Monthly trends, determinants, and forecasting of perinatal mortality in Ghana: a comparison of ARIMA, BPNN, DLNN, and GRNN models

Predicting Neonatal Mortality Using Demographic and Health Survey Data: Comparing Logistic Regression and Machine Learning under Severe Class Imbalance

Neonatal mortality continues to be a significant public health problem, especially in low and middle-income countries, where improvements in neonatal mortality have been slower than progress in overall child survival. Logistic regression has been applied in the past to determine the factors associated with neonatal mortality from the Demographic and Health Survey (DHS) data, but machine learning techniques have been gaining popularity as alternative methods for predictive modelling. Evidence comparing these approaches based on nationally representative household survey data, however, is limited, especially when considering the case of severe class imbalance. This study evaluated alternative strategies to address class imbalance and examined the effect of outcome-endogenous predictors to compare the predictive performance of logistic regression, random forest and gradient boosting classifiers with respect to the prediction of neonatal mortality. Secondary analysis was performed with data from 18,978 live births with 422 neonatal deaths (2.22%) and 27 predictors related to the mother, household, and child. The data were split with stratified sampling into training set (75%) and testing set (25%). Three strategies of handling imbalanced data were used in this study: baseline (no adjustment), class weighting, and Synthetic Minority Oversampling Technique (SMOTE). The area under the receiver operating characteristic curve (ROC-AUC), the area under the precision-recall curve (PR-AUC), the sensitivity, the specificity, the precision, the F1 score, the calibration curve and the permutation feature importance were used to assess model performance. Class-weighted gradient boosting had the best overall predictive performance (ROC-AUC = 0.921, PR-AUC = 0.311, sensitivity = 83.0%) and logistic regression had similar discrimination ability but was easier to interpret (ROCAUC = 0.920). Explicitly dealing with class imbalance significantly outperformed unadjusted models for sensitivity across all of the algorithms. The number of living children under five years, number of births in the last five years and gestation length were the most important variables identified by the feature importance analysis. When the mechanically outcome-related predictor, "number of living children under 5 years," was removed, the ROC-AUC dropped from 0.921 to 0.700, suggesting that a significant amount of the apparent predictive performance was due to information leakage and not to the predictor itself. These results show that the solution for class imbalance problem is more important than using different classification methods, and emphasize the importance of selecting predictors for temporal and definitional relationship with the outcome. The study gives methodological directions for creating accurate predictions to neonatal mortality and other rare health phenomena based on survey data nationally representative.

Predicting the neonatal, infant, and under-five mortality outcomes in the Philippines using Count TS GLM and GLARMA

Mortality statistics are often considered indicators of overall population health, highlighting the need for accurate forecasting to support public health planning and policy development. While various statistical models have been applied to mortality forecasting globally, research on the use of count time series modeling for mortality trends in the Philippines remains limited. Addressing this gap, this study provides the first systematic comparison of Count Time Series Generalized Linear Models (TS GLM) and Generalized Linear Autoregressive Moving Average (GLARMA) models for forecasting Philippine neonatal, infant, and under-five mortality. The predictive performance of these models was evaluated for the period 2023–2027. Prior to model implementation, stationarity was assessed using the KPSS test, after which predictive models were developed and compared, with the best-performing specifications selected based on the Akaike Information Criterion (AIC) and Mean Absolute Error (MAE). Results show that TS GLM consistently outperforms GLARMA in predictive accuracy, with the negative binomial distribution yielding the best results. The optimal TS GLM models for neonatal (1,2), infant (1,2), and under-five mortality (1,5) surpass their GLARMA counterparts. Projected trends indicate rising absolute numbers of deaths across all subgroups. These findings provide valuable guidance for government agencies, health institutions, and insurance providers in monitoring child mortality, anticipating healthcare demand, improving actuarial projections, and informing targeted strategies, while also serving as a reference for future research on mortality forecasting using count time-series models.

