Predicting high-risk preterm birth using artificial neural networks

Predicting preterm birth using explainable machine learning in a prospective cohort of nulliparous and multiparous pregnant women

Preterm birth (PTB) presents a complex challenge in pregnancy, often leading to significant perinatal and long-term morbidities. "While machine learning (ML) algorithms have shown promise in PTB prediction, the lack of interpretability in existing models hinders their clinical utility. This study aimed to predict PTB in a pregnant population using ML models, identify the key risk factors associated with PTB through the SHapley Additive exPlanations (SHAP) algorithm, and provide comprehensive explanations for these predictions to assist clinicians in providing appropriate care. This study analyzed a dataset of 3509 pregnant women in the United Arab Emirates and selected 35 risk factors associated with PTB based on the existing medical and artificial intelligence literature. Six ML algorithms were tested, wherein the XGBoost model exhibited the best performance, with an area under the operator receiving curves of 0.735 and 0.723 for parous and nulliparous women, respectively. The SHAP feature attribution framework was employed to identify the most significant risk factors linked to PTB. Additionally, individual patient analysis was performed using the SHAP and the local interpretable model-agnostic explanation algorithms (LIME). The overall incidence of PTB was 11.23% (11 and 12.1% in parous and nulliparous women, respectively). The main risk factors associated with PTB in parous women are previous PTB, previous cesarean section, preeclampsia during pregnancy, and maternal age. In nulliparous women, body mass index at delivery, maternal age, and the presence of amniotic infection were the most relevant risk factors. The trained ML prediction model developed in this study holds promise as a valuable screening tool for predicting PTB within this specific population. Furthermore, SHAP and LIME analyses can assist clinicians in understanding the individualized impact of each risk factor on their patients and provide appropriate care to reduce morbidity and mortality related to PTB.

Status of cognitive frailty in elderly patients with chronic kidney disease and construction of a risk prediction model: a cross-sectional study

OBJECTIVE

To investigate the risk factors of cognitive frailty in elderly patients with chronic kidney disease (CKD), and to establish an artificial neural network (ANN) model.

DESIGN

A cross-sectional design.

SETTING

Two tertiary hospitals in southern China.

PARTICIPANTS

425 elderly patients aged ≥60 years with CKD.

METHODS

Data were collected via questionnaire investigation, anthropometric measurements, laboratory tests and electronic medical records. The 425 samples were randomly divided into a training set, test set and validation set at a ratio of 5:3:2. Variables were screened by univariate and multivariate logistic regression analyses, then an ANN model was constructed. The accuracy, specificity, sensitivity, receiver operating characteristic (ROC) curve and area under the ROC curve (AUC) were used to evaluate the predictive power of the model.

RESULTS

Barthel Index (BI) score, albumin, education level, 15-item Geriatric Depression Scale score and Social Support Rating Scale score were the factors influencing the occurrence of cognitive frailty (p<0.05). Among them, BI score was the most important factor determining cognitive frailty, with an importance index of 0.30. The accuracy, specificity and sensitivity of the ANN model were 86.36%, 88.61% and 80.65%, respectively, and the AUC of the constructed ANN model was 0.913.

CONCLUSION

The ANN model constructed in this study has good predictive ability, and can provide a reference tool for clinical nursing staff in the early prediction of cognitive frailty in a high-risk population.

Developing and validating a risk prediction model for preterm birth at Felege Hiwot Comprehensive Specialized Hospital, North-West Ethiopia: a retrospective follow-up study

OBJECTIVE

To develop and validate a risk prediction model for the prediction of preterm birth using maternal characteristics.

DESIGN

This was a retrospective follow-up study. Data were coded and entered into EpiData, V.3.02, and were analysed using R statistical programming language V.4.0.4 for further processing and analysis. Bivariable logistic regression was used to identify the relationship between each predictor and preterm birth. Variables with p≤0.25 from the bivariable analysis were entered into a backward stepwise multivariable logistic regression model, and significant variables (p<0.05) were retained in the multivariable model. Model accuracy and goodness of fit were assessed by computing the area under the receiver operating characteristic curve (discrimination) and calibration plot (calibration), respectively.

