[Voiceover] National Data Archive on Child Abuse and Neglect
[Erin McCauley]
all right everyone it is noon so we'll start us off although I assume usually we get a little bit of a burst of people coming in in the next few minutes but out get us started in respect of time. So this is the last presentation of the our series for the Summer of NYTD. Again this series was hosted by the national data archive on child abuse and neglect in the Bronfenbrenner Center for Translational research at Cornell University. This has been our summer series so this is actually our seventh and final presentation. I know you guys are probably bored of hearing me congratulations on completing the series. For people who missed any as I said the videos will be posted toward the end of the month. And so without further ado I will let our esteemed guest take it over this is going to be our research presentation 2 focusing on published research and then some tips for the trade from a very, very famous and respected user of our data which we are very lucky to have here today Svetlana. Want to take away?
[Svetlana Shpiegel]
thanks Erin. So hi everybody, my name is Svetlana Shpiegel I'm an associate professor at Montclair State University in New Jersey and I've been doing research on youth's aging out of the foster care system for some time now. And I've been using NYTD for the past I would say three or four years, yeah I started in about 2015 so about 3 1/2 years now. I have to say I am a big proponent of NYTD I'm hoping to see more researchers and also child welfare administrators really use this data set and I have published with this data set in several journals. I published in The Journal of Adolescent Health, I published in Children and Youth Services Review, in Journal of Child Welfare, so really in a couple of journals that are pretty well-established in the field so NYTD data is publishable and that's really what I want to talk to you guys about and this is my key message here. So I want to talk a little bit about but I want to start this off talking about the advantages of using NYTD I came to use this data set really for two reasons. I was looking for a large data set on youth who age out of the foster care system and that was pretty difficult task back a couple of years ago when I started it because I also wanted to data set that wasn't sort of used to death right? Okay I wanted something that has the information on my target population but not something that everybody has used multiple times so I've explored several different data sets, have worked with McMillan's data I worked with the Chafee data but these are not national and they had some limitations for me especially because I was interested in some groups that were generally smaller groups so in smaller databases I couldn't get the numbers that I needed. So that really brings me to kind of using NYTD and that's how I got there I was able to see that this is a large national data set which had advantages for me especially because of the size and also at that point when I started using NYTD and even today the data set has been really then we used. It almost hasn't been used I think my 2015 article that I'm going to talk to about today are pretty sure it was either the first or one of the first articles that was published with this data. So that at some advantages for me and I'll discuss that a little bit later on. There are some other really good things about NYTD that I think sort of stressing that helped me to publish this data. So this data can be combined with other data sets. I've combined it with AFCARS but it could also be combined with NCANDS so really when you combine this data with other data that is something that gives a little bit more information the NYTD by itself could provide. The biggest advantage for me was that it includes large enough samples for analysis of groups that are generally small. So I in particular was interested in teen parents, males and females who had children either earlier in adolescence or as they transitioned to adulthood and in the other data sets that I've looked at there really weren't enough kids to do meaningful analysis. So I could see 15, 20, 30 of them but generally it wasn't big enough for me to look at especially if I also wanted to look at males and not just females and some other things. So that's a really good and significant advantage of these data is that you could look at that. And teen parents is just one group but there are other groups, and I'll reference those later on, that you could look at that in other data sets it's really difficult to find. The other thing that's good about NYTD is you could connect service data to outcome data so for those of us who are interested in how Chafee funded services might or might not relate to outcomes later on, this data set provides the ability to do that with some limitations but everything has limitations in life. So that wasn't a significant stopping point necessarily for me. This data is also longitudinal because every cohort is followed up so there are three waves of each cohort basically age 17, age 19 and 21, so that's also an advantage for those of us who are looking at questions that lend themselves to longitudinal analysis. And also I personally believe that this data set is useful for policy research. I think I've just seen I can't remember if this was a published article or a presentation, that looked at how state policies in terms of tuition waivers and ETB funds I believe, how it relates to variations in outcomes in terms of post secondary educational attainment so those kinds of stuff could be done with NYTD and not necessarily could be done with other data sets. But as I've said before there are also some challenges of using NYTD and I have over the years trying to publish this data encountered many of these. It's worth to note and it that distinction really needs to be made in all the publications using this data is that it's national but it's not nationally representative so those of you who are familiar with the data set you know that essentially youth self select to participate in NYTD. They are trying to fill out as many as they can but it's not a nationally representative data set. There is also the added complication that response rate