Annual EMSC Grantee Meeting

National Pediatric Trauma Registry for Children Grant Presentation

June 20 – 22, 2006

KAREN GUICE, MD: Thank you, Dan. It’s a pleasure to be with you again this morning, and specifically, to present some work that we’ve done on a National Trauma Registry for Children Project funded through EMSC. When I was asked to talk about this project, the project of creating a blueprint for a creation of a National Pediatric National Registry can be a fairly dry topic, and I wanted to make it a little bit more interesting, so what I thought I would do is kind of go through some our thought processes along the way, and how we arrived at some of the recommendations that will be coming out of this particular project.

Can I get the first slide up, please? Thanks. This project started from a couple of reports from the IOM. Much as we heard yesterday, they make recommendations and agencies act upon them. This particular project came out of two reports. One is the 1993 IOM report that stressed the importance of program evaluation, and called for states to collect data. This was a report specifically on emergency medical services for children, and that was followed by the ION report of 2001, which was quality chasm report. They called for applying evidence to healthcare delivery systems, stressed the need for an information infrastructure, and called for monitoring and tracking processes to evaluate safety, effectiveness, patient centeredness, timeliness, efficiency, and equity.

The project also came about because of the National Pediatric Trauma Registry, which in many ways is kind of a legacy system for this new concept of a National Trauma Registry for Children. I think many of you in the audience are familiar with the NPTR. The NPTR began in 1984. Phase one, phase two, and phase three were the different parts of the NPTR prior to its, kind of going away, in 2000.

Phase one was implemented in 1985. Data forms were filled out on paper--and faxed in, and they were entered in a fairly rudimentary, now; at that time it was pretty state of the art, but entered into an electronic data system or a computer. Phase two added some additional data elements, as did phase three. However, funding ceased in 2000 leaving sort of a perceived void about national data regarding injured children. The National Trauma Registry for Children Project, according to the grant guidance, was to provide a blueprint for the creation of a national registry for children. It was not actually implemented. It was merely to make recommendations about how it should be structured, what it should include, data elements, kind of the nuts and bolts of what a data registry or data information system might look like. I’m not going to read this particular slide to you, but keep in mind, the project was to create a blueprint.

The project was also divided into two particular components. One was the data identification used in collection component, which was awarded to the Medical College of Wisconsin, and I was the principal investigator on that particular piece. The other component was the registry design and technology component that was awarded to the University of Pittsburgh, and Doctor Laura Cassidy was the PI on that particular component. While they were two separate components, it was quite clear from the guidance that we were to work together as one unit, and kind of combine our forces, use some economies of scale, and use one advisory council. The advisory council was formed of individuals from the continuum of care for injured children from the scene through rehabilitation. We also had several subcommittees that were formed out of the advisory council based on interest of the members and expertise, and those were data privacy issues, HIPAA was a big deal; data outcomes; health outcomes; registry parameters; and technology. This is a list of the advisory council members and as you can see, the names are on one side, maybe backwards to you but on my left are the members’ names and on the right are kind of the areas where they had specific expertise.

So, we had individuals from neurosurgery, pediatric neurosurgery, pediatric orthopedic surgery, pediatric surgery, pediatric critical care, pediatric ED. We also had some adult representation, people who primarily worked on the adult side. But as we know, most injured children are seen in adult centers. So, it was very important to have their input as well as just pediatric specialties. We also had health economists, health service researchers. I see Mike Dean in the back; he was a member of the advisory council. Diana Fendya has served as sort of our liaison with the national EMSC resource center. Clay Mann was also a member of the committee.

Well, in designing a blueprint for the creation of a National Trauma Registry For Children, we’ve tried to follow a guideline of how databases are actually built and come together. So, the first thing was sort of to determine a need for the system and we had a lot of that in our grant guidance. We wanted to define the goals and objectives for the system. We wanted to define the functions, which dictated the data elements that would be included in the system. Finally, then, what you do should take those data elements and those functions and you convert them to system requirements. And then, you kind of get into the nuts and bolts of actually creating the database or the data registry.