SETTING AND PARTICIPANTS

This retrospective study was conducted among 1260 pregnant women who did prenatal care and finally delivered at Felege Hiwot Comprehensive Specialised Hospital, Bahir Dar city, north-west Ethiopia, from 30 January 2019 to 30 January 2021.

RESULTS

Residence, gravidity, haemoglobin <11 mg/dL, early rupture of membranes, antepartum haemorrhage and pregnancy-induced hypertension remained in the final multivariable prediction model. The area under the curve of the model was 0.816 (95% CI 0.779 to 0.856).

CONCLUSION

This study showed the possibility of predicting preterm birth using maternal characteristics during pregnancy. Thus, use of this model could help identify pregnant women at a higher risk of having a preterm birth to be linked to a centre.

Blood pressure trajectories during pregnancy and preterm delivery: A prospective cohort study in China

Women's blood pressure (BP) changes throughout pregnancy. The effect of BP trajectories on preterm delivery is not clear. The authors aim to evaluate the association between maternal BP trajectories during pregnancy and preterm delivery. The authors studied pregnant women included in the Born in Guangzhou Cohort Study in China between February 2012 and June 2016. Maternal BP was measured at antenatal visits between 13 and 40 gestational weeks, and gestational age of delivery data was collected. The authors used linear mixed models to capture the BP trajectories of women with term, and spontaneous and iatrogenic preterm delivery. BP trajectories of women with various gestational lengths (34, 35, 36, 37, 38, 39, 40 weeks) were compared. Of the 17 426 women included in the analysis, 618 (3.55%) had spontaneous preterm delivery; 158 (.91%) had iatrogenic preterm delivery; and 16 650 (95.55%) women delivered at term. The BP trajectories were all J‐shaped curves for different delivery types. Women with iatrogenic preterm delivery had the highest mean BP from 13 weeks till delivery, followed by those with spontaneous preterm delivery and term delivery (p < .001). Trajectory analysis stratified by maternal parity showed similar results for nulliparous and multiparous women. Excluding women with pre‐eclampsia and gestational hypertension (GH) significantly attenuated the aforementioned association. Also, women with shorter gestational length tend to have higher BP trajectories during pregnancy. In conclusion, Women with spontaneous preterm delivery have a higher BP from 13 weeks till delivery than women with term delivery, while women with iatrogenic preterm delivery have the highest BP.

Unsupervised Learning Applied to the Stratification of Preterm Birth Risk in Brazil with Socioeconomic Data

Preterm birth (PTB) is a phenomenon that brings risks and challenges for the survival of the newborn child. Despite many advances in research, not all the causes of PTB are already clear. It is understood that PTB risk is multi-factorial and can also be associated with socioeconomic factors. Thereby, this article seeks to use unsupervised learning techniques to stratify PTB risk in Brazil using only socioeconomic data. Through the use of datasets made publicly available by the Federal Government of Brazil, a new dataset was generated with municipality-level socioeconomic data and a PTB occurrence rate. This dataset was processed using various unsupervised learning techniques, such as k-means, principal component analysis (PCA), and density-based spatial clustering of applications with noise (DBSCAN). After validation, four clusters with high levels of PTB occurrence were discovered, as well as three with low levels. The clusters with high PTB were comprised mostly of municipalities with lower levels of education, worse quality of public services-such as basic sanitation and garbage collection-and a less white population. The regional distribution of the clusters was also observed, with clusters of high PTB located mostly in the North and Northeast regions of Brazil. The results indicate a positive influence of the quality of life and the offer of public services on the reduction in PTB risk.

Nomogram Incorporating Multimodal Transvaginal Ultrasound Assessment at 20 to 24 Weeks' Gestation for Predicting Spontaneous Preterm Delivery in Low-Risk Women

Background

The majority of women who experience spontaneous preterm delivery (SPTD) have low-risk, asymptomatic pregnancies with a cervical length (CL) ≥25mm and no clear risk factors. Despite the fact that cervical elastography is a potential tool for predicting SPTD, there is currently no feasible solution to make a reliable prediction for preventing SPTD.

