Presented at the Neonatal Society 2010 Spring Meeting.
Santhakumaran S1, Statnikov Y1, Gale C1, Manktelow B1,2, Modi N1 on behalf of the Neonatal Data Analysis Unit
1 Section of Neonatal Medicine, Department of Medicine, Chelsea & Westminster Campus, Imperial College London, UK
2 Department of Health Sciences, University of Leicester, UK
Members of the NDAU Steering Board: Abbott J (BLISS), Brocklehurst P, Costeloe K, Draper E, Kemp J, Majeed A, Modi N, Wilkinson A
Background: The Neonatal Data Analysis Unit (NDAU) has previously investigated the extent to which intensive care is provided by non-level 3 neonatal units (1). It was found that while intensive care represented a small proportion of the total care provided by most non-level 3 units contributing data to the NDAU, they collectively provided 43% of total intensive care days. For babies born in a non-level 3 unit, it may be appropriate to provide short term intensive care, but transfer is usually considered desirable for additional intensive care. Prior to the introduction of managed neonatal networks, the UK Neonatal Staffing Study Group showed that risk-adjusted mortality was associated with high unit occupancy, but did not vary with unit size, and recommended centralisation of care of the sickest infants (2).
Aim: The aim of this study was to evaluate whether mortality varies with unit level and transfer status for preterm babies (≤ 27+6 weeks) requiring intensive care for at least 2 days after delivery, before and after adjustment for case-mix.
Methods: The NDAU has Caldicott Guardian approval from contributing neonatal units to receive anonymised data entered onto neonatal.net. These electronic records were analysed for all babies born in 2008 at ≤27+6 weeks gestation and admitted to contributing units. All babies who required intensive care (BAPM 2001) for the first two days after birth and did not die during this period were included. For each baby, the daily electronic records were used to determine whether there had been a transfer in the first week after birth.
A logistic regression model was constructed with mortality as the outcome, and neonatal unit level and transfer status (yes/no) as the predictor variables of interest. An interaction term between unit level and transfer status was used to establish whether the effect of unit level on mortality is different in transferred and non-transferred babies. Gender, gestational age, birthweight, multiplicity and use of antenatal steroids were included as covariates in the model. The CRIB II score was not used for risk adjustment as base excess, which is required to calculate the score, was only available for 241/1047 babies. An interaction between gender and birthweight was included to adjust for gender-specific birthweight. Robust standard errors were used to account for clustering by neonatal unit.
Results: There were 1216 babies born in 2008 at ≤27+6 weeks gestation recorded by 86 contributing neonatal units. Of these, 1100 survived the first 48 hours after birth and received intensive care during this period; 53 babies were excluded because the mortality outcome was unknown due to transfer to a unit not contributing data to the NDAU, leaving 1047 babies. Of these 568 were born in a level 3 unit of whom 22 were transferred out; 479 were born in a non-level 3 unit (222 transferred). Before adjusting for risk factors, the odds ratio for mortality for non-transferred babies born in a level 1 or 2 unit compared to (i) transferred babies born in a level 1 or 2 unit was 0.37 (95% CI 0.23 to 0.59, p<0.001) and (ii) non-transferred babies born in a level 3 unit was 0.44 (95% CI 0.27 to 0.69, p<0.001).
These differences were not statistically significant after risk adjustment: (i) OR 0.80, 95% CI 0.50 to 1.28, p=0.35 (ii) OR 0.73, 95% CI 0.46 to 1.15, p=0.17. There was no statistically significant difference in mortality between transferred and non-transferred babies born in level 3 units, but only 4% (22/568) of babies born in a level 3 unit were transferred. Of the babies receiving care in non-level 3 units who were transferred within one week, most (193/222) were transferred within 12 hours. There was no evidence (p=0.9) that those who are not transferred within 12 hours had different mortality outcomes.
Conclusion: Significant differences in mortality are not observed after adjustment for risk factors. This suggests that differences in mortality may be attributable to differences in case mix and not to unit characteristics or transfer status.
Acknowledgements: SS and YS are funded by NIHR programme grant, Medicine for Neonates.
1. Statnikov Y et al. Neonatal Society 2009 Winter Meeting 2009
2. The UK Neonatal Staffing Study Group. The Lancet, 2002; 359: 99-107