Well, in determining a need for a system, I think everybody in this room understands there’s a need for better information about injured children. We just don’t have all of the information that we would really like to have. We know that injury is a big deal for kids. And we know that in the year 2002, for the individuals, one to 21 years of age, there were almost 22,000 deaths due to injury. About a third of those were due to intentional and two-thirds due to unintentional injury. That information comes from one database and that’s the National Vital Statistics database.

In the year 2000, which was the year where I could get summary data, for these individuals who’s same years of age, an estimated 11 million emergency departments visits were injury related and approximately a quarter of a million children were hospitalized for treatment of their injuries. These two bits of data come from a separate database, come from the NICE system and both of this systems are web-based, you can go and you can define the data however you like to, you can slice and dice it and look at the ICT9 codes; they’re quite wonderful. But they’re separate.

So, if you want to create a picture, a national picture of injured children in the United States right now, you would have to go to this data set and that data set and a hospital discharge data set or the NTDB. Some of the emergency department information, there’s very little in rehabilitation and try to kind of package it together and fit it together and it wouldn’t fit quite right but you get kind of snippets of it. You wouldn’t get a nice cogent picture. And that’s what I think we would all like to see eventually. It’s a long way off but it would be really great to be able to go to one resource, a kind of a data warehouse if you will, and you will be able to pull out a nice picture, a complete picture of injured children in the United States.

This kind of describes the various points of the continuum of injury for children. There’s the injury event, there’s the EMS scene evaluation and triage and treatment, the ED evaluation and treatment, hospital admission and treatment, rehabilitation and recovery, and up in the corner, there’s risk because that’s also an interesting area to explore. All of those are niduses for data collection. And currently, there are data collection efforts in each one of those areas but they are, by no means, unified. We need the data because we want to prevent injury, we want to improve the care that we provide to injured children as currently and we want to make sure that we optimize the outcomes. And that’s why data is such an important piece. And it’s important to pull it from all of these areas so we get a, like I said, a nice picture, a complete picture of injured children in the United States.

Through a series of four advisory council meetings and a consensus-based process, we developed a mission statement for National Trauma Registry for Children Information System. And the mission is to develop and maintain a contemporary information system, which will efficiently and effectively support the health and safety of children through the promotion of research, education, and treatment evaluation. We also came up with specific objectives for the information system. And those were to collect and facilitate access to aggregate data for the assessment and analysis of pediatric injury in the United States; to develop and maintain a user-friendly system for the recording, storage, and retrieval of pediatric trauma data; to provide flexibility in order to accommodate future needs, in other words stay up with technologies, stay up with what’s current; collect data elements useful in the formulation of pediatric trauma related public policy; resource allocation; prevention strategies; program development and finally (can’t get the slide to change) to provide reports on the status of injured children in the United States.

We were very pleased to be actually to come up with the mission statement and objectives. For those of you who participated in that kind of a process, you know it can be sometimes be quite painful. Wording is very important and it needs to be very specific.

Now, what we had to do was we had to take this mission statement and these specific objectives and turn it into something a little more concrete. So we came up with a list of data functions. What do we want the data to do? What do we want to be able to know about injured children in the United States? This is a fairly complete list. At least, we thought it was fairly complete for what we wanted or what we thought were important functions for the data. Advocacy. This is alphabetical so it’s not in any kind of other priority. It’s just alphabetical. Advocacy is important. Benchmarking for hospitals, it’s what they want to do with data. Data linkage so that we could actually tie some of these data pockets together. It wouldn’t have to recreate things all along, all of the time. Education and training is an important function for data. Guidelines and clinical pathway to improve clinical care, hospital accreditation, injury prevention, surveillance, outcome measurements, policy development, process evaluation, quality assurance and improvement, reimbursement research, resource allocation and review. So that’s a fairly complete list.

