Application of machine learning to identify risk factors of birth asphyxia

Despite rapid technological advances, under-five deaths among children remain high. A significant proportion of these deaths worldwide are due to BA. Several studies have been conducted using traditional statistical analysis techniques to identify risk factors for BA. For example, a meta-analysis conducted by Desalew et al. found that maternal illiteracy, prepartum hemorrhage, cesarean section, instrumental delivery, duration of labor, pregnancy-related hypertension, induction of labor, parity, low birth weight, preterm birth, non-cephalic delivery, and meconium staining were significantly associated with BA [14]. Our study was conducted to identify the various factors leading to BA in neonates delivered in a hospital in Bandar Abbas, Iran. According to the findings using the ML approach, Random Forest Classification was found to be the best model to predict BA with an AUC and an accuracy of 0.99. The analysis of the importance of the variables showed that maternal chronic hypertension, maternal anemia, diabetes, drug addiction, gestational age, newborn weight, newborn sex, preeclampsia, placenta abruption, parity, IUGR, meconium amniotic fluid, mal-presentation, and delivery method were considered to be the weighted factors.

Sociodemographic factors were not associated with BA. Among maternal comorbidities, chronic hypertension and diabetes were found to be correlated with BA. Hypertension can lead to a reduction in blood flow and thus asphyxia [15], while diabetes causes intrapartum hypoxia by developing placenta insufficiency [16]. Anemia was also found to be a risk factor for BA, as also observed in previous studies [17, 18]. Maternal anemia is a common pregnancy problem that disrupts maternal and fetal oxygen transport. The disorder may cause fetal hypoxia inside the womb, resulting in BA [19].

Another factor linked to BA was drug addiction. Drug addiction was demonstrated in the current study by declarations from mothers themselves. The actual number of addicts is always several times greater than the number of those identified; however, detecting addicted women is further complicated by their proclivity to conceal and deny the problem. Infants born to addicted mothers are more likely to have prematurity, low birth weight, and IUGR, all of which can contribute to BA [20]. We have found a significant association between gestational age and risk for BA. Preterm birth was found to be one of the most important risk factors for BA, as reported in previous studies [21, 22]. This could be due to the fact that preterm infants face multiple morbidities, including organ system, immaturity, and especially lung immaturity, which leads to respiratory failure [23]. However, some studies have shown that BA increases with gestational age [24, 25]. According to our findings, newborn weight was associated with BA. Low birth weight newborns were at higher risk of developing BA. A potential confounding factor for this could be the fact that mothers of low birth weight babies are often associated with complications such as maternal hypertension and diabetes that occur before conception or before birth [26]. Indeed, many LBW neonates are more likely to be preterm, unable to produce sufficient surfactant, and prone to multiple morbidities, including organ system immaturity, inability to initiate breathing, challenges with cardiopulmonary transition, and eventually developing BA. Fetuses with IUGR who experience growth restriction inside the uterus do not reach their full growth potential for a given gestational age and are at an increased risk of perinatal mortality and morbidity. In IUGR, the reduced rate of fetal growth is essentially an adaptation to an unfavorable intrauterine environment, and it can result in long-term changes in metabolism, growth, and development [27]. Fetuses with IUGR who have intrauterine hypoxia are more vulnerable to asphyxia. BA was observed in 34.4% of IUGR neonates in the clinical setting [28].

Our study showed that parity is associated with BA. The incidence of BA was higher in primiparous mothers. This is consistent with previous studies [29, 30]. Primiparous mothers are more likely to be younger, and they are more likely to have mal-presentations and prolonged obstructed labor. As a result, BA is expected to be more common in these women than in multipara women [31].

Preeclampsia is significantly associated with an increased risk of BA. The finding is in agreement with the evidence [32, 33]. This may be due to the reduction in blood, nutrient, and oxygen supply to the fetus, which may increase the risk of restriction of intrauterine development leading to BA [32].

Placenta abruption was also found to be associated with BA in our study, which was in contrast to previous studies [34]. The association of placenta abruption with BA can be explained by the fact that blood flows from the placenta to the fetus is restricted, leading to hypoxemia and thus asphyxia or stillbirth if maternal transfusion is delayed at the time of delivery [32].

It has long been known that non-cephalic fetuses are at greater risk during the birth process, including asphyxia, birth trauma, and death. This may be because non-cephalic fetuses are more likely to have other problems, such as cord prolapse and head entrapment, that predispose them to BA [35].

Newborns delivered via cesarean section and assisted vaginal birth had a higher rate of BA than those delivered via spontaneous vaginal delivery. This finding is consistent with previous research [36]. This is because either most mothers arrived late due to labor complications or the decision to have a cesarean section was delayed, increasing the burden of BA [37]. Another possibility is that the fetal chest is pressed as the newborns pass through the birth canal, causing secretion to be evacuated. This reduces the likelihood of developing BA, but this physiological benefit is not seen in cesarean section deliveries [38]. Furthermore, both vacuum and forceps extraction exert pressure on the newborn’s brain, which may cause the brain to bleed on the cranium, which contributes to intracranial hemorrhage and BA [39]. This finding suggests that interventions should be carefully evaluated and decided upon during intrapartum care to reduce unnecessary indications for an assisted vaginal birth and cesarean section to reduce the magnitude of BA.

In terms of newborn sex, our findings show that male infants are more likely to develop BA. This is consistent with previous research [40]. Our findings support previous findings of an association between congenital malformation and BA [41]. Although CNS anomalies might be expected to be associated with BA, the presence of other non-CNS birth defects raises important questions about the etiology of BA in these infants.

In line with several previous studies [41, 42], meconium-stained amniotic fluid was associated with BA. Grade III or IV meconium staining has been considered an indicator of a prolonged or severe episode of asphyxia [43]. One possible reason for this could be the inhalation of meconium-stained amniotic fluid, which causes irritation and inflammation of lung tissue or can obstruct the airways, leading to hypoxia and asphyxia at birth. In healthy, well-oxygenated fetuses, this diluted meconium is readily expelled from the lungs by normal physiologic mechanisms, but in a few cases, a meconium aspiration syndrome occurs [44].

The strength of our study is that we used a high-quality registration system in accordance with birth records. We studied both BA after vaginal birth and after cesarean section. We also examined a wide range of clinical factors associated with BA that may not be easily found in registries. Our study was retrospective, which is another limitation. The database did not allow us to determine the exact timing of the different events during pregnancy. For some variables, such as body mass index and weight gain during pregnancy, there was a lack of other data that might influence BA.