Friday, November 30, 2007

Forecasting in medicine


Nassim Taleb and Spyros Makridakis are co-editors of an upcoming special issue of the International Journal of Forecasting. The theme for this issue is intriguing: "Decision Making and Planning Under Low Levels of Predictability." On his website, Taleb has issued a call for papers from all disciplines and fields, including medicine. But why medicine? Forecasting in medicine occurs all the time, often with tremendous consequences.

Doctors act as forecasters when they diagnose patients. Developing a diagnosis requires gathering information, processing that information, and communicating a conclusion. Uncertainty infiltrates this process quickly. For example, physicians depend on their patients' abilities to communicate their symptoms and histories. When patients are poor historians or struggle to communicate their physical symptoms into words, doctors must gather information from other sources. Given our technological advances with diagnostic tests, doctors usually collect information from images, labs, vitals, and various other tools. Processing the results of these tests provides another source of uncertainty - radiologists misread X-rays for certain conditions in up to 30% of cases and most tests suffer from false-negative and false-positive results, for instance. Plus, physicians' cognitive biases - anchoring, confirmation bias, and various other heuristics - cloud these diagnostic landscapes even further. Finally, when physicians prescribe treatments based on their diagnoses, they make forecasts once again. Doctors predict how treatments, often drugs, will interact with and affect their patients' bodies - predictions marred by uncertainties such as adverse drug reactions and side affects, to name a few.

Clearly, when doctors collect and integrate information to generate diagnoses and treatments - forecasts subject to errors - they must employ tools or "tricks of the trade" that they either have learned from others or have developed from experience. These trade tricks are exactly what Taleb and Makridakis want to publish. I have a lot of respect for doctors who search for effective ways to deal with the uncertainties involved in diagnosing and treating patients; these physicians embrace reality, and, in the end, embrace efforts to improve patient care.

In part, the art of medicine is the art of decision making under uncertainty. In the increasingly complex and recursive Information Age that we live in today, I feel that physicians must be trained - as Dr. Jerome Groopman contends - to confront, deconstruct, and act in the face of uncertainty, and this training must teach physicians practical ways to avoid cognitive blunders. Clinical medicine is a thinking art and science, so I am hopeful that members of the medical community will submit papers to this special issue of the International Journal of Forecasting - many doctors are excellent thinkers, and other disciplines and fields could learn from their insights.

Here is the noble challenge that Taleb and Makridakis have posed:

The critical question that this special issue aims to address is what we can do if we accept the serious limits in our predictability and the huge uncertainty surrounding our future decisions and plans. It is, therefore, critical to consider, and provide practical solutions, on how we can live with such uncertainty without being paralyzed by hesitation and not falling victims of the illusion of control, by wrongly believing that we are able to forecast and by pretending that uncertainty does not exist.


Creative responses to this commendable challenge provide tangible ways to balance the paradox of searching versus acting.

No comments:

Post a Comment