Conference Webcasts - 11th Annual Maternal and Child Health Epidemiology Conference - Making Methods and Practice Matter for Women, Children, and Families

Using Geographic Information System (GIS) to Analyze MCH EPI Data

MCHB/EPI Miami Conference — December 7 - 9, 2005

Innovative Research on Teen and Unintended Pregnancy Prevention — Transcript

ROY OMAN : Thank you, Trish. I'm here today to talk about the Youth Asset Study, as Trish mentioned. And I must confess right up front I have too many slides. I'm trying to do too much. So I'm going to--what I really want to talk about today is a little bit unusual. I want to talk more about the study itself and although there's a lot of results already coming out of the study I might not get to those. And so I'll mostly talk about how the study is designed and about its importance and what we intend to find out using this study.

This is a CDC funded project, as Trish mentioned, and investigators in the study, I have some co-investigators, Vesely, Aspy and Tolma. Rodine is one of my community partners. And then Lorrie Gavin is the technical officer on the project. And then Trish also has worked as both fellow, and then, I think she's with the human subjects now for CDC.

All right, so we know, I think most of you have heard this term 'Youth Development' and this is basically a positive way of looking at youth and risk behaviors, and at these things called youth assets, whatever they might be, can help prevent youth from participating in risk behaviors. And we have now quite a bit of evidence showing that the youth who have these assets are less likely to participate in risk behaviors. But all the data to this point has been cross-sectional, so all we have are correlational relationships. And we don't know for sure that the presence or absence of assets cause an increase or decrease in risk behaviors. We can't do that.

So that's really the purpose of this, of the Youth Asset Study, to test this youth asset risk behavior relationship, by following youth over time and seeing how their risk behaviors change and how the presence or absence of assets contribute to these risk behavior changes. And it's actually a very interesting study. There's a lot more to it than that and that's what I want to go over in the next few minutes.

So one thing I want to do is we have to measure assets. If we're going to see how they're related to risk behaviors we need to make sure we're measuring them appropriately. And that, in itself, is an interesting task. We can't have 10, 20, 30 questions measuring each asset. You need a small number of items to measure each asset and they have to have reliability and validity, et cetera. And then we want to not only measure assets and how they're related to risk behaviors, but look at it by youth age, by youth gender by ethnicity. Assets that are important for one group of youth might not be important for others. The measurement issues might change also. There's all kinds of things to think about when you start looking at youth assets. And then of course the last point is to test this assumption that assets are causally related to reduction in teen sex and related risk behaviors.

So the purpose then, again, is to describe the Youth Asset Survey; if I get to it, present some selected baseline results; and then some practical implications. The original study that led to the Youth Asset Survey was initiated as part of a program evaluation effort. And we started with a needs assessment. We did not start with a theory and working as professors and looking for literature. Instead we started with a real grassroots effort where we talked to youth and youth team leaders and had them tell us what they thought was important for them to be successful in life and to grow up and avoid risky behaviors. And so it was really, again, a grassroots approach, starting with the needs assessment, starting with focus groups, and not a top-down approach.

These assets then were the basis for interventions that worked in what was called the HEART of OKC project, Healthy, Empowered and Responsible Teens of OKC. And that was the original study that has now led to the Youth Asset Survey or study. This is a five year longitudinal study where there are six assessment points. And of course we're following these youth and parents over time and so there's a very strong emphasis on keeping them in the study.

We asked the youth a lot of questions. They're summarized here but it's about 200 questions. When we go in, and these are in-home face-to-face interviews, we ask about 120 questions related to youth assets and then 12 items about sexual behaviors that we use as outcomes: contraceptive use about six items; intentions, what are called the happy, wanted and trying to get pregnant questions, about 15 items of males or females; stages of change, or they're loosely based on the stages of change that Lorrie Gavin presented on Wednesday, three sets of items there; and then other ones related to sexual activity and et cetera, education issues, et cetera.

All right, here's the original assets listed here that came from the original study. We measured them again in the current study. We also then, based on literature, we did not go back to youth our second time around. We actually looked at the literature and said what other possible assets might youth have that the literature suggests might be protective risk behavior. So we came out with this additional list. And so we measured all these as well in the current study. But at the time that we submitted our abstract for this conference we did not have the full data in and we did not do the work necessary to come up with a construct. So these new assets are not in today's results.

All right, so the parents. We also asked them about 50 questions related to SES, about 30 questions about their community and also what's called the broken window survey, in which trained observers rate the neighborhood on a number of variables such as the amount of graffiti, the amount of vacant, broken-down residences, et cetera, like that. And so we kind of have an idea of how the neighborhood is, the condition of the neighborhood. Is it very affluent or is it very poor? And how that might relate to assets, which in turn probably relate to risk behaviors. So have multi-level data. We have the parents' perceptions of the neighborhood. We have the broken window survey, which is independent of the parents, by trained observers. We also will bring in census data, whatever we need there. And then we have the youth level data as well.

In wave two we added a few items. We have the social status, these ladder questions indicate, parents indicate where they would be on this ladder and then where they believe others see them on this same ladder. We have the life events, stressful life events, how that might change behavior. And presence or absence of assets. And then we have the abstinence pledge. The abstinence pledge has been in the news a lot, especially when we were developing wave two items. So we went into an extensive line of questioning about abstinent pledges to find out from these youth if they'd taken them. Are they personal or public pledges? Were they written or just oral? And things like that. And we're going to find out over time how they relate to the outcomes and to assets.