Objective

The aim of this study was to construct a nomogram including multimodal transvaginal ultrasound parameters during the second trimester to predict SPTD in low-risk women.

Methods

This multi-center study enrolled 1260 women with singleton pregnancies between 20 and 24 weeks’ gestation. CL and cervical elastography data were obtained when they were undergoing the second-trimester anomaly scan. Univariate and multivariate Logistic regression were utilized to screen predictors independently related to SPTD from the maternal characteristics and multimodal ultrasound data. Then construct a nomogram to determine the likelihood of SPTD in pregnant women.

Results

A total of 66 pregnancies in the training cohort (7.8%, 66/842) and 37 pregnancies (8.9%, 37/418) in the validation cohort ended in SPTD. Age, uterine curettage, CL, and strain in the anterior lip of internal os were the independent predictors of SPTD (P < 0.001, < 0.001, = 0.007, and < 0.001, respectively). These predictors constituted a nomogram to predict the probability of SPTD for a pregnant woman in her second trimester. It showed good discrimination (C-index = 0.898 and 0.839), calibration (P = 0.258 and 0.115), and yielded net benefits both in the training and validation cohorts.

Conclusion

The nomogram including data of multimodal transvaginal ultrasound at 20 to 24 weeks’ gestation is expected to identify women with SPTD in the low-risk, asymptomatic population.

Prediction of preterm birth in nulliparous women using logistic regression and machine learning

OBJECTIVE

To predict preterm birth in nulliparous women using logistic regression and machine learning.

DESIGN

Population-based retrospective cohort.

PARTICIPANTS

Nulliparous women (N = 112,963) with a singleton gestation who gave birth between 20-42 weeks gestation in Ontario hospitals from April 1, 2012 to March 31, 2014.

METHODS

We used data during the first and second trimesters to build logistic regression and machine learning models in a "training" sample to predict overall and spontaneous preterm birth. We assessed model performance using various measures of accuracy including sensitivity, specificity, positive predictive value, negative predictive value, and area under the receiver operating characteristic curve (AUC) in an independent "validation" sample.

RESULTS

During the first trimester, logistic regression identified 13 variables associated with preterm birth, of which the strongest predictors were diabetes (Type I: adjusted odds ratio (AOR): 4.21; 95% confidence interval (CI): 3.23-5.42; Type II: AOR: 2.68; 95% CI: 2.05-3.46) and abnormal pregnancy-associated plasma protein A concentration (AOR: 2.04; 95% CI: 1.80-2.30). During the first trimester, the maximum AUC was 60% (95% CI: 58-62%) with artificial neural networks in the validation sample. During the second trimester, 17 variables were significantly associated with preterm birth, among which complications during pregnancy had the highest AOR (13.03; 95% CI: 12.21-13.90). During the second trimester, the AUC increased to 65% (95% CI: 63-66%) with artificial neural networks in the validation sample. Including complications during the pregnancy yielded an AUC of 80% (95% CI: 79-81%) with artificial neural networks. All models yielded 94-97% negative predictive values for spontaneous PTB during the first and second trimesters.

CONCLUSION

Although artificial neural networks provided slightly higher AUC than logistic regression, prediction of preterm birth in the first trimester remained elusive. However, including data from the second trimester improved prediction to a moderate level by both logistic regression and machine learning approaches.

Data-Driven Modeling of Pregnancy-Related Complications

A healthy pregnancy depends on complex interrelated biological adaptations involving placentation, maternal immune responses, and hormonal homeostasis. Recent advances in high-throughput technologies have provided access to multiomics biological data that, combined with clinical and social data, can provide a deeper understanding of normal and abnormal pregnancies. Integration of these heterogeneous datasets using state-of-the-art machine-learning methods can enable the prediction of short- and long-term health trajectories for a mother and offspring and the development of treatments to prevent or minimize complications. We review advanced machine-learning methods that could: provide deeper biological insights into a pregnancy not yet unveiled by current methodologies; clarify the etiologies and heterogeneity of pathologies that affect a pregnancy; and suggest the best approaches to address disparities in outcomes affecting vulnerable populations.

