AMCHP 2005 ANNUAL CONFERENCE
DELIVERING RESULTS, IMPROVING PREGNANCY & BIRTH
February 19-23, 2005
MARY LEE: Back in 1995, when Connecticut 's managed care program went from fee for service to managed care, the Connecticut state Legislature was very concerned about children's health services in the Medicaid program in managed care. And they created an independent performance monitoring entity called the Children's Health Council, and we actually were quite fortunate to have a lot of resources for running a children's health info line, doing community‑wide education and training on managed care and also doing performance monitoring in the program using enrollment and encounter data from the entire managed care program. That was done under a contract with the state Medicaid agency, but independent of the agency itself.
The Children's Health Council subcontracted for the data management and data analysis with Maximus Incorporated, a national healthcare consulting firm that's located in Restin, Virginia or McClain, Virginia, and we've been working with Maximus ever since. We had a brief interruption in funding for the performance monitoring back in '03/'04 when the Connecticut state budget was quite challenged and we have recently resumed the performance monitoring under the auspices of the Connecticut Voices For Children, who I work for, and voices 12 has subcontracted with Maximus again. So Amanda and I are up and running again monitoring children's health in the Medicaid program.
What we're going to do today is describe to you the methods that we have used to link the data not only within the Medicaid program, that is, the enrollment and encounter data, but also to link the Medicaid data with the birth data from the state of Connecticut . But after describing how we brought the data together, we'd like to talk a bit about how we have used the data to inform policy discussions in the state of Connecticut , and how it's been very, very useful to have the linked data set as the basis for that.
I'm getting a little ahead of myself. First, just a very brief description of Connecticut 's HUSKY Program. The basketball fans in the room know that our basketball team is the Yukon Huskies . Okay. And but HUSKY really is an acronym that stands for Healthcare for Uninsured Kids and Youth. It's a little tortuous but, nevertheless, that's the name for the program. HUSKY‑A is the Medicaid managed care program. And it's this program for which we have the data. HUSKY‑B is actually our state children's health insurance program. We are not going to be talking about that at all today. In Connecticut , children under 19 in families under 195% are in HUSKY‑A. This Medicaid managed care program. Their caretaker relatives and parents can be in family incomes below 100 percent and pregnant women are eligible with income below 185% of the federal poverty level. It's a statewide program with mandatory enrollment for these groups who are in Medicaid. And currently there are just over 300,000 people in the program. Mostly children.
The pregnant women who are qualify for Medicaid are almost all enrolled in the Medicaid managed care program. There are a few who enrolled during the third trimester, whose providers don't participate, who can get an exemption. But that's a very few births in the state that are not in Medicaid managed care. The emergency Medicaid births of course are also covered for women who are qualified for it at birth but never came into the program during pregnancy.
In Connecticut , we don't have any asset test. And while we do have ‑‑ we say we have presumptive eligibility, and in fact it's an expedited eligibility determination process. Women who were in the program get full Medicaid benefits, including dental care, for example, and they're covered for 60 days postpartum.
I wanted to say just a little bit about what we see as the value of having independent performance monitoring in a Medicaid managed care program like this. This is a function that has been supported by the Connecticut General Assembly, even under extraordinary pressure during the economic downturn in our state budget over the last two years, and I might add with a lot of pressure from the budget director under the previous Governor. The general assembly has committed to having this independent function, and because they value the information that comes out.
The majority of the work is state funded, under a contract with the Department of Social Services. It's not part of the Title V Bloc grant. We have no formal relationship whatsoever ‑‑ no contractual relationship with the Department of Public Health, which is the Title V agency in Connecticut . So this is work that goes on entirely under this contract with Department of Social Services. We do undertake special studies and some ad hoc studies with funds from private foundations, as well.
And what we see and what the Legislature has told us that they see in this independent performance monitoring capacity is some ability to account for how federal and state dollars are spent in the state and what the outcomes are.
We are able to show and to hold the program accountable for program change. For example, we cut back on parents' eligibility last year from 150 to 100% of the federal poverty level. We were able to use the data from the program to show the differential impact in different parts of the state as we cut off nearly 20,000 parents off of coverage.
We've also been able to, because of our independent role, to bridge the gap between the two agencies and to promote some data sharing that we believe is good for the policy development. And finally the work that we've been able to do has contributed to the data quality in this state, because using the data on a regular basis and using it very publicly holds the program up for scrutiny. Holds the health plans accountable for their data and creates an additional incentive for the data quality and completeness to improve. And we have felt very firmly committed to using the data that we get from the beginning even when the plans were telling us that, oh, the well child screening rate is low because the data are not as complete as they should be.