Now, we wanted, once we had our list of what we thought a data information system should be able to respond to, we wanted to know what people thought of that and if they had specific priorities for that. We created two surveys, one to a group of stakeholders, which was a statistically based survey. We got names from a bunch of organizations such as the Emergency Medicine Physicians’ Public Health. We sent this survey out. We also at the same time, asked our advisory councils to sort of rank order these particular functions. And these are the two lists we came up with. They’re both in order of importance for that particular group and you can see that they are all a little different. A little different which was interesting and probably it relates to the type of responders that we had for each group. Well, what do you do with all of this information?

If you start to create data elements for these things, you realize that pretty soon it’ll probably just be easier to fax the patient’s medical records and send it in. So we had to figure it out how to use these functions and what people wanted to get out of the data, what they wanted to do with the data and kind of figure out how to package that into a data registry or a data collection. So it was very helpful. At this point, we did this little exercise where we took these functions and we put them under categories of public health and hospital, more hospital interest. And then a category in the middle of, they’re really both. They’re both sort of public health interest and they’re both kind of hospital interest. Mind you that we also had in our grant guidance, this specific chart to, and remember, this is going to be a voluntary data collection.

So, you have to make sure that hospitals will want to send you the data voluntarily because we don’t think we want to pay for that. So, we had to be very aware of what hospitals wanted because that’s where a big part of the data collection resides in terms of hospitalized kids and we had to make sure that the hospital needs for data; we were well aware of those and kind of maintain those as part of our focus. Once we made this sort of table, we realized that one data collection wasn’t going to make this happen. And what we realized, we really needed a national injury surveillance system for children.

So for the first time, we would actually be able to say what rates were, really a surveillance tool. And then we needed something else for a little bit deeper data delving. A little bit more of the benchmarking. A little bit more of people wanting to compare how they were doing against someone else. That would really entail more data elements and be a richer data source but it would be much more of a case contribution part. The national injury surveillance system would be based on a statistical sample of hospitals that care for children, so that we would actually get national data. Usable national statistically valid data about injured children.

We also had to remember that there are very different levels of information. The biggest user of data elements is going to be someone doing a research project. And for those of you in the audience who’ve done clinical research or health services research, you know that you need a lot of data elements. We recently conducted a burn study for the APSA Outcomes Center and we have about 250 data elements. Some of those are in the chart, a lot of them aren’t but they are very important for that particular project. So we need a lot of data elements to conduct a research project. A hospital is going to need a lot of data elements as well because they have a lot of different uses for that data. They are not only concerned about meeting JACO standards or their trauma accreditation standards. But they are also worried about patient volume and where the patients going and are we making our in the hospital transit times reasonable.

So there you have a lot of data elements that really apply only to that hospital, that are unique only to that hospital and wouldn’t make sense to collect in a national data collection. Then we recognized that there also is a level of data elements that are specific to a community and they are very important for a community. For instance, if the accident happens at the corner of Fourth and Elm, it’s important to know it’s Fourth and Elm if it’s been like the fifteenth accident there. Now on national scale, that’s probably not that important to know that it’s Fourth and Elm. It’s probably more important to know that those are all motor vehicle accident related and they we’re all failure for airbags. That’s probably what you’d want to know at the national level. But at the community level, you’d want to look at that and say maybe we need a stop light instead of that stop sign or maybe we need a crosswalk if there are kids getting run over by cars. So there is another specific data element need for a community.

Similarly, there’s a specific data need for states and they may want to collect certainly a lot of the data elements that are collected below, but they have specific needs as well. Data elements that rise to a national level though are a subset. And that is why this is drawn this way. This is drawn to show that the national would be sort of the core that would come up through all of those but really is a smaller data collection than any of the ones that are listed behind it.

We also had some discussion about injuries versus non-injuries. People that access emergency services aren’t all injured. Some kids have asthma. Some kids have diabetic problems. But we wanted to focus just on injuries for this particular project. But keep in mind that if you see this as a data warehouse, that injuries could be a module, non-injuries could be a module. And so that eventually what you wind up with if the system works well, you can add these particular components along as people gets used to submitting data, as the system gets up and running and all the kinks get worked up.

