Obstructive hypertrophic cardiomyopathy (HCM) is a genetic heart disease characterised by unexplained left ventricular (LV) hypertrophy (thickening of heart lower chamber walls), that affects patients’ daily functioning and quality of life. Patients with obstructive HCM typically experience fatigue and/or shortness of breath, which interferes with their ability to participate in daily activities. Obstructive HCM has been associated with increased risks of atrial fibrillation (irregular and rapid heart rhythm), stroke, heart failure and sudden cardiac death. Current treatments are limited to those that provide symptomatic relief and have limited efficacy.
Obstructive HCM is underdiagnosed—patients with the disease often present early to general practitioners (GP) with non-specific symptoms (i.e., tiredness, shortness of breath, dizziness, chest pain, and palpitation). This, in turn, is thought to result in misdiagnosis or delayed referral to specialist cardiology services and can lead to increased likelihood of serious clinical events and thus worse quality of life. Therefore, there is a need for a deeper understanding of the profile, including the path to diagnosis, of patients with obstructive HCM in the United Kingdom (UK) to accelerate the pathway to diagnosis and optimal patient care. This study will be conducted in two phases. In Phase I, the aim is to assess: (a) characteristics of patients diagnosed with obstructive HCM; (b) clinical pathways; and (c) healthcare resource utilisation leading to diagnosis. Phase II will use machine learning to build a model to predict the patients still undiagnosed with obstructive HCM to aid diagnostic tools, thus reducing delays in diagnosis and misdiagnoses.
Hypertrophic cardiomyopathy (HCM) is a genetic, progressive myocardial disease. Community-based studies indicate the prevalence of HCM is one in 500, with many individuals remaining undiagnosed. HCM can affect people of any age, race, or sex. Obstructive HCM is the most common form of HCM, with an estimated two-thirds of patients affected. Obstructive HCM is characterised by unexplained left ventricular (LV) hypertrophy, which is associated with dynamic LV outflow tract (LVOT) obstruction; it is defined by the presence of either a resting or provoked LVOT peak gradient ≥30 mmHg. Obstructive HCM is commonly underdiagnosed which can delay referral to specialist cardiology services for an echocardiogram. This study will be conducted in two phases and aim to provide insights to accelerate the pathway to diagnosis and optimal patient care.
The study will utilise data from population-level electronic medical records for patients with a diagnosis of obstructive HCM in England. Specifically, the Clinical Practice Research Datalink (CPRD) [AURUM dataset] linked with Hospital Episode Statistics (HES) Accident and Emergency, HES Admitted Patient Care, HES Diagnostic Imaging Dataset, and HES Outpatient data will be utilized. Phase I will be descriptive and involve a cohort design, to assess the following outcomes: (a) patient clinical and demographic characteristics at diagnosis of obstructive HCM; (b) cardiovascular disease-related clinical pathways up to diagnosis; and (c) healthcare resource utilisation leading to diagnosis (i.e., GP, inpatient and outpatient visits and associated length of stay, diagnostic test and procedures, and prescribed treatments). Phase II will involve the development of a predictive model using machine learning methods and building an at-risk patient profile. The model outcomes will be measured in terms of accuracy of obstructive HCM prediction, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), F1 score and area under the receiver operating characteristics (AUROC) curve.
Primary Outcomes - Phase I: Descriptive Study
A. Patient clinical and demographic characteristics at diagnosis of obstructive HCM:
• Clinical characteristics
• Demographic characteristics
B. Clinical pathway up to diagnosis of obstructive HCM
• Cardiovascular disease-related clinical diagnoses
• Symptoms
C. Healthcare resource use (HRU) up to diagnosis of obstructive HCM, assessed via number and proportion with at least one use of resources, and number of resources per patient per year, including:
• General practitioner (GP) consultations/nurse visits
• Outpatient specialist visits
• Accident and emergency (A&E) visits
• Hospital inpatient admissions and associated length of stay
• Diagnostic tests and procedures
• Prescribed medications and non-therapeutic interventions
Secondary Outcomes - Phase II: Machine Learning Model
The machine learning model will be developed to predict whether patients are a obstructive HCM case (i.e. true) or are not an obstructive HCM case (i.e. false).
Outcomes for the machine learning model such as random forest, gradient boosting trees, neural network will be a will be measured in terms of the predictive model performance (i.e., measures for how well the model can predict true cases of obstructive HCM diagnosis). Performance will be assessed based on the following metrics:
• Accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), F1 score and area under the receiver operating characteristics (AUROC) curve.
Michael Hurst - Chief Investigator - Bristol-Myers Squibb Pharmaceuticals Limited - UK ( BMS )
Sonia Halhol - Corresponding Applicant - Evidera Ltd - UK
- Collaborator -
Alison Booth - Collaborator - Evidera, Inc
Amanda Pulfer - Collaborator - Evidera Ltd - UK
Ashwin Rai - Collaborator - Evidera, Inc
Belinda Sandler - Collaborator - Bristol Myers Squibb - Europe ( BMS )
Carla Zema - Collaborator - Bristol-Myers Squibb Pharmaceuticals Limited - UK ( BMS )
Dimitra Lambrelli - Collaborator - Evidera, Inc
Kevin Pollock - Collaborator - Bristol-Myers Squibb Pharmaceuticals Limited - UK ( BMS )
Michael Papadakis - Collaborator - St George's, University of London
Robert Donaldson - Collaborator - Evidera Ltd - UK
Yue Zhong - Collaborator - Bristol-Myers Squibb - USA ( BMS )
Zahra Chambers - Collaborator - Bristol Myers Squibb - Europe ( BMS )
HES Accident and Emergency;HES Admitted Patient Care;HES Diagnostic Imaging Dataset;HES Outpatient