Associations of maternal motor vehicle crashes during pregnancy with offspring’s neonatal birth outcomes

Abstract

Background

Adverse events in fetuses are well researched but studies on the follow-up health outcomes of infants exposed to maternal motor vehicle crashes (MVCs) during pregnancy have yielded inconsistent results. This study aimed to investigate the association of maternal exposure to MVCs during pregnancy with the risk of adverse neonatal outcomes.

Methods

This population-based cohort study used data from birth notifications in Taiwan. A total of 19 277 offspring with maternal exposure to MVCs during pregnancy and 76 015 randomly selected comparison offspring without such exposure were selected. Neonatal adverse outcomes were identified from National Health Insurance medical claims data. Conditional logistic regression was used to estimate the unadjusted and adjusted odds ratios (aORs) of neonatal adverse outcomes.

Results

Offspring exposed to maternal MVCs during pregnancy had a higher risk of birth defects (aOR, 1.21; 95% CI, 1.04–1.41) than offspring without such exposure. This positive association was sustained with exposure to an MVC during the first or second trimester. A dose–response relationship (P = 0.0023) was observed between the level of injury severity and the risk of birth defects.

Conclusions

In the early stages of pregnancy, maternal exposure to MVCs may entail a risk of birth defects in the offspring. The potential mechanisms for the associations of maternal exposure to MVCs with birth defects need further investigation.

Introduction

Motor vehicle crashes (MVCs) during pregnancy are associated with many adverse maternal outcomes, including splenic rupture, uterine rupture, pelvic fracture, placental abruption and maternal death.^1–3 However, the association between maternal exposure to MVCs and adverse outcomes for the fetus remains inconsistent. One study showed possible central nervous system (CNS) damage to surviving fetuses, such as localized vascular infarctions, haemorrhage, hydrocephalus and global brain damage, after maternal exposure to MVCs during pregnancy.⁴ Sauber-Schatz et al. did not find an increased risk of birth defects from various maternal injuries (e.g. abdominal trauma, abuse, MVC and gunshot wound).⁵ Tinker et al. found that maternal injury during the pre-conceptional period (the month before pregnancy to the end of the third month of pregnancy) may be associated with selected birth defects but the study did not provide sufficient information for identifying injury specifically caused by MVCs.⁶

CNS development in a baby’s brain begins with cell proliferation and neurogenesis from 8 weeks after conception to Week 16 of gestation. The brain continues to develop throughout pregnancy and after birth.⁷ Any disruption or injury to brain development during this period can lead to neurological problems.⁸ Additionally, the severity of traumatic brain injury has a dose–response relationship with poor outcomes in children.⁹ Despite these findings, to the best of our knowledge, the risk of neonatal outcomes in association with injury severity from maternal MVCs during pregnancy has not been assessed.³

The role of the people involved in MVCs (e.g. driver, passenger or pedestrian) and the vehicle type are important determinants for the severity of MVCs.¹⁰ However, these MVC-related factors have rarely been considered in prior studies.^11–13 In addition, prior studies using data from traumatic registries or emergency rooms could suffer from potential selection bias because not all victims involved in MVCs require medical care at the time of the accident.¹⁴ This study aimed to investigate the association of maternal exposure to MVCs during pregnancy with risks of adverse neonatal outcomes by using a cohort study design and linking all victims involved in MVCs at accident scenes to medical claims that could provide information about injury severity and study outcomes. We hypothesized that maternal exposure to MVCs during pregnancy is associated with an increased risk of adverse neonatal outcomes.

Methods

Data source

Four national data sets supervised by the Health & Welfare Data Science Center were used. These were vital records (birth notifications and the death registry), the Taiwan Maternal and Child Health Database, medical claims of the National Health Insurance (NHI) programme and the Police-Reported Traffic Accident Registry (PTAR) collected by the National Police Agency in Taiwan. Birth notifications provide information on demographic characteristics, mother’s place of residence at time of birth and birth characteristics.^15,16 The Taiwan Maternal and Child Health Database provides parent–offspring linkages for all infants.¹⁷ NHI data sets cover the medical claims of nearly all Taiwanese residents (>99%) and include their outpatient, emergency and inpatient medical claims.¹⁸ The PTAR, which is recorded by vehicle accident investigators, contains general information about MVC victims, including the date of the accident, role of the road user, sex and vehicle type. Details of the PTAR are given in our previous study.¹⁹ This study was approved by the Institutional Review Board of the National Cheng Kung University Hospital (No. B−EX-109–088). This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.