really varies across states so there are some states with fantastic response rate and there are other states where response rate is actually pretty low. And the first cohort of NYTD was established in 2011 has more of that limitation than the second. In the second one the response rates are a little bit better but still there's still a lot of attrition between the baseline and in the follow-up later on and in even in the baseline there is certainly some variation in how many Jews the states are able to survey. The other difficulty I found with NYTD in terms of using the service data. So again I'm sure this has been gone over in the previous part of this seminar series but NYTD has to separate components the outcome data in the service data so the service data sometimes is difficult to work with because there's some differences by states and how they define certain services and also how they entered data and even though there's some guidance in terms of how to define services and all that kind of stuff but still the variation persists and it's really difficult to know how many services youth received and what did this actually mean. The other thing that reviewers often pointed out is that outcome data is pretty basic, right? It's mostly yes no questions and it doesn't necessarily include some important detail. So specifically for childbirth for instance there is a question there do you have children, right? But there isn't any question about how many children do you have? Do you currently parent your children? Are they placed with you or are they placed somewhere else? All of this is not accounted for and that obviously presents some limitation in terms of wanting to work with this data in a meaningful way. There's some missing data that even when youth respond they sometimes don't respond to certain questions even though I found that to be a little bit less of an issue than attrition and differences in response rate. But really I would say that the biggest challenge especially in the early years when I tried to publish this data was that reviewers and people in my field who do this work on aging out of youth were not familiar with this data set and they specifically said "well I don't know if I can trust this administrative data and what it means." So all of this is something that I have devoted thought to and really tried to respond in meaningful ways every time I tried to publish with this data so I'm going to talk about some of this and how I approached overcoming this. So to do that I want to give some examples of research that I have published with NYTD and I've chosen to focus all the studies that I present here today on this sort of adolescent parenthood topic because I think there those three studies that I'm going to talk about are pretty well connected to one another and I think they give they are in different journals and they give a good overview of kind of the publication process and what were some of the challenges there. So the first study I really want to talk about is that first initial study that I published with NYTD and that study is on adolescent parents in the first wave of the National Youth in Transition Database. I'm going to give some information about the study but all of the studies are published so I'm not going to in great depth about what I've done in each of them feel free to look it up. But I do want to mention some key things that I wanted to do. So with this particular study this was sort of the beginning of my interest in adolescent parents who are aging out of care and I wanted to see and there's some research on this topic even by 2015 there is more now but even at that point in 2015 that was already some research available on this issue but the research that was available had some substantial limitations. So some studies on this topic were small and qualitative, right, so they were very helpful in terms of trying to understand kind of how youth experience parenthood and what are the challenges that they experience when they become parents but they really couldn't give any meaningful information about rates and some of these kind of larger questions, risk factors and things like that. There were also some other studies and other sort of just data out there from the Midwest study and from some others but those were all states specific or included a couple of states and there was really nothing national at this point about how many foster youth are actually parents in adolescents. So that drew me to NYTD and I figured that this data set, even though it has some limitations in terms of response rate but it still gives a national perspective on this topic and that was attractive to me. And also this other big limitation that I found in the literature on this topic is that there is really nothing almost nothing about males. So data about male parents are really infrequent it was included in passing in some articles or reports but there really wasn't any systematic look at this issue of male parenthood and one of the reasons for that is because most of the other data sets I looked at didn't include enough males that are parents to do any meaningful kinds of analysis. So again with that in mind the goals of this particular study was to look at the number of males and females who had children by age 17 so, how many? How many youth are having kids? Right and this was an important question for me. I also wanted to, me and my co-author, we wanted to examine some of the differences between male and female parents on some indicators that the outcome component of NYTD has so we wanted to look at things like homelessness and incarceration and substance use and school enrollment, right those are important indicators that we wanted to look at and this is something that we were able to do. And then we wanted to do this sort of simple fairly simple multivariate analysis trying to figure out what factors are associated with teen parenthood for males and for females. So we wanted to see if there are different factors if they are similar and just in general what's important. So in terms of methods for this study I used only the baseline data for the 2011 cohort of NYTD so age 17 data only at that point the other data was not yet available and we really only focused on this