added 4 research items
Objectives: The purpose of this study was to compare short-term outcomes in children born between 24 and 34 weeks’ gestation, according to observed antenatal corticosteroids (ACS)-to-birth intervals. Research question: ‘Is there a difference in short-term outcomes between observed ACS-to-birth intervals across a range of gestational ages at birth?’ Methods: Cohort study assessing differences in incidence of short-term neonatal outcomes according to the observed interval between the last administration of ACS and birth. Linear, non-weighted GEE models with an independence working correlation structure were fitted to infant level data providing valid point estimates for either incidence or rate differences (binary outcomes) or average differences (continuous outcomes). Results: Of 886 children, 35.9% were born within 2 days after the last administration of ACS, 32.2% within 2 to 7 days, 14.1% within 8 to 14 days, and 17.8% more than 14 days after. Across gestational ages at birth, there were no differences in birth weight between children born at an ACS-to-birth interval of 7 days or less compared to more than 7 days, nor were there differences in respiratory outcomes, cerebral outcomes, or composite outcome. Conclusion: Drawing conclusions on the importance of the ACS-to-birth interval is difficult due to the post-hoc nature of the variable. In the absence of tools to better estimate if and when PTB will occur, it might not have any value in daily practice, regardless of whether there is an optimal ACS-to-birth interval or not.
This paper contributes to the pursuit of leveraging unstructured medical notes to structured clinical decision making. In particular, we present a pipeline for clinical information extraction from medical notes related to preterm birth, and discuss the main challenges as well as its potential for clinical practice. A large collection of medical notes, created by staff during hospitalizations of patients who were at risk of delivering preterm, was gathered and analyzed. Based on an annotated collection of notes, we trained and evaluated information extraction components to discover clinical entities such as symptoms, events, anatomical sites and procedures, as well as attributes linked to these clinical entities. In a retrospective study, we show that these are highly informative for clinical decision support models that are trained to predict whether delivery is likely to occur within specific time windows, in combination with structured information from electronic health records.
PurposePreterm birth (PTB) can be categorised according to aetiology into: spontaneous preterm labour (SPL), preterm prelabour rupture of membranes (PPROM), and iatrogenic (iatro) PTB. Outcomes could differ between these groups, which could be of interest in counselling. We aimed to explore differences between aetiologic groups of PTB in maternal demographics, obstetrical characteristics and management, and neonatal outcomes.Methods This is a cohort study (2012–2018) in Ghent University Hospital, Belgium, of deliveries from 24 + 0 to 33 + 6 weeks. We compared perinatal demographics, management, and outcomes between the aetiologic types of PTB. Point and interval estimates for differences between aetiologic types were estimated using a Generalised Estimating Equations approach to handle clustering due to multiple gestations.Results813 mothers and 987 neonates were included. Prevalences of different aetiologic types of PTB were similar. Maternal BMI was higher in the iatrogenic group (iatro-SPL: + 1.92 kg/m2, 95% CI 1.02, 2.83; iatro-PPROM: + 2.06 kg/m2, 95% CI 1.15, 2.96). There was an inversed sex ratio (0.82, 95% CI 0.65, 1.03), more growth restriction (iatro-SPL: + 22.60%, 95% CI 17.08, 28.13; iatro-PPROM: + 24.64%, 95% CI 19.44, 29.83), and a higher caesarean section rate in the iatrogenic group (iatro-SPL: + 57.23%, 95% CI 50.32, 64.13, iatro-PPROM: + 56.79%, 95% CI 50.20, 63.38) and more patients received at least one complete course of antenatal corticosteroids (iatro-SPL: + 17.60%, 95% CI 10.60, 24.60, iatro-PPROM: + 10.73%, 95% CI 4.52, 16.94). In all types of PTB, adverse neonatal outcomes had a low prevalence, except for respiratory distress syndrome. A composite of adverse neonatal outcome was more prevalent in the SPL- compared to the PPROM group, and there was less intraventricular haemorrhage in the iatrogenic group.Conclusion Additional to gestational age at birth, the aetiology of PTB is associated with neonatal outcome. More data are needed to enable individualised management and counselling in case of threatened PTB.Trial registration numberNCT03405116.
Information extracted from electrohysterography recordings could potentially prove to be an interesting additional source of information to estimate the risk on preterm birth. Recently, a large number of studies have reported near-perfect results to distinguish between recordings of patients that will deliver term or preterm using a public resource, called the Term/Preterm Electrohysterogram database. However, we argue that these results are overly optimistic due to a methodological flaw being made. In this work, we focus on one specific type of methodological flaw: applying over-sampling before partitioning the data into mutually exclusive training and testing sets. We show how this causes the results to be biased using two artificial datasets and reproduce results of studies in which this flaw was identified. Moreover, we evaluate the actual impact of over-sampling on predictive performance, when applied prior to data partitioning, using the same methodologies of related studies, to provide a realistic view of these methodologies' generalization capabilities. We make our research reproducible by providing all the code under an open license.
Preterm birth is the leading cause of death among young children and has a large prevalence globally. Machine learning models, based on features extracted from clinical sources such as electronic patient files, yield promising results. In this study, we review similar studies that constructed predictive models based on a publicly available dataset, called the Term-Preterm EHG Database (TPEHGDB), which contains electrohysterogram signals on top of clinical data. These studies often report near-perfect prediction results, by applying over-sampling as a means of data augmentation. We reconstruct these results to show that they can only be achieved when data augmentation is applied on the entire dataset prior to partitioning into training and testing set. This results in (i) samples that are highly correlated to data points from the test set are introduced and added to the training set, and (ii) artificial samples that are highly correlated to points from the training set being added to the test set. Many previously reported results therefore carry little meaning in terms of the actual effectiveness of the model in making predictions on unseen data in a real-world setting. After focusing on the danger of applying over-sampling strategies before data partitioning, we present a realistic baseline for the TPEHGDB dataset and show how the predictive performance and clinical use can be improved by incorporating features from electrohysterogram sensors and by applying over-sampling on the training set.
Preterm birth is the leading cause of death among children under five years old. The pathophysiology and etiology of preterm labor are not yet fully understood. This causes a large number of unnecessary hospitalizations due to high–sensitivity clinical policies, which has a significant psychological and economic impact. In this study, we present a predictive model, based on a new dataset containing information of 1,243 admissions, that predicts whether a patient will give birth within a given time after admission. Such a model could provide support in the clinical decision-making process. Predictions for birth within 48 h or 7 days after admission yield an Area Under the Curve of the Receiver Operating Characteristic (AUC) of 0.72 for both tasks. Furthermore, we show that by incorporating predictions made by experts at admission, which introduces a potential bias, the prediction effectiveness increases to an AUC score of 0.83 and 0.81 for these respective tasks.