All right, so we wanted to get a high number of youth from different socioeconomic status as well as from various races. So what we did is we selected--there are 200. We wanted 200 African-American of higher SES, 200 African-American lower-SES, white high SES, white lower-SES and then Hispanic all levels. And the number at the very far right is actually the number of census tracks that were sampled to get our final sample, which I'll describe in just a minute.

Here's an actual census track. This is a satellite photo. This is Oklahoma City and there's a turnpike there and that kind of a bluish green color, that's probably about a mile and a half of freeway or so. Down to the right is a little local airport. And so we from this census tract collected 188 interviews. It was one of our most productive census tracts. So that'll give you an idea of what a census tract looks like from the air.

It was a lot of work to get this data because when you get the census data it doesn't tell you at the individual house level who is in there and is a business or is it a home with a residency. You have to go in and enumerate every single house and every single census block or tract that we randomly select. If the block had more than 70 households, then we randomly selected 70. So then, like I said, you go in there, you have to, once you identify households, then you have to hopefully complete an interview or you find out that there is a residence but there's no youth between the ages of 12 and 17, et cetera. You can also get a refusal--whatever. You have to go through each one and just keep working until you get a final disposition. And so you can imagine, it's a lot of work first to enumerate each household and then to get an interview or not.

Baseline interviews began in August of '03 and were completed in December '04. We completed 1,120 interviews. We had to throw a few out. We had like one innovative youth who did his survey twice. With 731 refusals. We had an extremely low missing data rate. This is all done on computers and so we had, I think, less than one percent missing data rate, even for the sensitive questions. One hundred sixty-five interviews were done for the parents in Spanish. No youth interviews in Spanish.

So the way we did this, one parent, one teenager from each household randomly interviewed. They're done at the same time, simultaneously, separate rooms, and then when we got to the risk behavior questions we turned the computer over to the youth.

There's four outcomes here that I want to talk about because I knew I wouldn't have time to go through them all. Sexual intercourse, tobacco use, alcohol use, and drug use. And so these four outcomes. But I can tell you right upfront, here's our same assets that we're looking how these assets are related to those four outcomes and we do a logistic regression, controlling for important demographic variables that make a difference. You know, as youth age they're more likely to have had sexual intercourse, et cetera, so you control for age and other factors, 1,111 households as mentioned, here's the mean age: 14.3, 53 percent female. Race ethnicity, we got a good breakdown there, distribution. Not surprisingly, most of the parents were female. We usually randomly choose when you go in there and there's more than two people willing to participate but we always took the father. The father said he'd participate we'd automatically choose the father because we had such low participation rate from fathers. Family structure: 70 versus 30 percent two parents. Here's our youth age. It's a really nice youth age to follow over time, 165 12-year olds, 233 13-year olds, et cetera. If you look at them by have you ever had sexual intercourse, that is our primary outcome. You see that just seven individuals of the 12 year olds had sexual intercourse. And of course there's a longitudinal trend there, yeah, a trend. Not a longitudinal trend but a linear trend. There we go, linear trend. As youth get older they're more likely to participate in sexual intercourse, with about 50 percent of the 17 year olds.

Okay, not surprisingly, these assets are associated with all these outcomes, and I'm going to go through these really quickly. Seven assets, seven of the nine significantly related to never having participated in sexual intercourse with odds ratios 1.5 to 2.22. And here they are, I apologize for not spending more time on them. I've listed only significant assets, only ones that were significantly related. What this indicates is that for example, non-parental role models, youth with a non-parental role model were about 1.5 likely to never have had sexual intercourse compared to youth without that particular asset. There was an interaction on race with other and whites being about four times less likely to have had sexual intercourse with a peer role model compared to youth of the same race.

Tobacco. Eight assets significantly related. There they are. Alcohol. All of you have handouts so you can go through these yourself. Non use of alcohol, seven assets. Here they are, listed again, non-use of drugs, eight assets. So for our four outcomes it was either seven or eight assets related to our outcome.

All right, so this confirms original study results. These are very similar to our first study results so it seems to be strengthening our belief that there is a strong association between assets and risk behaviors. But this longitudinal design will help us to suggest or test the causality relationship between assets and risk behaviors.

Implications, then, and that to this point it does seem that strengthening or increasing the number of assets youth possess may be an important or effective intervention strategy. Some assets seem to go across risk behavior. So if you had to pick certain assets, you have limited resources as practitioners, things like family communication, aspirations for the future, responsible choices, are significantly associated with reduction in all four risk behaviors. So that seems to be if you wanted to concentrate on certain assets, those are the ones you might pick. You know, though, as practitioner the challenge is to develop effective programs. It's not easy to deliver these kinds of programs that are going to make a difference in family communication or develop whatever it takes--aspirations for the future or to help youth make responsible choices. So that's still a real challenge, regardless of whether we find out that this is a causality relationship or not. Thank you.

TRISHA MUELLER: And if anyone has any questions for Dr. Oman , we have about five minutes for questions.

ROY OMAN : Yes?

UNKNOWN SPEAKER: I'm wondering about the lack of role of men answering the survey and if you have any thoughts or comments about that. Is it a concern or you didn't infer anything from that?

ROY OMAN : Well, it is a concern that only, I think it was 19 percent of our sample is--are fathers responding. From a lot of this analysis, all we do with the parent data is look at parent education, parent income, and we can adjust for those factors because we know they're related to the outcomes. So for the data I presented today, for example, it's all, except for those two variables--income and education--it's all asked of the youth. So it really doesn't have much of an effect. But it's not surprising that I think when the phone rings it's more likely that a mother or the female will answer it. Same with the door, when the front doorbell rings that just the female's more likely to participate for whatever reason. But, yeah, like I said, we don't randomly select. If a father will participate we'll take the father.