I started writing content for D4 a few months ago. At the time, my knowledge of the area was sparse and to be honest was limited to actually working within a pharmacy rather than pharma itself. Over the past few months, I have begun to understand the area a lot better and dare I say it, find it interesting. However, the one thing I have noticed is an apparent dislike amongst some of the community towards certain ‘buzzwords’, such as AI and blockchain. I never really understood why, until now.
Use cases?
At the Festival of Genomics and Biodata 2021, I was watching a panel with Paul Agapow (Health Informatics Director, AstraZeneca), Victor Neduva (Senior Principal Scientist, MSD) and Natalie Gavrielov (Director of Medical Writing, BioForum) about machine learning and how it can unlock the power of real-world evidence (you can read more here). During the panel, an audience member asked: What are the exact use cases/examples of machine learning for RWE? The panel laughed and I assume hoped that their video connection would timely drop out or that they turned into a cat. Why? Because they couldn’t really think of areas where this is actually being implemented and/or works in practice. This I have noticed is a recurring theme.
In the past few months, I have read (and admittedly written) multiple articles about how machine learning can be applied within radiology. Several algorithms have claimed that they can outperform a human radiologist and then the papers end saying something about improving efficiency and patient care. While I don’t doubt that these algorithms are impressive, and in theory they could revolutionise healthcare, in reality they don’t actually work because it’s a much more complicated issue. Real data is a lot messier than that used to train the models.
The Hype Problem
Derek Lowe (Director in Chemical Biology Therapeutics, NIBR) recently wrote a commentary in Science about a paper that explored hype and hypocrisy within science. The commentary titled ‘The Hype Problem’ discussed issues of publicity-mongering and how incentives have worsened the problem. While pride and excitement about your own work is expected, deliberate hype or as Lowe referred to it – “sitting down and saying ‘All right, how can we generate the biggest headlines?’” – is not. It is clear that some authors are throwing buzzwords into their titles just for clickbait and overclaiming the value of their methods. Once academia does this, then it gets into the hands of the media who let’s face it start making all sorts of claims.
During the panel, Agapow humorously said that “scientists are black boxes” and it’s funny because it’s true. Experts are striving for transparency from these black box algorithms, but we ourselves are not transparent.
The world is advancing – but our ability to act on our advances and implement these into practice is lagging behind. For example, I learned about pharmacogenomics while I was at school and how it will personalise healthcare and reduce side effects. But even today it still isn’t being harnessed as effectively and widespread as it should be. Now, you may be thinking “it’s not that simple” – and you would be accurate – it’s not. But maybe if I wasn’t sold the dream that pharmacogenomics was going to save lives imminently then I wouldn’t be so disappointed now.
We need to stop hyping things and start managing expectations. We also need to identify the barriers to clinical adoption and staring working on these! Because otherwise we will end up like Back to the Future fans waiting for hoverboards to exist.
Image credit: Photo by naomi tamar on Unsplash