Another thing that we had to remember is that trauma is a subset of injury. Also, under injury are poisoning, drowning, late-effects, and complications. And we wanted to focus on this because in thinking on the modular concept, we wanted to do something initially that would start out where there was already some history. And we know that data on trauma has been collected for a long time in many state and many hospitals. So we thought if we started with capturing the trauma piece first, really defining that well, having that up and going smoothly that then you can add things such as poisoning or drowning, again, as a modular concept.

So, here’s that modular concept. What we thought was we would take trauma domain first; we know we have other diagnoses that maybe important such as asthma, kids at Access Emergency Services. We know that we have other injuries that we would eventually like to add to our National Injury Surveillance System but we thought we’d make it easy and start with something that we didn’t have to sort of recreate the wheel from the get go. We developed inclusion criteria for our National Trauma Registry Information system and that really will look familiar to most of you who’ve used or know of your trauma registries in your states, or your hospitals and it’s fairly standard that the patients with really trauma-related would be included first, and here’s sort of a graphic of that, that we have the Trauma ICT9 codes collected first and all of the others potential modules for adding on later.

Now, we thought it was going to be that, okay, well, states collect a lot of data, hospitals collect a lot of data; can’t we make this easy? Now, we sort of said we wanted a national, consistent trauma data collection. We know they’re collecting data elements, so, wouldn’t it be great if we could make it easy and push the easy button? And we’ve found out that it wasn’t going to be easy. So, this was the concept, we thought, well, states collect data and it would be great if we could just collect those little oranges out of each of those state data buckets and put them in our national data bucket. That would be great, be easy, be done. So, we did an extensive look at state trauma registries and this was in the year 2004, and as most of you know, this is a fairly dynamic thing, so things have changed since 2004, for the good. So, we did a map showing which states had trauma registries and you can see that in the light purple, most states do. A lot of states were planning and very few states said, “no, we’re just not going to do that right now.”

Now, that was the good news because the bad news started when we asked about what is your data submission mandatory or voluntary? And mandatory is in the light purple but voluntary is in the yellow. And as we learned, even the mandatory data submission, if the rules weren’t written with sticks and punitive things that sometimes the mandatory looks a bit more voluntary submission but that gave us kind of a “Hmm, I’m not sure what we do with some of that” and then what really gave us pause was who submits? Some of the states say that all acute care facilities submit data, some say only their designated trauma centers submit data. And there are a few states that do a little bit differently; Ohio is the only state that requires rehabilitation facilities to submit data. Mississippi, the hospitals who participate in the reimbursement scheme to pay for this are the ones who submit data. So, that’s got to be a little more trouble. And then, when we went one layer deeper, we said we actually got the data dictionaries from all of the state trauma registries and did a data element mapping for this project. And here’s what we’ve found: this is just one very simple data element, the intent of the injury.

From three states, the data element name is intentionality, injury intent, and intent of injury. The definition, and you can kind of see that the definition is a little bit variable; one state didn’t have a definition; probably thought it was obvious, you know, intentional and not intentional. The other two states did but then the values, you got one state that records it as one or two, you’ve got one state that records it as accidental, assault, or self inflicted. And you’ve got one state that records it as a whole different menu. So right there we started realizing, well, everybody kind of collecting similar data, maybe. Working with Clay Mann who was doing a related project, which we will talk about in a minute. We also had a lot of variability case definition in the State Trauma registries. For example, the inclusion and exclusion criteria: abuse was in specifically included in four state trauma registries, and specifically excluded in three. Drowning was specifically included in 13 data registries and excluded specifically in 15.

