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

Methods Can Help — Transcript

SANDY FAWBUSH: Hi, I'm Sandy Fawbush, and I'm a nurse consultant and the program coordinator for the Newborn Screening Program in Kentucky . And Charles Mundt, from the University of Louisville School of Public Health, is also going to doing part of this presentation. And we want to recognize Laurie Chestnut, who's also in our division, who is the grant administrator for the SSDI grant that funded this project.

The State System Development Initiative is a grant through HURSA and NCHB that Kentucky was awarded, and as I said Laurie's the administrator of, and we decided to use that money to look at some of the newborn screening data for Kentucky --to look at record linkage, to look at system analysis of the newborn screening in general.

And why did we pick newborn screening? The biggest reason was it's of great significance to Kentucky --to any state because all your newborns are screened should be. And in Kentucky that's about 54,000 newborns a year. So the population's huge, and it's mandated by our legislation. And the data had never had any focused analysis, no attention paid to it. And if newborn screening is done well, we'll improve the health of the infants that are born in our state, so that's why we chose newborn screening. Go ahead.

Just to give you a little background of newborn screening in general, it is a significant public health activity, and there's about four million babies per year that are screened through newborn screening throughout the nation. Go ahead.

It originally started in the late '50s, early '60s, with Dr. Robert Guthrie and--I don't know how many are familiar with the PKU card. But the actual test for PKU that began in the mid-60s--go to the next one. By the mid-60s most every state in the United States was screening for PKU, so it's been around for a long time. Every state, however, has a different panel of tests that they screen for. Some states screen for four, six; some screen for thirty. It varies throughout the nation, and that is a problem. And so there's a group--the American College of Medical Genetics actually convened an advisory committee to look at screening and to make recommendations as to what should be added to newborn screening panels across the nation. And in September of '04 they came out with a recommendation of--originally it was 30 disorders; now it's 29. They took one back off. And in March of '05, the March of Dimes concurred with their recommendations and also is making the same recommendation that every state screen for these 29 disorders. As part of the criteria for the screening, the disease has to be treatable, the testing method has to be basically simple and inexpensive and a fairly easy screening method and those were the things that they considered when they gave us the list of the 29 that they recommend that we screen for. Okay, we'll go to the next.

And I told you already that Kentucky 's law requires that every newborn receive a newborn screen. And it's basically just a little heel prick. They put--fill those six spots on that card and mail to our state lab. And it's virtually painless. I mean, it's a little stick, but it's not too bad. Go to the next one.

That's just a diagram showing where to stick the baby's heel. And as I said, it takes six drops of blood. Time is of the essence with this screening--go to the next one--because the conditions that we're screening for are fairly rare, but they do have detrimental outcomes for the babies if they're not treated early. So as quickly as we can get this test to the lab and the results back to the primary care physician and the doctors and the parents, the quicker they can begin treatment. Go ahead.

Part of the push for newborn screening at the current moment is the new technology, the Tandom Aspect Trauma Team, which has been around since I believe the early '90s but was very, very expensive originally, and like everything else, you know, there's newer and better things, so now this is not as expensive as it originally was and most states can afford to buy the equipment and do the testing. The Tandom Aspect Trauma--that one piece of equipment can test for up to over 20 disorders just from one blood spot. So Kentucky is in the process of expanding its screening, as are many are states.

As of January 1st, we will go from six disorders to 29. We will be doing the full panel of testing that's recommended by the American College of Medical Genetics. And I didn't list every disorder. There's the categories and then the few extras that we've added. Go ahead.

The newborn screening program actually has five components: we have the screening that's done at the lab, follow-up that's done at the state, definitive diagnosis, treatment, and management is done at the universities, and then were going to have evaluation that we're looking at evaluating our program. And I forgot to tell you at the beginning that there's a few--in the handout there's a few extra slides that we've deleted from the presentation that are there for your information.

There are some factors that can affect newborn screening and their results, and those are the TBN feedings, prematurity, an inadequate specimen, or a blood transfusion. And what most states require if the infant does receive blood transfusion is to get a sequential repeats at different periods; every state's different. Based on the results, if we have a presumptive positive, the lab calls us. We call the primary care physician at the universities and refer the baby for definitive diagnosis and testing at the universities. The clinics that we use at the universities--you can go to the next one--would be Metabolic Genetic, Endocrinology, Pulmonary, and Hematology for the different disorders that we test for. And then the universities or whoever--if the PCP feels comfortable, he can see the child, but they fill out a case report and fax it back to us so we'll eventually have a complete circle of newborn screening that we actually know what the outcomes are, which we've not been able to track in the past. Go ahead.

This is the system that we actually are in the process of developing, or it's been done, but we have never gotten the feedback from the universities or the primary care physicians as to what the final status for the outcome for the child. The newborn screening involves hospitals, physicians, parents, professional organizations. We've also used WIC and Medicaid for getting infants referred. Go ahead.