Study design and participants

This study was a retrospective cohort study involving data on live births from 2007 to 2016 in Taiwan’s birth notifications data set. Records were excluded for mothers aged <18 or >50 years at the time of birth (n = 6822), mothers with multiple births (n = 66 207) and mothers with a missing mother’s ID (n = 3). Among the 1 970 155 remaining singletons born to mothers aged 18–50 years, 20 844 experienced MVCs during pregnancy. The control group comprised 76 015 comparison offspring randomly selected from all offspring whose mothers did not experience MVCs during pregnancy (n = 83 274). On the day of a crash involving an MVC mother, we randomly selected four non-MVC mothers of the same age (in years) and gestational weeks on that day. The study further excluded mothers who failed to link their offspring through the Taiwan Maternal and Child Health Database, leaving 19 277 and 76 015 mother–offspring pairs for the MVC and non-MVC groups, respectively (Figure 1).

MVC exposure measurement

The independent variables of interest in the study included the trimester of MVC occurrence, level of injury severity (uninjured, mild, severe), role of the road user (driver, passenger, pedestrian) and vehicle type (car, scooter). The severity level of an injury was generated using R software and was based on the International Classification of Diseases Programs for Injury Categorization (ICDPIC) from the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) or 10th Revision, Clinical Modification (ICD10-CM) codes for each patient’s clinical registry.²⁰ The ICDPIC programme was used to determine the maximum abbreviated injury scale (MAIS) from the inpatient and outpatient claims data within 3 days after an MVC. Severe injury was considered a MAIS score of ≥3, mild injury a MAIS score of 1 or 2 and absence of injury a MAIS score of 0. We also conducted a sensitivity analysis by using ICDPIC to generate an approximate injury severity score (ISS), known as RISS (ISS calculated using ICDPIC-R).²⁰ This analysis aimed to assess the robustness of the association between the level of injury severity and neonatal outcomes. We categorized a RISS score of ≥16 as a severe injury and a score of 1–15 as a minor injury.

Neonatal outcomes

The neonatal outcomes of each offspring were confirmed according to the NHI claims. The neonatal outcomes assessed were selected on the basis of findings from prior studies. Schiff and Holt indicated a high risk of fetal distress, hypoxia and respiratory distress syndrome when infants were exposed to maternal MVCs.¹ Tinker et al. reported that maternal injury during early pregnancy may be associated with some birth defects.⁶ Therefore, the neonatal outcomes assessed in this study included the following diagnoses in the NHI claims data: fetal distress (ICD-9-CM code 656.3, 768.2–768.4; ICD-10-CM code O77.9), intrauterine hypoxia and birth asphyxia (ICD-9-CM code 768; ICD-10-CM code P84), respiratory distress syndrome (ICD-9-CM code 769; ICD-10-CM code P22) and birth defects (ICD-9-CM code 740–759; ICD-10-CM code Q00–Q99). Any of the above diagnostic codes was counted when it appeared in the medical claims at the time of first-time discharge after birth. We further performed analyses of specific birth defects.

