WHAT IS ALREADY KNOWN ON THIS TOPIC

Motor vehicle crashes (MVCs) are the leading cause of traumatic injury during pregnancy.

The risk of maternal outcomes among pregnant women involved in MVCs remained inconclusive in hospital-based design studies.

WHAT THIS STUDY ADDS

Pregnant women involved in MVCs had significantly higher risks of placental abruption, prolonged uterine contractions, antepartum haemorrhage and caesarean delivery; the increased risks varied with injury severity level and type of vehicle.

HOW THIS STUDY MIGHT AFFECT PRACTICEIntroduction

Motor vehicle crashes (MVCs) are the leading cause of traumatic injury during pregnancy.1 For some 50–70% of pregnant women who had a trauma, it was caused by vehicle collisions.2–4 Injury during pregnancy may result in adverse maternal outcomes, such as splenic rupture, uterine rupture, pelvic fracture, placental abruption and maternal death.5 6 Severe trauma has also been associated with a higher risk of premature delivery or fetal death.7 8 Injured pregnant women have a huge impact on maternal outcomes.

However, limited studies have used population-based nationwide data to investigate the association of trauma with adverse pregnancy outcomes. The majority of previous studies have applied hospital-based data and tended to underestimate the occurrence of MVCs, because pregnant victims who sustained no or mild injuries may not be willing to seek medical care at a hospital for fetal assessment, or even did not know they were pregnant status at the time of the MVCs.5 Also, using hospitalisation data following MVCs may result in an over- or under-representation of severe injuries, because women involved in minor injuries may not present themselves to hospitals. Hospital records might also preclude those MVCs that result in immediate death.9 Thus, there is a potential selection bias for the studies entirely based on the MVC data reported to the hospital.

In addition, the vehicle-type distribution among victims in Asia is different from that in Western countries. In Taiwan, 70% of driver victims were motorcyclists at the time of MVCs and 20% of driver victims were car drivers.10 Compared with Taiwan, a previous study in the USA has shown that pregnant women involved in an MVC were more likely to be involved in a collision with a motor vehicle (88.93%) but less likely with a motorcyclist (0.25%) and bicyclist or pedestrian (4.31%).11

Thus, the potential adverse effect of MVCs in pregnant women requires further research. In contrast to people in Western nations, people in Taiwan use more scooters than cars for transportation, which along with environmental factors and road conditions, may also be associated with the severity of the MVC and type of trauma. This study was performed to measure more comprehensively the overall maternal outcome of MVCs among affected pregnant women by comparison with a control group of pregnant women who had not been involved in an MVC.

MethodsData sources

Four population-based databases in Taiwan were used—namely, the National Birth Notification (BN) Database in 2007–2016, the National Health Insurance (NHI) Database claims in 2007–2016, the Police-reported Traffic Collision Registry in 2006–2016 and the Death Registry (DR) in 2007–2016. The BN database was used to create a cohort of births from pregnant women. Information on the newborns and the mothers, including gestational age, birth weight, birthplace, residential area and age of the mother, are recorded in the BN.12 The NHI comprises various sub-datasets, including the beneficiary registry and inpatient/outpatient medical claims. The beneficiary registry includes the demographic characteristics, insurance premium and personal identification number (PIN) of each individual covered by Taiwan’s NHI programme. Inpatient and outpatient medical claims contain all information related to patients, such as emergency room visit, PIN, date of birth, sex and date of inpatient/outpatient visit, with a maximum of five leading diagnostic codes.13 We used those medical claim data to capture the maternal outcome in our study. For the encounter of MVCs, we linked the data from the Police-reported Traffic Collision Registry by the National Police Agency in Taiwan. After a road traffic crash is reported to the police, certified police investigators examine the crash scene and complete the crash reports, which include information relevant to the MVC.10 Those data are also linked to the DR for vital status, causes and dates of death to ensure the fatality.14 We linked and combined data from each database using the encrypted PIN. The linkage of numerous datasets was done by the first author (Y-HC). All coauthors would review step-by-step the data management and analytical results to assure the accuracy of data linkages and analyses.