Prevention of preterm birth: Proactive and reactive clinical practice-are we on the right track?

Preterm birth remains the major cause of death and disability among children under the age of five. In developing countries antenatal preterm birth prevention clinics are set up to provide cervical length surveillance and/or treatment modalities such as cerclage or progesterone for those women with identified risk factors such as previous cervical treatment or preterm birth. However, 85% of women have no risk factors for PTB and currently there is no biomarker to screen women early in pregnancy. Women will present unexpectedly in threatened preterm labour and we have no choice but to adopt a re-active approach to their care by using predication and preparation strategies such as fetal fibronectin, tocolytic therapy and steroids. Despite these strategies approximately 15-20% of these women will give birth preterm before 34 weeks. There is a urgent need to re-design primary, secondary and tertiary prevention strategies for spontaneous preterm labour (sPTL) in singleton pregnancies aimed at identifying and addressing key gaps in clinical practice and research.

Evaluation of Cervical Elastography for Prediction of Spontaneous Preterm Birth in Low-Risk Women: A Prospective Study

OBJECTIVES

The aim of this study was to determine whether cervical elastographic parameters in addition to cervical length (CL) during the 3 trimesters of pregnancy would be predictive of spontaneous preterm birth (sPTB) among low-risk women.

METHODS

This work was a prospective nested case-control study evaluating cervical elastographic parameters and CL in low-risk women during the 3 trimesters of pregnancy. A binary logistic regression analysis was used to calculate significant covariates for prediction of sPTB. The area under the curve of the prediction model was calculated by using a receiver operating characteristic curve.

RESULTS

There were 286 women (26 cases and 260 controls) included in the analysis. The parameters of cervical elasticity became softened and heterogeneous during the 3 trimesters of pregnancy in both women with and without sPTB. The differences in the mean strain value at the internal os of the cervix (IOS), ratio (strain ratio of the internal os to the external os) during the second trimester and the IOS during the third trimester between the groups had statistical significance (P < .01; P = .01; P < .01, respectively). The CL had no association with sPTB during the 3 trimesters. The IOS during the second trimester was a better predictor of sPTB, with an area under the curve of 0.730, and sensitivity was 72.73%.

CONCLUSIONS

We observed multiple elastographic parameters and demonstrated the physiologic changes in the cervix during the 3 trimesters of pregnancy. Furthermore, we found that the IOS during the second trimester can be helpful in predicting sPTB. However, the CL had no association with sPTB during the 3 trimesters of pregnancy.

Using an innovative stacked ensemble algorithm for the accurate prediction of preterm birth

Objective:

A birth before the normal term of 38 weeks of gestation is called a preterm birth (PTB). It is one of the major reasons for neonatal death. The objective of this article was to predict PTB well in advance so that it was converted to a term birth.

Material and Methods:

This study uses the historical data of expectant mothers and an innovative stacked ensemble (SE) algorithm to predict PTB. The proposed algorithm stacks classifiers in multiple tiers. The accuracy of the classiffication is improved in every tier.

Results:

The experimental results from this study show that PTB can be predicted with more than 96% accuracy using innovative SE learning.

Conclusion:

The proposed approach helps physicians in Gynecology and Obstetrics departments to decide whether the expectant mother needs treatment. Treatment can be given to delay the birth only in patients for whom PTB is predicted, or in many cases to convert the PTB to a normal birth. This, in turn, can reduce the mortality of babies due to PTB.