sort of baseline age 17 data. And we've done some bivariate analysis and some logistic regression analysis to answer the questions that I've talked about before. So what did we find we had some interesting findings there. We first of all we found in terms of rates about 10% of females at about 4% of males reported that they had children at age 17. As I previously said NYTD doesn't include information on whether or not they are actively parenting their children or even how many kids they have at this point. We don't know that information based on NYTD but we do know that 10% of females and 4% of males at that point were parents. We also found that there weren't necessarily that many differences between the mothers and fathers on functioning indicators or service use there's some exceptions to that that I could talk about but generally they were pretty similar. And our multivariate analysis which is really what I've included here we found that sort of the usual suspects related to parenthood with some differences though between males and females. So as you can see the similarities are there such that race nonwhite race and Hispanic ethnicity was associated with higher likelihood of parenthood for males and for females. That is not surprising we know that from research on the general population and we also know that from prior research in foster youth. We also found that youth who were not enrolled in school had less likelihood of becoming parents, right and remember this is all cross-sectional so at this point these are just associations there is nothing we could talk even try to hypothesize about causation but nevertheless there was this association between school enrollment and decreased likelihood of childbirth and that was sort of present for males and females alike. However in terms of this risk behavior so we've looked at homelessness, substance abuse referrals and incarceration and what we found is that for females having been incarcerated and this is all lifetime in baseline data so those are all sort of lifetime homelessness, lifetime substance abuse or lifetime incarceration, so we have found that females who have been incarcerated were more likely to have children but homelessness and substance abuse referrals did not matter for females. However for males all three of these were associated with higher likelihood of childbirth or parenthood I should say for males and also the odds ratios were much higher in general the whole model was a better fit for males that it was for females to this really led us to conclude for girls there are probably other unexamined factors that are more important in terms of their relation relationship with childbirth than the ones that we have included. In terms of just kind of some of the publication challenges I would say that the main challenge with this particular article was that it was as I said before either the first or one of the first articles that was published with NYTD so reviewers were not familiar with it at all they were asking well what is this? We haven't heard of it. What does it mean? And all of these sort of questions that needed a lot of explanation and response on our part to make sure that reviewers are convinced that this is actually valid data. They were also concerned about response rates and generalizability so response rates for the first NYTD for the baseline cohort of 2011 NYTD was about 53%, 50+ percent baby was 53 1/2, 54 but in that range and that was obviously a concern for some reviewers that the response rate is really low and how does that potentially influence the study findings. And then the other thing that they brought up and they continue to bring up even in subsequent studies as well is this issue of lack of detail in key variables. So they were particularly concerned about not knowing more detail about childbirth, right so not knowing how many kids they had, when exactly they had children. Before age 17 could be 16 1/2 or it could be 14 right so none of that data is available and the we really made sure to provide as much detail as possible about the data set and really kind of point out each of the concerns and balance it against some of the benefits of the data set. So we emphasized strengths and we said you know the data set is not perfect however it doesn't mean it can't be used and in fact it provides this national perspective that really none of the other data sets have. So that was a big selling point right there isn't better national that we are not using right. NSCAW has a little bit especially the second NSCAW that goes a little bit later in terms of ages has some information on on youth who would be kind of overlapping with this aging out population but it's small the analysis for some of these variables can't be done in that same way so it really didn't it wasn't a competition here right. We also stressed that yes even though response rates were only about 50+ percent it's actually not that dissimilar from other high risk populations right so we look at population such as homeless youth or youth with histories of delinquency or mental health issues. Some of the response rates tend to be kind of similar right so that and we provided some citations for that and that was helpful for our reviewers to really have that context. We have spent a lot of time saying look the findings that were seeing here are pretty similar to findings of other research both on foster youth and also on the general population so we basically said he know there was a study a specific study in Maryland that was done and the rates were also about 9% there at similar ages we, the Midwest study found that at age 17 males had about 7% of reporting having children so that was sort of similar to what we've has, so we've emphasized those numbers to people so emphasized that prior research found that nonwhite race and Hispanic ethnicity is associated with higher childbirth and in both males and females so we've stressed that to kind of showed the reviewers look you don't know this database and we understand that but this really the results that were finding are not very dissimilar from other research and that has assuaged some of the concerns. We also capitalize on the fact that NYTD has demographic information on responders and nonresponders