Study population and design

This study was a retrospective cohort study which aimed to appraise the maternal outcomes of pregnant women who encountered MVCs in Taiwan. The BN in 2007–2016 covered the information of 2 043 187 newborns and mothers. We excluded records if the mother had an age of less than 18 or above 50 years at the time of delivery (n=6822), delivered multiple births (n=66 207), or had missing PIN value (n=3). The eligible study(analytical) units were 1 970 155 births from women who were 18–50 years old between 2007 and 2016 (online supplemental figure A1).

After linking the BN and Police-reported Traffic Collision Registry from 2006 to 2016, we determined the MVC events based on whether a pregnant woman had collision records during the pregnancy from the Police-reported Traffic Collision Registry. The total number of births from pregnant women exposed to MVCs was 20 844. The date of the MVC was considered as the index date.

The control women were selected by matching exposed women on the maternal age at MVC and length of gestation at MVC, with an MVC/non-MVC ratio of 1:4. In addition, the controls had to be alive on the index date. A total of 83 274 control births from women were successfully selected and the index date for the control group was the same as that of the matched exposed woman (online supplemental figure A1).

Outcome variables

To evaluate the maternal outcomes, we retrieved the diagnoses of interest from the NHI in 2007–2016. The diagnoses were based on the International Classification of Diseases-ninth revision, clinical modification code (ICD-9-CM) or the ICD-10-CM for the hospitalisation, ambulatory and emergency claims prior to the date of delivery.

Maternal outcomes analysed included the following diagnosis in the NHI: hypertonic incoordinate or prolonged uterine contractions (ICD-9-CM code 661.4/ICD-10-CM code O624); premature rupture of membranes (PROM) (ICD-9-CM codes 658.1, 658.2/ICD-10-CM code O42); placental abruption (ICD-9-CM code 641.2/ICD-10-CM diagnosis code: O45); induction of labour (ICD-9-CM codes 659.0, 659.1/ICD-10-CM diagnosis code O61); antepartum haemorrhage (ICD-9-CM code 641/ICD-10-CM code O44) and postpartum haemorrhage (ICD-9-CM code 666/ICD-10-CM code O72). Any of the above diagnostic codes was counted if it appeared once in the medical claims after MVCs between the index date and end of delivery. The other three maternal outcomes were retrieved from the BN—namely, caesarean delivery, gestational age <37 completed weeks and birth weight <2500 g. The maternal death was ascertained by the DR.

Covariates

Several sociodemographic characteristics were considered as covariates, including maternal and gestational age at MVCs, calendar year at MVCs, geographical area of residence, urbanisation of residence, median family-income quartiles and maternal nationality. Gestational age at the time of the MVCs was estimated by the difference (in weeks) between the date of the MVCs and the date of conception. Comprehensive evidence suggests a link between those variables and birth outcomes.15 16 In addition, previous studies also reported an increased risk of traffic injury in people with risky behaviour or unhealthy lifestyle.17 18 Thus, the risky behaviours (including smoking, drug addiction and alcoholism) during pregnancy were also included as covariates in our analysis model.

Injury severity score

The injury severity scale was generated by the software programme ICDPIC from the ICD-9-CM or ICD-10-CM codes for each clinical patient registry.19 For each injured pregnant woman, the ICDPIC programme determined the maximum abbreviated injury scale (MAIS) from the health insurance database within 3 days after the crashes. The MAIS included the three most severely injured forms and seven body regions (ie, head/neck, face, thorax, abdomen, extremity and external). Subjects with a MAIS score ≥3 were considered severely injured and those with MAIS score of 1 or 2 were mildly injured.

Statistical analysis

We compared the characteristics between women with and without involvement of maternal exposure to MVCs by using χ2 tests. In addition to the overall association between maternal exposure to MVCs and various maternal adverse outcomes, we also examined the study outcomes in relation to severity, role of road user (drivers, passengers, pedestrians) and vehicle type involving MVCs. Crude and covariate adjusted odds ratios (aORs) and the corresponding 95% confidence intervals (CIs) of various adverse outcomes in association with MVCs were estimated with the generalised estimation equation method to the logistic regression model to account for matching subsets. Statistical analyses were conducted with SAS (version 9.4; SAS Institute, Cary, North Carolina, USA), and the level of significance was set to α of 0.05.

