Ninth Annual Maternal and Child Health Epidemiology Conference / December 10-12, 2003

Pregnancy, Injury and Birth Outcomes: Linking Birth/Fetal Death to Hospitalization Data

DR. MELISSA SCHIFF:  Thank you, Hank.  Today I’m going to be presenting some work that I did a couple years ago on pregnancy injury and birth outcomes, linking birth and fetal death certificates to hospitalization data.  And first I wanted to begin by thanking my co investigators at the University of Washington, Victoria Holt as well as Janet Dailing were very helpful in this study.  I work at the Harbor View Injury Prevention and Research Center and that’s where this work was done.  Just to give you a little bit of background, Hank already mentioned some of these things, as I think as you all have heard, injury is one of the leading causes of morbidity and mortality during pregnancy.  And it’s estimated to occur in about one out of 12 injuries. 

Injury mostly in the past has been evaluated during pregnancy in studies of maternal mortality and there really have been few population based studies of adverse outcomes after non fatal injuries occurring during pregnancy.  Some of these studies have evaluated pregnancy outcomes following injury and that’s an area where I’ve been very interested to find what really occurs to women in their pregnancies after they’ve experienced an injury.  Many of these studies have been case series and they’ve come out of either trauma registries or from obstetric services at large or medium sized hospitals.  The case series have included patient numbers from an N of 27 up to 476.  Most of the articles that have been published using these case series have really focused on the appropriate clinical management of these patients. 

So when a pregnant woman is evaluated in an emergency department or by a trauma service, what really should be going on in terms of her clinical management?  Some of them have, some of these cities have also tried to evaluate predictions of pregnancy outcomes but even though the problem affects quite a number of women the outcomes that occur are relatively small, so if you can imagine doing some prediction on fetal deaths that occur just rarely in a population of 100 or 200 women is somewhat difficult to do.  The outcomes that have been accessed in these studies have included maternal and fetal death, placental abruption, as well as preterm labor and delivery.  But at the point in time when I was doing these kind of literature review and evaluations, there hadn’t been any studies at that point comparing pregnant women who were uninjured to pregnant women who were injured and looking at what the pregnancy outcomes were.  And that’s where I decided to start focusing my research attentions. 

So the aims of my study that I’m going to present today were number one, to describe the mechanisms of injury during pregnancy, it’s kind of more of a descriptive piece because we really don’t have good information on population basis of what the mechanisms of injury are.  Some of the case series that discussed already, they’ll focus either on falls or domestic violence or motor vehicle crashes, so we wanted to kind of get a big overall picture of what mechanisms we were talking about.  In addition I wanted to evaluate the risk of adverse maternal and infant outcomes following injury during pregnancy. 

So next I’d like to spend a couple of minutes talking about the methodology that we used in Washington State.  What we did was performed a retrospective cohort study of pregnant women who were hospitalized for injury in Washington between 1989 and 2001.  And we evaluated these women by doing a hospital linkage of hospital discharge data to our birth and fetal death certificates and we did this using personal identifiers such as name, birth date and some residence information.  And we did this using deterministic linkage procedures.  The subjects that we included in our study, we had an exposed and unexposed group.  Our exposed group of women who were hospitalized for injury during pregnancy and we found these in our hospitalization discharge data using *ICD9 codes 800 through 959, or we also were looking for external causation codes, or E codes. 

In terms of the wide range of injury that can be coded in a discharge diagnosis chart, we decided to exclude certain types of injuries such as what are termed medical or surgical misadventures, which is problems that happen during medical care.  We also excluded poisonings, which is one of the areas that Hank did look at in his study.  We also included some foreign bodies and some other types of what are classically termed injuries, but really we weren’t interested in looking at those areas in particular.  When we did this linkage we found that our ICD codes identified about 73 percent of our cases and our *E codes identified about 90 percent.  So if we used just one or the other, you can imagine, we wouldn’t pick up all of the patients that we had in our study.  And then we compared these women to an unexposed or uninjured group and these were randomly chosen pregnant women who had experienced a singleton fetal, singleton live birth or fetal death and had no injury hospitalization and just to make them comparable we also looked at just singleton deliveries among the exposed group, so we didn’t look at multiples.  The pregnancy outcomes that we looked at, we had access not only to the birth and fetal death certificates, but also to the hospitalization discharge information, so we ascertained the pregnancy outcomes using both of these sources.  So we looked at the checked boxes and some information on the fetal death certificates and birth certificates as well as specific pregnancy ICD codes that we evaluated. 

