A
RAND
The Effects of the Florida
Medicaid Eligibility Expansion
for Pregnant Women
Stephen H. Long and M. Susan Marquis
December 1995
Health Care Financing Administration and March of Dimes Birth
Defects Foundation
RAND is a nonprofit institution that seeks to improve public policy through research and analysis.
RAND's publications do not necessarily reflect the opinions or policies of its research sponsors.
Ill
CONTENTS
LIST OF APPENDICES v
FIGURES Vii
TABLES ix
EXECUTIVE SUMMARY xi
ACKNOWLEDGMENTS xv
INTRODUCTION 1
BACKGROUND 2
Previous Studies 2
Expansions for Pregnant Women in Florida 4
DATA AND METHODS 6
Data 6
Methods 7
RESULTS 16
Aggregate Analysis 16
Person Level Analysis of Outcomes 25
DISCUSSION 31
Findings From the Florida Expansion Compared to Other Studies ... 31
Policy Implications 32
Directions for Further Research 34
REFERENCES 37
LIST OF APPENDICES
A. SELECTION OF THE STUDY STATE
B. IMPLEMENTING THE MEDICAID EXPANSIONS FOR PREGNANT WOMEN: THE
EXPERIENCE IN FLORIDA
C. EVALUATION OF A MEDICAID ELIGIBILITY EXPANSION IN FLORIDA:
DEVELOPING THE DATABASE
D. ESTIMATION FOR AGGREGATE ANALYSIS
E. EXPLANATORY VARIABLES AND REFERENCE POPULATION
F. REGRESSION RESULTS
Vll
FIGURES
Figure l--Distribution of Deliveries by Primary Payer 17
Figure 2--Distribution of Prenatal Care Visits Among Medicaid
Women by Site of Care 19
Figure 3 --Distribution of Source of Financing by Type of Hospital
and Study Period 21
Figure 4--Percent of Medicaid Expansion Population Enrolling in
First Trimester, by Time Period 26
IX
TABLES
Table 1. Predicted Values for Comparing Insurance Groups 15
Table 2 . Deliveries by Primary Delivery Payer 16
Table 3. Ambulatory Prenatal Visits by Primary Delivery Payer 18
Table 4. Maternal Hospital Admissions by Primary Delivery Payer 22
Table 5 . Flow of Payments for Maternity Care 24
Table 6. Prenatal Care and Birth Outcomes for Low- Income Women
by Insurance Status 28
Table 7 . Prenatal Care and Birth Outcomes for Medicaid Women
Using Different Delivery Systems, 1991 30
XI
EXECUTIVE SUMMARY
OBJECTIVES
This study investigates the effects of a Medicaid eligibility-
expansion on the financing of maternity care, on the use of prenatal
care services, and on birth outcomes for newly entitled women. It also
examines its effect on the role of other government programs that
provide prenatal care--specif ically, the public health system. We study
the experience in Florida. This is a populous state with a large
Medicaid population. It introduced a substantial eligibility expansion
in 1989, raising the income threshold for pregnant women from 100
percent of poverty to 150 percent of poverty. Florida has traditionally
relied heavily on county health departments to serve low-income women,
and thus is a good state in which to study interactions between Medicaid
financing changes and the publicly financed direct-delivery system.
DATA AND METHODS
We study prenatal care and birth outcomes in Florida for deliveries
in the period July 1988 through June 1989 and in calendar year 1991.
This includes the twelve month period just prior to the income
eligibility expansion and a full calendar year beginning eighteen months
after the expansion was first implemented. Data for the analysis come
from a variety of sources. These include: the Florida birth and death
certificates, the hospital discharge abstracts, Medicaid eligibility and
claims files, individual encounter records for personal health services
provided by each of the county health departments in the state, the
American Hospital Association annual survey files for Florida hospitals,
and the 1990 Census. The birth and fetal death records for all women in
.the state, regardless of payer, define our study universe. We developed
algorithms to match records in the other databases for women who
delivered in one of our study periods to the record for that birth in
the vital statistics database.
Our evaluation of the expansion consists of two parallel analyses.
First, we investigate aggregate changes in the quantity of care,
Xll
delivery provider, and sources of financing maternity care in the two
periods. Then we compare the amount and timing of prenatal care and the
birth outcomes for a woman covered by the Medicaid program and an
uninsured woman in each period to estimate the effect of extending
Medicaid coverage. Finally, we contrast birth outcomes for Medicaid
women who use the county health departments for their prenatal care with
outcomes for women using another delivery system.
RESULTS
Aggregate Analysis . The Medicaid expansion in Florida led to a
substantial shift in the source of payment for deliveries between the
baseline period and 1991. The number of births covered annually by
Medicaid rose by 47 percent over this time period. Most of this
increase represents women who otherwise would have been uninsured for
their prenatal care and delivery.
