The Cardiology Data Evolution: Advancing Women's Health

Introduction

The CDC (Center for Disease Control and Prevention) reports heart disease as the leading cause of mortality in women in the United States, responsible for about 1 in every 5 female deaths. Despite this, only about half (56%) of women recognize that heart disease is a significant threat. About 1 in 17 women aged 20 years and older (5.8%) have coronary artery disease, the most common type of heart disease in the United States. 

When it comes to heart disease and heart attacks, women commonly face misdiagnoses and delays in receiving care. These medical missteps mean the mortality rate for women is twice as high as men after having a heart attack. 

Big data can help uncover relationships between diseases and/or co-morbidities since they tend to co-cluster. It can help contribute to a better understanding of the so-called system or network medicine, helping physicians diagnose cardiovascular disease including heart failure that frequently coexist with other medical conditions. 

Cardiology data at scale has the potential to revolutionize the diagnosis and treatment of heart disease, particularly in women, by providing new perspectives and enabling precision medicine approaches. However, even today, the “proportion of contemporary heart failure trials do not adequately report sex/gender-specific subgroup data”, which regulatory bodies, like the NIH and FDA, increasingly realize “that this lack of sex- and gender-specific data in biomedical research severely limited the generalizability of research and quality of care.” Moreover, the under-representation of women in clinical trials have resulted in misinterpretations of under-powered subgroup analyses of women. 

This article discusses the rising role of big data in cardiology and advancing heart disease research. We emphasize the increasing need for privacy and security in the collection and analysis of cardiology data at this key turning point for big data.

Importance of Big Data in Cardiology

Thanks to the latest scientific and technological improvements, cardiology, and medicine at-large, is entering an unprecedented epoch characterized by the production and release of a vast amount of data termed “big data”. 

Cardiology can benefit from big data, which is characterized by its ability to be measured in real-time, the magnitude of storage, processing, and analytics, the variety of data sources, its accuracy, and its sheer raw value. 

In cardiology, big data is generated and released by different sources and channels, like epidemiological surveys, national registries, electronic health records (EHR), claims-based databases (epidemiological big data), wet-lab, and next-generation sequencing (molecular big data), smartphones, smartwatches, and other mobile devices, sensors and wearable technologies, imaging techniques (computational big data), non-conventional data streams such as social networks, and web queries (digital big data), among others. Big data is increasingly more relevant, being highly ubiquitous and pervasive in contemporary society and paving the way for new, unprecedented perspectives in cardiology.

Cardiology data can help uncover relationships between heart disease and comorbidities that can help lead to diagnoses. Investigating the comorbidome could allow the implementation of “precision cardiology” by devising ad hoc multi-dimensional interventions targeting the specific patient sub-population. This is particularly relevant for heart disease in women, which could benefit greatly from increased research, and therefore increased data volume, to help diagnose and treat women with heart disease. 

For example, healthcare provision delivery has changed dramatically in the last decades. New models and pathways of managing and treating diseases have emerged. A new biomedical approach termed “P4 medicine” (preventative, predictive, personalized, and participatory) describes the shift from a “one-size-fits-all” theoretical framework to one in which the individual signature of the disease matters. This biomedical approach further emphasizes the need for large-scale data.

Cardiology Data and Artificial Intelligence (AI)

Big data analytics allow researchers to analyze large datasets encompassing diverse populations of women. It can enable the identification of gender-specific differences in cardiovascular risk factors, symptoms, disease progression, and treatment responses that may not be apparent in smaller studies or in datasets primarily focused on men. Analytics of cardiology big data can enhance diagnostic accuracy by identifying patterns and markers specific to women's heart disease. 

For example, machine learning algorithms trained on large datasets can detect subtle differences in symptoms, biomarkers, and imaging findings that may indicate heart disease in women, even when presentations are atypical. 

Cardiologists have developed an algorithm to detect atrial fibrillation, an irregular heart rhythm, a month before it happens. One example of AI finding patterns the human eye can't see has been developed at the San Francisco VA Medical Center, where deep learning and machine learning can learn from seeing millions of ECGs. AI can go through all the ECGs and identify complicated relationships. 

In this study, the goal was to identify who is at risk of atrial fibrillation. Researchers trained the machine to assess the electrocardiograms (EKGs) of patients who had experience an irregular heart rhythm in the last month, compared to those who had not to look for subtle differences. It essentially takes in an ECG, and then it makes an informed estimate based on previous 20,000 EKGS numbers. The algorithm then learns whether that guess is right or wrong, and then it adjusts its model to make a better guess next time.

Cardiology Big Data and Complications

While big data-based studies can offer a different point of view, some conflicting findings of randomized controlled clinical studies and small, well-conducted investigations can be found.

Such discrepancies could be due to the unique nature of the database used in the study including:

• Each cardiological database significantly varies in the methods deployed to collect and capture data and the population(s) it specifically represents. 

• Format of the database (structured vs. unstructured) could impact data quality. For instance, Hernandez-Boussard et al. mined a dataset inclusive of 10,840 clinical notes and found lower recall and precision rates (51.7 and 98.3%, respectively) in the case of structured electronic health records (EHRs), concerning unstructured EHR (95.5 and 95.3%, respectively), warranting the routine measurement of recall for each database/registry, before proceeding with data processing and analysis.

Big data repositories, registries, and databases are increasingly common in the field of cardiological practice and clinical research. However, there are significant considerable variations in socio-demographic characteristics, co-morbidities, and major complication rates between individual (single- or multi-center) and database-based studies, and even among patient registry-studies themselves (e.g.,, clinical vs. administrative database).