Well, we count what we have. And we report what we have and we believe that that creates an incentive for them to make sure that it gets better.
So in the interest of monitoring children's health services in the program, we became very interested in the antecedents of children's health as well. We began looking around at what the existing data were to tell us something about maternal health and prenatal care and birth outcomes and what we found is that in Connecticut the birth certificate did not have a field for payer, so up until the work that we did, Connecticut public health agencies and planners could not determine who was on the Medicaid program using the birth certificate data alone. Now, the health plans that participate in the program report regularly to the Department of Social Services on the births that occur, but they have difficulty getting information from all of the providers and from the women themselves. So the data are often incomplete. In addition, the health plans that they submit report on prenatal care, for example, only for the time that the woman's actually in the program. They are using a modified HETUS (phonetic) measure which is a report of the percentage of women who had first trimester prenatal care among those who were enrolled in the first trimester. So it doesn't tell us about all of the prenatal care, some of which the woman may have received before ever enrolling in the managed care program.
And finally, and Amanda can say a little bit more about this, we have all of the encounter records, claims data for the managed care program, but it's not as easy as one might think to link the maternal and infant records using just the encounter data. And so we found that when we tried to count births by looking for mother's records and fee diagnosis and procedure codes that would indicate prenatal care, for example, that we couldn't match it always with the newborn records. And so we felt like we had a dearth of information with which to assess the maternal and infant health, and we wanted to know more.
So what we did was to try and bridge what we saw as a pretty wide gap between the two agencies. This can't be true only in Connecticut . This is ‑‑ from what I hear it's pretty typical that the state agencies are somewhat proprietary about their data and they don't always cooperate in ways that would give us the kind of information that we need.
So we began to work with staff from the Medicaid agency and the state public health agency to develop an algorithm and a plan for matching the Connecticut birth data with the HUSKY enrollment data. And we consulted with experts along the way, Paul Busher is in North Carolina is one who has had a lot of experience with this and he helped us out. And we developed a study protocol and a design for matching the data and how we would use the linked data set and obtained permission from the human investigations committee at the Department of Public Health for the release of the birth data.
The birth data were released to the Children's Health Council, they did not go to the Medicaid agency. Again it was this independent entity between the two agencies doing the work. And then after CDC, HRSA and HCFAA published their model data sharing agreement a number of years ago, the two agencies did begin to work together to try to develop an MOU to share data. In fact, there was legislation passed requiring the two agencies to cooperate.
What the two agencies have done since then is to develop an MOU with a specific addendum that covers sharing the birth data. And so after we've created the linked birth data set we now can return copies of that to both agencies so the agencies themselves have the linked data sets and can work on it. To use it for additional work like linking to WIC data, for example.
I'd like to turn this over now to Amanda to tell you a little bit more about the nitty‑gritty of the linkage.
AMANDA LEARNED: I'm here to tell you more about how we actually did the methods part of the linkage.
After being so involved with Medicaid enrollment and encounter data for five years, we've really thought we understood what that had to offer. Once we got the birth certificate records and really analyzed that, we realized these data elements were available for the linkage. It involved mother's Social Security number, the mother's exact first name, the mother's exact last name, the mother's exact date of birth, and off the birth certificate, the baby's date of birth. We used that in order to verify the mother's enrollment in Medicaid at the time of the delivery.
Our first step was that we had this file of birth data. It included about 43,000 births in Connecticut for 2000.
We also had the HUSKY‑A enrollment data, which was a little under 300,000 people ever enrolled in Medicaid for 2000. Our first step was to link the two data sets on the mother's Social Security number. And verify the mother's enrollment on the baby's date of birth.
Of that, about 9100 records linked. For those that still were not linked off the mother's Social Security number, we looked at matching the mother's name, again exact first name, exact last name, and the mother's date of birth and also verify that the mother was enrolled at the time of delivery. And of that about 457 records matched. So the total linked database that we had was about 9600 births. Then, in order to do an evaluation of the matches, we wanted to really look and see, okay, of those records that matched, were they true matches? This was done in two steps. I kind of messed up it's a lot better on the handout. After the first step, again, linking just on mother's Social Security number, we wanted to look at those matched records. Of those, 240 records or about three percent of those records, did not match on mother's date of birth. And then we actually looked at those records and we found that the majority of those records, the mother's date of birth on the birth certificate matched the baby's date of birth on the Medicaid enrollment data. So we still counted it as a match. It's just the data wasn't as clean as one would have hoped.
