Ninth Annual Maternal and Child Health Epidemiology Conference / December 10-12, 2003

Approaches to Using Data Linkage

ANGELA NANNINI:  Good afternoon, everyone.  How many here are on the data side of this issue?  Raise your hands.  And how many are on the program side?  And then how many are hybrids?  I'm kind of a hybrid.  Welcome.  You know, last night I was thinking about coming here, or how long it was to get to Tempe and how we feel so isolated, but you've got to consider the alternatives.  This is what I left in Boston on Tuesday, so now we may feel a little isolated out here in the desert, but we could be in Boston.  This has been an unusual fall for us, but also as usual we go through our usual October slump, which is around our sports team, and this was in a recent New Yorker.  It says, "Okay, calm down.

 Move away from the windows and tell me if you are a Cubs fan or a Red Sox fan."  But on that note I would like to begin my presentation and first give you an overview of data linkage pre-PELL and then talk about the State perspective of PELL, about how we built the partnerships, the collaborations, talk about PELL start-up, and then talk about how taking an innovative idea and evolving it to a sustainable idea.  And then maybe talk about tips for other States thinking about linking.  And at that point we'll have the discussion and I think a lot of you are probably in different phases of this Project and you can add your own two cents about this issue.  So, first of all, data linkage is not new to DPH.  I mean really since the 1980's we've been doing the birth, infant death linkage.  At least 1980, as far as I can tell, but maybe even farther back.  And that's really been a sustainable linkage that we've done.  And I got involved in data linkage actually with pregnancy-associated mortality in 1998, when we started trying to find maternal deaths.  And then in 1999 with the CDC and BU School of Public Health we launched a uterine rupture, where we did linkage with hospital discharge data and birth certificate data.  And then on and off we've out had outside researchers actually come to DPH and do linkage on-site, then de-identify the data and then take their data off-site and do the research.  So, this is not an exhaustive study of our past linkages, but just to say that we've had a lot of past projects, so we have this institutional recognition of the value of linkage. 

And I think that's very important and that you should appreciate that in all of your States.  And then plus coming to conferences like this one year after year, I've been very inspired by some of the other States that have done linkages, Michigan, Washington State, of course, some of California, other places, I can't mention them all, but coming here and seeing what other States have done, you know, you go back to your State and think, "Oh yeah, I can do that."  So, first you try to apply for funding on your own and we did try to apply on our own a few years ago and got rejected, so we had that behind us.  And then we also have had various attempts among the program data to try to link clients and that's actually inter-agency linking.  And we actually have an ongoing program now called Steps, which we're trying to do that, but that's much bigger than DPH.  It's clients through all the publicly funded programs.  But just knowing that there is this kind of recognition of linkage is very important when you start out.  In terms of our collaboration, I think Milt alluded to the strength of our partnership and it really wasn't built in a day.  We started with the uterine rupture study and, actually, at least half of the core team was involved in that. 

And then Milt joined us for PELL, Kay joined us for PELL, now Wanda Barfield, I'm happy is at the Department of Public Health with us as well.  So we've had this reconfiguration of the team, but the core team from the uterine rupture study has still stayed together.  And, see, this is what I see as the benefits, is that we get this outside expertise from local researchers from BU and also from Kay from the CDC, and that connection with the CDC gives us that, kind of, national arena and we know what other people are doing around the country.  For me it's been helpful working with BU because of the insulation from the political changes because since we've started the PELL Project we've had a change of Governor, we've had a change of Commissioner.  And our budget is just probably like your budget, it's been disseminated in public health areas, so just being in this outside funded Program has been very important, because working in State government you're always one heartbeat away from being changed to another job. 

From the mission totally being changed and so my involvement in PELL might end, but it's good to have this outside support.  And also, I can't say enough about the outside funding helping us with getting the expert programming, which we have some wonderful programmers at the Department of Public Health, but sometimes it's hard to get the level of expertise that we have in Jane Lazar and also keep her dedicated to that task and not in budget times getting diverted to another crisis.  On the negative side, I think that when you go into a public/private partnership, now you have shared ownership, so you share control of the project.  So you have to relinquish the control that you've had before. 

