The most nerve-wracking thirty minutes of my life was spent in the clinic of one of Australia’s top specialists. He put my pregnant wife through a series of ultrasound examinations in ominous silence. At his computer, he measured the various bones in the body of our unborn baby on the screen. When he was done, he finally turned away from his machines and spoke to us.
“His long bones are too short. Definitely some kind of abnormality.”
The remainder of the pregnancy passed for us in a fog of anxiety. When our son was born, the doctors whisked him away. There were CAT scans, x-rays and blood tests. He seemed healthy and happy, but the experts recommended more testing. The geneticist came back to have a look, but could not find any “obvious abnormalities” this time. Eventually we went home to start life with our new baby boy.
A happy ending, and we were relieved to have a healthy child. But our personal story of a wrong diagnosis illustrates a fundamental flaw in modern medicine – the overreliance on statistical data to treat an individual patient.
To understand what I mean, consider the process of diagnosis. Over three months, our baby’s bones were measured at weekly intervals and compared to standard growth charts. When they dropped too far away from the average, that was taken as evidence of a medical condition. He was reduced to points on a graph and sent via email. Sometimes, it felt like we were dealing with something abstract, not a human being. In consultations there was little discussion of our families, personal histories or opinions. Doctors spent much of their time looking at screens and everyone used the jargon of statistics – “percentiles”, “standard deviations” and “z-scores”.
But doctors often don’t understand the limitations of their data. All statistical data has inherent biases and flaws. For example, the growth charts used by our doctor were from Caucasian populations, but we are of South Asian background. In the case of bone lengths, the racial differences are significant, and therefore the data was of limited relevance in our case.
Even when relevant, data may not be effective in diagnosis. Measuring bone lengths is a crude way of finding skeletal abnormalities, but our doctors did not mention the fact that most such diagnoses are false positives. There was no explanation of the process and its reliability. Every measurement comes with a margin of error, but we were presented only with former and not the latter.
No one questioned the numbers. And no one asked – are they misleading us?
This is the deeper problem underlying these issues. In a world awash with data, some branches of medicine now play a dangerous game of reductionism. An individual with thousands of highly complex mental and physical traits is reduced to a small set of numbers. The numbers are easy to manipulate and analyse. But they are misleading because they are stripped of context. Unfortunately many processes discard qualitative knowledge about the individual and deal only with the data.
Statistical measures are useful, but they are not real. No one is exactly “average”. The average is an abstract, mathematical construction. It suffers from the same biases as the population data. Any statistical distribution can vary depending on where and how you get your data. Sometimes they vary a lot. Putting an individual on such a curve and then using it for decision making is problematic.
Numbers are seductive. They provide certainty. But people are not numbers.
Mistakes will always happen in medicine, and doctors are generally caring and competent. But the misuse and overuse of data is a systemic issue and reduces the scope for medical professionals to exercise judgement. An improved system would give doctors an incentive to spend less time on their computers and more time getting to know their patients.
Share
