On June 19th 2018, Prof. Tito Poli gave a speech on "Clinical and cultural challenges ahead for the efficient use of Big Data in healthcare" in the frame of the Workshop “Digitalisation and Big Data: Implications for the health sector” organized by the by the Policy Department A, Economic & Scientific Policy for the Committee on the Environment, Public Health and Food Safety (ENVI) Working Group Health of the European Parliament.
Prof. Poli's speech illustrated how, since early 2000s, Machine Learning (ML) and Big Data analytics have been developed to respond to the growing need to achieve an integrated analysis of ever-increasing volumes of medical knowledge and patients’ data and to develop easy-to use IT tools assisting researchers, physicians and pharmaceutical industry in new knowledge discovery, development of targeted drugs, personalized diagnosis and treatment decision making.
However, Prof. Poli highlighted that, despite the recognized potential of Big Data, the skepticism of physicians and researchers in adopting such “digitalized revolution” in their daily activity has created a series of clinical and cultural challenges that hinder the efficient use of such models.
In Prof. Poli's opinion, the strongest resistance is cultural and derives from the very different approaches to produce evidence that characterize Evidence-Based Medicine (EBM) and Big Data Analytics. While traditional EBM is hypothesis-driven and validated through scientifically assessed Randomized Controlled Trials (RCTs), Big Data analysis is data-driven and may produce "unexpected" results. Physicians are accustomed to make their decisions independently, using their own clinical judgment and considering contextual data, rather than relying on "black-box" -derived protocols from Big Data analysis.
The main pitfall of Big Data analytics is considered the inscrutability of many algorithms, i.e. the rationale for the outputs generated by these models is inexplicable not only by physicians but also by the engineers who develop them.
So, the final question is: the choice of one method excludes the other? Despite their differences, EBM and ML can assist one another and hybrid algorithms are emerging.
There is a growing need in the clinical field for a specific training of medical personnel (more simple for the so-called “digital natives”, i.e. students born with the new technologies of the digital age) and adequate ITC support at the hospital level to prevent that Big Data stored locally still remain an opportunity not adequately exploited.