Covariates

The following risk factors were considered covariates in the analysis: demographic characteristics including maternal age at MVC (17–24, 25–29, 30–34, 35–39, 40+ years), calendar year of MVC (2006–08, 2009–10, 2011–12, 2013–14, 2015–16), gestational age at MVC (<28, 28–32, 33–36, 37+ weeks) and offspring sex (boy, girl).^1,12,21 Socio-economic risk factors included the mother’s birthplace (Taiwan, other), geographical area of residence (north, central, south, east, other), urbanization of residence (urban, satellite, rural) and median family income quartile (minimum–first quartile, first–third quartile, third quartile–maximum). Clinical and lifestyle risk factors associated with developmental adverse neonatal outcomes included risky behaviour (smoking, alcohol consumption and substance use) during pregnancy (no, yes),^22–24 gestational diabetes mellitus (no, yes), gestational hypertension (no, yes), type of birth (vaginal birth, caesarean section) and offspring birthweight (<1500, 1500–2500, 2501–4000, >4000 g) were considered. We considered that the calendar year should be included as a potential confounder in this study because our data showed an increase in the number of MVCs over the study period and the trend for neonatal outcomes might change over the years.²⁵

Statistical analysis

Continuous variables are presented as mean (standard deviation) and categorical variables as number (percentage). The characteristics of offspring in the MVC and non-MVC groups were compared using the chi-squared test for categorical variables. Crude and covariate-adjusted odds ratios (aORs) and corresponding 95% CIs for various neonatal outcomes in association with maternal exposure to MVCs during pregnancy were estimated using a conditional logistic regression model. The model considered the study participants in the same risk sets and adjusted the demographic characteristics, socio-economic factors and clinical/lifestyle risk factors. Those covariates were based on a directed acyclic graph (Supplementary Figure S1, available as Supplementary data at IJE online). In addition to the analysis of neonatal outcomes in association with overall MVCs during pregnancy, we also performed restricted analyses of the neonatal outcomes associated with MVCs that occurred during various trimesters, with different levels of injury severity, involving different vehicle types and with different roles of road users. We compared those offspring exposed to maternal MVCs under restricted conditions with all unexposed offspring. We then calculated the E-value, defined as the minimum strength of association, on the risk ratio scale. An unmeasured confounder needs to have both the exposure and the outcome to fully explain away a specific exposure–outcome association, conditional on the measured covariates.²⁶ Statistical analyses were conducted in SAS (version 9.4; SAS Institute, Cary, NC) and the level of significance was set to α = 0.05.

Results

Compared with the non-MVC group, offspring exposed to maternal MVCs during pregnancy were likely to live in southern (29.8% vs 24.3%) and rural (30.0% vs 26.9%) areas and less likely to reside in city districts or townships with high median family income quartiles (21.8% vs 25.9%). In addition, mothers in the MVC group were more likely to be born in Taiwan (96.9% vs 93.6%) and to have a slightly increased prevalence of gestational diabetes mellitus (12.5% vs 11.8%) and gestational hypertension (5.1% vs 4.4%). The MVC-exposed offspring had a slightly higher prevalence of birth by caesarean section (34.8% vs 33.5%) and low birthweight (<2500 g, 7.1% vs 6.5%) than the non-MVC group (Table 1).

Compared with the controls, the offspring with maternal MVCs had a higher incidence of fetal distress (1.7% vs 1.5%), respiratory distress symptoms (0.8% vs 0.7%) and birth defects (1.2% vs 0.9%) but a lower incidence of intrauterine hypoxia and birth asphyxia (0.29% vs 0.31%). Nonetheless, only the increased incidence rate of birth defects reached statistical significance, with a crude odds ratio of 1.25 (95% CI, 1.08–1.45). After covariate adjustment, the significance in the increased aOR of birth defects was sustained, at 1.21 (95% CI, 1.04–1.41). The full models are shown in Supplementary Tables S1–S4 (available as Supplementary data at IJE online). The trimester-specific analyses indicated that offspring exposed to maternal MVCs during the first and second trimesters experienced an obvious increase in birth defects, with aORs of 1.31 (95% CI, 1.04–1.65) and 1.32 (95% CI, 1.03–1.68), respectively (Table 2).

