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

Prevalence and Risk of Pregnancy-Associated Injury Hospitalizations:
A Population-Based Approach

HAROLD WEISS: I want to acknowledge the supporters of this work, the National Institute of Justice and it’s based itself on methods pioneered by *Greenblat, *Dannonberg, and Johnson and also Karen *Zurry who helped a lot with the data.  The objectives of this study were to estimate the incidence of pregnancy associated injury hospitalization in order to better understand the causes of serious hospitalized pregnancy associated injury and thus fetal threats.  And it was also to get at the interesting and important questions of whether pregnancy increases or decreases injury risks, controlling for age, race, and most importantly, severity.  These are the states that the data was collected from.  These are 19 states; we collected all of the hospital discharge information from these states so we ended up with about half the U.S. population covered in this convenient sample.  There were 19 states, there were 36 million women of reproductive age living in these states and again representing 52 percent of the U.S. age sex group.  They also covered 1.9 million live births in 1997.  In terms of the hospital and patient coverage we got data from over 2,000 hospitals, it included 13 million discharges for all age, sex, and causes, 1.2 million injuries and 175,000 injuries to women ages 15 to 49. 

And a lot of these states were chosen because they had the E code, they had the mechanism of injury in the data and that’s why many other states were excluded because the E code rate wasn’t high enough.  Quickly, the advantages of hospital discharge data they do look at severe injuries, much more than those that are not hospitalized, those that are just seen in an emergency department.  Because we can now get multiple hospital discharge databases around the country, very large population base coverage is available and the data is pretty well standardized across the states.  Also, this captures cases that are missed by surveys.  Have you ever thought about that when you’re doing pregnancy risk assessment often those are surveys that are done by telephone.  If someone is hospitalized at the time, you don’t get them.  And unlike *ED visits, the patient has a longer time to relate the underlying cause.  There’s more time to enter the documentation so it’s probably better and more accurate and more complete than emergency department data.  The methods I’m going to go over quickly. 

There are handouts for this and I’d be happy to send people the *SPSS code we used.  But it was really quite simple to look at pregnancy association we used these *ICD9 codes for complications of pregnancy, childbirth and any of the *ICD9 diagnosis fields.  And we looked for the *V codes for normal pregnancy and high risk and a few others.  And then we looked for a code for concomitant injury and that was defined as either an injury diagnosis in the first three fields or an injury *E code somewhere in the record.  The rate calculations are calculated per 100,000 person years and this requires a little explanation.  In the pregnant population the denominators are derived from the live births adjusted for the nine months gestation and the difficulty detecting pregnancy during the first two months.  So you can see we multiplied times 7/12.  All injured women the denominator was from the population itself.  And then we constructed rate ratios, those are the rates between the pregnant women compared to all injured women ages 15 to 49.  And you need this background to understand the next slide or two.  Our injury selection filter, again from the 18 million cases, 1.2 million injuries when we look just for females of reproductive age we had 176,000 total and 8,266 that had a pregnancy associated code.  We excluded rehab hospitals and we could only include those patients that had an *E code because otherwise we didn’t know the mechanism.  So we were left with, and we also include only residents so that we could compute population rates, so we’re left with almost 5,500 pregnant women in this data. 

This shows the distribution of the proportion of pregnancy associated hospitalized injuries by age.  And note that of all the women then entering into the hospital for an injury in the peak reproductive age group, 8 to 10 percent of them were pregnant.  Some of the selected comparisons between the pregnancy associated cases and all women ages 15 to 49 who were hospitalized for an injury, just pulling out some of the more important things I think, notice that the rate of the pregnancy associated cases is a fair amount higher, the average is younger, the average length of stay though is less, the mean severity is less.  And the median charge per visit is less.  So some things seem to be going on.  The rate is higher but the severity seems to be less.  Let’s look at some of the leading mechanisms of injury by pregnancy status comparing all women 15 to 49 to the pregnancy associated cases.  You can see that poisoning, which is an under appreciated cause of young women hospitalization, fell from 34 percent to 17 percent.  The transportation injuries went from 22 percent to 32 percent.  Falls went from 18 to 22 and struck by, which is often where a lot of the domestic violence and assault cases are coded, went up from four percent to eight percent.  So there is a difference in the hospitalization pattern, injury patterns, by mechanisms for pregnant versus non pregnant women. 