The second step we looked at for those records that did match on Social Security number, what did the mother's exact first name look like. Of that, we found 499 records, or about five percent of the matched records, did not match on exact mother's first name. Again, looking at those 499 records, we saw that the majority of the time they did not match because the names were spelled differently. For example, Maxine was spelled with an E in one record, no E on the other record. So looking at the records, we still considered it a match.
Under the same step, we looked at the exact, the mother's exact last name. Under that, about 893 records or about 10 percent of our matched records did not match on mother's exact last name. After looking at that, we found that the majority of those, they did match on the birth certificates mother's maiden name was equal to the Medicaid enrollment mother's last name. So, again, the Medicaid enrollment data was just a little bit dirty.
And then out of the 9100 that matched, only 41 records, or under one percent, did not match on date of birth, last name and first name. Under the second step where the mother's Social Security number did not match but it did match on date of birth, mother's first name and mother's last name, we did an evaluation of looking at those Social Security numbers to see why they didn't match. And 14%, or 63 records, the Social Security number was missing on the birth certificate. For example, they're all zeros or they're all 9s so they did not match. 394 or 86% of the Social Security numbers didn't match, the numbers were just inverted. We looked at all those records. So we still considered it a match. The numbers were just a little dirty.
In order to evaluate even further, we've asked the health plans to provide us diskettes of all their births for 2000. Two of the health plans gave us the births with Social Security numbers. And what we found was about 97 of those births were found in our linked database. The other two health plans gave us mother's name and the maternal date of birth. And out of those about 85% matched in our linked database. And I'll give it back to Mary now.
MARY LEE: I'll say briefly what we found and for those of you in maternal and child health there are no surprises here. As expected we found a higher teen birth rate among those who were in the Medicaid program, a higher low birth rate and preterm birth rates and lower rate of first trimester prenatal care.
There were no surprises, but it was the very first time that we had the information on the maternal health and birth outcomes in the maternal, in the Medicaid program in Connecticut . And Medicaid, we also knew for the first time that Medicaid was covering one in four to one in five births in the state of Connecticut . So the impact of the program was quantifiable for the first time.
Having the linked data set also allowed us to quantify the impact of the program in various geographic areas or subgroups in the state. What we found is in the three large test cities in Connecticut , which some of you may know are among the most economically challenged in the nation, that half or more of all the births in those cities were covered by Medicaid. In fact, nearly half of the births in five other towns, including the Governor's hometown, were covered by Medicaid. We can use that information to generate support for the program with individual legislators by being able to show them that the program does affect the health of their citizens right from the time of the birth.
The information is also very useful for local planning and needs assessment.
So since we've begun this linking project, we've linked the 2000 and the 2001 births, we have had, as I said, an interruption in our funding and have just resumed again. We're poised to link the 2002 and 2003 births within the next year. And the way we have used the data from the linked data set have been focused mainly on trying to show how this publicly funded program has contributed to maternal and child health in the state of Connecticut . We can also take a look at the health risks of the low income women who are in the program and identify some opportunities for increasing access to care and coordination with other publicly funded programs like WIC.
So the next couple of slides go over how we've addressed some questions that have been raised in various policy making arenas in the state. Connecticut , I'm sad to say, is one of 32 states in the United States that does not cover tobacco dependence treatment in its Medicaid program. In the last couple of years, there have been several attempts to bring forth legislation that would mandate coverage in the program but none of these bills have passed. But the question has come up about what the extent of the problem is, and what we have been able to show with these linked data is that overall the smoking rate among mothers giving birth while they're in the Medicaid program is about four times higher than it is for other mothers in the state and is actually quite high among white non‑Hispanic women, almost 30% of them are reported to have smoked during their pregnancy.
With the new information that's available from the CDC report on the consequences of maternal smoking, we now can take these figures and quantify the costs to the state of maternal smoking. And hopefully use that information in the current legislative session to bring about a change in covered benefits.
We have another oversight council that pays a lot of attention to what goes on in the program, and one of the obstetrician gynecologists who sits on that council raised the question two years ago about whether mothers who were in the Medicaid program, the husky Medicaid managed care program were getting access to the appropriate level of care for high risk pregnancies, and so what we were able to do was to look at the percentage of all births and then the percentage of very low birth weight births that were born in the subspecialty hospitals in the state. And what we found was that the very low birth weight babies born to mothers in the Medicaid managed care program were equally likely as other very low birth weight babies to deliver in these specialty hospitals, about nine of them. And this information was used as background for some discussions at the Department of Public Health about improving the regionalization of perinatal care in the state.
Two years ago, during the lows in our economic, our state budget crisis, Connecticut proposed ‑‑ the Governor proposed and the Legislature went along with the idea of asking CMS for permission to charge premiums in the Medicaid program, including premiums that would be charged to pregnant women with family income as low as 50% of the federal poverty level. And the tone of my voice probably betrays my sentiments on this.
