CDSU Data Access Committee

MDS/24/GA/001: Novel Artificial Intelligence Methods for the Early diagnosis of Acute Bowel Ischaemia

Principal Investigator: Mr Anthony Chan

Summary: Acute mesenteric ischaemia (AMI) is a life-threatening emergency where bowel is starved of oxygen due to a blockage in their blood supply. AMI kills 50-90% of patients unless emergency surgery is performed. Diagnosing AMI is difficult as symptoms are often vague and not always obvious on a CT scan even to an experienced radiologist. There is a need to explore better ways to diagnose AMI earlier to improve outcomes for these patients.

Artificial intelligence (AI) learns by using lots of information to recognise certain patterns. One example of this might be an AI program looking at a chest x-ray and predicting if that patient has COVID.

We want to train an AI program to use a patient’s blood results, clinical observations such as blood pressure and temperature, together with CT scan images to predict whether that patient has AMI. If successful, this could lead to an earlier diagnosis.

MDS/24/RS/003: Creating a Virtual Cohort for Chronic Obstructive Pulmonary Disease (COPD): Combining Data Exploration, Clinical Coding Systems, and Natural Language Processing for Enhanced Monitoring and Insights.

Principal Investigator: Anuoluwapo Adetayo

Summary: The study goal is to create a virtual group of people who have (COPD). To do this, there is a need to first analyse a large amount of data and understand how doctors and hospitals code information on COPD. Once this information is gathered, it will be organized to build a clear image of the COPD community, emphasising key health data. To gain deeper understanding of the group, advanced computer techniques will be used to extract additional information from textual records. 

The goal is to have detailed understanding of this group’s health, which will benefit doctors and researchers in the future. A mechanism to monitor changes in the group’s health will be developed over time, which will help review how the disease progresses and how well treatments work.

MDS/24/RS/004:: Derivation and validation of a novel model incorporating PET-CT for predicting malignancy in screen detected lung nodules – The PRECISE Study

Principal Investigator: Dr Haval Balata

Summary: Screening CT scans are used to look for early stage lung cancers in healthy people without any lung symptoms. This is predicted to pick up thousands of patients with lung nodules (spots on the lungs), some of which will be benign but some will be lung cancers. Once we see a nodule, we undertake further tests, one of which tests is a PET-CT scan. Once we have done the PET-CT scan, we apply a cancer risk prediction to help decide what is the appropriate next step.

Anonymous records describing patient features, scan findings, and cancers that are found will be analysed to develop an improved cancer risk model that is developed specifically for people who have nodules that are found during screening.

Overall, this work should help make us better at knowing which lung nodules might be cancer and which can be left alone.

MDS/24/CA/002: Pre-hospital point of care cardiac troponin testing to maximise efficiency for patients calling 999 with chest pain

Principal Investigator: Professor Rick Body

Summary: Currently it can be difficult for paramedics to accurately determine whether a patient suffering from chest pain is having a heart attack. The Chest Pain Diagnosis project aims to improve emergency ambulance efficiency for acute chest-pain patients by utilising a point-of-care test within ambulances that identifies the presence of a biomarker called Troponin that is released when a patient is having a heart attack. The results of this test (and other clinical observations) will be input by paramedics into a computerised decision aid called T-MACS which calculates an individual patient’s probability that they are having a heart attack. This innovative way of working will enable paramedics to accurately identify patients that are suffering from a heart attack and transport them immediately to the most appropriate location for further treatment and care.

MDS/24/NO/001: Fetal scalp blood sampling during labour: real-world data for improved safety evaluation

Principal Investigator: Dr Victoria Palin

Summary: Fetal scalp blood sampling (FSBS) is a test used during labour to measure whether a baby is receiving enough oxygen. The test consists of a small scratch made on the baby’s scalp to draw blood, which is quickly analysed. If a test result is positive then birth is expedited, for example via emergency caesarean section, to prevent harm to the baby.  Studies have suggested that FSBS can lead to harm to mother and baby, such as infection from the scratch. However, the frequency of these harms is unclear and the accuracy of the test for predicting poor pregnancy outcomes is unknown but could be improved. This study will analyse historic hospital data to determine the safety of FSBS in labour, and how often it leads to harm for mother and/or baby. We will also see if the accuracy of the test can be improved for future clinical use.

MDS/22/IF/001: The impact of introducing a digital technology intervention on outpatient appointment no-shows

Principal Investigator: Professor Dawn Dowding

Summary: Outpatient appointment no-shows pose a significant challenge to the NHS and are associated with inefficiencies in the delivery of healthcare services, as well as substantial financial cost.

In September 2022, MFT implemented a new Trust-wide Electronic Health Record (EHR), known as Hive, which is expected to bring efficiencies in the scheduling and management of outpatient appointments.

Our main aim is to assess the impact of the implementation of Hive on the proportion and management of outpatient appointment no-shows at MFT.

We will also consider whether there was any variation in the impact of Hive across different patient groups, including level of deprivation, age, sex, and ethnicity.

MDS/22/ON/008: Machine Learning Models in the Assessment of an Acute Abdomen

Principal Investigator: Mr Anthony Chan

Summary: Acute abdominal pain (AAP) accounts for about 10% of visits to A&E, and the diagnosis can be challenging as

pain can be caused by a broad range of conditions, such as acute appendicitis. A missed diagnosis or misdiagnosis can result in delays in treatment, worse outcomes or, in some cases, permanent impairment or death.

