Abstract:
Late-onset neonatal sepsis is an increasing problem, and an effective and simple clinical prediction tool is needed for treatment decisions. Therefore, a simple bedside scoring was developed and validated. T The study was performed at Ramathibodi Hospital, Bangkok, Thailand. The derivative phase was a retrospective cohort study in which data were derived from medical records of 1,824 inborn neonates who were hospitalized for more than 72 hours during 1998-2000. The validation phase was performed in seventy three neonates in whom sepsis was suspected between June 2002 and June 2003. The incidence density of late-onset sepsis was 1.5/1000 hospital days/infant. The score was developed based on coefficients of Coxs Proportional Hazards Model. The predictive score consisted of duration of umbilical venous catheter usage (0, 1-7, and > 7 days; score=0, 2, and 4), temperature abnormality (score=3), respiratory insufficiency (score=2), hypotension (score=4), presence of neutrophil band form more or equal to 1% (score=2), and platelet count less than 150,000/uL (score=2). Receiver operating characteristics curve analysis demonstrated good predictive performance in the derivative study with area under the curve (AUC) of 0.99. The AUC in the validation study was 0.78. At the score of 4, the sensitivity was 88.9% (95%CI 70.8- 97.6) and specificity was 58.7% (95%CI 43.2- 73.0). The probability of late-onset sepsis was 0.15 if the score was less than 4. In conclusion, a simple bedside predictive score for late-onset neonatal sepsis was developed. Preliminary validation suggests good performance of the score.