Predictions of Preterm Birth from Early Pregnancy Characteristics: Born in Guangzhou Cohort Study

Preterm birth (PTB, <37 weeks) is the leading cause of death in children <5 years of age. Early risk prediction for PTB would enable early monitoring and intervention. However, such prediction models have been rarely reported, especially in low- and middle-income areas. We used data on a number of easily accessible predictors during early pregnancy from 9044 women in Born in Guangzhou Cohort Study, China to generate prediction models for overall PTB and spontaneous, iatrogenic, late (34–36 weeks), and early (<34 weeks) PTB. Models were constructed using the Cox proportional hazard model, and their performance was evaluated by Harrell’s c and D statistics and calibration plot. We further performed a systematic review to identify published models and validated them in our population. Our new prediction models had moderate discrimination, with Harrell’s c statistics ranging from 0.60–0.66 for overall and subtypes of PTB. Significant predictors included maternal age, height, history of preterm delivery, amount of vaginal bleeding, folic acid intake before pregnancy, and passive smoking during pregnancy. Calibration plots showed good fit for all models except for early PTB. We validated three published models, all of which were from studies conducted in high-income countries; the area under receiver operating characteristic for these models ranged from 0.50 to 0.56. Based on early pregnancy characteristics, our models have moderate predictive ability for PTB. Future studies should consider inclusion of laboratory markers for the prediction of PTB.

Prognostic models in obstetrics: available, but far from applicable

Health care provision is increasingly focused on the prediction of patients' individual risk for developing a particular health outcome in planning further tests and treatments. There has been a steady increase in the development and publication of prognostic models for various maternal and fetal outcomes in obstetrics. We undertook a systematic review to give an overview of the current status of available prognostic models in obstetrics in the context of their potential advantages and the process of developing and validating models. Important aspects to consider when assessing a prognostic model are discussed and recommendations on how to proceed on this within the obstetric domain are given. We searched MEDLINE (up to July 2012) for articles developing prognostic models in obstetrics. We identified 177 papers that reported the development of 263 prognostic models for 40 different outcomes. The most frequently predicted outcomes were preeclampsia (n = 69), preterm delivery (n = 63), mode of delivery (n = 22), gestational hypertension (n = 11), and small-for-gestational-age infants (n = 10). The performance of newer models was generally not better than that of older models predicting the same outcome. The most important measures of predictive accuracy (ie, a model's discrimination and calibration) were often (82.9%, 218/263) not both assessed. Very few developed models were validated in data other than the development data (8.7%, 23/263). Only two-thirds of the papers (62.4%, 164/263) presented the model such that validation in other populations was possible, and the clinical applicability was discussed in only 11.0% (29/263). The impact of developed models on clinical practice was unknown. We identified a large number of prognostic models in obstetrics, but there is relatively little evidence about their performance, impact, and usefulness in clinical practice so that at this point, clinical implementation cannot be recommended. New efforts should be directed toward evaluating the performance and impact of the existing models.

Predicting risk of spontaneous preterm delivery in women with a singleton pregnancy

BACKGROUND

Prediction of a woman's risk of a spontaneous preterm delivery (PTD) is a core challenge and an unresolved problem in today's obstetric practice. The objective of this study was to develop prediction models for spontaneous PTD (<37 weeks).

METHODS

A population-based register study of women born in Sweden with spontaneous onset of delivery was designed using Swedish Medical Birth Register data for 1992-2008. Predictive variables were identified by multiple logistic regression analysis, and outputs were used to calculate adjusted likelihood ratios in primiparous (n = 199 272) and multiparous (n = 249 580) singleton pregnant women. The predictive ability of each model was validated in a separate test sample for primiparous (n = 190 936) and multiparous (n = 239 203) women, respectively.

RESULTS

For multiparous women, the area under the ROC curve (AUC) of 0.74 [95% confidence interval (CI) 0.73, 0.74] indicated a satisfying performance of the model, while for primiparous women, it was rather poor {AUC: 0.58 [95% CI 0.57, 0.58]}. For both primiparous and multiparous women, the prediction models were quite good for pregnancies with comparatively low risk for spontaneous PTD, whereas more limited to predict pregnancies with ≥30% risk of spontaneous PTD.

CONCLUSIONS

Spontaneous PTD is difficult to predict in multiparous women and nearly impossible in primiparous, by using this statistical method in a large and unselected sample. However, adding clinical data (like cervical length) may in the future further improve its predictive performance.