The additional prenatal care financed by Medicaid as a result of
these expansions was accommodated almost entirely by county health
departments. The number of prenatal visits provided by the county
health departments over this period increased by 100 percent. Thus, the
Medicaid expansion did not substitute for care already provided by
county health departments, but rather the Medicaid financing appears to
have provided the financial resources to expand the capacity of the
counties to deliver prenatal care to low-income women.
The Medicaid expansion had a much smaller effect on the type of
hospital women used for their care; although there was a small shift in
care away from public hospitals toward both voluntary and proprietary
hospitals. Although there was little change in the number of admissions
of pregnant women to public and voluntary hospitals, there was a sizable
decrease in the share of patients who are self -pay. As a result,
hospitals benefited from the expansions. Their maternity-related
revenues grew by 5 percent, whereas the admissions for maternity care
were fairly constant .
Person Level Analysis of Outcomes . We find evidence that women
enrolled in the Medicaid expansion use more prenatal care and have
better birth outcomes than they would if they remained uninsured.
Xlll
Although providing Medicaid benefits to low-income women serves to
improve their access to care, there remains a substantial gap between
Medicaid enrollees and those with private insurance in use of prenatal
care and in birth outcomes .
The county health departments may have been a significant factor in
the improved outcomes for the expansion population. As noted above,
most of the additional Medicaid financed prenatal care was provided by
the counties. In addition, we find that Medicaid women using the county
health departments for their prenatal care experienced significantly
better health outcomes than women using another delivery system.
DISCUSSION
We come to stronger conclusions about the benefits of the
expansions than most of the earlier literature. Methodological
differences may account for some of this. The large sample size and our
use of area income to identify the subset of women who are most likely
to be uninsured and eligible under the expansion are methodological
improvements over other studies. In addition, we rule out selection as
the sole explanation of our results by also studying differences between
Medicaid women enrolled because of their AFDC participation, as opposed
to their pregnancy, and the uninsured. On the other hand, the Florida
experience may have differed from that in other states. The county
health departments were a significant factor in the Medicaid expansion;
improvements probably would have been more modest without it.
Our findings have significance for current policy discussions.
Congress is considering major changes in federal funding for Medicaid
and public health. Our results suggest policymakers should be cautious
about cutting back on the eligibility for the expansion population,
because the expansions appear to have had a beneficial impact. Second,
our results emphasize the inter-relationship of expanding insurance
coverage and providing a delivery system to accommodate peoples' needs.
In particular, the findings suggest that financing Medicaid eligibility
expansions by contractions in the public health system and other
policies to shift care sources to the private delivery system may have
unintended, unfavorable effects on birth outcomes for low-income women.
DIRECTIONS FOR FUTURE RESEARCH
The findings from our study of the 1989 Florida eligibility-
expansion raise a number of important questions about the effect that
new directions in Medicaid will have on access to care for low-income
pregnant women and on their birth outcomes. These new directions
include further eligibility expansions for higher income women,
increasing physician fees to encourage use of office-based care, and
greater emphasis on managed care enrollment.
• Will eliminating financial barriers have the same effect on
access to care and outcomes for the near poor--who are the
subject of the more recent expansions in Florida and other
states?
• Does Medicaid eligibility improve access and outcomes for
important subgroups of pregnant women--especially teenagers and
women at high risk for poor birth outcomes?
• What are the likely effects on birth outcomes if increasing
physician fees leads to a shift to more prenatal care delivered
by office-based physicians? What are the likely effects on
birth outcomes for Medicaid beneficiaries of the greater
emphasis on managed care?
• Will the effects of the expansions change when the public
health system cannot expand further to meet the increased
demand?
• Does providing care directly to uninsured women through the
public health system have the same effect on prenatal care use
and birth outcomes as providing public insurance to pay for
care received in the private sector?
Florida remains a good candidate state for study to answer these
questions. The state has experienced a number of changes in Medicaid
eligibility and in the delivery system over the period 1988 through
1994. The questions that we posed above can be addressed with further
analysis of the database that we have constructed for 1988 through 1991
or by extending the database to 1994 using the technology that we
developed in this work. Such a database would be an invaluable source
for answering the critical questions that are facing policymakers about
reforming the Medicaid program.
XV
ACKNOWLEDGMENTS
This research was supported by a cooperative agreement from the
Health Care Financing Administration (18-C-90113/9-01) and a grant from
the March of Dimes Birth Defects Foundation. Any views expressed herein
are the authors' and should not necessarily be attributed to the
sponsors or RAND. The authors are grateful to their project officers--
Marilyn Hirsch and Herb Silverman at HCFA and Kay Johnson at the March
of Dimes--for a great deal of encouragement, assistance, and patience
over the life of this project. Many people in Florida assisted in
providing data and helping us understand the f iles--including Meade
Grigg, Phil Street, and Dan Thompson of the Department of Health and
Rehabilitative Services; Fred Roberson and Jack Shi of the Medicaid
program; and Chris Augsburger of the Agency for Health Care
Administration. Ellen Harrison performed expertly in constructing the
analysis files; Nancy Allen assisted with the Medicaid data. Finally,
we are grateful to many people who commented on an earlier draft,
particularly Feather Davis, Genevieve Kenny, Charles Mahan, and Sarah
Rosenbaum.