And then, plus, being on the PELL team as a State person I have the PELL interest to think about, but also I'm representing the State and they're not always the same.  I have some challenges and solutions here for the PELL start-up and I know that Milt went over a lot of the confidentiality issues, but just to say that this was a new kind of approval because there was really no big analysis connected with it, but it was a lot of data and people were nervous about that.  And it was just for a linkage.  And getting the Program from BU sitting at the Department of Public Health was good for the Project too, because it gave the Project visibility at the Department of Public Health.  Sometimes you work with outside researcher and they go away and, you know, they just come back for their presentations.  But with having Jane there day-to-day, PELL has a face at DPH.  And I think Milt talked about the feasibility runs that, you know, there was a concern too that, "Oh, this outside partner's going to have all this data.  What if they start looking at all this data that they don't have permission to look at all this data?"  So we established this process for feasibility runs.  We also talked about some of the caveats.  DPH didn't own all the data, we had the inter-agency agreements and then we had this Program administrative data that extra protections.  And the other issue is that we had funding for all the linkage in this Project, but there was no funding for all the Program staff on the data side to work with us, so sometimes when we'd ask for data the staff were busy.  So how we solved that problem was we got the linker variables to get the data set links and then we can download the claims data at another time. 

So at least the Project goes forward not waiting for (inaudible) staff who are under-resourced right now.  So these next few slides is kind of thinking about how this innovative project can go from being a good new idea to sustainability and what happens.  This is kind of like technology diffusion and adaptation, is in the process of being institutionalized what happens to the idea and how it evolves over time.  And I think the first level of this is what's PELL?  Jane shows up, "Who's PELL?  What's PELL?"  And kind of explaining it.  And we had those kind of PELL 101 for the stakeholders at the Department of Public Health.  And those of you that attended Jane's training earlier in the week, the Link Pro, we had her do Link Pro demonstrations of the computer program that she used and, you know, it kind of got people excited about PELL and started to generate some excitement there.  And then, plus, having Jane around, you know, there were other people thinking about linking, or even she's a great programmer so she actually gives a little technical consulting on the side and technical assistance just because she's around.  A lot of people probably don't even know she's funded by BU, she feels so much like part of our staff.  So that's a good thing for our Project. 

So we went from this, "What's PELL?" to this was my role, "Have you thought about using PELL data?"  Because even though we had told people about PELL, I was talking to the Director of Research in another division whose doing this managed care analysis and said, "Well, you know, I've been thinking about linking this data."  And they were going to do this birth link to hospital discharge data.  And I said, "Well, PELL has done that.  Have you thought about using PELL?"  So then there's this whole level you have to keep raising the conscious and awareness in your agency about that PELL exists and PELL can be used.  This is kind of like PELL 202.  So we gave key individuals little in-services about the new data from PELL, we demonstrated the quality of the linkage and we also started doing validation of the linkages, because when you start putting together two data sets, you don't get the same answer all the time, so that's important.  Then the next major step was when we started to hear from the constinguency of DPH, "Let's link more data to PELL." 

All of a sudden we had started out with these ideas of what data set we wanted to link, but then there were all these other data sets that people wanted to link and that's really exciting because then people start to think about us as a living Project.  But this is where PELL begins adaptation, because we really hadn't planned for these databases to come in and they challenge us in certain ways, but we have to adapt to that.  But the more we adapt and the more we integrate these new databases the more it becomes usable for DPH.  This is just a couple, there's been some other ones, care coordination database for special healthcare needs and the (inaudible) data.  And we're working on these databases and trying to see if their feasible, but these are the way the Project is changing right now, so we're going to have to adjust what we're doing.   So then after, "Let's link more data to PELL," we have "We need PELL data now." 

So all of a sudden people realize that you've been linking data and you have this data.  And you know in the State, it's not that we need it now, it's really we needed it yesterday.  So you have this need all of a sudden and this challenges us, because this was retrospective and we were spending a lot of time in getting this good data set together.  But all of a sudden you're in everybody's consciousness and there's questions to be answered around early intervention.  Wanda's helping with a Program evaluation on early intervention.  This summer we needed to get early intervention data set compiled and start to do an analysis that was different than the other IRB's that we had been doing.