so we've done some comparisons and we reported that. In this particular article we only reported that in the discussion because we've done that after reviewers raised that question and then subsequent articles this was something that  we've done routinely. We compared demographics of responders and nonresponders just to sort of show that yes there's some differences we know them we could talk about how they might kind of influence our findings and they are also not that big. So so we've done all of that to really assuaged some of those concerns. And then the next study data that's we've done still on this sort of topic of childbirth. So this was actually born from the limitation of the previous study which basically because it was cross-sectional we couldn't really say much about the chicken and the egg, right? So for instance we were really interested in this issue of incarceration right so does that precede kind of what happens at precedes childbirth or is that a consequence of childbirth so we've done a series of studies to try to answer that question and so I'll talk about this one and the next one and how it answers them. In this study we really wanted to so by that time the second wave of NYTD four 2011 became available and we wanted to capitalize on that to look at birthrates now between the ages of 17 and 19. So before we only looked at prior to age 17, now we're looking at ages 17 and 19 and wanting to see well how did the rates change. We know from research that they're supposed to be a big increase because all prior studies basically say that youth who are in foster care have this significant jump in the rates of pregnancies and birth as they enter the. Of transition to adulthood. So we wanted to see is this the case in this national data as well. So here we kind of wanted to document the rates of birth and we also wanted to look at this issue of repeat births right so if youth had children before age 17 how likely that they are going to have an additional birth between the ages of 17 and 19. This question was really significant because if you think about it aging out with one child is difficult enough and if you age out and you have multiple children how hard that is so it has some specific implications for prevention and for services and other things so we really wanted to look at that. And then we wanted to take our prior study one step further and we wanted to look how risk and protective factors prior to age 17 right relate to childbirth between the ages of 17 and 19. So sort of adding a longitudinal aspects to this. In this particular study we also combined AFCARS so we've looked at AFCARS 2011 which was collected approximately at the same time as the baseline NYTD data for 2011 so at age 17 and we wanted to see how certain child welfare characteristics and experiences might relate to childbirths. So that for that we need needed AFCARS because NYTD does not necessarily have that. In terms of analysis was still sort of simple relatively simple analysis logistic regression and bivariate and all that kind of stuff. So what did we find? We found that the cumulative rates of childbirth by age 19 so including the those who had children before age 17 and between the ages of 17 and 19 was 21% and we did see that jump between prior to age 17 and between ages 17 and 19. So about 17% were having children between ages 17 and 19. So that jump that we anticipated to find we actually have seen. We also saw that repeats childbirth was very very common in this population so if we look at those females and this was by the way only focused on females so we look that those females who had children at age by age 17 and wanted to know how many of them had children between ages 17 and 19 again and we found that about 60% of them did. So that's a pretty big number. And then in our multivariate analysis you could see the results here so again Hispanic ethnicity, black race were associated with higher rates of childbirth. So not unanticipated right? We also found some differences by foster care placement. So if a youth was in relative foster home at age 17 they had higher likelihood of childbirth and if they were considered to be a runaway and age 17 they had a much higher likelihood. Also those in trial home visits had higher likelihood of births. We as expected found that earlier exit from foster care was associated with higher likelihood of births. That was already sort of a finding in some prior data of Dworsky and Courtney and some others. So really if they emancipate and not all of these exit from care is not necessarily synonymous with emancipation but most of these youth likely emancipate and the earlier they exit care, that sort of has more influence on giving birth data run. And then in terms of protective factors those that had employment skills and were enrolled in school at age 17 had slightly less likelihood of childbirth between the ages of 17 and 19. Incarceration, consistent with our prior study, was associated with higher likelihood of childbirth. But really I would say that the most significant finding of this study was that prior childbirth is associated with subsequent childbirth. Now it's not anything unexpected we know past behavior predicts future behavior, but the magnitude of this was pretty substantial and that really had some implications for possible prevention services and interventions with youth who are already parents to potentially prevent subsequent pregnancies especially if they're not in the situation where they can handle that. Right so we've talked a lot about that in this article in terms of how high the rates really are. Some publication challenges here: there were mostly, so Journal of Adolescent Health is not a child welfare Journal so they had less concerns about some of this child welfare stuff and more concerns about response rates and also lack of detail in terms of childbirth and some associated variables. So they brought up things like well you know what does incarceration mean it could be anything from spending one night in jail or being convicted of something and spending much longer time in jail. Right so they were worried about some of those kinds of things and you know the strategies to address that were pretty similar to what we've done