Analysis of hospital infection surveillance data.

How often infection rates should be calculated and how large a change is required for "significance" are pertinent questions in nosocomial infection surveillance programs. A method is presented which establishes outbreak threshold infection frequencies. Comparison is direct and immediate: computation of rates or use of electronic data processing is not required. We have validated this method, using computer systems, by comparing the distributions of mean weekly incidence and prevalence statistics for each ward by nosocomial infection site in an acute care general hospital against both our theoretical outbreak threshold limits and the distribution of proven infection outbreaks. Sensitive and specific distinction between random variation or sporadic cross-infection and true persisting outbreaks requiring intervention is obtained. This approach provides a simple and timely alternative to intuitive after-the-fact interpretation of infection patterns which is applicable to infection surveillance and cost-effective infection control in hospitals of all sizes.