Big Data

For Better Hearts


BigData@Heart latest publications

Estimating the Effect of Reduced Attendance at Emergency Departments for Suspected Cardiac Conditions on Cardiac Mortality During the COVID-19 Pandemic

Published: 20 December 2020

Authors: Michail Katsoulis, Manuel Gomes, Alvina G. Lai, Albert Henry, Spiros Denaxas, Pagona Lagiou, Vahe Nafilyan, Ben Humberstone, Amitava Banerjee, Harry Hemingway, R. Thomas Lumbers


The wider health impacts of the coronavirus disease 2019 (COVID-19) pandemic are of increasing concern, with an increase in rates of non-COVID-19 excess mortality observed. In England and the United States, the early pandemic was accompanied by a decline in patient visits to Emergency Departments (EDs), including those for cardiac diseases. The decline may have been influenced by patient’s reluctance to visit hospital due to the public health messages to protect National Health Service capacity, concerns about the risk of coronavirus infection, or difficulties in accessing medical care. The impact of delayed or nonpresentation to EDs with suspected cardiac disease on cardiac mortality is unknown. In this study, we used instrumental variable analysis to estimate the effects of reduced ED visits on cardiac mortality in England.

Read the full paper here


Association between renin-angiotensin-aldosterone system inhibitor use and COVID-19 hospitalization and death: a 1.4 million patient nationwide registry analysis

Published: 22 November 2020

Authors: Gianluigi Savarese, Lina Benson, Johan Sundström, Lars H. Lund


Aims Renin-angiotensin-aldosterone system inhibitors (RAASi) improve outcomes in cardiorenal disease but concerns have been raised over increased risk of incident hospitalization and death from coronavirus disease 2019 (COVID-19). We investigated the association between use of angiotensin-converting enzyme inhibitors (ACEi), angiotensin receptor blockers (ARBs) or mineralocorticoid receptor antagonists (MRAs) and COVID-19 hospitalization/death in a large nationwide population. 

Methods and results Patients with hypertension, heart failure, diabetes, kidney disease, or ischaemic heart disease registered in the Swedish National Patient Registry until 1 February 2020 were included and followed until 31 May 2020. COVID-19 cases were defined based on hospitalization/death for COVID-19. Multivariable logistic and Cox regressions were fitted to investigate the association between ACEi/ARB and MRA and risk of hospitalization/death for COVID-19 in the overall population, and of all-cause mortality in COVID-19 cases. We performed consistency analysis to quantify the impact of potential unmeasured confounding. Of 1 387 746 patients (60% receiving ACEi/ARB and 5.8% MRA), 7146 (0.51%) had incident hospitalization/death from COVID-19. After adjustment for 45 variables, ACEi/ARB use was associated with a reduced risk of hospitalization/death for COVID-19 (odds ratio 0.86, 95% confidence interval 0.81-0.91) in the overall population, and with reduced mortality in COVID-19 cases (hazard ratio 0.89, 95% confidence interval 0.82-0.96). MRA use was not associated with risk of any outcome. Consistency analysis showed that unmeasured confounding would need to be large for there to be harmful signals associated with RAASi use. 

Conclusions In a 1.4 million nationwide cohort, use of RAASi was not associated with increased risk of hospitalization for or death from COVID-19. 

Read the full paper here


Prediction of vascular aging based on smartphone acquired PPG signals

Published: 12 November 2020

Authors: Lorenzo Dall’Olio, Nico Curti, Daniel Remondini, Yosef Safi Harb, Folkert W. Asselbergs, Gastone Castellani, Hae-Won Uh


Photoplethysmography (PPG) measured by smartphone has the potential for a large scale, non-invasive, and easy-to-use screening tool. Vascular aging is linked to increased arterial stiffness, which can be measured by PPG. We investigate the feasibility of using PPG to predict healthy vascular aging (HVA) based on two approaches: machine learning (ML) and deep learning (DL). We performed data preprocessing, including detrending, demodulating, and denoising on the raw PPG signals. For ML, ridge penalized regression has been applied to 38 features extracted from PPG, whereas for DL several convolutional neural networks (CNNs) have been applied to the whole PPG signals as input. The analysis has been conducted using the crowd-sourced Heart for Heart data. The prediction performance of ML using two features (AUC of 94.7%) – the a wave of the second derivative PPG and tpr, including four covariates, sex, height, weight, and smoking – was similar to that of the best performing CNN, 12-layer ResNet (AUC of 95.3%). Without having the heavy computational cost of DL, ML might be advantageous in finding potential biomarkers for HVA prediction. The whole workflow of the procedure is clearly described, and open software has been made available to facilitate replication of the results.

Read the full paper here


A semi-supervised approach for rapidly creating clinical biomarker phenotypes in the UK Biobank using different primary care EHR and clinical terminology systems

Published: 19 May 2020

Authors: Spiros Denaxas, Anoop D. Shah, Bilal A. Mateen, Valerie Kuan, Jennifer K. Quint, Natalie Fitzpatrick, Ana Torralbo, Ghazaleh Fatemifar, Harry Hemingway


Objectives The UK Biobank (UKB) is making primary care Electronic Health Records (EHR) for 500,000 participants available for COVID-19-related research. Data are extracted from four sources, recorded using five clinical terminologies and stored in different schemas. The aims of our research were to: a) develop a semi-supervised approach for bootstrapping EHR phenotyping algorithms in UKB EHR, and b) to evaluate our approach by implementing and evaluating phenotypes for 31 common biomarkers.

Materials and Methods We describe an algorithmic approach to phenotyping biomarkers in primary care EHR involving a) bootstrapping definitions using existing phenotypes, b) excluding generic, rare or semantically distant terms, c) forward-mapping terminology terms, d) expert review, and e) data extraction. We evaluated the phenotypes by assessing the ability to reproduce known epidemiological associations with all-cause mortality using Cox proportional hazards models.

Read the full paper here

Published on: 02/10/2021