Low birth weight (LBW) is a global concern associated with fetal and neonatal mortality as well as adverse consequences such as intellectual disability, impaired cognitive development, and chronic diseases in adulthood. Numerous factors contribute to LBW and vary based on the region. The main objectives of this study were to compare four machine learning classifiers in the prediction of LBW and to determine the most important factors related to this phenomenon in Hamadan, Iran.
We carried out a retrospective cross-sectional study on a dataset collected from Fatemieh Hospital in 2017 that included 741 mother-newborn pairs and 13 potential factors. Decision tree, random forest, artificial neural network, support vector machine, and logistic regression (LR) methods were used to predict LBW, with five evaluation criteria utilized to compare performance.
Our findings revealed a 7% prevalence of LBW. The average accuracy of all models was 87% or higher. The LR method provided a sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and accuracy of 74%, 89%, 7.04%, 29%, and 88%, respectively. Using LR, gestational age, number of abortions, gravida, consanguinity, maternal age at delivery, and neonatal sex were determined to be the six most important variables associated with LBW.
Our findings underscore the importance of facilitating timely diagnosis of causes of abortion, providing genetic counseling to consanguineous couples, and strengthening care before and during pregnancy (particularly for young mothers) to reduce LBW.
Infant survival, physical and mental growth, and maternal health status and history are associated with a vital health indicator: birth weight [
Recently, machine learning (ML), an important branch of artificial intelligence, has been widely used in many fields. In particular, breakthroughs have been made with ML methods in medical diagnosis and outcome prediction [
In this retrospective cross-sectional study, we selected a random sample of 800 infants born at Fatemieh Hospital in the city of Hamadan. For data collection, we used a researcher-designed questionnaire based on case records available on the Iranian Maternal and Neonatal Network (IMaN Net) in 2017. The IMaN Net was designed by Iran’s Ministry of Health and Medical Education to evaluate the maternal and neonatal health status in Iran. After collecting the data, we excluded multiple pregnancies, stillbirths, and infants who died for any reason before discharge from the hospital or who had at least one abnormality. After this exclusion, 741 infants were included in the study, and the associated information was extracted as follows: (1) maternal data included place of residence (urban or rural); maternity insurance (insured or uninsured); delivery type (cesarean section or vaginal); maternal age at delivery (< 18, 18–35, or > 35 years); gestational age in weeks; preterm delivery (yes [< 37 weeks] or no [≥ 37 weeks of gestation]); consanguinity (yes or no); pregnancy risk factors such as chronic blood pressure, hepatitis, thyroid disease, cardiovascular disease, and preeclampsia/ eclampsia (yes or no); gravida; parity (i.e., number of previous live and non-live births); number of abortions; and number of previous live births. (2) Neonatal data included sex (male or female) and birth weight in grams.
Infants were classified using a binary outcome (1 for LBW and 0 for NBW), with a birth weight of 2,500 g as a threshold.
Before the analysis, the data were evaluated to ensure that no outliers were present. However, some missing values were found for seven variables, ranging from 0.13% to 5% of the dataset. The mean and median were used to impute quantitative and qualitative variables, respectively.
In classification, class imbalance and a bias toward the majority class may lead to misclassification. Thus, the data were first evaluated for imbalance, and the Synthetic Minority Oversampling TEchnique (SMOTE) was then used for the ML methods as an efficient algorithm for data balancing [
A simple ML technique, DT learning generates a tree-like structure by repeatedly splitting the dataset based on a criterion that maximizes the separation of the data [
The RF algorithm is based on an ensemble of large, correlated decision trees and combines the decisions of individual trees to produce accurate, stable results [
An ANN is a mathematical model designed to simulate the structure and function of biological neural networks in the brain [
SVMs use a decision boundary termed the hyperplane to separate classes. The hyperplane is located at a maximum distance from the closest data points of each class. These points are known as support vectors [
LR, a special case of generalized linear modeling, is extensively used for binary outcomes in epidemiology and medicine. By fitting data to a logistic function, the probability of an occurrence may be predicted [
To compare the performance of the utilized classifiers, we divided the data into training (70% of the data) and test (30%) sets and repeated this process 10 times. Then, the performance of the trained models was evaluated using the test set based on criteria of sensitivity (or recall), specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and accuracy as follows:
A false positive (FP) indicates NBW neonates that were incorrectly identified as LBW, a true positive (TP) indicates LBW neonates that were correctly diagnosed as LBW, a true negative (TN) indicates NBW neonates correctly identified as NBW, and a false negative (FN) indicates LBW neonates incorrectly identified as NBW.
To find the optimum values of the hyperparameters for the methods (DT, RF, ANN, and SVM), we applied a 10-fold cross-validation strategy to the training set. Hyperparameters were chosen when the maximum value of the receiver operating characteristic was observed. We repeated this process 10 times with different training and test partitions.
