Use of the Pediatric Risk of Mortality score to predict nosocomial infection in a pediatric intensive care unit.

Abstract

To define infection rates in patients with Pediatric Risk of Mortality (PRISM) scores greater than and less than 10 on admission to the pediatric ICU (PICU). Descriptive. An 18-bed PICU admitting patients of all ages except nonsurgical neonates; within a 585-bed tertiary care pediatric hospital. Patients admitted to the PICU from July 1987 to February 1988 inclusive. Of 685 admitted, 480 were followed for greater than or equal to 72 hr. The baseline state of the patients on admission was determined by a designated intensivist using the PRISM score. Other variables included age, length of stay, and hospital day of onset of infection. Infections were identified by a designated intensivist who undertook prospective daily bedside observation, chart, radiographic, and laboratory review. Equal portions of patients had PRISM scores less than and greater than 10. Significantly more infections occurred in the high PRISM population (10.8% vs. 3.4%, p less than .001). This association held through age, service, and length of stay. Sensitivity, specificity, positive and negative predictive values of a PRISM score greater than 10 were 75%, 53%, 11%, and 97%, respectively. Bacteremias accounted for 36% of infections, skin/eye/drain site 22%, respiratory 16%, wound 15%, and urine 9%. The most prevalent organisms were coagulase-negative staphylococci (32%), Pseudomonas aeruginosa (23%), Candida sp. (20%), and S. aureus (9%). A PRISM score greater than 10 on PICU admission characterizes a population within the PICU at increased risk of infection. However, 93% of patients did not develop infection and thus, a negative predictive value of 97% yields little additional information.