And then this is just a diagram of the system the way everything that feeds into the newborn screening program. It originates in the lab, comes to us. Right now ours is a paper process, but we're in development, and Charlie's going to show you some of that. We also try to link it with the birth and death certificates--is what we're going to try to do in the future. It's not done at the moment. And then we refer to the UK U of L Clinics. And KBSR is our Birth Defects Registry, which once we have a positive case that's diagnosed, the child's entered into the Birth Defects Registry and will be tracked along with the other (inaudible) that are tracked in that system.

And now Charlie Mundt is going to take over.

CHARLES MUNDT: Good morning. I guess I was trying to figure out what a successful presentation was going to be this morning, and one of the first things was not to fall off the stage. I don't know if you saw me wobbling over there, but I was getting pretty close. But anyway, Sandy sort of described about what the newborn screening is, a little bit of its history, and the importance that Kentucky sees to the newborn screening efforts. And clearly the Newborn Screening Program requires a lot of data integration and has a lot of internal and external data systems--next slide--that really looks rather complex when you start looking at the number of files and activities and different--let's face it--in state government the different silos that exist in terms of how the information is developed and tends to be used within a particular program, the vital statistics folks, and the lab people. But the newborn screening is one of those programs that needs information from a wide variety of sources, including the lab systems, the birth certificate systems--we use, but not as part of production--the death certificate information, and need to and report out to the various university clinics. University of Kentucky , University of Louisville abbreviated there.

What really hasn't also been done very much is have some time to really analyze the data and look at the information. And so one of the activities that we were engaged in is to look at the information and try to capture it in a way that we can look at it with (inaudible) and graphic mapping programs or GIS, and other linkage programs, Link Plus and SPSS. So that was what we were really trying to do systematically. Next slide.

But why? Why are we doing that? And clearly it is to answer some key questions that have impact to the results of the screening activity. It's nice to have integrated data and real-time online, but we really do have some essential questions that need to be answered. Are the babies being screened? Are they being screened in a timely manner that the results are of any value? And so we were somewhat driven not only to develop better ways of getting the information together, but better ways of looking at the information.

One of the big problems that we do have--next slide--is how do we pull some of these files together. There are some futures I think will make life a lot easier. Some of the states have been successful in re-looking at how their data is collected within the state government regarding newborns and individuals. But historically we've been sort of looking at data trying to look for certification, birth certificate numbers, or something magical--Social Security numbers don't work at this stage of the game--but to sort of link the records together, and we call that deterministic. We're looking at a couple of keys and we're trying to link them together.

We also have tools available to us that are called probabilistic. And a couple years ago I think at Tempe they discussed some of the more advanced linking tech technology at the NCA (inaudible) Conference. And probabilistic allows you to sort of look at the probability of two records being the same. And we chose--next slide--in one of the key--two of the key file relationships that we were interested in was really the birth and the death--newborn deaths, and also--go right ahead to the next slide--and also to look at the birth and the death and also the birth and the screening results. The idea of looking at quality in this environment is to focus on not only the structure of the programs but also the process and ultimately the outcome. And so much of what we do is in a sort of a state of dynamic change. And sometimes it's hard to pin down in production systems and requires the Newborn Screening Program to do side analytic methods. Next slide.

The Link Plus is a program that is available from CDC. It was originally developed for their tumor and cancer registry activities. It is free. It is downloaded from the internet. The reference there is in the handout. And if you are interested in record linkage, I recommend taking a look at it. It's certainly like all programs has strengths and weaknesses. The magic word is that in terms of prototyping and trying things out and exploring hypothesis, a free program is a nice way to start. And there are more advanced things to use. Not to minimize the sophistication that is in that program. It is a fairly sophisticated program and quite effective for what we were trying to do. Next slide.

Include there the sort of almost offhanded suggestion there in the last sentence that, hey, you can maybe use this for wider things. And we said, ah-ha, let's give it a try. And we have seen a couple of other states using it for a wide diverse set of uses. We thought we'd try it in this program. Next slide.

This is the main working screen of Link Plus. And what we are showing there is just a sample. And hat we have done is linked newborn deaths with the birth file. And it allows you to identify a number of variables from each of the files. They can be named different things. They can have some different characteristics, not much, and it will not only look for exact matches but come up with a scoring methodology to look for transcriptions and changes in each of those fields. And we use birth certificate number, first name/last name, birth date, and we can also set over here in the right written in red the cut-off value, which is set unusually low there for purpose of demonstration that we wanted. You can adjust some of these parameters to tune your data because it is hard. It is an art. There is some science involved, but it is an art to do a record linkage. Next slide. Next slide.

