In this study, six parameters were selected discreetly in the estimation of the overall risk of iatrogenic preterm birth: shorter maternal height, extremely low or advanced IPI, greater vaginal number of vaginal bleeding during pregnancy, higher parity, hypertension during pregnancy and placenta previa. By combining these factors, the risk of iatrogenic preterm birth can be well predicted.
Many studies have attempted to establish a simple way to predict preterm birth, but most focused on spontaneous preterm birth. In our study, many variables were used to predict iatrogenic preterm birth. Most of the underlying variables have been previously reported to impact preterm birth, while myomectomy and dysmenorrhea were included tentatively to assess their relationship with iatrogenic preterm birth. Maternal and foetal indications are the direct reasons for obstetricians to consider a patient at high risk of iatrogenic preterm birth or to tend to terminate the pregnancy. Severe complications during pregnancy, such as uterine rupture, are likely to lead to iatrogenic preterm birth. Although the prevalence of such complications is usually low, quick treatment is needed for these complications. Meanwhile, common foetal reasons for iatrogenic preterm birth, such as foetal distress, also require quick treatment given their sudden occurrence. Therefore, we mainly focused on chronic pregnancy complications, such as placenta previa and GDM, when we included complications into the model.
A shorter height has been associated with a progressive increase in the odds of having an infant born preterm [17, 18]. Our study shows the same result by using a Chinese population, and a smaller maternal pelvic size may be the underlying mechanism due to evolutionary adaptation.
Placenta previa is a risk factor for preterm birth [5]. The formation of the lower uterine segment and cervical dilation will cause a certain degree of spontaneous placental separation, which may result in severe haemorrhage and can indicate preterm birth [19]. As a clinical indicator of iatrogenic preterm birth, placenta previa accounts for 14.1% of all cases, which is much larger than the prevalence in China. To our knowledge, this is the first report of the morbidity of placenta previa in a large-scale, Chinese, scarred uterus population.
IPI is defined as the time from the most recent prior birth to conception of the index birth by Mckinney and his coworkers [20]. An IPI of 0 to < 12 months accounted for only a small portion of the data (n = 35,1.5%) because most of the women with IPIs less than 12 months were recommended for delivery of the baby. The 12 to < 24 months category was chosen as the reference group based on Mckinney’s study, and an IPI of 12 to < 24 months was associated with the lowest risk of preterm birth in both Mckinney’s and our studies.
Self-reported vaginal bleeding during pregnancy is predictive for preterm birth. The odds ratio of vaginal bleeding was 2.7 (95% Cl 2.03–3.70) in our study, with an incidence of 10.9%, smaller than that in other studies [21, 22]. Our study used number of vaginal bleeding during pregnancy instead of the presence of bleeding over the three trimesters or bleeding volume because we considered it to be easier for the patients to recall.
Women with a parity of less than two composed the majority of our data (n = 2171, 93.8%) and in the general population. This demographic characteristic is quite different in China due to the singleton policy that had been in place over the past years. Nulliparous and highly multiparous women are at higher risk of adverse pregnancy outcomes than those with low multiparity [23]. In our study, advancing parity showed a higher risk of adverse pregnancy outcomes. However, the nulliparity included in our study corresponded to women who had received myomectomy. Thus, there is no conflict between these two studies because of the different inclusion criteria.
Hypertension during pregnancy increases the risk of preterm birth. A recent meta-analysis including 55 studies found that women with chronic hypertension had high pooled incidences of preterm birth [24]. Preeclampsia was also found to be associated with high rates of preterm birth and puerperal complications, while gestational hypertension was only found to be related to preterm birth [25].
Apart from these six parameters, some factors were not included in our model and are thought to be important in the prediction of preterm birth. Age has always been considered a significant variable in preterm birth prediction. We have tried many ways to categorize age, including dividing it into three groups based on the report of a U-shaped relationship with preterm birth [26]. Sadly, none of these attempts give us a statistically significant result. Likewise, factors related to infection are important variables related to preterm birth. However, screening for infection requires sequential tests, including vaginal secretion cultures and inflammatory indexes, which are not fully covered by medical insurance in China. Therefore, factors related to infection were not included given their low cost-efficiency and data integrity.
After screening the above six factors among all the variables, a model was built and validated. The results of the discrimination and calibration tests are shown above. Overall, our model shows good discrimination and calibration. However, discrimination and calibration alone cannot capture the clinical consequences of a particular level of discrimination or degree of miscalibration [27,28,29]. To justify the practical applicability of our model, decision curve analysis was applied in this study. The results of the decision curve analysis indicate a worth-expecting practice in the clinic. With the threshold probability between 15 and 60%, the use of a nomogram in our study to predict iatrogenic preterm birth adds more benefit than either a treat-all-patients scheme or a treat-none scheme.
The distribution of medical resources in China is not even. Many hospitals in rural areas lack neonatal intensive care units, which leads to adverse outcomes for the newborns. Sometimes mothers with pregnancy complications cannot be treated effectively. Reasonable and efficient referral can improve this situation. Our model improves maternal and child outcomes by assessing the risk of iatrogenic premature birth in patients, thus assisting in referral-making and helping in the rational allocation of medical resources. A correct prediction can provide patients better medical care, thus improving the prognosis of the mother and foetus, while an incorrect prediction would waste local medical resources or delay the opportunity to treat patients. We recommend that doctors use lower cutoff values in districts with rich medical resources or high economic capability. Note that the applicable population for this model is the same as the inclusion criteria of this study, which means that this model is only suitable for pregnant women with more than 20 weeks of gestation. Meanwhile, new problems may arise at any time, so we recommend using this model sequentially during pregnancy.
Strengths and limitations
The prediction model we developed is novel. To our knowledge, this is the first model to predict iatrogenic preterm birth in a scarred uterus population using the population from Northeast China. In our study, we selectively collected individual information that can be easily acquired during consultation or through some basic examinations. Thus, it is a convenient model that can be easily practised in the clinic for the recognition of high-risk populations and for making referrals, which means that it can be widely applied in primary health care institutions in rural areas of China that lack sufficient medical resources.
The shortcomings of this study are as follows: (1). The influence of maternal weight on preterm birth is complicated. We decided to collect all the data on weight and weight gain for all trimesters at first. Due to data loss, only weight before delivery was included in this study. (2). A total of 2315 cases were divided into a training set and a validation set; the nature of internal validation indicates one of the weakness of our study. (3). Due to the retrospective nature of this study, the data we collected, except for age and height, were all acquired before delivery. Therefore, the model is only valid for pregnant women with more than 20 gestational weeks.