before. We again in terms of response rates we compared responders and nonresponders and sort of capitalized on the strengths of the dataset in that way. I would say the biggest selling point was the fact that there was virtually nothing on repeats births at that point. So and the fact that the finding was so strong in terms of the rates of repeat childbirth I think this was a big selling point and we certainly weren't shy about emphasizing that. The reviewers liked that we combined AFCARS to NYTD because they felt that it provided a little bit more depth to the data especially with respect to these placement and other characteristics we also looked at placement stability and some other things. So that's something that also played kind of in our favor. And then you know we were honest about the limitations so you know the lack of detail about childbirth and about some of these other variables is nothing that we could necessarily overcome easily. It is what it is, right? But we really kind of played on the strengths of the data and said well just the fact it doesn't negate the benefits right? It's still a snapshot of what's going on with this youth on this particular topic. So yes the limitations are there we are aware of them, were not trying to overstate our findings but we still believe these are important things to read and kind of consider. And then the last one that just came out last year was this article that was in Children and Youth Services Review which looked at the impact of childbirth on other outcomes. So by that point the age 21 data came out which was great and what we really wanted to look at is kind of getting to that question that I talked about before in terms of the chicken and the egg, right? So is childbirth independently associated with some of these outcomes at age 21 when you control for some of the prior things? So we wanted and not only did we want to do that we also wanted to look at the timing, right? So so if childbirth is related to say worse outcomes at age 21, is it any childbirth or is timing something of importance, right? Does timing matter? So we wanted to look at that as well. So here we basically only used NYTD but we included baseline, first follow-up, and second follow-up so ages 17, 19, and 21 and did logistic regression analysis and bivariate. And you could see I mean obviously there are a lot of things that we included here and you can see some of the details in the published article but what we found is first of all in terms of rates, by age 21, over 40% of females , this article also only included females, and over 40% of females reported childbirth at some of these ages, either by age 17, between ages 17 and 19, or between ages 19 and 21, at some point they had a child. So that's those are big substantial rates right? So that's something to talk about and the increase that we have discussed it was evident here as well so the increase was actually even larger between ages 19 and 21 than it was between ages 17 and 19. And some of it is not surprising obviously the youth are older but still it's much higher than what we would expect for general population youth at the same point. We also sort of found that if you look at this bivariate rate, right? So if you look at childbirth and whether or not it relates to say homelessness or incarceration at age 21 the answer will be yes. But if you do multivariate analysis and you control for prior risk indicators right? So does childbirth related to, say, incarceration at age 21? After you control for prior incarceration experiences, right, before age 19, and the answer was sort of more complex than that. So you could see the results here: essentially what you see is recent childbirth matters most, right? So childbirth before age 17 really didn't matter much once you control for the subsequent birth, right? So it wasn't independently associated with worse or better outcomes it was sort of a moot point. Childbirth between the ages of 17 and 19 mattered for school, right? So for educational attainment. So those youth who gave birth between the ages of 17 and 19 and also between the ages of 19 and 21 were less likely to get a high school diploma, GED, or higher certificate than those who did not have children at this time point. So again that's not really surprising obviously. We also know that foster youth tend to get even high school diploma a little bit later so this sort of ages of 17 and 19 this is likely when most of these kids are trying to finish school and get their high school diplomas or GED's and if it's interrupted with birth obviously that could result in a little bit less educational attainment. And also in terms of employment we found that those kids those girls who had given birth between ages 19 and 21 had less likelihood of employment at age 21 which is also not surprising because it's fairly recent childbirth, right? So if you've given birth recently it does make sense that potentially you're not employed right away, right? Because you're either caring for your baby or if it's very recent you might still be recovering from birth so that was our finding. In terms of public assistance and we only included housing assistance and financial assistance, we did not include food assistance here because food assistance also includes WIC, and WIC is obviously specifically for those girls who have children, right? So we excluded that but if we look at either financial or housing assistance than those youth who gave birth between the ages of 19 and 21 had higher likelihood of receiving that assistance but earlier ages were unrelated and again it does make sense. Before age 17 they don't even qualify for it, right? So I mean it does make sense. In terms of but the interesting finding was that homelessness, substance abuse, incarceration were not related independently to childbirth, right? So those females who had given birth were necessarily more likely to be homeless, to have a substance abuse referral or two report incarceration at age 21 once you account for these experiences before, right? So this kind of points out that it's possible that these risk factors are they're actually related such that youth who have them are more likely to give birth rather than the other way around, right? So if the youth's