One of the lessons I learned in politics is that systems rarely fail because of bad people. They fail because of bad structures. Put the wrong incentives in place, divide responsibility into narrow silos, and even the most dedicated professionals will end up serving the system rather than the individual.
Unfortunately, modern healthcare is beginning to look uncomfortably familiar.
Over recent months, I’ve been examining the way treatment decisions are made for conditions such as atrial fibrillation. What I found was a system increasingly driven by population level algorithms, where individual patients risk being reduced to statistical categories. The most striking example is the CHA₂DS₂ VASc score — a tool now used internationally to determine whether someone should be placed on lifelong anticoagulant medication.
On the surface, it appears sensible: a simple checklist that assigns points for age and certain medical conditions. But the more closely I looked at it, the more uneasy I became. The score treats all patients of a given age as identical, regardless of their fitness, metabolic health, lifestyle, or biological variation. A healthy, active 65 year old is given the same risk weighting as someone with multiple chronic conditions. This may be convenient for large scale modelling, but it is a poor substitute for personalised care.
What troubled me most was not the score itself, but the structural incentives surrounding it. Once a tool like this is embedded in clinical guidelines, it becomes the gateway to treatment pathways that are difficult to question. Clinicians are judged on compliance whilst pharmaceutical companies benefit from the expansion of “at risk” categories. And the patient — the person who must live with the consequences — becomes the least influential voice in the process.
This is not a criticism of individual doctors. They work under immense pressure and within systems they did not design. But the system they operate in has drifted into the same siloed thinking that plagues government. Cardiology focuses on stroke prevention. Primary care focuses on guideline adherence. Each part optimises its own metric, yet no one is accountable for the whole patient.
Nowhere is this more evident than in the treatment of side effects. Preventing a stroke is, of course, a vital goal. But preventing one life threatening condition by causing another is not an ethically neutral trade off. Too often, side effects are treated as unfortunate but acceptable collateral damage — a cost absorbed by the patient, not by the system that prescribed the medication.
When I raised concerns about this, I was struck by how quickly the conversation returned to the score. Not my individual physiology. Not my previous adverse reactions. Not my personal risk tolerance. Just the score. It was as if the algorithm had become the patient, and I had become the supporting data.
This is the heart of the problem. A system designed for efficiency has begun to mistake uniformity for safety. But safety is not achieved by treating everyone the same. It is achieved by treating each person as an individual. I am not rejecting medicine. I am rejecting the idea that a crude statistical tool should override personalised clinical judgement. I am asking for a healthcare system that sees patients as people, not probabilities.
Until that happens, I — and many others — will continue to question pathways that prioritise institutional metrics over lived experience. And we will continue to ask a simple, reasonable question: When did the system stop seeing the patient? Because if we don’t ask it now, we may not like the answer later.
