To compare pediatric intensive care unit (ICU) mortality risk using models from two distinct time periods; and to discuss the implications of changing mortality risk for severity systems and quality-of-care assessment.
Data Sources and Setting
Consecutive admissions (n = 10,833) from 16 pediatric ICUs across the United States that participate in the Pediatric Critical Care Study Group were recorded prospectively. Data collection occurred during a 12-mo period beginning in January 1993.
Data collection for the development and validation of the original Pediatric Risk of Mortality (PRISM) score occurred from 1980 to 1985. The original PRISM coefficients were used to calculate mortality probabilities in the current data set. Updated estimates of mortality probabilities were calculated, using coefficients from a logistic regression analysis using the original PRISM variable set. Quality-of-care tests were performed using standardized mortality ratios.
Risk of mortality from pediatric ICU admission improved considerably between the two periods. Overall, the reduction in mortality risk averaged 15% (p < .001). Analysis of mortality risk by age indicated a large improvement for younger infants. The mortality risk for infants <1 mo improved by 39% (p < .001). Mortality risk improved by 28% (p < .001) for infants between 1 and 12 mos. Analysis of mortality risk by principal diagnosis indicated substantial improvement in respiratory diseases, including respiratory diseases developing in the perinatal period. The mortality risk for respiratory diseases improved by 45% (p < .001). The improvement in mortality risk substantially deteriorated the calibration of the original PRISM severity system (p < .001). As a result of changing mortality risk, the standardized mortality ratios across the 16 pediatric ICUs demonstrated substantial disparities, depending on the choice of models.
This study documents differences in pediatric ICU risk of mortality over time that are consistent with a general improvement in the quality of pediatric intensive care. Despite continued widespread use of the original PRISM, recent improvements in pediatric ICU quality of care have negated its usefulness for many intended applications, including quality-of-care assessment. (Crit Care Med 1998; 26:1737-1743)