Objective: To investigate the use of the formula of Rhame and Sudderth for the interconversion of prevalence and incidence data on the frequency of nosocomial infections.
Design: Comparison of observed and calculated incidence data and prevalence data.
Setting: One 8-week incidence investigation in the surgical and intensive care units of eight medium-sized hospitals; three separate point-prevalence studies in the same units.
Results: The overall prevalence observed after the three prevalence studies in 2,169 patients was 6.8% (95% confidence interval [CI95], 5.7-8.0). In 2,882 discharged patients observed during the incidence study, the mean hospitalization was 9.8 days; patients with one or more nosocomial infection had a mean hospitalization time of 22.3 days and a mean interval of 8.2 days from admission to the first day of infection. Based on these data, the overall calculated incidence was 4.7%, whereas the observed incidence was 4.3% (CI95, 3.6-5.2). Vice versa, an overall prevalence of 6.2% was found when calculated from the observed incidence data. The incidence data calculated from prevalence data also were within the confidence interval of the incidences observed for urinary tract infections and surgical-site infections. (However, it was not possible to convert the data for two of the eight hospitals.)
Conclusion: The approximate mathematical relationship between the prevalence and incidence data of nosocomial infection is confirmed by this study. However, although it is theoretically possible, we would not recommend the conversion of prevalence into incidence data or vice versa.