Letter: Harnessing Big Data: The Need for Datathon Research in Neurosurgery

Abstract

To the Editor: As the field of neurosurgery progresses, “big data” is increasingly being used to shape clinical practice.1 Big data encompasses large-scale population databases with manifold information types including, but not limited to: demographic, clinical, radiographic, pathological, and genetic data for each subject. Publicly available datasets currently exist, which record neurosurgical outcomes,2 but there are also those that are subspecialty specific3 with other registries and data repositories, expected to be available over the coming years.4 Machine learning and artificial intelligence tools are well-suited for the analysis of high-dimensional clinical datasets and are becoming more readily available. However, what is limiting their use is the availability of operators who are not only proficient in these advanced statistical techniques, but also have intimate knowledge of the underpinning neurosurgical sciences and who can translate findings into clinical practice. The “datathon” is a research model that is ideally placed at this intersection and is already rapidly accruing momentum within intensive care research.5 Previous datathons have, for example, studied the impact of critical illness and ARDS on mortality.6,7 The datathon represents a recent variation of the “hackathon”: an event whereby teams of programmers together with designers and subject-matter experts generate creative solutions in an intense, time-limited competition.8 In contrast, datathons assemble clinical experts, data scientists, statisticians, and other researchers together with the aim of investigating healthcare problems in a collaborative environment through analysis of pregathered electronic patient datasets. Datathons establish a unique cross-disciplinary platform.5 They permit discussion and exchange of ideas and skills between a diverse group of specialists who would otherwise be geographically separated. Yet datathons are much more than this. They are high-yield research events that can gain significant traction in analyzing and solving healthcare problems in a very short amount of time. By leveraging large numbers of participants who can work synergistically under time pressure, and by having peer review and specialist feedback in real-time, pre-defined objectives can be met and often exceeded. Still, there are limitations with this approach. Datathon organization requires significant manpower, physical resources, and IT infrastructure capable of storing high-dimensional datasets and permitting their analysis. It also requires willing participants who are not technologically naive or are unfamiliar with data science. In the previous datathons, some of these requirements have been offset by recruiting financial sponsors, utilizing cloud-based servers for data storage, and asking participants to bring their own laptops and contribute toward fees. Despite these drawbacks, the attraction of performing high-quality research with an enviable network of collaborators over just a single day or weekend would be considerable for the academically minded neurosurgeon. With patient data accumulating at an accelerated rate, the datathon represents an effective and novel means by which neurosurgical big data can be analyzed and information translated from the “algorithm to the bedside.” Disclosures The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article.

Ähnliche Arbeiten