Mild or severe maternal injury from MVCs was associated with a high risk of birth defects, with aORs of 1.31 (95% CI, 1.07–1.60) and 2.43 (95% CI, 1.11–5.30), respectively, and a dose–gradient relationship was observed (P for trend= 0.0023; Table 3). Although such a dose–gradient relationship was also observed for the association of injury severity level with intrauterine hypoxia and birth asphyxia (P=0.0272) and respiratory distress symptoms (P=0.0026), the aORs of neonatal outcomes evaluated for exposure to various levels of injury severity showed no substantial association. When performing RISS analysis to determine injury severity and examine the relative risk estimates, we found that infants born to mothers who experienced severe injury (RISS ≥16) from MVCs had a higher risk of birth defects and respiratory distress syndrome than infants born to mothers not involved in MVCs (see Supplementary Table S5, available as Supplementary data at IJE online).

The results of the analyses according to the road user role and vehicle type in MVCs are shown in Supplementary Table S6 (available as Supplementary data at IJE online). Increased risks of fetal distress (aOR, 1.36; 95% CI, 1.05–1.77) and respiratory distress syndrome (aOR, 1.56; 95% CI, 1.08–2.25) were seen in infants of mothers who were passengers in MVCs. An increased risk of birth defects was associated with maternal exposure to MVCs as car drivers (aOR, 1.32; 95% CI, 1.04–1.67) or scooter riders (aOR, 1.25; 95% CI, 1.02–1.54).

We analysed infants with birth defects (714 offspring without MVCs and 226 offspring with MVCs) and stratified them into birth defect categories. The results showed that the offspring exposed to MVCs during pregnancy had a higher proportion of congenital anomalies of the nervous system than offspring without such exposure (9.06% vs 6.78%).

Discussion

A moderately elevated risk of birth defects was observed in offspring with maternal exposure to MVCs during pregnancy where the mother was a driver. Specific analyses further showed that offspring could have an even higher risk of birth defects when the MVC occurred in the first or second trimester. In addition, mothers who sustained mild to severe injuries were found to have a higher risk of birth defects in their offspring.

The association between maternal exposure to MVCs during pregnancy and birth defects may be related to damage to the CNS of the fetus.⁴ Maternal trauma might be associated with CNS damage to the fetus, especially hydrocephalus.⁴ The main evidence from animal models showed that if the traumatic event results in maternal haemorrhage or shock, with a decline in systemic arterial blood pressure, vasoconstriction may occur, which may lead to uterine hypoperfusion.²⁷ Once a mother experiences hypoperfusion, placental perfusion becomes unstable and affects the fetal extraction of oxygen depending on uterine blood.²⁸ Finally, congenital defects can develop because of a hypoxemic maternal situation.⁴

The results of this study are consistent with those of earlier case–control studies and case reports that suggested a possible association of maternal injuries during pregnancy with hydrocephalus and other defects of the CNS.⁴,29–31 The National Birth Defects Prevention Study in the USA analysed 16 074 infants with birth defects; 6328 controls without major birth defects did not have an increase in odds of birth defects and maternal injuries resulting from MVCs.⁶ However, nearly 30% of injuries were not classified according to their intent and 30% of injuries were not classified according to cause because of insufficient details and lack of specific information about injury in the interviewer’s report.⁶ A population-based retrospective cohort study used the Texas Birth Defects Registry (TBDR) and included 59 750 live-born infants to assess the association between injury during pregnancy and birth defects related to the nervous system; a total of 4144 infants (6.94%) were diagnosed with any form of nervous system birth defect.⁵ Data on maternal injury during pregnancy were retrieved from the TBDR registry by reviewing the medical charts of each participant. No association between injury during pregnancy and the risk of nervous system birth defects was observed (aOR, 1.00; 95% CI, 0.63–1.56). However, the study did not examine the effect of injuries specifically caused by MVCs.

The effect of the timing of maternal MVCs on the risk of adverse health effects remains controversial. In this study, the odds ratio for birth defects was substantially higher in the first and second trimesters, but not in the third trimester. Therefore, early exposure of fetuses to MVCs might entail more harm to their development, although some studies have suggested that trauma in late pregnancy poses greater risks. Maternal pelvic fracture leads to fetal head injuries when the fetus is in the vertex position during the late stage of pregnancy.^32–34 Despite inconsistency in the findings, the adverse effect of maternal injuries on the health of a fetus is likely to be timing-dependent.