Results

A total of 20 844 births by 20 739 mothers involved in MVCs during pregnancy were recorded from 2006 to 2016 (20 634 women were involved once and 105 were involved twice or more). Compared with the women were not involved in MVCs during pregnancy, pregnant women with MVCs were more likely to reside in central and southern Taiwan as well as in rural areas. The incidence of MVCs was higher in the most disadvantaged family income group. Pregnant women involved in crashes also tended to engage in risky behaviours during pregnancy (table 1).

Table 1

Comparison of characteristics between pregnant women with and without motor vehicle crashes

Compared with control pregnant women, pregnant women with MVCs were at significantly increased risk of uterine contractions (aOR=1.31, 95% CI 1.11 to 1.53), placental abruption (aOR=1.51, 95% CI 1.30 to 1.74), antepartum haemorrhage (aOR=1.19, 95% CI 1.12 to 1.26) and caesarean delivery (aOR=1.05, 95% CI 1.02 to 1.09) (table 2). There were three and two maternal deaths, respectively, observed in pregnant women with and without MVC. We did not perform the relative risk of maternal death in our study because it was a relatively rare event.

Table 2

Maternal outcomes in relation to MVCs during pregnancy

Pregnant women who had visited emergency room or were hospitalised within 3 days after the MVCs were at increased risk of placental abruption at 1.77 (95% CI 1.46 to 2.15) and 5.89 (95% CI 3.95 to 8.80), respectively. Significantly increased aORs of antepartum haemorrhage, caesarean delivery and low birth weight were also observed among pregnant women with MVCs and emergency room visits as well as in those with MVCs and hospitalisation (online supplemental table A1).

Table 3 compares the risks of maternal outcomes in pregnant women involved in MVCs who have various levels of injury severity, with outcomes for women not involved in MVCs. The results showed that severely injured pregnant women had a 4.01-fold increased risk of placental abruption, 1.61-fold increased risk of caesarean delivery, 1.80-fold increased risk of preterm delivery and 1.84-fold increased risk of low birth weight. Pregnant women with minor injuries caused by MVCs were at significantly increased risk of placental abruption (aOR=1.70), labour induction (1.54), uterine contractions (1.34), antepartum haemorrhage (1.18) and caesarean delivery (1.08) compared with those not involved in an MVC. Pregnant women involved in MVCs without documented injury (MAIS=0 but clinic visit) were also at a marked increased risk of premature rupture of membranes, placental abruption and antepartum haemorrhage (table 3).

Table 3

Maternal outcomes in relation to MVCs during pregnancy with the injury severity level

The analyses were further stratified by the role of road users (ie, driver, passenger and pedestrian) at the crash scene (online supplemental table A2). Compared with women with no record of MVC, pregnant woman involved in MVCs as drivers had significantly higher risks of placental abruption (aOR=1.59, 95% CI 1.35 to 1.86) and other adverse outcomes, including uterine contractions, antepartum haemorrhage and caesarean delivery. Similarly, pregnant woman as passengers had a higher increased risk of uterine contractions (aOR=1.58, 95% CI 1.08 to 2.32) and antepartum haemorrhage (aOR=1.25, 95% CI 1.07 to 1.45) following their crashes. For pregnant pedestrians involved in MVCs, no significant association with adverse outcomes was found in this study (online supplemental table A2).

Table 4 shows that a pregnant woman who was a car driver during the MVC had an increased risk for antepartum haemorrhage (aOR=1.20, 95% CI 1.08 to 1.32) and caesarean delivery (aOR=1.07, 95% CI 1.01 to 1.13). However, these risks did not persist for delivering preterm or low birth weight infants. The woman involved in crashes as a scooter rider had significantly higher risks of placental abruption, labour induction, uterine contractions, antepartum haemorrhage, low birth weight and caesarean delivery, with an estimated aOR ranging from 1.06 (caesarean delivery) to 1.83 (placental abruption).