In this study, depending on the ML model, different methods were used to compute variable importance on a numerical scale from 0 to 100. For DT, it has been stated that “an overall measure of variable importance is the sum of the goodness of split measures for each split for which it was the primary variable, plus goodness (adjusted agreement) for all splits in which it was a surrogate” [
SPSS version 25 (IBM Corp., Armonk, NY, USA) was used to calculate descriptive statistics, after which R software (version 4.2.1; R Foundation for Statistical Computing, Vienna, Austria; packages: caret, themis, rpart, randomForest, nnet, and kernlab) was used to apply ML classifiers to the dataset and measure the evaluation criteria.
The data were collected from the IMaN Net. Therefore, a waiver of informed consent was awarded for this study. All methods were carried out in accordance with relevant guidelines and regulations, and the study was approved by the Ethical Committee of the Hamadan University of Medical Sciences (No. IR.UMSHA.REC.1401.779).
Among the 741 neonates, 51 (7%) had LBW (birth weight < 2,500 g). The mean ± standard deviation of the birth weight was approximately 3,138 ± 511 g. Overall, 51% of the neonates were male, and 10% were delivered before 37 weeks of gestation. Approximately 55% of mothers lived in urban areas, 95% were insured, and 74% had a vaginal delivery. The maternal age at delivery was between 18 and 35 years in most cases (80%). The mean ± standard deviation of the mother’ s gestational age was approximately 38 ± 2 weeks. Other maternal and neonatal characteristics are presented in
The importance levels of variables calculated using each ML model are shown in
Different classification approaches for LBW have been utilized in several studies. In the United Arab Emirates, Khan et al. [
Numerous studies have indicated that low gestational age is one of the most critical risk factors for LBW [
The present study had some limitations. First, we had no data regarding several important maternal characteristics, such as prenatal care, nutritional status, body mass index, interpregnancy interval, and financial status. Second, we could not include some features, such as maternal education, in the analysis due to a high percentage of missing values. Third, because considering each pregnancy risk factor (such as chronic blood pressure and cardiovascular disease) separately led to extremely unbalanced factor distributions, we considered all pregnancy risk factors as a single combined factor. Fourth, this study may have been vulnerable to potential bias in the evaluation of performance criteria, as the data were obtained from a retrospective registry-based study.
The results of this study showed that LR outperformed the other ML classifiers. Using promising classifiers to identify key LBW-related factors can allow medical practitioners to take preventative steps to minimize LBW. Based on the results, facilitating timely diagnosis of causes of abortion, providing genetic counseling to consanguineous couples, and improving care before and during pregnancy (especially for young mothers) can play an important role in reducing LBW. Additionally, the results of this study could be used to design an online mobile application to predict LBW risk in pregnant women. This would assist healthcare practitioners in the timely detection of mothers at high risk of giving birth to LBW infants and help provide them with appropriate interventions. In addition, researchers should consider the factors noted in the limitations section in further studies.
For the technical support, we are grateful to the Vice-chancellor of Education of Hamadan University of Medical Sciences. We also would like to appreciate the staff of Fatemieh Hospital in Hamadan for providing appropriate facilities for data collection.
This study was supported by the Hamadan University of Medical Sciences (Grant No. 140110138637).
No potential conflict of interest relevant to this article was reported.
Variable importance based on machine learning methods: (A) decision tree, (B) random forest, (C) artificial neural network, (D) support vector machine, and (E) logistic regression.
Maternal and neonatal demographic characteristics
