This paper addresses the problems associated with defining and classifying events as nosocomial infections, discusses the methods by which rates of nosocomial infection are calculated and their rationales, and presents some specific rates useful in nosocomial epidemiology. Previously unpublished data demonstrate important differences between antibiotic susceptibility tallies produced by clinical laboratories and similar tallies derived from nosocomial infection surveillance data.

Conversion of real world events into categorical data presents formidable difficulties. Surveillance personnel must classify a given series of clinical events as 0,1, or more infections and make a determination as to whether each infection is nosocomial or community acquired. High-quality research studies to validate these efforts should compare the sensitivity and specificity of methods used to some “gold standard.” The gold standard is usually a review of medical records or patients by an infectious diseases physician. But even the standard is flawed. In clinical practice this flaw presents less of a problem because therapy for infectious diseases is generally quite safe and may be instituted when the probability of infection is 10%, 5%, or even lower. For surveillance purposes a higher standard is required, which is particularly important when surveillance information is used to provide feedback data to physicians who understandably bridle at overestimates of infection rates in their patients. The overestimation of infections based on weak evidence under-cuts feedback efforts.