In other words, it’s written specifically, we’re not collecting this data. So you can see there that already we’ve got not only data element differences, but what actually gets into the registry is different. And then within each of those, for instance, variability and same level fall, you can see that same level fall, and if the age is 55, the age is greater than 65, and it’s sliced and diced a million different ways. So, while the concept is really a good one that you can take what’s currently being collected at the states, and take all those data elements that are important for a national picture of trauma, it’s really hard to do because of the disparate ways that the data is defined and collected and cleaned at the state. This leads to just a brief mention of another project, which started the same time which our project did, which was the National Trauma Data Standardization Project. This is a project from the trauma program at HRSA, funded the American College of Surgeons Committee on Trauma to kind of have start to grapple with this. We’re very pleased to be part of this and included in actually kind of working through this, realizing that we had this huge problem out there and if we could bring some kind of consistency, it would be a really good thing.

So in the couple of minutes that remain, I’m going to just kind of go through one component of our information system in the National Entry Surveillance System for children, just briefly kind of scroll through some slides. And they’ll be a quiz on how many of you remember the data elements at the end. It’s part of this brain age thing that’s this new computer game that my kids have, memorizing 24 words in two minutes and then you have to repeat them back. Again, this is kind of a unique concept, a National Entry Surveillance System for children. This is designed to be a statistically based system that the two different ways of looking at it, to actually collect the statistically valid information are to base the sampling on a stratified cluster design on a 10 percent of hospitals, included in the American Hospital Association, which represent the country, the hospital size, and type, or to use a cluster sample with a probability proportional size to estimates of pediatric hospital admissions.

That’s all I’m going to say about statistical sampling, Lars, the design person on that, and really she is much more in favor of the second approach, using the proportionality construct. The data elements we did; we decided on data elements through a series of consensus using both our stake-holders’ survey to score what they thought was important and multiple, multiple iterations for the advisory council and knowing that we wanted for the National Surveillance System, we only needed a small number but very important data elements to include. And I’m pleased to say that most of these were unanimous. Among the advisory counsels when we wound up with it, I’ll point out the couple that were not unanimous, but even though if it’s not unanimous it was just one or two people at most who didn’t necessarily disagree with the concept but thought maybe we should do it a little differently.

But we think it is very important to capture information about the hospital and the trauma center level. We think it’s important to capture some information about the demographics of the individual. The ethnicity data variable that is one that we didn’t get a 100 percent, we had an 18 out of the 20 responding advisory council members and mostly because they thought that the way that the OMB defines, which is ethnicity, which is really just Hispanic or not Hispanic, wasn’t complete enough. And they would like to see more ethnic groups included on a pick list for that particular thing, not that ethnicity is not a bad thing to collect; it’s an important thing to collect.

The data elements for the injury, the site of the injury are included here, and again were consensus based elements. The ED data elements are listed here and are the notable ones are that GCS really does need to have the pediatric levels and not just the adult levels. And we decided that really it’s the first GCS that should capture as well as the first vital signs that are measured in the ED. If you look at what‘s actually happening, what gets recorded may or may not be the actual first, and the way that the trauma registries define that are quite variable.

The hospital data elements are listed here and the two that were really not consensus were the hospital charges basically because charges: that's a real hornet’s nest once you get into cost charges, discounted charges but the concept is that we need to collect some sort of financial information. We need to know how much it costs and how much it costs not only individuals, but hospitals and the nation. That's an important variable; exactly how to do it, this is probably the best we can get right now. The other outcomes are very important to capture but when you start talking to people about what do you mean about outcomes and what tools should we use. Outcome tools are proprietary, number one, so it means mandating that hospitals bias particular type of outcome tool. So we didn’t reach a lot of consensus in exactly what kind of tool to use. But the outcomes are very important and that it was actually an area that may be ripe for some additional work and funding for doing the work.

The data elements that are proposed for the other side: the data collection, the case contribution. There are more of those data elements so you can collect more data elements on the case contribution basis and they are a little bit more diverse but they have the same core in them. And finally I’d like to thank the program staff at EMSC for their help, the advisory counsel members, all of the people I’ve annoyed with stake-holder surveys, and the staff that contributed to these project and I’ll be happy to answer any questions. Thank you.