This study uses artificial intelligence to develop a machine learning model to predict a patient’s diagnosis and anticipated outcomes in hospital based on routine measurements such as blood pressure, heart rate, temperature, and blood results. Data from previous patient admissions will be used to train this model, and it is hoped that the model can be integrated into MFT’s Hive Electronic Patient Record system to automatically predict and help clinicians determine their patient’s diagnosis as well as other metrics such as predicted length of stay in hospital.

MDS/23/IN/002: Measures of shock reversal and clinical outcomes in critically ill patients with septic shock: An observational cohort study

Principal Investigator: Dr Jonathan Bannard-Smith

Summary: Sepsis is a life-threatening infection, with septic shock being the most severe form. In septic shock, the infection is so severe that drugs (vasopressors) are needed to maintain blood pressure in order to maintain flow to the body’s organs.

We think that reversing septic shock quickly is a good thing that will help patients to get better quicker and suffer less complications. However, no one knows the best way to measure when septic shock has been reversed.

We will look at three ways to measure this: ‘time on vasopressors’, ‘time vasopressor free’, and ‘average vasopressor dose over time’. We will work this out in a group of patients with septic shock and see which measure best describes when septic shock has reversed.  We will explore which measure is best linked to important outcomes for patients.

We will use data from patients who have been cared for in the intensive care unit at Manchester Royal Infirmary.  Only data which has already been collected will be used.  We will look at other characteristics of the patients (age, gender, other health problems) and their illness to see if any other factors affect how long it takes for septic shock to reverse.

MDS/23/CA/002: ECG-X – Explainable automated ECG interpretation for Long QT Syndrome – validation with further data

Principal Investigator: Dr Alaa Alahmadi

Summary: Every year about 100,000 people in the UK die of sudden cardiac death, often without having had any recognisable symptoms. Some of these deaths are caused by a condition called Long QT Syndrome (LQTS) which can lead to the heart beating irregularly and subsequently failing. LQTS can be acquired (e.g., caused by certain medications or having other cardiac and non-cardiac conditions including diabetes) or congenital (when someone was born with this syndrome caused by certain genetic mutation) and is often not discovered on routine electrocardiograms (ECG) as it is hard to spot visually and need careful manual measurement of QT-interval. We have now developed an algorithm that highlights LQTS through applying colour to the ECG waves and therefore make it easier to interpret and diagnose. Our algorithm was tested successfully as a proof of concept, but we now need further clinical data to test, verify, and optimise our algorithm. Hence, we’d like to access MFT’s data to develop a clinically sound and useful algorithm that can reliable detect LQTS and save lives. On top of that, our algorithms are explainable which means we always how the algorithm came to that decision (e.g. patient is at risk because of very prolonged QT) and how it came to its decision based on reading the visualised coloured ECG. The explainability is based on clinical rules and the project included important stakeholders from early design to implementation

MDS/23/IF/034 Improving Transparency of Processes for Accessing Health Data for Research Purposes

Principal Investigator: Claire MacDonald

Summary: CDSU will work with our Research Communications department to develop a public-facing website on already existing site building and hosting infrastructure. The website will meet the Data Transparency Standards as set out by Health Data Research UK (HDRUK) working groups and will display information on how researchers can apply for access to MFT data. There will be brief descriptions of the role of our Data Access Committee and our security and governance approvals, along with a list of projects approved through the Committee.

MDS/23/HO/001: Optimal Blood Culture Timing in the ICU

Principal Investigator: Mr Kenny Wong

Summary: Blood cultures are a test used to check for bacteria in the blood and confirm their type if they exist, however, they are very prone to error. Although many papers have looked at what factors could possibly influence the result, there has been little research into how the timing of taking these tests affects the result. In this project, I aim to look at how timing affects blood cultures and, if possible, find the best time to take samples to consistently produce correct results.

MDS/23/PR/005: Investigating The Potential Value of Pre-Emptive Pharmacogenomic Testing Through a Trust-Wide Assessment of Prescribing Practice

Principal Investigator: Dr John McDermott

Summary: Medicines play a crucial role in healthcare, but their effectiveness and safety vary between individuals. Some people receive ineffective medication while others experience adverse reactions. This has negative effects on individuals and society. One solution is to use a person’s genetic information to improve medicine selection and dosing, known as pharmacogenetics. The aim is to gather data to understand the benefits of this approach on a larger scale. However, it is challenging to measure the long-term impact of this approach using standard data capture methodologies. To address this, the proposal suggests using a clinical dataset from the

Manchester University NHS Foundation Trust to assess the value of pharmacogenetic testing.

MDS/23/OB/002: Routine data to investigate risk prediction and implementation science in maternity

Principal Investigator: Professor Jenny Myers

Summary: Routine pregnancy care includes a number of detailed risk assessments which are carried out a different times in pregnancy. Whilst some risk assessments are supported by high quality evidence, some assessments could be improved by testing how they perform in routine care and adjusting them where necessary to improve accuracy. In addition, we know from our previous analyses that some adverse pregnancy outcomes are more common in some of the communities we serve, but we don’t fully understand why. This project will use routinely recorded pregnancy information to refine risk assessments in pregnancy and to understand how we can improve care and outcomes for women across our diverse communities.