingdom) 10.1136/gutjnl-2021-BSG.355 Gut 2021;70:A190-A191. This abstract presented data demonstrating that patients with indicators of primary biliary cholangitis were not investigated in line with guidance, and even those with appropriate tests in primary care were frequently not referred.
• DELAYED DIAGNOSIS IN HEREDITARY HAEMOCHROMATOSIS – POTENTIAL UTILITY OF ‘CASE-FINDING’ APPROACH
  1 Angus Kitchin*, 2 Sophie Stretch, 1 Amin Abdulgader, 1 Katharine Hutchison, 1 Timothy Jobson. (1 Somerset NHS Foundation Trust, Taunton, UK; 2 Bristol University Medical School, Bristol, UK) 10.1136/gutjnl-2021-BSG.354 Gut 2021;70:A190. This abstract demonstrated the delay in referral and diagnosis for genetic haemochromatosis – indicating the potential for a ‘case-finding’ approach.
• A CUMULATIVE LIVER DAMAGE INDEX (CLDI) IDENTIFIES PATIENTS AT RISK OF SIGNIFICANT LIVER DISEASE, Wesley E, Matull W, Kitchin A, Hutchison K, Madge S, Jobson T. Gut 2022;71:A4. Presented at British Association for Study of the Liver (Leeds, September 2022) and at the American Association for the Study of Liver Diseases (Washington DC, November 2022). This abstract demonstrated the potential utility of longitudinal analysis of time trends in laboratory data as a powerful tool for case finding. The ability to perform these searches rapidly is a unique feature of the system developed.
• YOUNG ADULTS ARE AT INCREASING RISK OF CHRONIC LIVER DISEASE AND CAN BE IDENTIFIED BEFORE THE DEVELOPMENT OF ADVANCED DISEASE USING A NOVEL CASE FINDING DATABASE – Creamer JM, Mohamed A, Jobson T. Presented at Digestive Diseases Week (Chicago, May 2023) and the British Society of Gastroenterology (Liverpool, June 2023). This abstract shows that the system developed can demonstrate clearly that young adults are frequently under investigated for liver disease, and also that they can be individually identified and recalled for treatment.
• ASSOCIATION OF A CUMULATIVE LIVER DAMAGE INDEX (CLDI-ALT) AND THROMBOCYTOPENIA WITH ADVANCED LIVER DISEASE AND THE POTENTIAL UTILITY IN A CASE FINDING APPROACH Owen C, Gormley S, Maggs D, Mohamed A, Wesley E, Jobson T. Presented at Digestive Diseases Week (Chicago, May 2023) and the British Society of Gastroenterology (Liverpool, June 2023). This abstract demonstrates that the system developed can identify patients at high risk of end-stage liver disease – these are patients that have a significant risk of presenting to secondary care with irreversible / fatal liver disease in the next few months to years.
• PATIENTS WITH PERSISTENTLY ABNORMAL LIVER BIOCHEMISTRY ARE UNDER-INVESTIGATED AND CAN BE RAPIDLY IDENTIFIED USING A NOVEL CASE-FINDING DATABASE Mohamed A, Owen C, Gormley S, Khan W, Wesley E, Jobson T. Presented at the European Association for the Study of the Liver (Vienna, June 2023), and the British Association for Study of the Liver (Brighton, September 2023). This abstract showed that the system can identify patients on the basis of longitudinal data with persistently abnormal liver function tests who have not had a ‘non-invasive liver screen’ – panel of blood tests to identify laboratory-definable liver disease.
• PATIENTS WITH HEPATITIS B INFECTION ARE HIGHLY REPRESENTED IN DEPRIVED AREAS, LESS ENGAGED WITH HEALTHCARE AND CAN BE RAPIDLY IDENTIFIED USING A NOVEL CASE-FINDING DATABASE TO HELP ACHIEVE THE CHRONIC HEPATITIS B ELIMINATION 2030 GOAL
  Mohamed A, Owen C, Creamer J, Wesley E, Meron D, Jobson T. Presented at the British Association for Study of the Liver (Brighton, September 2023).and the American Association for the Study of Liver Diseases (Boston USA, November 2023). This presentation demonstrated the utility of the case-finding approach to target patients in deprived areas with Hepatitis B.
• A NOVEL CASE-FINDING DATABASE CAN BE UTILISED TO IDENTIFY AND TARGET INDIVIDUALS IN AREAS OF DEPRIVATION AT RISK OF ADVANCED LIVER DISEASE. Owen C, Mohamed A, Creamer J, Wesley E, Jobson T. Presented at the British Association for Study of the Liver (Brighton, September 2023).and the American Association for the Study of Liver Diseases (Boston USA, November 2023). This abstract showed that the system can use the Index of Multiple Deprivation to target case-finding in areas of deprivation.