(233) Development of a multivariate prognostic model to predict mortality in neonatal sepsis patients.

Background: The elusive signs and symptoms are one of the chief hurdles for the unsatisfactory prognosis of neonatal sepsis. Developing a prediction model will support clinicians in building firm and improved decisions on disease management and supervision.

Objectives: To identify the risk factors and develop a prognostic model to predict mortality in neonatal sepsis.

Methods:

Design: Single-centre, observational, prospective, cohort study. Neonatal sepsis patients admitted to the intensive care unit were chosen as the cohort group and followed till discharge or death. Patient data and maternal risk factors were obtained from hospital records. Risk factors responsible for mortality were identified by multiple logistic regression, and based on the identified predictive variables, a mortality prediction model was developed.
Setting: The study site was Kasturba Hospital, a tertiary care teaching hospital in Manipal, India, for a period of three years. The sample size for the early-onset sepsis (EOS) category, 101 and the late-onset sepsis (LOS) category, 235, were enrolled after signing informed consent.
Main outcome measures: The primary outcome measures include mortality or recovery.
Statistical analysis: Data entry and statistical analysis were done using SPSS version 20.0. Nominal data was expressed and described in percentage and frequency, while parametric data were denoted as mean ± standard deviation.

Results: A total of 335 patients were recruited in the entire study, of which 100 were from early EOS, and 235 were from the LOS category. The mean age of the patients in the EOS and LOS groups was 1.87 ± 0.87 and 6.70 ± 1.68 days, respectively. 63% of the patients survived in the EOS category, while 67.23% survived in the LOS category. Extreme low birth weight (OR= 6.85, p= 0.03), very low birth weight (OR= 4.31, p= 0.04), extreme preterm (OR= 6.83, p= 0.01), very preterm (OR= 4.03, p= 0.03), moderate preterm (OR= 2.90, p= 0.00), PPROM (OR= 8.00, p= 0.001), maternal fever (OR= 5.38, p= 0.001), maternal UTI (OR= 6.32, p= 0.04), spontaneous preterm delivery (OR= 2.63, p= 0.03), surgical intervention (OR= 6.49, p= 0.00) Invasive ventilation (OR=5.098, p=0.001), Central catheter line (OR=5.431,p=0.004), Delay in breastfeeding (OR=4.001,p=.003) Acute kidney injury(OR=1.664,p=0.046) were the independent risk factors identified. The sensitivity and specificity of the developed model for EOS were 87.025% and 85.285% and for LOS were 90.089% and 86.763% respectively.

Conclusions: Most maternal risk factors showed significant results in the EOS model; meanwhile, the LOS model showed significant factors pertaining to neonatal risk factors. Low birth weight and low gestational age were the main risk factors responsible for mortality in both EOS and LOS.