Neonatal intensive care unit: predictive models for length of stay

• Published in: Journal of perinatology Vol. 33; no. 2; pp. 147 - 153
• Main Authors: Bender, G J, Koestler, D, Ombao, H, McCourt, M, Alskinis, B, Rubin, L P, Padbury, J F
• Format: Journal Article
• Language: English
• Published: United States Nature Publishing Group 01.02.2013
• Subjects: Benchmarking; Birth size; Birth Weight; Buildings and facilities; Censuses; Cohort Studies; Data collection; Datasets; Female; Gestational Age; Goodness of fit; Health aspects; Hospitals; Humans; Illnesses; Infant, Newborn; Infant, Newborn, Diseases - diagnosis; Infant, Newborn, Diseases - therapy; Intensive care; Intensive Care Units, Neonatal - utilization; Length of stay; Length of Stay - statistics & numerical data; Male; Management; Models, Statistical; Morbidity; Mortality; Multivariate Analysis; Neonatal intensive care units; Neonates; New England; Newborn babies; Obstetrics; Pediatrics; Physiological aspects; Physiology; Prediction models; Predictive Value of Tests; Regression Analysis; Retrospective Studies; Risk Factors; Severity of Illness Index; United States; New England; United States; Rhode Island; United States > US
• ISSN: 0743-8346; 1476-5543
• DOI: 10.1038/jp.2012.62

Hospital length of stay (LOS) is important to administrators and families of neonates admitted to the neonatal intensive care unit (NICU). A prediction model for NICU LOS was developed using predictors birth weight, gestational age and two severity of illness tools, the score for neonatal acute physiology, perinatal extension (SNAPPE) and the morbidity assessment index for newborns (MAIN). Consecutive admissions (n=293) to a New England regional level III NICU were retrospectively collected. Multiple predictive models were compared for complexity and goodness-of-fit, coefficient of determination (R (2)) and predictive error. The optimal model was validated prospectively with consecutive admissions (n=615). Observed and expected LOS was compared. The MAIN models had best Akaike's information criterion, highest R (2) (0.786) and lowest predictive error. The best SNAPPE model underestimated LOS, with substantial variability, yet was fairly well calibrated by birthweight category. LOS was longer in the prospective cohort than the retrospective cohort, without differences in birth weight, gestational age, MAIN or SNAPPE. LOS prediction is improved by accounting for severity of illness in the first week of life, beyond factors known at birth. Prospective validation of both MAIN and SNAPPE models is warranted.