The RSS Emerging Applications and Medical Sections held a joint afternoon of talks at Errol Street on the subject of: 'Analysing observational data on early life survival' held on 7 December 2016, at Errol Street.
These talks included discussion of challenges around analysing observational data on early life survival (including capturing information about terminations, still births, and infant deaths).
The afternoon kicked-off with a talk from Dr Brad Manktelow from the University of Leicester, on his work with MBRRACE-UK (Mothers and Babies: Reducing Risk through Audits and Confidential Enquiries in the UK). MBRRACE-UK is a national surveillance project covering all stillbirths, neonatal deaths and fetal loses from 2013 onwards, with over 650 reporters of these data (eg trusts and health boards, local authorities, neonatal networks). Terminations are not included. In the yearly national report mortality rates are stablilised (ie shrunken to reduce influence of small organisations) and adjusted (for maternal factors). Brad discussed how these rates are more robust measures of the underlying mortality rate than crude rates, and that trends over time in these stabilised adjusted rates may be studied in future.
Next, Ania Zylberstejn from University College London’s Institute of Child Health gave us an overview on her PhD work, which aimed to determine the factors driving the great difference in mortality rates, to inform mortality prevention policy. Ania used routinely recorded national hospital and mortality data for this work. She discussed issues with making sure that the data were robust and comparable: for example, socio-economic status was recorded differently, and hospital/mortality data for England (Hospital Episode Statistics) were only considered comparable to the gold standard from 2003 onwards. She also described how information was very incomplete for birthweight, gestational and maternal age, and she overcame this problem through linkage of babies’ records to those of their mothers. Ania found that the difference in survival between England and Sweden was greatly attenuated after adjustment for birthweight and gestational age. She concluded that preventative policies should focus on the health of women and girls before and during pregnancy.
Dr Ulla Sovio, from Obstetrics & Gynaecology at the University of Cambridge, presented the methods and some early results from the Pregnancy Outcome Prediction (POP) study. The POP study is an example of a ‘deep’ data study in which data from routine hospital, maternity systems and ultrasound scans have been collected together with biosamples including DNA, placenta and maternal blood from around 4500 pregnant women attending hospital in Cambridge. The aim of the study is to identify markers during pregnancy that could be used to predict and screen for adverse pregnancy outcomes, such as stillbirth and fetal growth restriction. Ulla talked about the challenges presented by the development of prediction models for this study (some of which are shared for other prediction models, and some unique), including choice of birthweight and fetal weight reference centiles, missing data on key outcome measures such as fetal head circumference, identifying external validation cohorts and the large number of potential predictors.
Finally, Professor Joan Morris from Queen Mary University of London gave an overview of her work in infants with one of two congenital chromosomal abnormalities: Patau syndrome and Edwards syndrome. These rare syndromes are associated with high rates of miscarriage and low survival rates among those born alive. Of key importance for parents is to know the fetal loss rate (that is what proportion of pregnancies with these syndromes end in a miscarriage or stillbirth) and the probability and average time of survival, in babies who are born with these syndromes. An important question is whether prenatal diagnosis improves survival for children with Edwards or Patau syndrome.
In order to address this important question, robust and reliable data on time of diagnosis, birth outcome (eg live, still or termination) and postnatal mortality is required. Joan presented the challenges in using data from the National Downs Syndrome Cytogenetic Register (NDSCR) to answer questions about outcomes following prenatal vs postnatal diagnosis. These included misclassification of live births (as a fetal loss, stillbirth or termination), and underrecording of anomalies for children who were diagnosed postnatally. She concluded that there is insufficient evidence from the NDSCR to say that outcomes are different for children picked up pre- or postnatally, and provided a note of caution about overinterpreting results on early life survival from data sources unless you are very sure about recording of key variables such as pregnancy outcome.
The afternoon was full of fascinating results and stimulating discussion. Thanks to everyone who came along!