homeless she might be more likely to become pregnant. If she had incarceration experience she might be more likely to become pregnant but not necessarily the reverse, okay? So these were so with this study again the challenges were the same: a response rates, generalizability, and lack of detail in especially in the exact timing of birth and we sort of dealt with those challenges in that same way. We looked at the importance of the research question: there wasn't much about outcomes of childbirth believe it or not there was much more about risk factors for childbirth out there than it is about what happens to mothers and fathers later on as they transition to adulthood. So that they can our favor and then we discussed the limitations as we've done in prior research as well and as I said before we've just kind of stressed this trade-off between depth and breadth. Yes limited data on some of the key variables but look at this big national data set that allows us to do this longitudinal analysis and also look at sort of the timing which many times we are unable to do in some of these other data sets. So with that in mind how do we sort of published research with NYTD? My personal perspective, and this is just my experience, is that you continue to emphasize strengths, right? You continue to emphasize it's large and it's national even though it's not nationally representative you continue to say that this allows for longitudinal analysis it can I can look at services, I can research those small groups that really need we need information about, that I can link it with other data sets, and I can really answer some questions that have not been answered, right? Or at least contribute to an answer to a question that really needs answering. And for me at least, these strategies have been effective. If you explain this sort of in a right way than reviewers tend to believe it. In terms of some other things or some strategies for successful publication I think focusing on new research questions probably is the best bet, right? Reviewers will overlook some of the limitations if you can contribute to an important question. I have not used weights for these articles, I've used them with another article that I haven't included in this presentation but they are available and I know you've talked about this in some of the previous sort of webinars in the series. So that helps, right? The use of weights could improve generalizability and really kind of assuage some of these issues related to response rate, so use them. Comparing demographics of responders to nonresponders, so if you don't use weights at least you could do that and that does help as well. And obviously combining NYTD with other data sets and that tends to give a little bit more data so it's still not depth but at least it's much more breadth, and that helps. Some other strategies I've considered you know one thing that you could do is limit the analysis to states with adequate response rates. So for instance if you are trying to publish this article and they really get hung up on the response rates issue, there are some states in their whose response rates are 80% and up. So you could limit your analysis to only those states and then say that in the limitation that yes it loses some of the national perspective but we gain in response rates and potentially that could assuage some of the concerns the reviewers have. I think being upfront about the data set's limitations helps, and not overstating the findings, right? So that something that I always try to be really careful about. Emphasizing some similarities to other research helped me a lot. So for instance with this repeat childbirth issue there was one article, I think it was Dworsky's article that said that about 40+ percent of girls who were pregnant reported a repeat pregnancy. So I used that, right? I use that finding and I said look what I'm finding is not necessarily dissimilar from what other people have been looking at. And I think also it helps to educate colleagues about NYTD. I go to conferences and I present with these data and I always try to convince people that it is worth looking at. It's good data even though it has limitations but it's still something that you could publish out of and you could provide important information. So I want to finish this with some suggestions for possible future research directions with NYTD. I think a focus on understudied sub groups is something that this data allows to do because it's so large. So for instance, there's really nothing in the literature that I could find a be I'm wrong about that but when I looked previously there wasn't much, on Native American youths who transition out of foster care. In most data sets that other data sets look at they just don't have enough Native Americans to even consider that analysis but you could do that with NYTD. I think linking services and outcomes is something that is still a relatively unexplored direction. With all the limitations of the service data it still usable data and I think it's important to look and this was really the original intent of the NYTD is to look whether or not the services really do anything. So that's an outstanding question and I think more research on this is needed. This data allows because you can combine several waves of AFCARS and several waves of sort of this other data like NCANDS you could really examine the child welfare histories in depth. And also you could do longitudinal or trend analysis. I am dying to see if the trends for childbirth parallel sort of the national trends with the general population that just continues to decrease. So I want to see and when we have several NYTD cohorts we could look at that. And also as I said before there are some policy analysis that could be done with this data as well. That's about it from me. If there any questions I'm happy to answer them.
[Erin McCauley] wonderful thank you so much for that presentation. And just quickly in case we lose any participants thank you guys so much for participating in in the series this is been such a great group and this presentation was absolutely wonderful. So if you have any questions just please send them in through the chat feature and then Michael is going to read the questions aloud.
