ANDREA MACKAY: Thank you. Can you hear me? Thank you. First of all, I’d like to acknowledge my co-authors listed here, and before I actually begin my own presentation, I wanted to give you just a brief background on maternal health in the United States – maternal mortality. As you can see, over the last century maternal mortality rates dropped significantly in the United States. In spite of that, maternal deaths continue to occur as Susan Eluded to and are a serious public health concern. I’m also going to give you a little background from the most recent MMWR surveillance summary report on pregnancy-related mortality from the division or reproductive health, and this was not my work, it was the work of some of the people in this room, but I wanted to share their latest findings with you.
The latest causes of pregnancy-related deaths from 91 through 99 were embolism, hemorrhage, hypertensive disorders of pregnancy and infection, and they have found some increases in deaths from cardiomyopathy, the distribution of pregnancy-related deaths by the outcome of pregnancy. The majority have a live birth outcome. The next largest group is unknown outcome of pregnancy, ten percent are undelivered and then the other areas you can see are somewhat smaller. These show the ratios deaths per 100,000 live births for white, black and all races, and as you can see, black is in pink and white is in green along the bottom, there is a three to four fold disparity by age between white and black women in pregnancy-related mortality.
All right. Now moving on to my own presentation. An assessment of maternal mortality in the United States 1995 to 1997. While women’s risk of dying from pregnancy complications has decreased dramatically over the past century, pregnancy mortality remains an important public health indicator. Nationally, two systems collect information on the number of deaths and the characteristics of the women who died. The National Vital Statistics System or NVSS at the National Center for Health Statistics at CDC and the Pregnancy-Related Mortality Surveillance System PMSS, through the Division of Reproductive Health at CDC. And over the past 20 years, these two systems have had a continuing disparity in the ratios they have reported, particularly since 1986 with the beginning of prospective data collection. Prior to that, it was retrospective back to 1979. Our objectives were to compare maternal deaths in NVSS, which I will refer to from here on out as the Vital Statistics System or Vital Statistics and Pregnancy-Related Deaths in PMSS or I will refer to that as the surveillance system, so that I don’t get all caught up in this alphabet soup and as Susan said before, remember that maternal deaths means vital statistics reporting pregnancy-related deaths or through the surveillance system. Those are those code words.
So our objectives were determine the nature and magnitude of reporting differences and to identify possible disparities in reporting by age, race, region and cause of death, and then to determine a combined pregnancy mortality ration based on information from both systems. Okay, what are these two systems? Vital Statistics or NVSS is the National Vital Registration System that’s been in place since the early 1900’s. It follows the very strict WHO guidelines for coding and deaths are coded by MICAR according to an algorithm. The rates are reported as maternal mortality rates. PMSS or the surveillance system is a voluntary surveillance system. It was initiated in 1986. It is a joint effort between CDC and the American College of Obstetrician Gynecologist. Deaths are reviewed and classified by medical epidemiologist based on deaths certificates and other information and they report pregnancy-related mortality ratios. One of the differences in these systems are the temporal boundaries.
Vital statistics reports a maternal death as occurring to a woman who is pregnant or within 42 days of pregnancy, where as the surveillance system reports the death of a woman if she was pregnant or within one year of pregnancy. And how did they ascertain these deaths? Determining the cause of death. Vital statistics uses only the information and the underlying cause of deaths on the death certificates – parts one and two. There’s three lines in part one and then additional lines in part two. No other information on the death certificate can be used. They use ICD-9 codes 630 to 676 to identify a maternal death and there are pregnancy check boxes on some death certificates and that information may be used as well if it’s a marked checkbox. The surveillance system in contrast uses the information from the death certificate in the underlying cause of death. They also use any and all other notes and information on the death certificate, notes written in the margin down in the bottom anywhere. They used matched live birth and fetal death certificates. This gives a temporal relationship, if there is a live birth certificate for that woman’s delivery. They use autopsy reports when available and maternal mortality review committee reports when available and the information in the pregnancy checkbox. Our methods.
The surveillance system received copies of death certificates for women of reproductive age from the 50 states, D.C. and New York City, and they used all available information to classify deaths if pregnancy‑related. We obtained the underlying cause of death or the ICD-9 codes for all pregnancy-related deaths in the surveillance system for the three years study period from the vital statistics system by the death certificate number. All pregnancy-related deaths assigned in ICD-9 code 630 to 676 by vital statistics were considered maternal deaths in vital statistics and were also pregnancy-related deaths in the surveillance system. So we called those both PMSS and NVSS. Those that were coded outside the pregnancy range by vital statistics were not considered maternal deaths, but they were pregnancy-related deaths. So we called PMSS only, and finally, the mortality files on a CDC wonder system were used to ascertain information on maternal deaths that were only in the vital statistics system and to calculate those deaths reported only in vital statistics.