Why has Nigerias neonatal mortality decline stalled? An ecological analysis of public health financing and macroeconomic instability, 1990-2024

Neonatal admissions and mortality: a statistical adjustment approach using birthweight-specific curves to address bias due to underreporting in 65 hospitals in four African countries

Underreporting of neonatal admissions and deaths – particularly among < 1000 g newborns – compromises the plausibility of in-patient neonatal mortality rate (iNMR) estimates. This limits progress tracking, since 2.3 million neonatal deaths occur annually, and 65 countries are projected to miss the Sustainable Development Goal (SDG) 3.2 target of fewer than 12 deaths per 1000 live births by 2030. Reliable iNMR estimates are essential for national and hospital-level monitoring. This study assessed birthweight-specific adjustment curves developed using simulated data, and extended the approach to routine data from 172,809 neonatal admissions across four African countries in the NEST360 network. Simulated data were generated using birthweight intervals of 500 g from 500 to 5000 g, with 1000 values per category representing 1000 hospitals. Admissions were drawn from a negative binomial distribution and mortality rates from a beta distribution based on published birthweight-specific iNMRs. Adjustment functions were developed using data from 500 hospitals without underreporting and tested on 500 with induced underreporting. This approach was then extended to real data from 65 NEST360 partnering hospitals. The adjustment functions were derived from 26 hospitals with plausible reporting (iNMR ≥ 700 per 1000 for neonates < 1000 g). In simulated data, adjusted iNMRs closely matched original values, with median differences ranging from − 4 to 9 deaths per 1000 across birthweight categories. In NEST360 dataset, unadjusted iNMRs were 121 (range 13–272) in Nigeria, 128 (60–214) in Tanzania, 132 (42–234) in Malawi, and 135 (60–250) in Kenya per 1000 admissions. After adjustment, estimates increased to 162 (127–333), 183 (128–233), 174 (120–268), and 172 (123–261), respectively. Variability across hospitals decreased, with standard deviation reducing from 13 to 5 deaths per 1000. Birthweight specific adjustment functions provide a practical and scalable approach to improving the plausibility of iNMR estimates where underreporting is common. These methods can be integrated into routine systems (such as Electronic Medical Record Systems and DHIS2) where key data are available, with modest computational demands. Their use can strengthen mortality surveillance and support more reliable policy and planning, while contributing to progress toward SDG 3.2, contingent on continued improvements in data quality and system capacity.

Impact of maternal and neonatal factors on mortality in the first six weeks of life: A comprehensive cohort study in Kinshasa, DR Congo

Results underscore the crucial importance of high-quality obstetric and neonatal care, as well as antenatal monitoring, for improving newborn survival and call for strengthened antenatal care, improved screening and management of at-risk newborns in healthcare facilities, and the integration of these indicators into strategies for reducing neonatal mortality in resource-limited settings.

Incidence of Neonatal Meningitis and the Premature Mortality in Eastern Sub-saharan Africa, From 1990 to 2023: A Global Burden of Diseases Study

Objective Quantify neonatal meningitis incidence and years of life lost (YLLs) in Eastern Sub-Saharan Africa (ESSA) from 1990 to 2023. Methods Secondary analysis of GBD data analyzed temporal trends by location, year and sex. Bayesian meta-regression estimated incidence, cause of death ensemble modeling calculated YLLs, and spatiotemporal Gaussian process regression smoothed trends. Estimates include 95% uncertainty intervals (UIs). Results In 2023, ESSA’s neonatal meningitis incidence rate was 2,542 per 100,000 (95% UI: 2,043–3,065), heavily concentrated in early neonates (5,415) versus late neonates (1,574). The regional YLL rate was 24,855 per 100,000, driven by early neonates (66,882). Somalia and South Sudan exceeded regional averages. Globally, YLL rates declined from 1990 to 2023 (Annual Rate of Change: -0.70%). Conclusion ESSA faces a severe neonatal meningitis burden skewed toward early neonates. Urgent strategies must scale up maternal screening, intrapartum antibiotics, and rapid diagnostics for fast-tracked treatment.