ABSTRACT
PURPOSE: To develop a new scoring method for predicting the mortality at 28 days of life in premature infants. METHODS: From January 2000 to August 2001 (period I), medical records of 138 premature infants less than 32 weeks' gestation admitted at Asan Medical Center were reviewed retrospectively, and a predictive model (named as KCRIB28) was developed based on several steps of regression analysis. From January 2001 to December 2002 (period II), KCRIB28 was validated for 99 premature infants admitted at Ulsan University Hospital using retrospective cohort study, and compared with results of CRIB score for those infants. The primary outcome variable was death at 28 days of life. RESULTS: One hundred twenty seven (92%) of 138 infants during period I and 86 (86.9%) of 99 infants during period II survived over 28 days of life (P=0.07). The mean gestational age and birth weight (+/-SD) were 29.1+/-2.0 weeks and 1.23+/-0.35 kg, respectively, during period I, and 29.6+/-2.0 weeks and 1.37+/-0.46 kg, respectively, during period II. Gestational age, birth weight, 5-min. Apgar score and peak inspiratory pressure divided by birth weight at 12 hours of age were selected as parameters of KCRIB 28. Areas under the receiver operator characteristic curves (AUCs) were 0.84 (SEM, 0.08; 95%CI, 0.77-0.90) during period I and 0.89 (SEM, 0.06; 95%CI, 0.81-0.94) during period II. No differences in AUCs were found between KCRIB 28 and CRIB for patients during period II (P=0.75). CONCLUSION: KCRIB 28 can be used as a valuable measurement to predict the mortality of premature infants at 28 days of life.