For people who know me by now, they know that I do not hold back on my opinions when it comes to the ability of the NHS to analyze data. This does not mean that I think that everyone in the NHS is terrible. I have met some very smart clinicians. Some of these are starting to code, implement machine learning, and work together to advance the lagging NHS. I’ve also met some established academics who are using advancements in other fields to further their ability.

This is great. Nothing brings me joy when I hear from a clinician or student who is learning how to program, teaching themselves math, or fraternizing with graduates from another discipline to get that edge. Luckily, I live in London so most of the time I get to meet them in person. However, there is a group of people in the NHS that are too big to ignore. These people are happy with the status quo. They say buzzwords words like innovation, leadership, entrepreneurship, but when you look at their efforts, they rarely amount to anything more than a simplistic tickbox result. If there’s anything that they do not know much about or something that requires them to learn something new, they scoff, dismiss, and say that it’s not important. This isn’t unique to the NHS, in every industry, there are people who chant the buzzwords, do their job and go home. They are not bad people for doing this. Some of them feel pressured by this weird culture where everyone should have a career development plan and want to change the world by it. We seem to have this unhealthy culture where it’s considered terrible if you just want to do your job and go home. No matter what the individual motivation is, there is a stubborn damaging practice that prevails in the NHS.

This is brushing off new approaches instantly without actually knowing what the new approach entails. I’ve even come across one senior replying to my points saying that he’s been doing this for years. I’m not making this up, that was his response to 4 detailed points that I laid out. The one deflection tactic that I’m seeing more and more of is the vague claim that the problem is “more complex than that”. Of course, there will be some cases where that is true, but in a lot of cases, they are accepting findings that utilize even more simplistic approaches to analysis. Most of the time this deflection tactic uses misplaced fancy sounding words, and little discussion is even offered. The issue they state that they have is that it will not take into account a certain variable and therefore will not be perfect. This would be relevant if I was claiming that machine learning would be perfect, or that the tools that they are already using are perfect. However, this is far from the truth. Most of the time a simple logistic regression algorithm out of the box will be an improvement to what they are using. For instance, I communicated with a medical professor who was measuring the effects of two approaches to sedation. He measured the systolic of both groups and compared the average. However, there were a few very high readings that were skewing the average. I did a probability distribution. It gave the opposite conclusion to what he was saying. The method he was proposing was twice as likely to drop the blood pressure. His response, he was going to stick with average. He may have had an emotional attachment to his proposed method but all the “research fellows” attached to this didn’t raise concerns of such a basic error. Another thing that has to be noted is that medical data is not the hardest data to analyze ever. There are some advanced techniques that have been developed in other fields. For instance, in financial tech, we suffer from selection bias on a big scale. Because we reject applicants with bad scores credit scores, as time goes on, the weighting of the credit scores reduces, as we do not have data showing that people will low credit scores will default or not, because we did not accept them. Hyperparameters for logistic regression is one solution. However, I have heard medics describe the same selection bias issue as if it’s just a problem medicine has, and that people in other fields have never seen, could not solve, and cannot possibly understand.

So, for the people advocating machine learning and computational methods to aid medicine, keep at it, you’re the future. Yes, machine learning isn’t perfect, but it’s a hell of an upgrade to the methods that many medics are currently accepting and practicing.