Navigating medicine through ‘noise’

Dr Ajit Bhagwat

Daniel Kahneman, who won the Nobel prize in economics in 2002, passed away recently. He was not an economist in traditional sense. He was a behavioural economist who studied the effects of human behaviour on all aspects of human life including economics. The ramifications of his research are profound. In his last book ‘Noise’, he discusses the fundamental flaws in human decision-making and how ‘noise’ can cloud judgment and adversely affect the outcomes. Here are some medical takeaways from his philosophy.

Clinical diagnosis in medicine is based on history, physical examination and investigations. Interpretation of history and relevant findings on physical examination are paramount in coming to right conclusions. However, interpretations are subjective and different doctors examining the same patient come to vastly different conclusions (interpersonal variation) and the same doctor examining the same patient at different times can come to different conclusions (intrapersonal variation). The same findings are observed when it comes to commenting on an X-ray or a CT scan by a radiologist. Kahneman calls this variation ‘noise’! The other term for noise is ‘judgment error.’ Noise is a cause of great concern for the patients and their families causing great confusion in decision-making and leading to a lot of waste of time, money and energy. There are three types of noises. 1. Level noise: Prescription of antibiotics typifies this class. Some doctors prescribe it at the drop of a hat (overprescribing) and some don’t do it until it’s agonisingly too late (under-prescribing). A skilled doctor would be the one who does minimal over or under-prescribing! 2. Pattern noise: A medical examiner may hate to listen to poor English by the student and gives him poor grades although the content of his presentation is very good! 3. Occasion noise: Outcomes of medical procedures performed by tired, sleep-deprived doctors at wee hours of night are poorer! Similarly, a study from Israel found that judges give harsher punishments when the verdict is declared just before lunch than after lunch.

To understand the magnitude of such noise in medicine, let’s take example of radiology. One large study found that frequency of false negative readings of mammograms ranged from 0% (radiologist was correct every time) to greater than 50% (radiologist missed the diagnosis of cancer in more than half cases). Similarly, false positives ranged from 1% to 64% (indicating that the radiologist overdiagnosed cancer in 64% cases when cancer wasn’t there). Isn’t that scary?

How can doctors steer away from these noises?

1. Acquiring skills: Nothing can substitute good training in medicine. Working under a master teacher for a few months is worth far more than reading books on the subject for years! He can make you see the nuances and the finer points in mammogram images that are barely discernible to an inexperienced eye!

2. Using Bayesian analysis: Tests are never interpreted in isolation. Predictive accuracy of any test depends largely on prevalence of the disease in the given population. A suspicious shadow on mammogram in a 60 years old woman with strong family history of breast cancer is far more likely to be a cancer than a similar shadow in a 25 years old woman without any family history.

3. Using artificial intelligence (AI): Using AI is the most reliable way of reducing noise in medicine. When human factors are taken out of decision loops, all types of noises are eliminated. The next decade is going to be full of exciting breakthroughs on use of AI in medicine.

A good doctor is the one who makes few judgement errors, and a better doctor makes fewer errors than him!

(The writer is senior cardiologist).