Methodologic issues in hospital epidemiology. III. Investigating the modifying effects of time and severity of underlying illness on estimates of cost of nosocomial infection.

Published estimates of extra cost and prolongation of hospital stay attributed to nosocomial infection obtained from epidemiologic comparisons are almost twice as large as judgements in studies based on subjective impressions. It is possible that this disparity may result from confounding by time and severity of underlying illness. Whether the effects of time and secondary disease diagnoses modified the results of an epidemiologic comparison of infected patients and comparison subjects matched on primary diagnosis and operation have been investigated. Whereas the average prolongation of hospital stay in a prevalence series of patients with nosocomial infection was 13.3 days, the average prolongation for the corresponding incidence series of infections from the same study population was only 7.3 days, or about one-half as long. No substantive changes resulted from adjusting for duration of exposure to hospital prior to infection. Five selected secondary diagnoses had the potential for substantial confounding effects on epidemiologic comparisons but had little overall effect on the estimates in this study. The large size of our estimates in both prevalence and incidence series is not the result of residual confounding by the effects of time or secondary disease diagnoses. Results from prevalence and incidence series must be clearly distinguished because the same events will be perceived differently in the two types of series.