Prediction of perinatal death using machine learning models: a birth registry-based cohort study in northern Tanzania

Abstract

ObjectiveWe aimed to determine the key predictors of perinatal deaths using machine learning models compared with the logistic regression model.DesignA secondary data analysis using the Kilimanjaro Christian Medical Centre (KCMC) Medical Birth Registry cohort from 2000 to 2015. We assessed the discriminative ability of models using the area under the receiver operating characteristics curve (AUC) and the net benefit using decision curve analysis.SettingThe KCMC is a zonal referral hospital located in Moshi Municipality, Kilimanjaro region, Northern Tanzania. The Medical Birth Registry is within the hospital grounds at the Reproductive and Child Health Centre.ParticipantsSingleton deliveries (n=42 319) with complete records from 2000 to 2015.Primary outcome measuresPerinatal death (composite of stillbirths and early neonatal deaths). These outcomes were only captured before mothers were discharged from the hospital.ResultsThe proportion of perinatal deaths was 3.7%. There were no statistically significant differences in the predictive performance of four machine learning models except for bagging, which had a significantly lower performance (AUC 0.76, 95% CI 0.74 to 0.79, p=0.006) compared with the logistic regression model (AUC 0.78, 95% CI 0.76 to 0.81). However, in the decision curve analysis, the machine learning models had a higher net benefit (ie, the correct classification of perinatal deaths considering a trade-off between false-negatives and false-positives)—over the logistic regression model across a range of threshold probability values.ConclusionsIn this cohort, there was no significant difference in the prediction of perinatal deaths between machine learning and logistic regression models, except for bagging. The machine learning models had a higher net benefit, as its predictive ability of perinatal death was considerably superior over the logistic regression model. The machine learning models, as demonstrated by our study, can be used to improve the prediction of perinatal deaths and triage for women at risk.

Prediction of neonatal deaths in NICUs: development and validation of machine learning models

“…Ravelli et al [ 43 ] developed the antenatal prediction of neonatal mortality in very premature infants on 13 variables. Mboya et al [ 5 ] considered 32 predictive variables for perinatal death prediction.…”

“…Predictions show that between 2019 and 2030, approximately 52 million children under the age of 5 will die, approximately half of whom will be neonate [ 2 ]. However, the under-5 mortality rate has declined around the world, but the neonatal mortality rate is still an alarming issue [ 5 ].…”

“…According to some studies, the use of machine learning methods in predicting the neonatal mortality was promising. For example, Mboya et al [ 5 ] showed that the predictive ability of perinatal death in machine learning algorithms was considerably superior over the logistic regression method. However, despite there are many studies in this field, most of them have been done on specific groups of neonates, such as premature or Very Low Birth Weight (VLBW) neonates or in general settings rather than specifically in the NICUs.…”

“…The results showed that there was no significant difference in perinatal death prediction between machine learning and regression models, except for bagging method. In addition, the machine learning algorithms had a superior net benefit and its predictive ability was greatly higher than regression model [ 5 ]. In a 2019 study, researchers in Bangladesh developed and evaluated regression models to predict the risk of neonatal death based on known characteristics in the beginning of pregnancy, beginning of delivery and five minutes after delivery.…”

“…Unlike the previous studies, which were performed on premature or VLBW neonates [ 3 , 13 ], or in general settings other than NICUs [ 5 , 6 , 12 ], the present study considers all neonates without birth weight limitation in NICU settings. Furthermore, the majority of previous models were not prospectively applied and evaluated [ 5 , 6 , 13 ]; however, in the current study, we prospectively evaluated our models in a NICU to better examine the performance of the models from a clinical perspective.…”