So that was something that was internally DPH staff wanted to do, so we had to adjust to that.  And plus this fall we've been doing a needs assessment for family planning for a RFR we're issuing and we thought, "Well, isn't it great we can add some PELL data to the needs assessment."  We're right now doing short inter-pregnancy periods in high-risk cities to add to that to our needs assessment.  But some of the issues this brings out about getting real-time data, or closer to real-time data, is related to mission delays and creating data files that we're streamline the process.  This is the last step we're out now, is how can we shape this next generation of PELL.  And this is really about how we institutionalize a project, and I guess I would say we haven't institutionalized it yet, but we're really at the next step to doing that.  And there's two components of that.  There's the need to plan for ongoing linkage and the need for an ongoing blueprint for the use of data.  So just a little bit about the need to plan for ongoing linkages.  We had these existing data and a lot of times with one-time linkages you're over with the data set.  But here we are another year's gone by, we need to get our approvals again, but plus we need to re-link all the new data that came, like, from early intervention, from birth data, from the core linkages. 

So it's an ongoing linkages, plus we're going to be adding WIC in this year and we have to think about that.  So there's all these real-time data merges going on.  And then we need to plan collaboratively about what kind of technical expertise at DPH do we need to develop that at our Agency to really take this Project further.  Because BU clearly can't do everything for us and as PELL catches on, there's more and more that we want to do with it. 

So, for instance, do we want to teach the people to download data?  At what point does DPH do some more of PELL on the technical side?  Milt alluded to streamlining access and approval process, I can't say anymore, but it needs to be done.  And then this ongoing blueprint, you need to think about this blueprint for the use of the data and thinking how PELL can be integrated into our activities for balance, program evaluation, program planning.  For instance, in the annual birth reports, do we want to have some PELL data elements there?  And then the plan for outside researchers and public access.  I don't know if any of you know about the MassCHIP System. We have this interactive data system available to the public with loads of data set.  Do we want to make PELL part of the interactive data on the Web so the public, communities doing needs assessments can download PELL data?  And these two parts will help us with the sustainability.  And then plus, two weeks ago we had this major reorganization at the Department.  So where are the opportunities in this reorganization?  Like, my Bureau of Family Community Health is now within the center of Community Health, and within that is AIDS and substance abuse. 

So are there opportunities there for us to expand our data linkages to those programs.  And then plus, Medicaid is really coming into the same Agency and a lot of States have Medicaid in the public health agency.  We have not had that.  So we're closer to Medicaid.  So are there opportunities there for linking in more Medicaid data?  So what are tips for getting started if you're thinking about linking?  Milt said a lot about IRB's.  It's not a small issue by any means, as you can see by his presentation.  The workforce issues; getting a skilled Programmer to get an early quality product is very important.  And I would really advocate to have people both from the program side and from the data side, the epidemiology staff, to work together at the State Department of Health.  Because you need the logistics, you need to understand the data linkage, how to do validation, et cetera, but you also need to identify those strategic opportunities for using the data system you're creating.  It's great to have research collaborators like we do, you know, the expertise and helping to insulate for some of the political changes.  You need to develop that initial blueprint for how to use the linkage.  But be aware that you're going to be adapting this plan all along.  As more people get excited about your data linkage project, they're going to want to use it in different ways.  But the more the people use it, then I think it becomes sustainable.  And early on you need to balance the linking with output. 

I mean, I love data and I love the thought of linking all these data together, but for the program people, they want to see output, they want to see how they're going to use the data.  So you need to balance that really from the beginning.  Keep communicating with the stakeholders, but try not to get their expectations too high, because you link data sets and sometimes you find it’s the Achilles heel of every data set when you link it together because it exposes all the weaknesses of each data set.  Resources, maybe you can think about the money you're using with other linkages to mobilize those resources under one roof, but I have no easy answer for that.  After start-up, pay attention to the core public health functions, again, surveillance programming planning and the sustainability of the link data, ongoing work plan.  It said that for us as well.  Communication; keep up with your political agendas and keep thinking about the resources that you need.  Establish your ongoing plan for linkage activity and really decide when you're going to use linked data and when not to. 

Because linked data systems are very complicated data sets and you may not always need to use linked data.  So just remember that.  Sometimes it's more difficult to explain.  Streamline your access.  And once you start linking the data you find, as I said, the Achilles heel of every data set.  You need to build in data quality improvement activities, because you're going to find what's wrong with the data.  Set up systems to get each data set in specific intervals.  And then I put, "Don't start thinking about tomorrow."  But that means don't stop thinking about how to make this a sustainable resource for you to use in your State.  So, I'd like to leave you with this image of Boston rather than snowstorm.  This was our new bridge that we have with the Public Works Project, the Lenny Zaken Bridge.  And this was the eclipse in November.