Cancer, Cardiovascular, Respiratory
Conditions
Keywords
Artificial Intelligence, Chest X-Ray, Lung Cancer, Heart Failure, COPD
Brief summary
SEARCH-ED is a research study which is running in Emergency Department (ED) of the Queen Elizabeth University Hospital. The aim of the study is to find out if using a computer programme can help doctors diagnose heart and lung problems from chest x-rays. We want to compare how many people are diagnosed with heart or lung problems for the first time when doctors have access to the computer programme results, in comparison to when they don't.
Detailed description
SEARCH-ED is a research study which is running in Emergency Department (ED) of the Queen Elizabeth University Hospital. The aim of the study is to find out if using an artificial intelligence (AI) computer programme can help doctors diagnose heart and lung problems from chest x-rays. The computer programme is made by Harrison.ai. It is approved for use in the United Kingdom (UK), United States of America (US) and the European Union (EU). Studies have been carried out previously to make sure it is safe to use and that it can detect signs of heart and lung problems. Many people who come to ED have a chest x-ray. Chest x-rays can show signs of heart or lung problems, which might be causing a patient's symptoms. All doctors can interpret chest x-rays. However, doctors who specialise in interpreting scans (radiologists) also provide an expert report for chest x-rays, describing what they have found. It can take a long time for chest x-ray reports to come back. Sometimes, doctors might miss signs of heart or lung problems. We want to see if using a computer programme to help doctors interpret chest x-rays could lead to more patients getting an accurate diagnosis. We want to compare how many people are diagnosed with heart or lung problems (Chronic obstructive pulmonary disease \[COPD\], heart failure or lung cancer) for the first time when doctors have access to the computer programme results, in comparison to when they don't. Patients older than 18 who have a chest x-ray in ED will be included. Patients with chest x-rays flagged by the computer programme for heart failure or COPD will be invited to an outpatient clinic for further assessment post-discharge, providing they have not been referred for testing or had testing previously. All patients with chest x-rays flagged for lung cancer will be reviewed and acted on by the study radiologist. The study consists of 1) a retrospective component; 2) a prospective live trial; 3) a qualitative evaluation of acceptability to patients and clinicians, and 4) a health economic analysis.
Interventions
DEVICEHarrison.ai Chest X-Ray Solution
The Harrison.ai CXR module is an AI-driven clinical decision support tool that is designed to augment clinical interpretation of CXRs. It is a Class IIb CE-marked device which is able to detect up to 124 findings on a CXR.
Sponsors
NHS Greater Glasgow and Clyde
University of Glasgow
Study design
Allocation
NON_RANDOMIZED
Intervention model
PARALLEL
Primary purpose
DIAGNOSTIC
Masking
SINGLE (Caregiver)
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
Unconsented Use of Harrison CXR Algorithm in Emergency Department (ED): * Frontal Chest X-Ray (CXR) (AP or PA) acquired in the Queen Elizabeth University Hospital (QEUH) ED * Patients aged 18 or over * Appropriate meta data (DICOM) to allow for Harrison CXR processing and secondary capture report provision. Patient Focus Groups: * Aged 18 or over * Able to provide written, informed consent in English. Clinician Focus Groups: * Aged 18 or over * Able to provide written, informed consent in English. * Working as a doctor, advanced nurse practitioner or advanced clinical practitioner in ED, radiology or downstream medical specialties * For post-implementation focus groups only, must have at least 4 months experience of working with Harrison CXR algorithm. Diagnostic Clinic: * Patients without terminal illness or advanced frailty * Usual healthcare provider based in NHS GGC
Exclusion criteria
Applies to use of unconsented CXRs: \- Patient has requested that they are removed from the study, or has objected to the use of AI in their routine clinical care and this has been subsequently upheld by the health board. Applies to invitation to combined diagnostic clinic: * Patients not available to follow up, including patients i.e. whose the patient's usual care (or onward care following index admission) is out-with NHS GGC. * Patients who have been referred to palliative care for end-stage disease, or patients with severe frailty (i.e. bedbound) will not be invited to the combined diagnostic clinic For Patient and Clinician Focus Groups: * Unable to provide informed written consent in English * Aged \<18
Design outcomes
Primary
| Measure | Time frame |
|---|---|
| Proportion of patients identified with a confirmed new diagnosis of heart failure, based on subsequent clinical assessment and guideline-based investigation. | 12 months |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Duration of admission during index hospitalisation | 12 months | — |
| Time to initiation of guideline-based, long-term therapy for Chronic obstructive pulmonary disease (COPD) and Heart Failure. | 12 months | For Chronic obstructive pulmonary disease (COPD), this will be defined as first prescription of combined long acting beta agonist (LABA)/long acting muscarinic antagonist (LAMA) inhaler or LABA/LAMA/inhaled corticosteroid (single or split) inhaler therapy. For Heart Failure , this will be defined as first prescription of either a) a renin-angiotensin system inhibitors, b) a beta blocker, or c) an SLGT2 inhibitor. |
| Time to diagnostic testing for Heart Failure, COPD and lung cancer (echocardiography, spirometry, CT). | 12 months | — |
| Time to inpatient or outpatient specialist review and confirmation of lung cancer, COPD or Heart Failure | 12 months | — |
| Acceptability of AI-supported interpretation of Chest X-Ray for Emergency Department clinicians pre and post intervention using Theoretical Framework of Acceptability (TFA) | Baseline and 12 months | We will ask clinicians what they think of using AI for Chest X-Rays |
| Readmission rate within 90 days | 3 months | — |
| Proportion of patients with new diagnosis of lung cancer detected by an AI-Chest X-Ray algorithm | 12 months | — |
| Proportion of patients with new diagnosis of COPD detected by an AI-Chest X-Ray algorithm | 12 months | — |
| Proportion of patients with clinically-confirmed known diagnosis of lung cancer, Heart Failure and COPD detected by an AI-Chest X-Ray algorithm | 12 months | — |
| Percentage of Chest X-Rays not identified by an AI-CXR algorithm that have a subsequent diagnosis of Heart Failure, COPD or lung cancer within 6 months of index imaging (Emergency Department Chest X-Ray). | 6 months | — |
| Statistical analysis of model performance e.g. sensitivity, specificity, positive and negative predictive value | 12 months | — |
Countries
United Kingdom
Contacts
CONTACTClea Du Toit
CONTACTDervla Carroll
PRINCIPAL_INVESTIGATORDavid J Lowe
University of Glasgow
Outcome results
None listed