AMCHP 2005 ANNUAL CONFERENCE
DELIVERING RESULTS, IMPROVING PREGNANCY & BIRTH
February 19-23, 2005
KRISTEN HELMS: Thank you, Elizabeth. Good afternoon everyone.
Today I'm going to discuss racial difference and risk status and in preterm delivery among very low risk women.
First I'd like to acknowledge my co-author,
Dr. (Inaudible) Whitehead, and also the PRAMS working group. This consist of one representative from each PRAMS state project, and we have all the participating PRAMS states represented here.
For background information, many geographic, economic and behavioral factor are associated with poor pregnancy outcomes. Black women are twice as likely to deliver preterm compared with white women. Age is an important factor as well. The very young and the older moms are very likely to have preterm deliveries as are women who smoke and women of lower economic statuses.
Racial and ethnic differences in preterm of many risk factors are well documented. A higher percentage of black women than white women give birth when they are teenagers or unmarried. Black and Hispanic women have less education and are more likely to not receive prenatal care in the first trimester. However little information on overall risk status of women giving birth is available and how that risk status may differ by race.
We examined racial and ethnic differences in women's risk status based on pre-pregnancy, demographic, economic and behavioral risk factors. The second part of the analysis assessed racial differences in preterm delivery among very low risk women. We'll determine if racial and ethnic difference in preterm delivery exist among the women who are at very low risk for preterm delivery before pregnancy based on our factors of interest.
We use PRAMS data for the analysis, and PRAMS is a population based surveillance system where participating states sample mothers two to six months after delivery from their state birth certificate files. The women are sampled two to six months after delivery, and higher risk women are over sampled and the stratification design does vary by stage.
The sample sizes run from 1600 to about 3,000 annually.
Data clerk for PRAMS is based on Don Dillman's Taylor Design Method. We have a self-administered questionnaire that goes out up to three times, and non-responses are followed up by telephone. The data are weighted for sampling design non-response and non-coverage of the frame file. That makes the data for PRAMS generalizable to state residence who delivered live births within the State.
These are the State's -- actually these are all states. Currently there are 31 states and New York City that participate in PRAMS, and for this analysis, we used data for 19 states, and those are the States represented here on this map.
We used PRAMS data for singleton births occurring between 1988 and 2000. In a State PRAMS (inaudible) and a state had a 70 percent response rate and collected data on maternal income for that year. The PRAMS questionnaire has a set of core questions all states ask and then a portion of the survey that states can choose questions even from a standard bank that CEC as available for States to develop or they can use their own. And states using income questions were those developed by the State. So only those that had those questions were included. And states included anywhere from one to 12 years of their data in this analysis. We did use Sudan to calculate the population prevalence
Now, low risk women were defined as those who met the following criteria: 18 to 34 years old, at least 12 years of education, they were married and prenatal care in the first trimester, four or less previous live births and those who had an interpregnancy interval greater than six months. All of those variables came from the birth certificate.
Very low risk women who met the above criteria, but also met the following from PRAMS: They were a non-smoker in the three months before pregnancy, they consumed three or less drinks per week, alcoholic drinks, no income from public assistance. They were at median or above-median income for the State that year, and had a pre-pregnancy BMI of 19.8 to 326.
This table shows lots of numbers. The represented population, total number sampled, and total number of respondent's to the PRAMS survey by race and ethnicity. The response rates were relatively high. We saw responses from 74 to 87 percent among women who were low risk using the birth certificate information only. In very low risk women, we would image that it would be even higher.
Overall, we have a 75 percent response rate and again, 85 percent low risk. We had to exclude 2,158 women because the race variable was missing, so that left us a total of 13,502 very low risk women and it went down by race and ethnicity as shown here. Very low risk was defined from birth certificate information and from PRAMS
Okay. This is kind of a busy slide here but I wanted to show the distribution of risk status by states. And this again is for the 19 PRAMS states we used. The green bar represents very low risk. The yellow bar, low risk, and the red bar represents those women at higher risk who didn't meet either of the birth certificate or PRAMS criteria.
The distribution does vary greatly by state. Prevalence of very low risk women range from 5.3 percent in Utah to 13.2 percent in Ohio. And the prevalence of higher risk women ranged from 50.0 percent in Maine to 75.3 in New Mexico.
These four graphs represent or present the risk status by the race and ethnicity. 9.6 percent of women who delivered a live birth in our sample were a very low risk. And the proportion of women who were low or very low risk did vary considerably among the different breaks we looked at. Greatest variance was seen among the very low risk women. 12 percent among Asian or Pacific Islanders and around 2 percent for the Native American or black women. Non-Hispanic were twice as likely than Hispanic to be classified as low risk.
Now again, a lot of data represented here, but this graph shows the determination of risk status and the exclusions from the very low risk classification of race and ethnicity. And the way it's listed here, age through body size, is the way that we did evaluate them. That's the order assessed.