Table 4

Maternal outcomes in relation to the type of vehicle at MVCs during pregnancy

DiscussionMain findings

In this study, pregnant women involved in MVCs had significantly higher risks of placental abruption, prolonged uterine contractions, antepartum haemorrhage and caesarean delivery than those who were not involved in an MVC during pregnancy. In addition, pregnant women who had clinic visits after crashes and sustained major injuries were more likely to have adverse pregnancies. Overall, the increased risk of placental abruption among scooter riders was higher than that of car drivers. The drivers and passengers had different adverse health risks in this study, and the drivers were apparently more vulnerable to adverse outcomes.

Clinical implications

Our finding was consistent with that of a previous study in which women who were involved in crashes during pregnancy had a higher risk of adverse pregnancy outcomes after the maternal demographic risk factors and risky behaviours were controlled.5 20 21 The results for women at greatest risk of placental abruption from vehicle crash were also consistent with earlier findings, in which women who were hospitalised due to crash injury had higher risks of placental abruption.5 The potential mechanisms of placental abruption caused by MVCs have shown that the shear forces of crash impact or contact with the steering wheel result in strain at the uteroplacental interface that could cause the placenta to separate from the uterus.22 23

Contrary to previous findings, the results in the current study showed that pregnant women involved in crashes were not significantly more at risk for premature delivery than pregnant women not involved in crashes.20 21 The possible reason for this phenomenon might be that we adjusted the gestational age during the MVCs. The previous study did not consider periods of exposure during which crashes occurring at specific periods in the fetal development might be more or less likely to result in particular outcomes. For example, crashes during the late gestational age are more likely to result in delivery than for pregnant women who are in an early gestational age and did not experience MVCs. We adjusted the gestational age at crashes because this age might affect the health outcome and risk of exposure. Therefore, this discrepancy from other studies could be attributed to the non-consideration of the timing of MVCs.20 21

Our findings revealed that pregnant women who received treatment at clinical institutions within 3 days after crashes had an increased risk of adverse pregnancy outcomes. The finding was consistent with those of studies that indicated that such women hospitalised or treated at an emergency department due to injury had a higher risk of adverse outcomes.5 8 24 Relevant literature has shown that pregnancy-induced termination is also associated with an increased risk of haemorrhage among pregnant women.25 However, pregnant women who did not seek clinical treatment had no significant risk of adverse pregnancy outcomes. A possible explanation for this phenomenon might be that such women had no complications and they s assessed themselves that they did not need any extra treatment after the MVCs.26

We found that pregnant women who visited the clinic but had no injury-related diagnosis after MVCs were also at increased risk of adverse pregnancy outcomes, including premature rupture of membranes, placental abruption, antepartum haemorrhage and caesarean delivery, compared with those who were not involved in MVCs. Our findings broadly support the work of a previous study, in which crashes resulting in no injury could increase the risk of adverse pregnancy outcomes.5 Although Schiff and Holt included only pregnant women who were hospitalised following MVCs, the potential for selection bias may exist.5 Furthermore, our findings of adverse outcomes among mildly injured pregnant women involved in crashes indicated that relatively minor trauma had a substantial impact on pregnancy outcomes. These results matched those observed in earlier studies.5 27 28

The driver had a higher risk of uterine contractions, placental abruption, haemorrhage and caesarean delivery compared with pregnant women without crashes. Most studies recruited the occupants or vehicle driver,5 20 21 29 and passenger-specific or pedestrian-specific risk studies are scarce. Our finding showed that the risk for pregnant women among passengers was also higher than for those without crashes. However, the pedestrian showed no significant risk of health outcomes. Previous studies have reported that the rear seat rather than a front seat is safer for people of all ages.30 However, we could not collect information on the passenger seating position.

