AMCHP 2006 ANNUAL CONFERENCE
EARLY CHILDHOOD: BUILDING THE FOUNDATION FOR LIFELONG HEALTH
March 4-8, 2006
WILLIAM SAPPENFIELD: Good afternoon. I'm Bill Sappenfield, and I started this project as the MCHFE team leader in CDC and have left CDC now and gone to Florida, but continue to work on the project, and was one of the conveners of the idea. And clearly at the rising infant mortality rate, we were concerned about trying to address it, and felt like the problem may not be one thing nationally, but it may be a series of things going on with states. And so in launching this initiative, we did it sort of in a two-pronged approach. One was we wanted to look at it collectively, and help do things collectively that would help states. And the toolkit that I'm about to talk about is part of that collective piece of what can come out of this that can be shared with others as a starting ground for what's to learn.
The other thing that we spent equally much as time, both of expertise time and meeting time and everything, was to work and consult with each of the five states separately, so that we could address them differently. And so we're going to start out with a toolkit of what we've had in common, and then talk about where each of the other states have gone from that process.
I do want to recognize Greg Alexander. He and I have been in co-lead in developing this toolkit, along with a series of other people in the collaborative. I think of our group Priscilla was one of the persons who was here working on the toolkit, but clearly had representations from both academic and some of the states in trying to pull the toolkit together. Mark, I think you were part of the group, as well. And I want to recognize everybody as we start to try and talk about this.
The toolkit was one of the main products we were trying to focus about for everybody, and it's clearly important to talk about what the purpose is. Clearly, we wanted to do something that was systematic. And by that, as CDC started to work with states on increasing infant mortality rates, we found that each of us came from a fairly different approach. When we talked about it with our academic colleagues, they too were approaching it differently. Which is good, but there was no place that sort of tried to capture all those different ideas, and in sort of a systematic way to look at it. So everyone was approaching it differently. And so our hope was to come up with sort of one approach that would lead to a variety of ideas being covered at one time, and in one process.
It was really targeted for state public health agencies. Population sizes that were large enough that more advanced EPI procedures could be used. And it's important to note we talked about not only increasing infant mortality rate, but also to look at high infant mortality rate. Because clearly there's still geographic disparity across this country. And then the idea is using current knowledge and available information. By that, we're not really able to explain everything about prematurity yet, and so we can't really expect to have a toolkit that can go beyond what we currently understand.
And then two, not every state has every possible data available, so we tried to use an approach that could be done in most states.
The toolkit has sort of a two step process. The first was sort of an overview investigation that we would encourage almost every state to look at with their infant mortality that would be fairly basic, and then upon the themes that would come out from the overview, would allow you to look at a more focused part.
On the overview piece we saw four major areas or four major drivers that states should be exploring. One is maturation. By maturation we mean either birth weight, gestational age, or both. Neither are perfect measures of physiologic, biologic maturation. But we really want to recognize the concept that we were trying to measure, here.
The next was what was the mortality for that maturation, or that birth weight or gestational age, and how that might be different.
Ages and causes of death might also give us a clue as to biologic pathways.
And then the fourth part that we really wanted to build in there, especially as we looked at state experiences, was data reporting. Because we were surprised at the role that data reporting might be having at changing trends in infant mortality, and that many states had not applied any sort of assessment of data reporting.
And then from there we could look at focus investigations into whether it was more maternal attributes, were they more environmental attributes, were they more related to health services, and could use sort of this overview to help us look at the hypotheses. Because as she said, we came up with 30 pages of different hypotheses, some probably more credible than others, but each could be very different in the state depending on what the overview showed you.
What I'm going to talk about today is more of the overview investigation, and some preliminary pieces of that. Because clearly, I covered this in a little over 60, 70 minutes at the MCHFE conference, and I have 15 today.
But where are we headed from this? I put this in here. Greg Alexander, part of this initiative, created a website up at UAB that really can help states look at their infant mortality. Not only does it give framework and references, it actually gives you SAS coding on how to run different analyses and variables. And our hope is to take some of what Greg did in his initial work, and build upon that for the toolkit to go from there. And that this would give a process that would be available to states. Because it's not only important to publish it, but it's important to make it readily available to states.
So let's talk about maturation. Some of the indicators that we're looking at, here were five of them. Percentage preterm under -- that were low birth weight. Percentage preterm that were greater than low birth weight. Percent that were term that were low birth weight. Percentage that were term that were small for gestational age. And percent preterm that were SGA, or small for gestational age.
These are different maturation and growth measures, and different indicators. So in other words, you may not be able to look at it only by birth weight or only by gestational age, and you might need to look at some combination of variables.
This is data from the U.S. using the National Center for Health Statistics files, comparing 1985 to '88, to 1995 and the year 2000. If you look at this, you can see that the percent of preterm low birth weight has actually gone up. The percentage of preterm that are normal birth weight has also gone up. Where the percentage of term low birth weight has actually decreased. The percentage of small for gestational age for term is also decreased. But our preterm SGA is going up.