In terms of the results of this particular program, it's good. It found identical matches, which is always a good sign in any linkage program. And this page right here is just really some data we mocked up from real data that shows on the bottom half of the screen pretty much identical matches. Notice over on the right-hand column it gives you a score and sort of says 22, that's a pretty good number here for an identical match, and indeed they are. The certification number's the same. First name, last name, birth date are all the same on this particular slide. Next slide.

However, down the line after you get past those high scores, you do have lower scores, and you do start to see things that you were hoping not to find, such as transcription errors in the certificate numbers on the bottom there that you see. Some one pretty much the same person and they have two different certificate numbers, one from the birth record and one from the death record. We have hyphenated last names. We have last name changes. We don't have birth certificate numbers on the death record on a couple of occasions. We have some transcriptions on the birth date. But through this we can see that these 13s and 12s as scored by the program were pretty good linkages. As you get down to the 5s, you can see that they're not very good at all. And that's why you set it higher than the 5. We use the program, not necessarily as this is the end result, but we use the results of this program as a bridging table between our birth certificates and our death certificates and our birth certificates and our newborn screening record. Next slide.

Why do we do this? You know, and this is one activity. We really looked at a number of the key variables in terms of here is the date--the number of days between the date of the birth and the date that the specimen was collected. Remember Sandy said with the new technology, mass spectrometry, has tighter restrictions in terms of you need to get the slides together and into the lab in a much quicker time period. You have validity problems with the specimen itself. And so we looked at the slides. We'll go through these rather quickly here. That we see a number of folks who have very late screenings. And sometimes in our old system, we allowed people to get out of the hospital without a screening and they were to be picked up in the clinics afterwards. And those clinics would be sort of the backup of the hospital screening. Not good anymore in terms of it's just too long and too late. We looked at this over periods of time. Next slide.

In 2001/2003, very little change. And then we start looking at our problem cases. Next slide. And there you see the number of screenings over four days from date of birth to collection. And we see that there are quite a few facilities that just were pumping out a lot of late collection and allowed us to take a look at this by facility. We didn't show that here, but we have it by facility. And we were able to go back and look at what's going on within the particular facilities. Next slide.

In particular, over seven days an especially critical problem when we're using Tandom Aspect for analyzing the specimens it just starts breaking down rather rapidly in terms of a number of these tests. And we see the same problem here, emphasizing the need to do something as the Tandom Aspect comes on board. Next slide.

One of the things we (inaudible) are we dealing with one facility, five facilities, how many facilities are we dealing with, and we sort of did some work by facilities and percent. Now the one on the far right is one or two clinics that were out of left field in terms over seven days. Well, those were the clinics; they weren't the hospital. They were the catch-up clinics and they rarely get to see the patients--the babies before a period of time and sometimes it's an actual annual visit after the birth, so they have really missed the opportunity to take advantage if we don't modify it--take advantage of the Mass Spectrometry. Next slide. Next slide.

I'm just going to go over it. We had the same problem with the distribution and delays in getting it from the hospitals and the facilities to the lab. Next slide. And here we were exploring using this data and say our lab is pretty much in the north-central part of the state and there's a geographic difference here. There's something going on within the postal system, how they're handling the mail, and we were able to get back and we started asking questions about are you using a third-party service on the blood specimens? You're probably doing it on the bills. Are you using a third party service to sort your mail? Are we losing days, and what needs to be done? And we were able to do a number of improvements in that regard. Next slide.

Clearly throughout this with results, there's a number of things that we need to be educated about, and the results showed it, but we also realized that there was a lot of provider education and preparation now that we were thinking that we'd have to worry about the new tests and educating about the new tests, but clearly the procedures and the tightening of the procedures needs a tremendous amount of provider education. We really have not--even in the current environment, we have hard--it's very difficulty to get at the actual diagnosis and results of the screening, and through our work we've been able to do that, and we need to really reduce a lot of these delivery times. Next slide.

What we have done is developed also a prototype system that pulls together the screening information which is on top, allows referrals--records the referrals to the University of Louisville and the University of Kentucky and start building the case management system. And while we have put this into production as a prototype, the State of Kentucky has taken the results here in the system and is building it into their standardize IT system. So there's a lot of opportunities and success, I think, in terms of getting involved in newborn screening. There's a lot of techniques and work that needs to be done--further work that needs to be done on newborn screening and record linkage. Next slide.

Fortunately or unfortunately, there's a lot of change taking place in newborn screening throughout the country. All of the states are undergoing tremendous amount of change as a result of these new technologies, and in the next few days, weeks, months--we got some systems that were coming on board while we were in flight to here--there's a lot of system changes that make some of this analysis difficult with changing formats and changing requests for information, but it is important that we feel that being able to look at this information and hopefully move to a better identification of these individuals within the state system is critical and paramount. Thank you.