Medicine's history has a rich track record of interventions. Since William Osler introduced the ideas of knowledge management and a learning organization into medical practice in the late 1800s, improvements to life expectancy have been consistent and remarkable. He was fond of saying, ""He who studies medicine without books sails an uncharted sea, but he who studies medicine without patients does not go to sea at all."

Throughout this period, diagnostic tests were at the core of all improvements. For test results to be useful, they have to be timely and reliable; to be reliable, they need to be well-defined and repeatable.

Medical records have been equally important, since they have enabled longer-term studies and decision-making about trends and the effectiveness of interventions. Biostatistics is the field where interventions are compared against outcomes (mortality and morbidity) for evaluation of the disease and treatment process. Causes of death vary widely across different settings, which reinforces the importance of considering contextual information when doing this analysis. Survival rates for patients once they have been diagnosed and begun treatment, are also significantly different, and are one of the primary causes for differences in life expectancy across countries.

To go back to Groopman,

A movement is afoot to base all treatment decisions strictly on statistically proven data. This so-called evidence-based medicine is rapidly becoming the canon in many hospitals. Treatments outside the statistically proven are considered taboo until a sufficient body of data can be generated from clinical trials. Of course, every doctor should consider research studies in choosing a therapy. But today’s rigid reliance on evidence-based medicine risks having the doctor choose care passively, solely by the numbers. Statistics cannot substitute for the human being before you; statistics embody averages, not individual.

Medical students are taught that the evaluation of a patient should proceed in a discrete, linear way: you first take the patient's history, then perform a physical examination, order tests, and analyze the results. Only after all the data are compiled should you formulate hypotheses about what might be wrong. These hypotheses should be winnowed by assigning statistical probabilities, based on existing databases, to each symptom, physical abnormality, and laboratory test; then you calculate the likely diagnosis. This is Bayesian analysis, a method of decision-making favored by those who construct algorithms and strictly adhere to evidence-based practice. But, in fact, few if any physicians work with this mathematical paradigm.

This tension between decision-making based upon facts and data, and decisions based upon intuition, is at the core of cultural resistance for many improvement efforts. Unfortunately, the facts and data tell us that intuition, while often necessary, is often wrong, due to flawed perceptions and thinking.

It is not enough to just track data flow through the system retrospectively, though; you need the ability to predict the future. Such predictive measures can be synthesized over time from other measures, once they are stabilized. An example of this is the APACHE II assessment used by most emergency rooms for patient triage, especially when a hospital's demand is greater than capacity.