Neonatal outcomes were also analysed in relation to the road user role in MVCs. For mothers involved in MVCs as passengers, the odds ratios of having a child with fetal distress or respiratory distress syndrome increased. Although several studies have investigated the associations between fetal distress and maternal MVCs, the results are neither comprehensive nor conclusive.^1,11,35 According to Zhai et al., the rear passenger seat in a car is associated with a higher risk of injury than the driver seat.³⁶ By contrast, in our study, scooters were mainly involved in MVCs. One study in Iran compared motorcycle riders and their passengers in terms of injury severity and characteristics. The results showed that, although riders had a higher level of injury severity, the passengers might differ in terms of injured body parts.³⁷

Strengths and limitations

Our results should be interpreted with caution in light of several limitations. Some unmeasured confounding factors may have compromised the validity of our findings, despite the adjustment for known risk factors associated with adverse neonatal outcomes. For example, gestational exposure to illicit drugs has been associated with injury during pregnancy and adverse fetal outcomes in previous studies.^38–40 A review indicated that 65–75% of birth defects are due to unknown causes with suspected multifactorial aetiologies, 15–20% are single-gene disorders, 5% of chromosomal abnormalities are the most comment genetic aetiologies and 10% of birth defects arise from environmental exposures.³⁸ In addition, the maternal injury level was roughly defined in this study and detailed information regarding the crashes was not available from the research data. Numerous studies have consistently highlighted a notable contrast in the health impacts of road traffic crashes between urban and rural areas. Specifically, less urbanized regions have been found to have higher rates of mortality associated with MVCs.⁴¹ This disparity can be attributed to several factors, including increased exposure to severe crashes,⁴² higher collision velocities⁴³ and limited access to healthcare services. In particular, emergency medical service response times and the duration of transportation to higher-level trauma centres tend to be longer in less urbanized areas.⁴⁴ Specific interpretations of our study findings are somewhat limited without the MVC-related information. Moreover, maternal occupational exposure to radiation may lead to birth defects.⁴⁵ In addition, pre-pregnancy diabetes and gestational diabetes have been found to be associated with certain forms of birth defects.¹⁶ To assess whether these unadjusted risk factors for adverse neonatal outcomes can account for the aOR of birth defects seen in this study, we calculated the E-value for these unadjusted risk factors. The calculated E-value was 1.71 with a CI of 1.24. An unadjusted confounder needs to be associated with maternal MVCs and birth defects with a relative risk estimate of ≥1.24 to nullify this association. A large E-value implies that a considerable unmeasured confounder would be needed to explain away an effect estimate. A small E-value implies that a small unmeasured confounder would be needed to explain away an effect estimate. An E-value of 1.71 with a CI of 1.24 was considered small in magnitude. We performed adjustment for a wide range of confounders in our analysis and known genetic risk factors for birth defects are less likely to have associations with MVCs. In addition, we considered gestational diabetes mellitus in our original analysis. To address the potential confounding by pre-pregnancy type 1 and type 2 diabetes, we retrieved information on medical visits by mothers for type 1 and type 2 diabetes in a 1-year period before pregnancy from the NHI claims data. Mothers with and without MVCs had nearly the same prevalence of pre-pregnancy type 1 diabetes (0.3‰, P = 0.7974). The corresponding figures for pre-pregnancy type 2 diabetes were 12.4‰ and 11.6‰, respectively (P = 0.3313). Therefore, pre-pregnancy diabetes was unlikely to introduce confounders to our study and adjusting for this variable would not have resulted in a meaningful change to our study results. The associations of the above unadjusted covariates with MVC exposure and birth defects are less likely to exceed the calculated E-values. Moreover, we used ICD-9-CM/ICD-10-CM diagnosis codes in medical claims data to determine the study outcomes, leaving room for disease misclassification. However, this problem would be encountered in both the MVC and non-MVC groups. Therefore, such disease misclassification is likely to be non-differential, which may lead to underestimation rather than overestimation of the true associations between MVC during pregnancy and adverse neonatal outcomes. As for information bias, our study lacked information on crash severity, which can also influence the level of maternal injury. We used the MAIS as an indicator of injury severity and there may have been a large variation in injury among those with the same MAIS.⁴⁶ Moreover, we noticed a probability of exposure of non-MVC mothers to other traumas. To prevent misclassification bias between MVC and non-MVC groups, we conducted additional analyses that excluded 30 383 infants whose mothers had hospital admission due to any-cause traumatic injuries. The 7869 infants born to the MVC mothers in the matched risk set were also excluded. Subsequently, we found an elevated relative risk estimate for birth defects (aOR, 1.33; 95% CI, 1.09–1.62). The result was reasonable because non-MVC-related traumatic injuries in non-MVC mothers can mask, at least to some extent, the true association between MVCs and neonatal birth outcomes. Additionally, although we adjusted for a collected measure of risky behaviour by mothers during pregnancy, including smoking, alcohol consumption and substance use, residual confounders might have appeared in our study because of the uncertain validity of the information from birth notifications. Lastly, the data from 2007 and 2016 were not updated and thus the association between current risk and maternal MVCs cannot be determined.