So you can see that the indicators for all these different measures are not all going in the same direction. And as the state looks at their infant mortality problem, they need to be aware of that.
Not only can do you that, you can also look at it by trying to adjust for different factors. If you take some of the similar categories here, they're not exactly the same, but you can take very low birth weight and low birth weight and small for gestational age and preterm in infant mortality rates, these are unadjusted odds ratio. This compares the earlier time period to the later time period.
If it was one, it means it's exactly the same. If it's greater than one, it means it's increasing. If it's less than one, it means it's decreasing. We are talking about millions of records, so you don't need to worry about P values. Any difference is significant. If you adjust for maternal age, race, marital status, education, parity, number of birth, prenup care utilization, there actually is no change in variable birth weight, looking at the adjusted rates.
There's actually no change in low birth weight. There's no change in small for gestational age. And not really much change at preterm. And mainly a drop in infant mortality rate. What that's trying to imply is that these factors that we've adjusted for might explain some of the national factors of what's going on. So as you look at it among your states, it's going to be very important to look at these range of factors to really understand what might be taking place with maturation.
And again, from a toolkit perspective, we'll try to make it more sequentially as to what you do and how you would do it.
On the maturation specific side, this is the birth weight mortality curve for the nation, same time periods. The yellow is the earlier time period in the '80s, the pink is that of the later time period. This is on a log graph, and what that means is the distance from here is a 10-fold difference, the difference from here is a 10-fold difference, the difference from here is. So that the further up you are, actually the smaller the change really is -- or larger the change really is.
The reason we do that is because mortality at the very low birth weight end is extremely high. Where at the lower end it's extremely small. And if we put it on the same exact curve, you wouldn't see those rates there off of the X axis, they would be almost one and the same. And you'd see huge gaps there.
But the reason analytically to do it is that this means the proportion change across time, if the distance is the same, is relatively the same. Meaning that the decrease in mortality, besides for the under 500-gram range group, roughly has the same distance. So that means that roughly the improvement in mortality is the same across each of the birth weights.
That's important to note nationally as you try to look for change, that it's not birth weight specific mortality that's driving this. And if you did gestational age, and we have that data, it looks essentially the same.
What we're trying to do now is figure out how to take both these maturations and mortality and look at indicators across time that may be easy to look at. And here is the U.S. data. These lines actually line up exactly not only by color, but by ranking to the term. So the red line is a prematurity term -- trend for that of the United States, the yellow is that for small for gestational age, the green is that for low birth weight, which you can see is increasing, and infant mortality rate there decreasing.
So preterm and SGA are going in different directions, that we talked about earlier. Low birth weight and infant mortality are going in slightly opposite directions. The mortality for very low birth weight babies, this is that birth weight specific mortality, this is the group that we think may drive many states' infant mortality and trends. For the nation as a whole it is going down, where for very low birth weight it's going up.
To give you a sample of why this might be useful to states we're going to talk about Hawaii, and Deliana is going to follow up on some of this. But clearly in Hawaii trends, they have the increase in maturity -- or preterm delivery. Their low birth weight is going up, but their infant mortality rate at this time period was going up, and going up fairly consistently. And their very low birth weight rate was going up, but the thing that was really different is their mortality among very low birth weight babies. So smaller babies was actually increasing, and may have a candidate for some of this trend, and something that a state may want to examine.
So this is how those indicators would be useful from a toolkits perspective.
Age and cause of death could provide some very different perspectives. We're now going to use a sample of data here from North Carolina. This compares an earlier time period and a later time period. The earlier in the '90s to the later 2000 time period. The aqua is the earlier, the blue is the later time period.
We show you first age at death across time, or fetal deaths at the far left going across the years of life, and then we give you some combined measures with infant mortality and fetal mortality. The reason this may be useful is one of the questions that comes out is, well, maybe your fetal deaths are now being recorded as infant deaths. Or especially in the lower ranges.
And so you can look at this and see where your fetal death rates are, compare to it your deaths less than a day, because if they're being recorded differently they would switch at those two time periods. But you could look across the time periods, and see what might be affecting by age, and give you some ideas or clues.
The next is to look at it by cause of death, and there are several classification systems out there. And the committee, looking at this, felt uncomfortable deciding which one was exactly best. They are using a modified Dalfus method, it has only nine categories, so it was fewer, so a lot better aggregation. We had some comparability ratios across coding of causes of death for two time periods. It did have a prevention focus and was very practical. And given that North Carolina was leading, was developed by a North Carolina researcher that had more experience, it was the one that the group had decided on. So there's some convenience there.
When you look at it by cause of death, this compares North Carolina to the U.S., you can see that it allows you to look at prematurity issues versus congenital anomalies versus SIDS or related conditions, birth asphyxia, parental infections, injuries, external causes. So it would let you look across at the potential biologic mechanisms that may be explaining, and were your trends improving all the way across the board, or is it an existing situation. Is there a disparity within North Carolina in almost all of the categories.
