Statistics from Altmetric.com
Many regional, national or international neonatal networks, such as the Vermont Oxford Network or the EuroNeoNet, collect data on the outcome of very low birth weight infants (VLBWI) for benchmarking in order to improve quality of care. Whereas physicians use these data to assure quality of the outcomes of interest and to target interventions and new treatments, data collected on a national level may be used by political or healthcare organisations to reorganise their healthcare system. Furthermore, pregnant mothers with threatened preterm delivery may use data from individual units to decide at which perinatal centre they want treatment to be provided. Specifically, pregnant mothers and their families may want to know the chances of survival to discharge and the risk for morbidity as both outcomes may affect their life dramatically. This aspect is extremely important in the situation of threatened preterm delivery at the threshold of viability for decision making on a provision of life support.
The obstetrical approach at the lower limit of viability may vary widely. In a prospective study comparing 10 different regions across Europe, some key obstetric interventions such as administration of corticosteroids, antenatal transfer to a level III perinatal centre and caesarean section for fetal reasons were provided very differently for pregnant mothers and their preterm babies at the threshold of viability.1 Furthermore, the rate of births declared as stillbirths were more than 10 times higher in some regions compared with others.1 Therefore, differences in obstetric and neonatal policies seem to have a pronounced influence on the decisions to prolong pregnancy, during labour and delivery and to provide life support after birth of VLBWI, especially for the extremely immature infants at the threshold of viability. Interestingly, the outcome for more mature preterm infants seems to be better in perinatal centres taking care on a larger number of extremely preterm infants at the threshold of viability.2 Using outcome data from the German national neonatal registry to compare the performance of individual units may be particularly prone to bias, as it has been shown that more than 30% of extremely preterm infants born alive (by official birth registry) may not be included in this registry.3
Smith and coworkers, in another paper in this edition of the journal, explored registration practices to quantify their impact on the performance of perinatal centres in England.4 They analysed the regional variation in the registration of preterm births <24 weeks of gestation and the impact on infant death rates for 147 out of 151 English Primary Care Trusts (PCTs) in an almost population-based cohort study. Registered births from 2005 to 2008 were linked to infant deaths to calculate infant death rate, rank of mortality for all life births and all life births with >24 weeks gestational age. The authors observed a sixfold difference in the 90th centile range in the very preterm live-born rate, which remained high even after adjustment for socioeconomic factors, ethnicity and maternal age. There was a large variation in the percentage of births with a gestational age <24 weeks recorded as born alive by the centres; the median was 52.6% with the 10th and 90th centiles ranging from 26% to 80%. Despite representing only approximately 1% of births, 19.5% of all infant deaths were related to births <24 weeks gestational age, and the 10th and 90th centiles were 6.7% and 31.9%. Excluding births <24 weeks gestational age led to significant changes in the infant mortality ranking of the PCTs. Comparing the performance of all PCTs, those institutions that registered less than half of preterm births <24 weeks gestation as live-born were ‘downgraded’ by a median of 12 ranks, whereas those PTCs that registered at least half of preterm births <24 weeks of gestation as live-born were ‘upgraded’ by a median of four ranks. The authors conclude that outcome data need to be corrected for the variable rate of preterm births <24 weeks gestation when comparing infant mortality between institutions in a benchmarking process, especially if the analysed data are being used for monitoring healthcare outcomes and to target interventions. Registration differences have a great impact on comparisons within a country, as well as between countries. It is very important to distinguish real variation (ie, true differences in quality of care) from artefacts, which obviously requires standardised inclusion criteria and complete registration of the subjects of interest.
A comprehensive systematic review on survival rates in extremely low birth weight infants suggests that failure to define specific denominators may result in substantial bias.5 The mean survival rate of extremely low birth weight infants in the studies included in this review was 45.0% using a denominator of all births, increasing to 60.7% using live births and to 71.6% using admissions to the neonatal intensive care unit only.5 Data from the Mosaic project suggest that including obstetrical care may be crucial and following the fetus alive at a certain stage of pregnancy has been suggested as baseline for benchmarking.6 These studies suggest combining perinatal and neonatal registries and defining the fetus alive in the uterus at 22 weeks gestational age as the standard inclusion criteria for a registry to be used for benchmarking. The pregnant mother with threatened preterm delivery and her family need to know data on the chances of the fetus to be born alive, to survive if born alive and to survive without substantial adverse outcome.
Based on the study results from Smith et al4 and the considerations mentioned above the following conclusions may be drawn:
The rate and the outcome of preterm infants born at the threshold of viability (ie, <24 weeks of gestational age) have a large influence on the outcome of VLBWI in individual centres. Comparing VLBWI outcome between institutions without correction for this variability is therefore particularly prone to bias.
Before using a dataset for benchmarking, it should be assured that the database is complete and includes all life-born infants. Using NICU admission data only is prone to bias. It seems crucial to include the obstetrical care and to define the fetus alive in the uterus at 22 weeks gestational age as the standard inclusion criterion for a registry used for benchmarking.
Using any database with a variable registration rate for infants born at the threshold of viability (for any reason) for benchmarking, excluding infants <24 weeks gestational age might be a second best choice to minimise the problem of variability in the obstetrical and neonatal approach and the variable registration rate of life-born infants at the threshold of viability. However, given this approach, one has to keep in mind that it will result in the exclusion of a large number of deaths from analysis. In the study of Smith et al,4 up to a third of infant deaths would have been excluded in some areas using this approach.
Any publication reporting survival of premature infants should clearly define the denominator used to define the population of interest.
Competing interests None.
Provenance and peer review Commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.