Despite the limitations, our study had several methodological merits. First, to the best of our knowledge, this study analysed the largest number of offspring to assess the association of maternal MVCs with risk of adverse neonatal outcomes. Second, this study used an unexposed group as controls in a cohort design, which is better than case series studies.^4,47 Third, we were able to specifically focus on motor vehicle trauma rather than on all traumatic injuries, achieving precise interpretations. Fourth, previous research that investigated maternal exposure relied on mothers’ recall of their MVC experiences during pregnancy, which might have involved some extent of recall bias.⁶ In this study, we determined exposure through traffic accident data reported by the police department, which greatly reduced the chance of information bias. Fifth, MVC data were retrieved from the PTAR, which was considered more comprehensive than data obtained solely from clinical settings. Analyses based on emergency data or trauma registries may exclude MVC victims who were uninjured from accidents or did not experience fatal injuries resulting in immediate death at the scene of the MVC.¹⁴ The utilization of databases beyond clinical data is essential in MVC epidemiological studies. Finally, we conducted specific analyses of MVC exposure according to injury severity level, trimester, road user role and type of vehicle. These analyses can be considered novel approaches.

Conclusions

This study reports a moderate but increased risk of birth defects in neonates with maternal exposure to MVCs, especially for mothers involved in MVCs as drivers and MVCs that occurred in the first two trimesters. Given that MVCs are the leading cause of injury in pregnant women,⁴⁸ the findings of our study are considered clinically meaningful for obstetricians, paediatricians, epidemiologists and public health practitioners, particularly those working in the field of injury prevention. These findings highlight the considerable role of the injury prevention community in preventing traffic injuries and developing effective healthcare treatment systems.

Ethics approval

This study was approved by the Institutional Review Board of the National Cheng Kung University Hospital (No. B−EX-109–088).

Data availability

The data underlying this article were provided by the Health and Welfare Data Science Center of the Taiwan Ministry of Health and Welfare by permission. Data cannot be shared publicly due to the current government regulations.

Supplementary data

Supplementary data are available at IJE online.

Author contributions

All authors were involved in the interpretation of the data. Y.H.C., Y.W.C., C.H.C., P.L.C. and C.Y.L. conceived the study methods and design. Y.H.C., T.H.L. and I.L.H. were responsible for data collection, data management and analyses. C.Y.L., I.L.H. and Y.H.C. drafted the initial manuscript, which was reviewed and revised by all authors. All authors read and approved the final manuscript.

Funding

This study was supported by a grant from the Ministry of Science and Technology (MOST 109–2314-B-006 –044 -MY3 and MOST 111–2917-I-006–002). The funder had no role in conducting and submitting this work.

Acknowledgements

We are grateful to Health Stat Science Center, National Cheng Kung University Hospital for providing administrative and technical support. The statistical consultation by Ms. Shang-Chi Lee during the revisions is highly appreciated.

Conflict of interest

None declared.

References

Author notes