The results showed that the scooter rider had more adverse health outcomes, such as prolonged uterine contractions, placental abruption, antepartum haemorrhage and caesarean delivery. This study supported evidence from previous reports in which higher proportions of injuries and other repercussions of these events were observed in motorcyclists.31 32 This expected result might have been caused by little protection for riders in a crash. Modern cars are equipped with various safety features, such as airbags, tire pressure monitors and collision warning systems, to reduce the number of fatalities on the road. Scooters are not equipped with many of these safety features. Motorcyclists rely only on helmets or high-visibility clothing to keep them safe.33

Strengths and limitations

The strength of the study was the use of a large, unselected nationwide population-based dataset that provided comprehensive information on the experience among pregnant women. The entire traits of pregnant women during the study period were captured by linking databases. In addition, we defined the MVCs based on the nationwide dataset, which included traffic injury victims who did not seek treatment or who died soon after crashes before receiving treatment in a hospital. Furthermore, many studies have neglected the early phase of fetus development because of the subjects’ unawareness of their pregnant status. In this study, we assessed the exposure from conception date, which is an advantage for the total evaluation of the risk of exposure to avoid underestimating the exposure to a crash.

Limitations of our pregnant women cohort lie in the difficulties of including those pregnant women who delivered in less than 20 weeks. However, no evidence was found that showed a positive relationship between the risk of crash and adverse maternal outcomes among women who delivered in less than 20 weeks. We infer that pregnant women in early gestational age and become involved in crashes would tend to have a preterm birth of less than 20 weeks. Thus, our study results would underestimate the association between women after a crash and adverse outcomes.

In addition, while the linkage rate of the study cohort to the maternal outcomes of interest in this study can be considered high because Taiwan’s NHI programme covers all residents of Taiwan and the law requires all births and death to be registered in 10 days, the linkage of study cohort to the Police-reported Traffic Collision Registry might have incurred certain degrees of error. Although the Police-reported Traffic Collision Registry covers nearly all MVCs in Taiwan, mostly for the purposes of insurances, it is possible that the registry could have under-reported some minor motor vehicle collisions that involved no injury or property damage or loss.34 However, no data are available regarding the magnitude of under-reporting by the Police-reported Traffic Collision Registry.

Also, we used the MAIS as a classified indicator that may be independent of the overall severity, especially for those people with the same AIS scores from different body regions. For instance, the injured victim with AIS 3 of the lower extremity may not indicate the same severity as AIS 3 of the head.35 Moreover, the mapping from ICD coding to MAIS might be subject to classification errors because the ICD codes related to MAIS have not been systematically validated in Taiwan’s NHI claims and also because Taiwan’s NHI claims provided only 3 and 5 leading diagnostic codes for outpatient and inpatient claims, respectively, which could under-code certain relatively minor disease/injury codes and result in measurement errors of MAIS. Although mapping from the ICD to MAIS conversion map is not a perfectly precise conversion tool, it provides a systematic method of assigning the severity for large administrative datasets that only have an ICD-9-CM or ICD-10-CM code available.36 Additionally, we used the diagnosis of ICD-9-CM/ICD-10-CM from the inpatient medical claim to determine maternal outcomes, which could result in disease misclassifications. However, such disease misclassification bias is likely to be non-differential, which tended to underestimate the true associations between maternal MVCs during pregnancy and various maternal outcomes of interest.

Lastly, we did not analyse risks of maternal outcomes for the pregnant women as passengers or pedestrians according to vehicle type, mainly because the number of passengers or pedestrians involving scooter crashes was so limited that reliable estimates of the risk of maternal outcomes in association with crashes with a specific vehicle type were not possible. Failure to separately analysed maternal outcomes for passengers and pedestrians associated with specific vehicle type could overlook the risk of maternal outcomes in certain groups of MVC victims.

Conclusion

The results of this study highlight the several maternal outcomes that would result from MVCs during pregnancy among women. These findings reinforce the fact that healthcare workers should be aware of these effects and consider providing pregnant women with educational materials about road traffic safety and choice of vehicle while travelling during pregnancy. Given the associations observed, a better understanding is needed in future research of the circumstance following crashes during pregnancy in order to develop effective management.

Data availability statement

No data are available. Data management and all analyses were performed onsite at the Health and Welfare Data Science Center of the Taiwan Ministry of Health and Welfare. Data are not available to the public and data sharing is prohibited under the current government regulations.

Ethics statementsPatient consent for publication

Not applicable.