Driving Factors of Preterm Birth Risk in Adolescents

Incidence and Risk Factors for Low Birthweight and Preterm Birth in Post-Conflict Northern Uganda: A Community-Based Cohort Study

BACKGROUND

Annually, an estimated 20 million (13%) low-birthweight (LBW) and 15 million (11.1%) preterm infants are born worldwide. A paucity of data and reliance on hospital-based studies from low-income countries make it difficult to quantify the true burden of LBW and PB, the leading cause of neonatal and under-five mortality. We aimed to determine the incidence and risk factors for LBW and preterm birth in Lira district of Northern Uganda.

METHODS

This was a community-based cohort study, nested within a cluster-randomized trial, designed to study the effect of a combined intervention on facility-based births. In total, 1877 pregnant women were recruited into the trial and followed from 28 weeks of gestation until birth. Infants of 1556 of these women had their birthweight recorded and 1279 infants were assessed for preterm birth using a maturity rating, the New Ballard Scoring system. Low birthweight was defined as birthweight <2.5kg and preterm birth was defined as birth before 37 completed weeks of gestation. The risk factors for low birthweight and preterm birth were analysed using a multivariable generalized estimation equation for the Poisson family.

RESULTS

The incidence of LBW was 121/1556 or 7.3% (95% Confidence interval (CI): 5.4-9.6%). The incidence of preterm births was 53/1279 or 5.0% (95% CI: 3.2-7.7%). Risk factors for LBW were maternal age ≥35 years (adjusted Risk Ratio or aRR: 1.9, 95% CI: 1.1-3.4), history of a small newborn (aRR: 2.1, 95% CI: 1.2-3.7), and maternal malaria in pregnancy (aRR: 1.7, 95% CI: 1.01-2.9). Intermittent preventive treatment (IPT) for malaria, on the other hand, was associated with a reduced risk of LBW (aRR: 0.6, 95% CI: 0.4-0.8). Risk factors for preterm birth were maternal HIV infection (aRR: 2.8, 95% CI: 1.1-7.3), while maternal education for ≥7 years was associated with a reduced risk of preterm birth (aRR: 0.2, 95% CI: 0.1-0.98) in post-conflict northern Uganda.

CONCLUSIONS

About 7.3% LBW and 5.0% PB infants were born in the community of post-conflict northern Uganda. Maternal malaria in pregnancy, history of small newborn and age ≥35 years increased the likelihood of LBW while IPT reduced it. Maternal HIV infection was associated with an increased risk of PB compared to HIV negative status. Maternal formal education of ≥7 years was associated with a reduced risk of PB compared to those with 0-6 years. Interventions to prevent LBW and PBs should include girl child education, and promote antenatal screening, prevention and treatment of malaria and HIV infections.

A Machine Learning-Based Prediction Model for Preterm Birth in Rural India

Preterm birth (PTB) in a pregnant woman is the most serious issue in the field of Gynaecology and Obstetrics, especially in rural India. In recent years, various clinical prediction models for PTB have been developed to improve the accuracy of learning models. However, to the best of the authors' knowledge, most of them suffer from selecting the most accurate features from the medical dataset in linear time. The present paper attempts to design a machine learning model named as risk prediction conceptual model (RPCM) for the prediction of PTB. In this paper, a feature selection approach is proposed based on the notion of entropy. The novel approach is used to find the best maternal features (responsible for PTB) from the obstetrical dataset and aims to predict the classifier's accuracy at the highest level. The paper first deals with the review of PTB cases (which is neglected in many developing countries including India). Next, we collect obstetrical data from the Community Health Centre of rural areas (Kamdara, Jharkhand). The suggested approach is then applied on collected data to identify the excellent maternal features (text-based symptoms) present in pregnant women in order to classify all birth cases into term birth and PTB. The machine learning part of the model is implemented using three different classifiers, namely, decision tree (DT), logistic regression (LR), and support vector machine (SVM) for PTB prediction. The performance of the classifiers is measured in terms of accuracy, specificity, and sensitivity. Finally, the SVM classifier generates an accuracy of 90.9%, which is higher than other learning classifiers used in this study.