This validated predictive model uses seven variables to discriminate between infants at increased or decreased risk of survival after periviable birth. All proposed variables can be known prior to delivery, and thus this risk calculator can be utilized by obstetricians and neonatologists in counseling women with threatening periviable birth. This model of seven variables achieved an AUC of 0.800 for the outcome of infant survival. The model demonstrated excellent calibration with the greatest deviation being 3.8% in the highest risk category (90–100% probability of survival). This model was created with the intent of being as inclusive as possible to the general U.S. population and providing contemporary survival outcomes on periviable infants. This model achieves comparable predictive accuracy (AUC 0.800) to the NICHD (AUC 0.751) periviable birth calculator.
As mentioned, the most commonly used tool for prediction of periviable birth is currently provided by the NICHD, which accounts for five variables when predicting the likelihood of infant survival in the setting of periviable birth. Interestingly, among the various permutations attempted in developing our predictive model, the best model included seven variables plus resuscitation, of which five comprise the NICHD survival calculator. Further confirmation of the validity is revealed by the close correlation of our predictive model with the NICHD calculator in the setting of vertex neonates delivered vaginally (Fig. 2).
When the NICHD calculator is used, in addition to the predicted survival, a range of hospital outcomes is given, reflective of the wide possibility of outcomes and various resuscitation practice patterns at the hospitals included in the NICHD studies on periviable birth. Interestingly, when comparing our calculator to the NICHD calculator, if the values entered for the 5 variables included in the NICHD calculator are equivalent in our calculator, regardless of the values entered into the final 3 (of 8) variables entered into our calculator, the predicted survival from our calculator always falls within the range of hospital outcomes seen with the NICHD calculator (Fig. 2). Furthermore, a difference between our calculator and that of the NICHD’s is the weight given to the birth hospital. Our calculator performs with a higher AUC without the same weight given to the birth hospital which is beneficial as the user cannot select the birth hospital when using the NICHD periviable birth calculator. Instead, our calculator gives weight to factors such as mode of delivery and presentation at the time of delivery. It is possible that by including these additional variables, we may be inadvertently shedding light on the underlying reason for such wide discrepancies in hospital outcomes.
Due to the added variables, our calculator may be able to provide the clinician and patient with a more accurate estimation of the rate of survival given the expansion of factors being considered with our model. Specifically, it may help guide patients regarding mode of delivery in the setting of non-vertex presenting neonates. There is mounting evidence that vaginal delivery of a non-vertex presenting fetuses in the periviable period is associated with worse neonatal outcomes, even in the most aggressively resuscitated groups4,6,7. However, it is also well established that a cesarean delivery in the periviable period carries significant maternal risks, including but not limited to: hemorrhage, intensive care unit admission and even death; and can carry significant implications on future pregnancies17,30. Our calculator may help guide clinicians to provide estimates for counseling patients on the risks and benefits of potentially performing a cesarean delivery in the periviable period with a non-vertex presenting fetus.
It is important to remember that while our survival calculator predicts infant survival at one year of life with good accuracy, our calculator does not currently have the ability to predict long term morbidity or neurologic outcomes. There is a difference between survival and intact survival, and this is an important distinction that our calculator cannot determine. Further research is needed into the development of more accurate models of long term neurological outcomes to continue to best counsel patients in the periviable period especially regarding mode of delivery. Furthermore, as technology advances, resuscitation efforts improve, and this model, as well as future models, will require further scrutiny in upcoming years to assure contemporary accuracy.
Limitations to our study are notable in that the data were collected retrospectively and therefore no causal relationship can be determined from the predictive factors on the observed outcomes. We also excluded infants with birthweights exceeding the 95th % or less than the 5th percentile for gestational age to limit the likelihood of gestational age misclassification and thus this calculator is not validated for infants with birthweights at the extremes of gestational age. This study is representative of all livebirths in the United States between 2009–2013 and therefore contains marked heterogeneity in practice patterns observed throughout the United States. While the model is generalizable to a broad population of periviable infants, there are other factors that may influence infant survival such as hospital of birth, hospital transfer before or after delivery, and provider intent for aggressive resuscitation. Furthermore, this calculator does not (nor does any available calculator) have the ability to factor in the increased rate of survival for infants born at hospitals with very low birthweight neonatal units. Additionally, this calculator does not include information on infant morbidities such as neurodevelopmental impairment which affects the majority of infants delivering at these gestational ages. From the publicly-available data source used, we are unable to ascertain the intent and actual use of palliative vs aggressive interventions. This study is also prone to limitations of vital statistics data, which include likely underreporting of maternal co-morbidities.31
Despite these limitations, this study is generalizable to the U.S. population and provides a contemporary model for predicting infant survival in the periviable period. In this study, over 57,000 infants were included. This tool has significant potential to provide quantitative, evidence-based counseling tool to obstetric and neonatal providers to best accurately counsel patients faced with an impending periviable birth.
This calculator may be used to predict infant survival utilizing the variables of fetal presentation and mode of delivery. All variables can be assessed prenatally, and those can be used to counsel a patient faced with the possibility of a periviable birth. It is freely available online at https://ob.tools/calc/elbw [note: Please access the calculator at https://stage.ob.tools/calc/elbw with username: [email protected] and password personalizeCare (case sensitive). This calculator is currently password protected until after peer review. Upon acceptance, we will publish a final, freely and publicly available version on the primary website] and can provide an individual risk assessment for health care providers counseling at risk women.