Prognosis, literally translated from its Greek roots, means "fore-knowing." It is one of the three classic responsibilities of a physician, the others being diagnosis and therapy. Although the breadth and detail of scientific evidence to support medical practice has expanded significantly in recent decades, there is a case to be made that evidence about prognosis has lagged behind evidence for physicians' other work.
Clinical questions in perinatal epidemiology demonstrate several important issues related to the conduct of prognostic research. Using examples from the study of prognosis for infants born with congenital diaphragmatic hernia, a condition for which estimates of survival vary widely, we illustrate the importance of selecting and specifying appropriate prognostic categories and contexts (e.g., time and place) to promote the appropriate interpretation of prognosis. In a study of extremely preterm birth--another condition with wide variation in available survival statistics--we show why decisions made by patients and physicians need to be accounted for in prognostic research.
By revealing several potential pitfalls of prognostic research, each of the projects described in this thesis also illuminates important opportunities for the better conduct and interpretation of such work. Making predictions about the future and providing this information to patients may always be difficult work for physicians. But better scientific evidence and interpretation of that evidence can at least make predictions more accurate. The aim of this thesis is to advance our knowledge about how to achieve such improvements.