There are a few things I would like to point out from this graph. The most common reasons for not being classified as low risk were marital status, 17 percent; maternal age, also 17 percent; maternal education, 16 percent; and income, 12. And that's overall, but when you look at it by race and ethnicity, there was some variation. The most common reason women were not classified as row risk was paternal age for non-Hispanic, white and Asian women, education for Hispanic women, and marital status among black and Native American women, 39 percent for black and 25 percent for the Native American women.
So who is low risk? Less than 10 percent of the women who had a live birth met the classification for very low risk. The proportion of women who were very low risk differed by race and ethnicity ranging from 2 percent for Native Americans to 12 percent for Asian, Pacific Islanders. And again, common reasons are listed here: Marital status, age, education and income.
Now, the primary reason for exclusion varied. For white non-Hispanic women it was age; Hispanic women, education; black and Native American women it was marital status; and Asian, Pacific Islanders, it was income.
Now we are going to focus in on our selected outcome for the analysis and that was preterm delivery. These graphs show the preference of preterm delivery of all women by race and by ethnicity. The prevalence varies from 8 percent in white women to 18.4 percent in African American women. And the percent preterm among Hispanic is just slightly higher than non-Hispanic women.
This graph those the prevalence of pre-term delivery by risk status. And as expected, women who did not meet the criteria for very low risk were more likely to deliver preterm than very low risk women for every group. The prevalence of preterm delivery varied by race and ethnicity even among the very lowest group, however. So among that group, the very lowest women, Native American women had the lowest prevalence of pre-term delivery and that is 3.4, while black women have the highest prevalence at 9.1. Women who were non-Hispanic had a slightly higher risk of pre-term delivery than Hispanic women, but that difference was not (inaudible).
We also looked at the distribution of completed weeks of gestational age among the very low risk women. So as you can see here represented by the different colors, Hispanic women had the highest portion of term births followed by Asian Pacific Islanders, Native American, white and then black women. You can see a hint of two secondary peaks. There's one at 33 and one and 29, just a hint. And so we magnified it, and you can see at the lowest gestational ages, there's a definite peak at 33 and one at 29. The peak at 33 was for all race groups, while the one at 29 weeks is limited to the black mothers.
And we have a further analysis schedule. We want to explore these differences in distribution using statistical techniques that test for the shape and the slope, we are going to use the Sloan Analysis for that and we will start looking at the results and it should be available fairly soon.
So for this analysis we see the prevalence of preterm delivery still exists or still differs by race in the very low risk women. The pattern of racial differences preterm delivery was not consistent among those two groups.
Lastly very low risk women were less likely to have a preterm birth than higher risk women in each of the racial and ethnic groups.
And I also need to mention some limitations for the analysis. While there are many strengths, we do realize that we were working with some limitations. We do not have data on several important risk factors. For example, chronic pre-pregnancy health problems we weren't able to look at, maternal drug use, nutrition and family history. We also had to do with a lot of missing information, specifically on income and pre-pregnancy BMI. This could have tarnished the results a bit. We could have an under-estimation of the proportion of the women who were very low risk or an over estimation of preterm delivery for some of the racial groups.
Now, as I said earlier about the income questions, they do vary by state because they were developed by PRAMS state and not all PRAMS state choose to include questions about income, so the ones that did, we had to do some work to include and look at this variable in the analysis. The questions could be collecting information on household income or family income. The time period, a reference period differed substantially.
Most states did use categorical data, so they had response options after they asked about income and women could check the different categories that they fell into. So that was challenging when we compared it to the US census bureau median household income level for each state, so some misclassification could have occurred there.
If you're interested, I'm not going to present these, but the last two slides in your handout provide some more information on the assessment of bias. We did some analyses just to show how the bias from non-response and missing information affected the analysis.
There are several points to take away from the analysis. Again, very few women who gave birth in our study were actually classified as very low risk at conception. The risk status and risk factors varied greatly by difference in race and ethnicity. And factors we examined in determining risk factors, they don't explain the increased risk of pre-term delivery among black women. They may explain the increased risk among Native American women, but again, we can't really be sure.
So these differences should be considered in research on racial disparities and pregnancy outcome, particularly preterm delivery and of claiming interventions.
We have better education of the implications of having a baby under less than ideal circumstances; therefore, an implication of preconception care. Identifying those women in less than ideal or with less than ideal risk profiles allows the health provider to A, help those who don't want to become pregnant not to become pregnant, and two to help those women who do want to become pregnant to improve their circumstances and their health if possible.
We, again, are taking this study further. We are going to do the spine analysis to look at gestational age. We will explore racial and ethnic differences in maturation. And we are also going to look at a relationship between race and preterm delivery among primiparas women and secondly women who have had previous preterm delivery.
And that concludes my presentation. Thank you.