The last category we don't want to -- we want to equally emphasize is that of data reporting. From a vital events perspective, there are several of them, but there are three that are really critical that the group thought for reporting that every state needs to examine pretty much on an annual basis. One is that of fetal death reporting, live birth reporting, and infant death reporting. And we're trying to come up with measures and approaches where states need to look at it, and I want to give you some examples.
Why could it be an issue, though? There are probably four main themes that we think affect states reporting. One is that you've actually changed your reporting regulation or your process or your training out in the field. And if you do that, your rates may change.
A special example of that is abortion reporting. If you now have enforcement of abortions in certain gestational ages, it may affect your reporting for other fact -- other events.
Reporting of type variability. Clarification of viability. Some states have tried to clarify who should be reported as live birth and who should not, and when you do that it affects your reporting.
And then the biggest thing that may be affecting changes in reporting is quality processing. As vital statistics agencies have less and less funds, they've been less able to do quality assurance in auditing and checking. And because of that, certain adverse effects have taken place.
Let me give you some example of reporting issues. This is fetal deaths. This is percent of all fetal deaths that are between 20 and 27 weeks, and this is how states report it. These are the number of states on the Y axis, and across the X is the percentages of all states.
If you look at this, the mode is at 18 and 45 to 50 percent. But this range varies quite substantially.
Then if you impose on this what the National Survey of Family Growth says this percentage should be, the National Survey of Family Growth says it should be 73 percent. If you take the states, not every state has the same definition of fetal death. If you take those states that report all fetal death events and limit it to 20 weeks or more, they estimate about 58 percent of births should be in this gestational age.
If that's true, that means many of our states are truly underreporting fetal deaths that should be appropriately reported. There's where our participating state showed up.
Why is this important? Let me give you some quick numbers, an assumption. Let's assume that the National Survey for Family Growth is correct and it should be 73 percent. Let's assume your state is reporting it at 35 percent, and that you only have 100 fetal deaths, so you only have 35 fetal deaths in this range. If you work the numbers on the percent and what it should be to reach 73 percent, it means your state missed 143 fetal deaths. I'll say that again. You reported 35, it means you are missing 143 fetal deaths.
Well, do you think your reporting in fetal mortality might change if these started to get to be reported? And you might call it an increase in fetal mortality, and it may be that your system is reporting them well. And there are several states in this category.
Next, let's look at live births. These are the percentages of live births. Again, the number of states are on the Y axis, the percentages of live births are on the X axis. The yellow is for under 500 grams, and the green is 500 to 749. These look fairly symmetrical, look like they're being reported at a random distribution of the two populations.
Now, if I were to tell you this, you would think that these two percentages, if you had a lot of under 500 gram births, you would think you'd have a lot of 500 to 749, won't you? And if you didn't have very many, you'd have thought you'd have fewer.
There is no correlation between these percentages. I'll say that again. There is no correlation between what states report for under 500 grams and between 500 and 749. In fact, states who have some of the largest percentage of low birth weight, especially because of their large percentage of African American births, the southern states, have very few under 500 gram births.
So in fact what it means is who we're reporting for birth weight is radically different. And again, if you start to report your under 500 gram births, who you should be recording, you might think you're having an increase in infant mortality. In fact, you may be finally reporting those births -- not deaths, but the births, who you should have been reporting who ultimately might have been deaths.
Let me give you an infant mortality situation. This is the trend for Louisiana. This is their infant mortality rate going from 1991 through to 2000. The pink line is that for the U.S, and the blue line is that for Louisiana.
And they were very excited when they got to their infant mortality rate in 2000 and talked about all of the rapid improvement they had had, but then they had two years worth of increases. And when they analyzed this, essentially what they found out is they had a remarkable drop in their mortality rate of under 500 gram babies, and it went from about 800 per thousand to about 500 per thousand. Amazing survival capability. And then all of a sudden it got worse again.
And what they figured out is that their vital records people had had a massive budget cut, they dropped their field staff, they did little auditing and checking back, and probably had underreporting of deaths on the under 500 grams babies. And you might think Louisiana is different, but the National Center for Health Statistics contract says that every state in that contract is supposed to be investigating all births under 750 grams to see if they survived by the time they left the hospital.
And what we found out is probably less than half the states are doing that. And if that is true, there are probably many states like Louisiana who might have changes in infant mortality rates that, again, relate to deaths. So as you can see, I've tried to pique your interest that we have a series of indicators and a series of issues that states need to be examining with their infant mortality rate, and each of these might have you look at things differently within your state.
As you -- what you should do next to investigate the problem. And our hope is that this toolkit will be out probably by this time next year, so that you all could be able to take advantage of some of the processes to looking at your high and increasing infant mortality rate.
And with that, I get to introduce Deliana Fuddy, and our partner in crime from the far west. Or east, depending which way you go.