Predictors of neonatal mortality: development and validation of prognostic models using prospective data from rural Bangladesh

Bann C, Bozdar MH, Raza J, Dhaded SM, Goudar SS, Gowdar G, Yasmin H, McClure EM, Goldenberg RL. Predictors of Preterm Neonatal Mortality in India and Pakistan: A Secondary Analysis of Data from PURPOSe Study

Objective. To create a prediction model for preterm neonatal mortality. Methods. A secondary analysis was conducted using data from a prospective cohort study, the Project to Understand and Research Preterm Pregnancy Outcome South Asia. The Cox proportional hazard model was used and adjusted hazard ratios (AHR) with 95% confidence intervals (95% CI) were reported. Results. Overall, 3446 preterm neonates were included. The mean age of preterm neonates was 0.65 (1.25) hours and 52% were female. The preterm neonatal mortality rate was 23.3%. The maternal factors predicting preterm neonatal death was any antepartum hemorrhage, AHR 1.99 (1.60-2.47), while neonatal predictors were preterm who received positive pressure ventilation AHR 1.30 (1.08-1.57), temperature <35.5°C AHR 1.18 (1.00-1.39), and congenital malformations AHR 3.31 (2.64-4.16). Conclusion. This study identified key maternal and neonatal predictors of preterm neonatal mortality, emphasizing the need for targeted interventions and collaborative public health efforts to address disparities and regional variations.

Development and Validation of a Clinical Prognostic Risk Score to Predict Early Neonatal Mortality, Ethiopia: A Receiver Operating Characteristic Curve Analysis

Background

Early neonatal death is the death of a live-born baby within the first seven days of life, which is 73% of all postnatal deaths in the globe. This study aimed to develop and validate a prognostic clinical risk tool for the prediction of early neonatal death.

Methods

A prospective follow-up study was conducted among 393 neonates at Debre Tabor Referral hospital, Northwest Ethiopia. Multivariable logistic regression model was employed to identify potential prognostic determinants for early neonatal mortality. Area under receiver operating characteristics curve (AUROC) was used to check the model discrimination probability using ‘pROC’ R-package. Model calibration plot was checked using ‘givitiR’ R-package. Finally, a risk score prediction tool was developed for ease of applicability. Decision curve analysis was done for cost-benefit analysis and to check the clinical impact of the model.

Results

Overall, 15.27% (95% CI: 12.03–19.18) of neonates had the event of death during the follow-up period. Maternal undernutrition, antenatal follow-up less than four times, birth asphyxia, low birth weight, and not exclusive breastfeeding were the prognostic predictors of early neonatal mortality. The AUROC for the reduced model was 88.7% (95% CI: 83.8–93.6%), which had good discriminative probability. The AUROC of the simplified risk score algorithm was 87.8% (95% CI, 82.7–92.9%). The sensitivity and specificity of the risk score tool was 70% and 89%, respectively. The true prediction accuracy of the risk score tool to predict early neonatal mortality was 86%, and the false prediction probability was 13%.

Conclusion

We developed an early neonatal death prediction tool using easily available maternal and neonatal characteristics for resource-limited settings. This risk prediction using risk score is an easily applicable tool to identify neonates at a higher risk of having early neonatal mortality. This risk score tool would offer an opportunity to reduce early neonatal mortality, thus improving the overall early neonatal death in a resource-limited setting.

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

OBJECTIVE

We aimed to determine the key predictors of perinatal deaths using machine learning models compared with the logistic regression model.

DESIGN

A 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.

SETTING

The 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.

PARTICIPANTS

Singleton deliveries (n=42 319) with complete records from 2000 to 2015.

PRIMARY OUTCOME MEASURES

Perinatal death (composite of stillbirths and early neonatal deaths). These outcomes were only captured before mothers were discharged from the hospital.

RESULTS

The 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.

CONCLUSIONS

In 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.