Heart failure affects nearly a million people in the UK. Patients experience unpleasant symptoms, like breathlessness, exhaustion and ankle swelling, and their life may be shortened. However, with the right treatment, symptoms can be controlled, and outlook improved, so getting a diagnosis is key. Currently, four out of every five people are admitted to hospital as an emergency to get a heart failure diagnosis. It would be better for patients and the NHS to diagnose and treat these people earlier in the community and avoid hospital admission.
In this project, we want to create a model which will predict which patients are most likely to develop heart failure in the next 12 months. We will use a database of anonymous GP records to look at which patient characteristics (e.g. age, blood pressure, previous heart attack) are more common in people who develop heart failure. We will then build the model and test it in a different part of the database to see if we can predict the people who get heart failure. If it works well, the tool could help GPs to understand which patients are most likely to develop heart failure so they can test, diagnose and treat them earlier.
Heart failure (HF) is a common, costly, and treatable condition. Evidence-based therapies can improve quality of life and survival once a diagnosis is made, but currently 80% of people are diagnosed on hospital admission. Rapid diagnosis in primary care could improve patient experience and outcome. We aim to derive and validate a robust risk prediction model (‘RiskHF’) to identify patients likely to develop HF.
The Clinical Practice Research Datalink contains demographic data and clinical codes for over 15 million patients, and is linked to Hospital Episode Statistics and Office for National Statistics mortality data. We will carry out an open cohort study from 1st January 2000 to 31st December 2020 and split the dataset into derivation and validation cohorts. The primary outcome will be a diagnosis of HF in the next 12 months. Predictor variables will include patient demographics, HF symptoms, cardiovascular risk factors, co-morbidities and prescriptions. We will use Cox proportional hazards to derive risk equations for men and women then evaluate the performance of the model in the validation cohort.
The RiskHF tool could be used in new HF pathways to identify patients for natriuretic peptide testing and referral to facilitate earlier diagnosis and treatment.
The primary outcome will be a diagnosis of HF in the primary care, hospital record or death certificate in the next 12 months. The index date will be the earliest from all three sources. Participants with a diagnosis of HF will be identified using diagnostic codes entered by GPs or hospital specialists to record new diagnoses in the primary or secondary care medical record, respectively, or by the clinician completing a death certificate using ONS mortality data. The NHS Clinical Terminology Browser and Quality and Outcomes Framework guidelines will be used to generate a comprehensive list of terms used to code for a diagnosis of HF. HF is a clinical syndrome and the diagnosis requires the presence of symptoms and objective evidence of a structural or functional abnormality of the heart. Patients with a clinical code of HF and/or an echocardiograph report or a record of HF in linked HES records will be classified as being a case of HF. Echocardiograms will be identified using clinical codes and entity code 342. Two distinct types of HF are now recognised by researchers and the clinical community based on the left ventricular ejection fraction: HF with reduced Ejection Fraction (HFrEF) and HF with preserved Ejection Fraction (HFpEF). We will search for these codes in the patient record, and ejection fraction results from echocardiogram reports to establish whether it is possible to report HFrEF and HFpEF rates separately. These codes have only recently been recorded in clinical practice so it may not be possible to distinguish between HFrEF and HFpEF, in which case a diagnosis of HF will be used.
Clare Taylor - Chief Investigator - University of Oxford
Clare Taylor - Corresponding Applicant - University of Oxford
Clare Bankhead - Collaborator - University of Oxford
José M. Ordóñez-Mena - Collaborator - University of Oxford
Kathryn Taylor - Collaborator - University of Oxford
Lucinda Archer - Collaborator - University of Birmingham
Maria Vazquez Montes - Collaborator - University of Oxford
Richard Hobbs - Collaborator - University of Oxford
Subhashisa Swain - Collaborator - University of Oxford
Andrea Roalfe - Collaborator - University of Oxford
HES Accident and Emergency;HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation;Practice Level Index of Multiple Deprivation