The maternal outcomes that we looked at included preterm labor, which we defined as less than 37 weeks of gestation, placental abruption, labor induction, cesarean delivery and maternal death.  Maternal deaths, just to let you know, we didn’t go into our death certificates, and we ascertained maternal death by status at the time of discharge from the hospitalization.  And then in terms of infant outcomes that we looked at we evaluated whether infants were born prematurely at less than 37 weeks, whether they were low birth weight, whether they had experienced any fetal distress documented in the hospital or the hospital discharge data set as well as fetal death.  Our analyses that we performed are summarized here.  We initially evaluated demographic and obstetric characteristic comparing our uninjured to our injured group.  We evaluated the mechanism of injury just among those who were injured.  We performed *(inaudible) regression analysis to estimate our relative risk and 95 percent competence intervals and we built models for each of the pregnancy outcomes that we looked at.  And we evaluated for many potential confounders but the ones that we wound up adjusting for that affected our relative risks included education, smoking, prenatally, as well as the trimester of initiation of prenatal care.  The next couple of slides just summarize some of our findings. 

Looking at our two different groups the uninjured and the injured group.  So as you note here our uninjured group approximated about, well, included 17,266 women and our injured group included 2,033 women.  And as you can see from the age distribution, the women who were injured, similar to in Hank’s study, were a little bit more likely to be younger, less than 20 years of age.  They were also more likely to have lower education with less than high school education.  We also evaluated marital status and they were more likely to be unmarried, the injured group were.  They were also more likely to smoke and use alcohol prenatally.  And then in terms of parody they were fairly similar.  There may have been some women among the injured group who had higher parody compared to those who were uninjured.  We also evaluated gestational age at delivery and among our uninjured group which reflects, I think, Washington State births, live births, we had about seven percent were delivered at preterm gestational age, about seven percent.  But among our injured group you can see about 11 percent of them were delivered prematurely or below 37 weeks.  And then this last area on our slide we just looked at among our injured group we wanted to see when the trimester of injury occurred, and about 15 percent were injured in the first trimester, 15 percent in the second and about 70 percent in the third trimester. 

This slide goes over the mechanism of injury that we found among our group.  And as you can see it’s similar to Hank’s that he presented earlier, the majority of our subjects had been in a motor vehicle crash, about 33 percent.  About 30 percent had experienced a fall and were hospitalized for that mechanism.  Ten percent had experienced assault and then 26 percent had experienced some other mechanism of injury.  We also looked at intentionality and found that about 12 percent, the injuries were intentional, so the majority was unintentional.  I also wanted to mention among our injured group we also evaluated what happened in terms of their hospitalization for injury with regard to the timing of their delivery and we found that about two-thirds of the women were hospitalized for their injury and then subsequently discharged undelivered and came back at a later date to deliver compared to about a third of women who were injured and then delivered at that injury hospitalization. 

The next couple slides show our results that we found in terms of our outcomes.  For maternal outcomes that we looked at, women who were injured were at an over threefold increased risk of preterm labor.  They were at a fourfold increased risk of placental abruption.  In terms of labor induction they really were not at any increased risk of being induced for their injuries or for other reasons.  And they were at a 30 percent increased risk of cesarean delivery.  I didn’t put any rate ratio calculations up for maternal deaths because our numbers were so small.  Among our over 17,000 uninjured women two women died in that group compared to two women who died among the injured group.  But in terms of just a percentage I think we had .01 percent of our uninjured group died compared to .15 percent among our injured group. 

And then with regard to infant outcomes, women who were injured were at increased risk of preterm delivery about 50 percent, about 40 percent increased risk of low birth weight as well as fetal distress.  And an almost threefold increased risk of fetal death.  Our study did have some limitations.  There likely is some misclassification in terms of the injury mechanisms possibly with an over reporting of falls and an under reporting of assault, part of this due to some social acceptability when you come to the hospital and report well how you got injured, it’s more socially acceptable to say that you fell down as opposed to that you were involved in some domestic violence or other type of assault.  So this likely has happened in our data but it’s really difficult to capture this with using these large data sets. 

We also had some limitations of our hospitalization discharge data for coning for exposures and outcome information because as you all know hospital discharge data may have some errors in it in terms of coding.  It’s not really used for research, it’s more for billing purposes and other things, so coding may not be as accurate as we would like.  But we really try to minimize this especially with our outcome information by using multiple sources.  So we used our hospital discharge data as well as our birth and fetal death certificates.  And one final limitation is that there may have been some differential misclassification of outcomes because women who are injured when they are pregnant undergo more surveillance and kind of get more attention, it’s likely that there’s an increased awareness of potential adverse outcomes among this group and may have resulted in some differential reporting of some more subjective diagnoses, things like placental abruption or fetal distress. 

In conclusion then what we found was that falls and motor vehicle crashes were our most common mechanisms of injury during pregnancy and that injury during pregnancy can result in adverse maternal and infant outcomes and I think this really warrants some recommendations that women who are injured during pregnancy need some careful monitoring in terms of evaluating the pregnancy and also pregnancy outcomes.  Thank you.