How Do We Know If Something Is an Important Risk Factor? Statistical and Clinical Significance

Abstract This notion, referred to as the “butterfly effect” in chaos theory (and often stated with different locations and different events), is ascribed to MIT mathematician Edward Lorenz. Making fun of his own concept that even tiny relationships among characteristics or events can effect huge changes, Lorenz later joked: “One meteorologist remarked that if the theory were correct, one flap of a seagull’s wings would be enough to alter the course of the weather forever.” Joking aside, Lorenz’s point is an important one to risk estimation. Some relationship, no matter how small, probably exists between any characteristic and a subsequent event. When a researcher identifies a risk factor—that is, establishes a “statistically significant” relationship between a factor and an outcome and demonstrates the precedence of the factor—the relationship may be so small as to have very little “real-life” significance. Few people would seriously suggest we capture butterflies to prevent unstable weather patterns. Nevertheless, newly discovered risk factors are often taken very seriously, sometimes without much consideration of how well the risk factor actually predicts the outcome.