So we worked on studying what the potential impact of this was going to be, and so with funding from the Connecticut Health Foundation, and in collaboration with researchers at Georgetown University, we used information from other states that had imposed premiums to estimate what the impact would be on enrollment, that is, people failing to enroll or dropping coverage because of being unable to afford premiums at various income bands. And knowing how many women were enrolled in the program was key to this. We were able to estimate that about 2,000 pregnant women would lose coverage and that, of course, then subsequently their infants would not be covered by Medicaid as well.
This information, along with additional information about the overall impact of premiums in the program, was key to leading the Legislature to repealing this proposal in the subsequent legislative session.
So luckily we're not charging any premiums and there's no cost, no other cost sharing for pregnant women.
The health plans that participate in the HUSKY Program frequently complain that women enroll so late in their pregnancies that the health plan doesn't have any chance to intervene and to improve birth outcomes and that this is really detrimental to their programs, their outcomes and their bottom line. And so what we did with the births that took place in 2000 was to take a look at the timing of enrollment in Medicaid managed care and how that affected not only prenatal care initiation but also the birth outcomes. So Amanda took the births from 2001 and added all of the enrollment data for each month during 2000 and 2001 to the mother's files. I got assistance from a colleague at the Department of Public Health who is an SBSS expert, to create enrollment vectors, and we were able to determine whether women were enrolled in the HUSKY Program in the month when they had their LNP and those who were not, we figured were most likely to have been uninsured. And in fact what we found was that 43% of the mothers who gave birth while enrolled in Medicaid managed care had actually been enrolled prior to ever becoming pregnant. So if you think ‑‑ and we would never know this just from counting the coverage categories for pregnant women. In fact, these were the teens who became pregnant. These were the second or third‑time mothers in low income families. These were parents of children in the program. And so many of these women were already in the program. And this 56% who were not in the program, that is uninsured at the time that they became pregnant, were more likely to be older, having their first babies, more likely to live in Suburban areas, and more likely to have a little bit higher, to be high school graduates. So the profile of these two groups and when they came into the program was pretty different in terms of their demographics. And what we felt this information pointed to was the, for those who were already enrolled, is the extraordinary importance of making sure that they have access to family planning services while they're in the program. And that they have access to early pregnancy detection and early prenatal care.
We also feel that for those 56% who were uninsured prior to the birth, the case finding early identification of those women and timely eligibility determinations are key for getting them in. And we've been able to use that information to work with policy makers who are interested now in fixing the eligibility determination process.
So just a comment on a couple of new opportunities we see coming up. As we've resumed performance monitoring, we have once again reestablished the database that includes the enrollment data and the encounter data for the Medicaid managed care program. So we have an opportunity now to do what we've been dying to do from the beginning, and that is to link the encounter data for the infant's first year with the birth certificate data and link the mother's health data during her pregnancy and after her pregnancy with the birth certificate data so that we can begin to study a number of questions about maternal health and about infant health, based on diagnoses and on care that mothers and infants get before and after the birth.
We also feel that there's a good opportunity for the Department of Public Health to use the linked data sets to examine WIC participation and other early intervention programs for children that they now can determine are in low income families. We would love to have the data for the HUSKY‑B births, that is, the births that occur to adolescents who were in the HUSKY‑B program. We're working ‑‑ that's kind of a long shot at this point, but that would be our goal, I suppose, at the end. And we also hope this time to be able to add in the fee for service births, because the Medicaid agency has become increasingly interested in those.
So the challenges facing us. The biggest one, of course, is ensuring there's continued funding for independent performance monitoring. And I'm glad to report that the political and fiscal environment for this has taken a turn for the better. We're not out of the woods, but there's a good deal of support in the Legislature for continuing this independent performance monitoring.
We believe that not only our resources but also the resources of the Department of Social Services and the Department of Public Health need to be enhanced in order to make better and full use of these linked data sets along the way.
We are mindful of HIPAA restrictions on the data, because the Department of Social Services, as a Medicaid agency, is a covered entity under the privacy rule, we have certain parameters around using, disclosing the information that we have, even in the aggregate. So we've been working a lot on that.
And, finally, over the next year, we hope to continue to use the data to examine issues around program change. As the Legislature expands or contracts eligibility, we want to be able to say how this is affecting the health and healthcare for mothers and infants in the state of Connecticut .
Our report, couple of reports on this are on the Connecticut Voices For Children website, ctkidslink.org, and I would urge you if you would like any additional information, to get in touch with Amanda or me directly. Thank you.