The way we currently conduct clinical research can be expensive and energy intensive. We ask a research question, and then we spend hundreds of thousands of dollars trying to set up a clinical trial to answer that question. The process involves organizing and training the sites, collecting generated data for review and analysis.
Meanwhile, the answers to many of the research questions we are posing might be in front of our eyes. The advent of electronic medical records and cloud-based information storage has made ECG database mining a possibility. By using ECG data analysis, we can answer clinical questions that have yet to be asked.
Data Mining Multiplies the Power of ECG
ECGs are some of the simplest tests conducted in the healthcare setting, and they can provide substantial information. ECGs not only include information about acute-care situations, such as myocardial infarction; they can also show subtle cardiac changes that may be the earliest manifestation of adverse cardiac remodeling, such as left ventricular hypertrophy. In addition, ECG can contain clues about hidden genetic diagnoses, such as hypertrophic cardiomyopathy.
If a single ECG can yield powerful insights, then a consolidated database of millions of annual ECGs on a standardized reporting platform could provide a rich source of information for researchers aiming to leverage data mining.
A Journal of Applied Research Review (JARR) article defines data mining as the analysis of observational data sets to derive relationships or extract hidden trends.1 This approach is especially applicable to ECG analysis for three key reasons:
1 . Unlike other electronic medical records, ECGs are standardized, always obtained in the same format, and always reported in the same way. These qualities make them amenable to artificial intelligence analysis and computerized algorithm review.
2 . ECGs can be readily stored in a cloud-based or centralized database in large quantities.
3 . As the JARR review notes, there are numerous validated and well-studied ECG data mining algorithms that can review and analyze thousands of ECGs in seconds.
In aggregate, these advantages create the perfect recipe for the application of data mining, a technique that has the potential to reveal trends and insights quickly and accurately.
In this video, Claus Graff, Associate Professor, Department of Health Science and Technology, Aalborg University describes some of his work with a large ECG database in Denmark.
The Trials of Tomorrow
The research designs of yesterday, which primarily involved expensive, outcomes-based, randomized clinical trials can now give way to new methodologies that employ existing databases to obtain novel insights.
The Wall Street Journal reported that Pfizer Inc., Johnson & Johnson and Amgen Inc. sought drug approvals via data-mining methods, which proved faster and more cost-effective than traditional approaches.2 A study in the Annals of Clinical and Laboratory Research on improving clinical trials in diabetic patients through data mining concluded that if the methodology is more widely adopted by researchers, "it will be much easier to collect, evaluate and analyze data and improve the quality of clinical trials all over the world."3
Importance of ECG Database Connectivity and Interoperability
Having an ECG database that is readily available anywhere at anytime accelerates data mining and enhancing its capabilities. To allow for effective ECG data analysis, though, the database must leverage connectivity and interoperability.
When ECG data is integrated into a hospital's electronic medical record, cardiologists can easily access ECG-specific information from their handheld device. Not only does an interoperable database provide one convenient location for patient data storage, but it also makes it easier to apply artificial intelligence algorithms for interpretations and abnormality identification. As a result, researchers can more readily recognize trends.
Still, there are still multiple barriers to achieving this level of ECG database connectivity. Some facilities still resist the adoption of new technologies, and others struggle with competing financial priorities between operational and clinical decision makers.
Effective cybersecurity remains a significant hurdle. Successful data protection requires multiple layers of security that can be integrated into a facility's existing infrastructure. Steps to overcome this obstacle can include two-factor authentication, encrypted virtual private networks (VPNs), antivirus software, and training providers to log out of connected devices.
Resources:
1. Tseles D, Vasileiadou S, Alafodimos C, et al. Data mining in ECG data. Journal of Applied Research Review. 2015 December; 15(2015): 70-76. http://journal.uniwa.gr/index.php/science/article/view/datam
2. Loftus Peter. "Drugmakers turn to data mining to avoid expensive, lengthy drug trials." The Wall Street Journal. December 23, 2019. https://www.wsj.com/articles/drugmakers-turn-to-data-mining-to-avoid-expensive-lengthy-drug-trials-11577097000.
3. Kaur, Chandeep, and Olufemi Muibi Omisakin. "Data Mining Methods to Improve Clinical Trials in Diabetic Patients." Annals of Clinical and Laboratory Research 06, no. 04 (2018). https://doi.org/10.21767/2386-5180.100266.
