Point of Care Artificial Intelligence Tool for Heart Failure Diagnosis

Abbreviations/acronyms:

DUO-EF = prediction of ejection fraction (EF) using the Eko-DUO digital stethoscope algorithm
HF = heart failure HFrEF = heart failure with reduced ejection fraction COVID-19 =
coronavirus disease 2019 Eko DUO = digital stethoscope device cMRI = cardiac magnetic
resonance imaging ECG = electrocardiogram

AIMS

To demonstrate DUO-EF can identify heart failure (HF) with reduced ejection fraction (HFrEF)
post-COVID-19 where diagnosis would otherwise be missed/delayed To demonstrate DUO-EF can
reliably and accurately diagnose new HFrEF in the primary care setting To further validate
DUO-EF diagnostic performance at-scale against gold-standard investigations (echocardiography
and cardia MRI) To measure if DUO-EF suggestive of HFrEF but with normal gold standard
investigations predicts future risk of developing HFrEF

Methods To demonstrate DUO-EF can identify heart failure (HF) with reduced ejection fraction
(HFrEF) post-COVID-19 where diagnosis would otherwise be missed/delayed To demonstrate DUO-EF
can reliably and accurately diagnose new HFrEF in the primary care setting To further
validate DUO-EF diagnostic performance at-scale against gold-standard investigations
(echocardiography and cardiac magnetic resonance imaging - cMRI) To measure if DUO-EF
suggestive of HFrEF but with normal gold standard investigations predicts future risk of
developing HFrEF

DUO-EF prediction of ejection fraction in patients attending COVID-19 follow up clinic and
comparison with:

subsequent DUO-EF at time of gold-standard investigation for HF ejection fraction as
calculated by gold-standard investigation

DUO-EF prediction of ejection fraction in patients where their GP suspects new heart failure
and comparison with:

subsequent DUO-EF at time of gold-standard investigation ejection fraction as calculated by
gold-standard investigation DUO-EF prediction of ejection fraction in unselected patients
attending for echocardiography or cardiac MRI, comparing DUO-EF predicted with gold-standard
calculated ejection fraction Telephone call follow-up at 24 months for all patients with
DUO-EF suggestive of HFrEF but normal gold standard investigations

OUTCOME MEASURES Area under curve (AUC) of DUO-EF calibrated for detection of EF below 40%;
classification accuracy Positive predictive value of DUO-EF in COVID-19 clinic and GP context
based on subsequent gold-standard estimation of EF; negative predictive value of DUO-EF in
COVID-19 follow up cohort; positive predictive value of DUO-EF at 24 months in those with
negative gold standard investigations Qualitative measurement of patient and clinical end
user acceptability of Eko DUO

POPULATION Group 1: Patients seen in the COVID-19 follow-up clinic (n = 400) Group 2:
Patients seen in primary care with symptoms newly suggestive of heart failure (n = 400) Group
3: All-comers to echocardiography departments across Imperial College Healthcare NHS Trust (n
= 1,500) Group 4: patients undergoing cardiac MRI investigation (n = 100)