[Michael Dineen] Oh here's one: when emphasizing findings in the literature to buttress your findings could that be construed as biased?
[Svetlana Shpiegel] well I mean is it biased? I don't know I think it helps to give context and really what I have tried to do is give context from prior research and kind of show how these findings may or may not be similar to what is already out there. The way I try to overcome bias is I also try to present some differences, right? So for instance there were some studies that reported much higher rates of childbirth, right? And I try to look at that as well and indicate that in the discussion section. So yes there are some similarities and I think it helps to provide context but there are also some differences and I try to hypothesize either in article or at least in my correspondence with reviewers. Some of it doesn't really make the cut for the actual article. But they do try to brainstorm sort of where the differences might be coming from. Is it because it's national and the other is not? Is it because of this attrition, right? And maybe there is some self-selection going on. So I tried to hypothesize and really be thorough about thinking this through to overcome some of this bias. So again I hope this makes sense and if you have any specific kind of question about that I'm happy to answer in more depth.
[Michael Dineen] okay another question. This is a follow-up question "those are helpful points thank you Svetlana" from the person who asks the first question. And then from Morgan E: "when combining data sets how do you decide which set of demographic data elements to use? AFCARS or outcomes?"
[Svetlana Shpiegel] oh that's a really good question Michael remember you and I were talking about that last time. So you could go in one of two ways. And Michael please feel free to contribute to this discussion since you and I have been working on this together many times. But you could either try to reconcile AFCARS in NYTD, right? So you could take a look at the differences and see potentially where they might be coming from and decide based on what you think is more reliable, right? So for instance you know in both these data sets the reports each race and ethnicity as a yes/no individually and then there's, at least in AFCARS, there's also a combined one. So sometimes you could take a look at that: are there any differences, any similarities between NYTD and AFCARS and you could just make an informed decision. Or, and I think that that makes sense as well, you could choose one realizing that there could be errors in data entry and it's sometimes hard to know kind of what error has been made. So in some of the studies I've just chosen NYTD, the NYTD race, ethnicity variables just because I figured, you know, potentially because of the compliance so they're supposed to be 100% accurate based on the NYTD final rules so I sort of said well maybe this is going to help a little bit so I just chose NYTD. But Michael please let us know if you have any more thoughts about this because I know you've done this as well.
[Michael Dineen] well I think it's more or less arbitrary because as you pointed out that the there could always be a data entry error, right, so data entry errors trump everything. But if you're assuming it's not a data entry error what often happens in NYTD is that at wave one they have one race, at wave 2 have a different race and then it wave three they are back to their other race or you know so you could say well if there's three waves plus AFCARS you have for data points and you could say which one has the most or which one is the most recent because it could be that you know it's been wrong all along in AFCARS in the first two waves and in the third wave they find the corrected it. So you could interpret it it seems like there's an argument for choosing anything but what I tend to do is just take whatever is in AFCARS and which is the opposite of what Svetlana just said but I think as long as you do something you know that you can give an argument for then you are okay because you just don't know if that was just an error the one that's different from everything else.
[Svetlana Shpiegel] yeah and I would say that you know at the end of the day I think what's more important is to identify what you are using, right? So if you are taking the race/ethnicity information from AFCARS, say that, and if you're taking it from NYTD say that. And sort of procedures for reconciling: explain what those procedures are. I think that's more important than actually deciding which one you're going to use.
[Michael Dineen] yeah I think so too. So that's all the questions, Svetlana. And Erin do you have closing remarks to make?
[Erin McCauley] I just want to thank everybody so much for participating in this series. It's been such a great experience and I look forward to the videos coming out at the end of the month.
[Michael Dineen] and Svetlana I want to thank you, that was a beautiful presentation and I learned a lot from it.
[Svetlana Shpiegel] thank you so much. [Erin McCauley] extremely helpful [Svetlana Shpiegel] and if somebody has any questions always feel free to email me and I am happy to answer from again my experience is just an N of one but if you could be helpful for somebody I am happy to answer any question.
[Erin McCauley] all right well thanks everyone and have a good day!
[Svetlana Shpiegel] thank you
[Michael Dineen] thanks
[Svetlana Shpiegel] bye-bye
[Erin McCauley] bye
[Voiceover] The National Data Archive on Child Abuse and Neglect is as project of the Bronfenbrenner Center for Translational Research at Cornell University. Funding for NDACAN is provided by the Children's Bureau.