This shows the hospitalization injury comparisons by intent and you can see that the self-inflicted for all women goes down quite a bit for the pregnancy, which is nice to see, the assault however goes up.  The unintentional, I think it goes down a little bit.  So this is the first slide that shows the rate ratios between the pregnant women versus women 15 to 49.  And this shows it by age group.  And this is again a way of seeing whether there’s an increased rate of hospitalization.  Now, this is a rate of hospitalization, not necessarily underlying incidents.  But you can see a very marked age related trend to the rate ratio.  So the increased hospitalization for pregnant women is almost entirely in the very young women and it’s actually below one or insignificantly different from one for the higher age groups.  When you look at the rate ratios by mechanism, what stands out are the struck by and again, and again this is probably where some of the domestic violence is hidden.  You see rate ratios significantly over one for motor vehicle occupants, for firearm related injuries and for falls.  Again, poisoning’s way down during pregnancy. 

And this is another, looking at the rate ratio by intent and you can see that the assault rate ratio is up quite a bit.  Now, this is important because this shows the rate ratios by severity.  And the trend that emerged very clearly was that the rate ratio is higher for mostly minor injuries.  So what we think is happening is pregnant women are much more likely to go into the healthcare system to be hospitalized for observation for the safety of the mom, and the baby, but not necessarily because they’re having more injury.  When you control for severity you actually find that for the more serious cases of injury the pregnant women rate ratio, pregnant to non-pregnant, is actually below one.  So if you’re going to look at this data and begin to get a clear picture you need to adjust for severity.  And when you do that this is sort of the big picture. 

This looks at the rate ratio for the leading hospitalized injury mechanisms it actually has both the rate and the rate ratio on the same axis so you can get an idea of what’s important over here, but what might be higher over on this axis.  This one is controlled for severity, so it takes out of the equation the women who may have just been there for observation.  And what emerges is motor vehicle occupants still has a rate ratio of one, they seem to be at just the same risk.  Fall seems to be at a little lower risk, but still is a very important problem.  Firearms, not very prevalent, but a significantly increased risk for pregnant women.  And this just is to bring home that the rate ratios between blacks and whites don’t change much.  The rates certainly do, the rates are the bars here, so you can see for example for the youngest age group for, this should be whites, it’s about 120, it’s about 80 percent greater for blacks.  But the rate ratio doesn’t change very much between whites and blacks.  It’s a little bit higher in blacks, but it’s still close to one.  

And this is the information just for assaults.  You do see an increased rate ratio in the very young age group for assaults in both blacks and whites.  You don’t see an increased rate ratio in the older age group.  The limitations of this work is that the quality and the completeness of the coding is done at the hospitals and really how good it is, is dependent on each individual state providing the good information.  It fails to identify early pregnancy and postpartum risks and there’s no information on the fetus in this data.  So the pros, if you were to try to go home and use this, I would suggest is that if you try to identify pregnancies just from diagnosis codes, it doesn’t require patient identifiers, you’re going to be able to get approval to do this very quickly.  And it’s easy to do and replicate.  And again I’ll be happy to send you the SPSS codes so you can use the exact same codes we used.  The ascertainment though is not 100 percent.  Some pregnant women will not have a pregnancy related diagnosis.  Maybe if we have time I’ll talk about the extra work that Melissa did for me to try to get it at that.  Also again, it’s not useful for looking at fetal child outcomes. 

So this study demonstrated the feasibility of characterizing and tracking the most dangerous threats to the fetus.  I think it’s a good first start for anybody that wants to get into this area.  It’s the first multi-state study to describe pregnancy associated hospitalization in a mostly *E coded hospital population.  Pregnancy lowers the hospital admission threshold for most injuries, so I like to think of pregnancy as a sensitive population when it comes to injuries.  Once suggested for severity, pregnancy is not related to an increase in overall injury risk.  And it highlights again, like the mortality did, the motor vehicle problem.  It did show a moderate increase in the rate ratio for assaults against young women, even after severity adjustment.  So I think I’ve probably gone a little bit long.  I’m going to self impose a question ban on my self and move ahead to the next presenter.  That’ll be Dr. Schiff.