The deaths in these three groups were analyzed by age group race, region and cause of death and the time interval between pregnancy and death. Chi square tests for statistical significance were performed and we calculated a combined pregnancy mortality ration based on our findings. All right. This pie shows how these deaths lined up. In the three year period, there 1,471 deaths due to pregnancy reported through one or both systems. About 54 percent were reported through both systems, the 795 deaths, 40 deaths were reported only in the surveillance system and 84 deaths, only in vital statistics, and how that occurred is that those 84 death certificates were never provided to the surveillance system, so the system did not have an opportunity to evaluate them and include them. Looking at first race on the left side, the yellow bars, Asian woman were much more likely to be coded as maternal deaths.
This is proportion of pregnancy-related deaths that were also picked up by the vital statistics system as maternal deaths, and about 80 percent of deaths among Asian woman, were coded as maternal deaths as compared to 54 to 59 percent among black and white women. Likewise by region, it was statistically significant deaths in the western region of the United States were more likely to be coded as maternal deaths, and we stratified race by region and these findings were consistent across all races and regions. Looking at causes of death, remembering that these bars represent, 100 percent represents all deaths in the pregnancy surveillance system and the colored bars represent what portion of those were actually identified in vital statistics. Embolism, hemorrhage and PIH were most frequently identified through vital statistics as maternal deaths. Deaths from infection, cardiomyopathy, cerebral vascular accidents and other medical conditions were much less likely, and one of the reasons we found for this is that the ICD-9 codes in the system have very specific causes for things like hemorrhage. You might have a placenta previa, placenta abruption, which would be hemorrhage, but there are not any specific ICD-9 codes in the pregnancy section for CVA’s, for example, or for other medical conditions, which affected a pregnancy.
Now this slide takes a little kind of, yet, get your head in it with me. This represents how all the deaths in the pregnancy‑related mortality system were identified in the vital statistics system. The blue bars represent all the deaths that both systems identified as pregnancy-related and maternal deaths. The green bars are the pregnancy-related deaths that were coded with cause group similar to the one in the surveillance system, but outside that pregnancy range 630 to 676, and as an example, let say a woman had a placenta abruption, which is a hemorrhage and it was listed on her death certificate simply as hemorrhage with no indication of pregnancy. Well, it was the same cause group, hemorrhage, but it wasn’t coded as a maternal death. Another example would be septicemia versus peritonitis. Peritonitis is listed in the 630 to 676 range. Septicemia is a general category in infection. So these would’ve been coded in the green range. They had the right idea. They just missed the pregnancy relationship because it wasn’t indicated on the death certificate. The ones in yellow had a completely different cause of death. The surveillance system may have identified embolism. Vital statistics may have coded it as hemorrhage and it is not clear how this can occur.
One of the considerations that has been suggested for the difference in this disparity was the different time interval. Remember I said 42 days versus one year. However, if you look in the right hand side, the last column. The number in blue 15.9 or 16 percent of the deaths that were picked up by the surveillance system only were beyond the limit. The other 65 percent, the first two columns were within the temporal limits and therefore, that was not a reason for them to have been excluded. The mortality ratios were 7.7 for the vital statistic system, 11.9 for the pregnancy-related mortality system and combining the deaths in both systems, we came up with a 12.6 combined mortality ratio. ICD-10 was implemented beginning in 1999 and there are changes in coding and selection rules, which should increase the number and in fact, did in the first year of 1999, increased the number of maternal deaths by 39 percent. The addition of the pregnancy status question on the revised U.S. standard certificate of death should also help improve maternal mortality reporting through vital statistics. I’m not going to repeat my conclusions.
I would rather take my last 30 seconds to say, why are so many of these deaths not picked up by vital statistics? Vital statistics often gets a bad rap. It’s not their fault basically. What it comes down to is that death certificates are often improperly completed. General cause of death rather than a specific cause of death and lack of information about the pregnancy relationship, as I indicated in some of the examples, and finally maternal deaths, pregnancy-related deaths in this country are still very rare. As bad as it is that they happen at all, they are rare. The clinitions who provide care for pregnant women have little experience having to fill out death certificates and therefore, this not a routine experience for them, and so that contributes to the lack of incorrect information on those underlying cause of death lines and that prevents vital statistics from picking up what they need to know to code these correctly. So the first line of improvement is education of clinitions and certifiers of death to properly complete death certificates. Thank you.