Precision treatment strategies using next-generation machine learning in multiple sclerosis

Context

MS-PINPOINT project aims to develop digital tools that can tell us about the future of MS and how it can be treated. To ensure that we can develop these digital tools using populations of patients that represent the diversity of all those living with MS, including under-represented patients, we aim to go back to hospital archives by 10 to 20 years and collect existing data from 12 hospitals.

Patient Notification MS-PINPOINT Project

Given the data for research is anonymised in the hospitals, the possibility of reidentifying people with multiple sclerosis (MS) is extremely low. We will exclude all NHS numbers registered in the National Data Opt-Out and have designed a project-specific opt-out process, allowing people living with MS to opt out from this project.

Introduction

This document aims to inform how we will use information from people living with multiple sclerosis fairly and ethically. This document will be made available publicly at www.ms-pinpoint.com and displayed in the MS clinics where the MS-PINPOINT project will gather information. Where possible, this page will be posted on the hospital websites. Data providers are hospitals listed in the study protocol, where data (MRI and electronic health records), including identifying information (NHS number, name, postcode, date of birth, gender), are removed. This document shows how the research team, led by Dr Arman Eshaghi, act as a trusted third party to use this information to link pseudonymised clinical data with MRI. We include broader information on the research project for context, though only anonymised data is used for the statistical analysis, and group-level results will be presented in scientific manuscripts. No identifiable information is shared outside the hospital, and no data from different hospitals will be centralised or transferred out of the hospital setting.

Motivation for the project

Multiple sclerosis is an incurable and disabling disease that costs the UK approximately £3 billion annually. We have more than 15 treatments approved for people with MS, which can only moderately slow disability worsening but cannot halt it. MS is, therefore, unpredictable and causes significant uncertainty in our lives. Additionally, selecting the right treatment for the right patient at the right time is so challenging that most MS patients will switch treatment at some point. NHS hospitals have gathered large data sets over the past few decades, which are used for patient care but not for research to help patients. Computers use artificial intelligence to make sense of enormous data sets to provide personalised recommendations. We aim to link electronic health records, that is, the information acquired from medications and characteristics of people living with MS at the national level with their medical images (brain and spinal cord MRI), build AI that can predict the future course of MS and group patients using their MRI in types that can tell us who can best respond to specific treatments, reduce adverse events, keep people with MS productive for longer, reduce costs of MS to NHS and eventually develop new treatments to stop MS fully.

Aims of the project and why linking health records with medical images of the brain and spinal cord is necessary

The overarching aim of the research is to build AI using electronic health records and brain and spinal cord images to predict the future course of MS, group people using their medical images into MS types, informative of treatment effect and recommend the right treatment for the right patient at the right time. The data linkage is essential to maximise the use of existing data and connect electronic health records with medical images. This linkage will enable us to combine information on age, ethnicity, and self-reported sex with medical images and build an AI model that can tell us about the likelihood of responding to MS treatments for people with specific brain images or clinical characteristics and the future course of MS.

Objectives and procedures

The research will gather information from health records of people living with MS in participating hospitals, link it with medical images and build AI models locally within the hospital setting. AI models across the UK will then share the learned knowledge without sharing or exchanging patient data across hospitals. Before feeding data into the AI model but after linking medical images and clinical information, patient-identifiable data, including date of birth, NHS numbers, names, and postcode, will be removed. Identifiable data will only flow within the hospital setting until medical images and electronic health records are linked and may be transferred to the affiliated university when computational power is necessary after pseudo-anonymisation (for example, when access to large-memory AI servers with special hardware is essential).

Benefits

Artificial intelligence is revolutionising every aspect of our life. It may one day become a decision-aid system to help people living with MS. Access to large data sets can help researchers build inclusive models (exposed to people from minority backgrounds and from diverse geographical locations), ensuring safe and equitable AI predictions. Predicting the future course of MS and those who are more likely to respond to specific treatments will allow prescribing the right treatments, delaying disability, keeping people with MS employed for longer, increasing their quality of life and saving costs for the NHS.

The identity of the researcher and their relationship with the University/Department

The principal investigator is Dr Arman Eshaghi, a National Institute for Health and Care Research (NIHR) Advanced Fellow at University College London. Dr Eshaghi holds (or will have) an honorary contract with the participating hospital in this research to access data. No data will be transferred to other hospitals. Pseudo-anonymised data may be transferred to the university affiliated with the same hospital. No data (identifiable or anonymised) will be exchanged outside this setting. The researcher nor anyone else will have access to identifiable data outside of the hospital setting. This study will lay the basis for a future prospective clinical trial which will consent patients to acquire data.

Opt-out procedure

National Opt-Out: Within each hospital, NHS numbers registered at the National Opt-Out server will be removed prior to the data flows, after which the data will be anonymised. MS-PINPOINT Specific Opt-Out: This can be done by contacting the researcher: Dr Arman Eshaghi, within University College London Hospitals Trust (a.eshaghi@ucl.ac.uk) or via a secure web form at www.ms-pinpoint.com/study-opt-out or by calling 020 3108 7420.