Variable | NBW (n = 690) | LBW (n = 51) | Total (n = 741) | |
---|---|---|---|---|
Neonatal sex | 0.558 | |||
Male | 354 (51.30) | 24 (47.06) | 378 (51.01) | |
Female | 336 (48.70) | 27 (52.94) | 363 (48.99) | |
| ||||
Place of residence | 0.406 | |||
Urban | 378 (54.78) | 31 (60.78) | 409 (55.20) | |
Rural | 312 (45.22) | 20 (39.22) | 332 (44.80) | |
| ||||
Maternity insurance | 0.276 | |||
Uninsured | 29 (4.20) | 4 (7.84) | 33 (4.45) | |
Insured | 661 (95.80) | 47 (92.16) | 708 (95.55) | |
| ||||
Delivery type | 0.538 | |||
Vaginal | 514 (74.49) | 36 (70.59) | 550 (74.22) | |
Cesarean section | 176 (25.51) | 15 (29.41) | 191 (25.78) | |
| ||||
Maternal age at delivery (yr) | 0.470 | |||
< 18 | 29 (4.20) | 4 (7.84) | 33 (4.45) | |
18–35 | 556 (80.58) | 40 (78.43) | 596 (80.43) | |
> 35 | 105 (15.22) | 7 (13.73) | 112 (15.12) | |
| ||||
Preterm delivery | <0.001 | |||
No (≥ 37 wk) | 651 (94.35) | 15 (29.41) | 666 (89.88) | |
Yes (< 37 wk) | 39 (5.65) | 36 (70.59) | 75 (10.12) | |
| ||||
Consanguinity | 0.133 | |||
No | 594 (86.09) | 40 (78.43) | 634 (85.56) | |
Yes | 96 (13.91) | 11 (21.57) | 107 (14.44) | |
| ||||
Pregnancy risk factors | 0.122 | |||
No | 573 (83.04) | 38 (74.51) | 611 (82.46) | |
Yes | 117 (16.96) | 13 (25.49) | 130 (17.54) | |
| ||||
Gravidity | 0.354 | |||
1 | 256 (37.10) | 21 (41.18) | 277 (37.38) | |
2 | 206 (29.85) | 10 (19.61) | 216 (29.15) | |
3 | 138 (20.00) | 11 (21.57) | 149 (20.11) | |
4 | 64 (9.28) | 5 (9.80) | 69 (9.31) | |
≥ 5 | 26 (3.77) | 4 (7.84) | 30 (4.05) | |
| ||||
Parity | 0.230 | |||
0 | 291 (42.17) | 24 (47.06) | 315 (42.51) | |
1 | 225 (32.61) | 10 (19.61) | 235 (31.71) | |
2 | 132 (19.13) | 12 (23.53) | 144 (19.44) | |
≥ 3 | 42 (6.09) | 5 (9.80) | 47 (6.34) | |
| ||||
Number of abortions | 0.095 | |||
0 | 555 (80.44) | 39 (76.47) | 594 (80.16) | |
1 | 110 (15.94) | 7 (13.73) | 117 (15.79) | |
2 | 19 (2.75) | 3 (5.88) | 22 (2.97) | |
≥ 3 | 6 (0.87) | 2 (3.92) | 8 (1.08) | |
| ||||
Number of previous live births | 0.320 | |||
0 | 299 (43.33) | 25 (49.02) | 324 (43.72) | |
1 | 227 (32.90) | 11 (21.57) | 238 (32.12) | |
2 | 133 (19.28) | 11 (21.57) | 144 (19.44) | |
≥ 3 | 31 (4.49) | 4 (7.84) | 35 (4.72) | |
| ||||
Gestational age (wk) | 38.90 ± 1.39 | 34.17 ± 3.87 | 38.57 ± 2.06 | <0.001 |
| ||||
Neonatal birth weight (g) | 3,226.36 ± 386.48 | 1,943.33 ± 503.77 | 3,138.05 ± 511.73 | - |
Values are presented as number (%) or mean ± standard deviation.
NBW: normal birth weight, LBW: low birth weight.
Factors associated with low birth weight in infants based on logistic regression
Variable | β | Exp (β) | Wald | |
---|---|---|---|---|
Intercept | 27.07 | - | 18.38 | <0.001 |
Neonatal sex (ref: male) | 0.88 | 2.42 | 4.08 | 0.043 |
Place of residence (ref: urban) | −0.14 | 0.86 | 0.12 | 0.723 |
Maternity insurance (ref: uninsured) | −0.29 | 0.74 | 0.08 | 0.772 |
Delivery type (ref: vaginal) | 0.38 | 1.47 | 0.69 | 0.404 |
Maternal age at delivery (yr) | −1.17 | 0.30 | 4.91 | 0.027 |
Preterm delivery (ref: no) | 0.97 | 2.65 | 1.84 | 0.174 |
Consanguinity (ref: no) | 1.13 | 3.12 | 5.50 | 0.019 |
Pregnancy risk factors (ref: no) | −0.24 | 0.78 | 0.21 | 0.647 |
Gravidity | −2.52 | 0.08 | 6.24 | 0.012 |
Parity | 1.95 | 7.03 | 2.47 | 0.115 |
Number of abortions | 2.35 | 10.50 | 8.25 | 0.004 |
Number of previous live births | 0.66 | 1.94 | 0.47 | 0.490 |
Gestational age (wk) | −0.71 | 0.49 | 19.91 | <0.001 |
The reference category for the outcome is normal birth weight.
Performance comparison of ML classifiers in the prediction of LBW on a test dataset with 10 repetitions
Model | Sensitivity | Specificity | PLR | NLR | Accuracy |
---|---|---|---|---|---|
Decision tree | 0.61 ± 0.16 | 0.93 ± 0.03 | 9.91 ± 3.42 | 0.41 ± 0.16 | 0.91 ± 0.03 |
Random forest | 0.44 ± 0.13 | 0.97 ± 0.01 | 15.27 ± 4.45 | 0.58 ± 0.13 | 0.93 ± 0.01 |
Artificial neural network | 0.71 ± 0.13 | 0.88 ± 0.03 | 6.37 ± 1.72 | 0.33 ± 0.14 | 0.87 ± 0.03 |
Support vector machine | 0.68 ± 0.10 | 0.94 ± 0.01 | 11.92 ± 3.11 | 0.34 ± 0.11 | 0.92 ± 0.01 |
Logistic regression | 0.74 ± 0.09 | 0.89 ± 0.03 | 7.04 ± 2.03 | 0.29 ± 0.10 | 0.88 ± 0.02 |
Values are presented as mean ± standard deviation.
ML: machine learning, LBW: low birth weight, PLR: positive likelihood ratio, NLR: negative likelihood ratio.