THE EFFECTS OF THE FLORIDA MEDICAID ELIGIBILITY
EXPANSIONS FOR PREGNANT WOMEN
Stephen H. Long and M. Susan Marquis
INTRODUCTION
The reduction of infant mortality became a policy priority for the
federal and state governments in the latter months of 1986. Since then,
publicly financed perinatal care and delivery systems have undergone
radical changes. In 1987, state Medicaid programs started to implement
a series of far-reaching eligibility expansions for low-income pregnant
women and young children. By July 1994, all states made Medicaid
benefits available to pregnant women and infants with income below 133
percent of the federal poverty level, and 33 states used optional
authority to set the income thresholds for eligibility at higher levels.
These expansions also stimulated states to simplify Medicaid eligibility
processing, implement outreach programs, and introduce enhanced prenatal
care benefit programs in an effort to improve access to prenatal care
for low- income women and, thereby, to improve birth outcomes and infant
health.
The Medicaid eligibility expansions for pregnant women and children
were the most important policy changes in the program in the 1980s. Yet
there are only a limited number of studies of the effect of these
expansions and it is not clear from these studies whether the expansions
led to an improvement in prenatal care and birth outcomes (Alpha Center,
1995) . Moreover, to understand the full effect of the interventions it
is also essential to understand how the Medicaid program changes
affected other government programs that deliver prenatal care and how
the expansions affected private payers. The effects on prenatal care
access and birth outcomes are likely to be quite different if Medicaid
financed care substitutes for care previously financed and provided
under other programs, such as Title V or private insurance, rather than
providing new coverage for those who previously lacked insurance or
access to other public programs. None of the previous studies, however,
addresses these substitutions.
Our objective is to investigate these interactions between the
Medicaid program and other sources of financing and providing maternal
health care, and, with this perspective, to study whether pregnant women
newly entitled to Medicaid coverage received more or earlier prenatal
care, and whether their birth outcomes were improved. We study the
experience in Florida. Florida is a good site for this study for a
number of reasons. It ranks fourth among the states in total
population, and there are about 200,000 births each year. Florida
significantly expanded Medicaid eligibility for pregnant women and also
aggressively implemented other strategies to ensure that women who were
made eligible by the expansions gained coverage under the program.
Florida relies heavily on county health departments to provide prenatal
care to its low-income women, and hence is a good place to study
interactions between the Medicaid financing changes and the publicly
financed direct-delivery system. (See Appendix A for a more complete
discussion of our choice of Florida as the state in which to carry out
this study) .
BACKGROUND
Previous Studies
Earlier studies measuring the effects of providing Medicaid
coverage to uninsured pregnant women have produced mixed results. Those
that measure the change in prenatal care and incidence of adverse birth
outcomes in a population after expanding the availability of public
insurance program to low-income women conclude that the expansion did
not result in improved care or birth outcomes (Piper, Ray, and Griffin,
1990), even when other temporal changes are accounted for (Haas,
Udvarhelyi, Morris, and Epstein, 1993). However, measures of the
effects of the increased access to insurance may be diluted in these
studies because the comparison groups include women who are not directly
affected by the expansions and because temporal controls are lacking or
limited.
Other studies making concurrent comparisons of pregnant women
enrolled in public insurance programs with uninsured women provide some
evidence of improved prenatal care access and birth outcomes for those
with public insurance coverage. Haas and colleagues (1993) report that
those enrolled in a state program for low-income pregnant women were
less likely to initiate care late in the pregnancy and less likely to
experience adverse birth outcomes than uninsured women. Braveman and
others (1993) find that Medicaid women have more prenatal care visits
than the uninsured, though they started care later than the uninsured.
These cross-section comparisons, however, may be biased if women
enrolling in the public insurance programs differ from those remaining
uninsured in ways that are not accounted for. In particular, the
studies control for demographic characteristics, but do not adjust for
income differences or health differences that are important correlates
of use and outcomes (Starfield et al . , 1991; Rosenzweig and Schultz,
1982) .
Moreover, expanding the availability of public insurance for low-
income women may not be enough to improve prenatal care use and birth
outcomes. Benefits of removing financial barriers to care can only be
realized if eligible woman enroll in Medicaid and if they enroll early
in the pregnancy. Piper and others (1994) , for example, found evidence
that simplifying the Medicaid enrollment process by establishing
presumptive eligibility led to earlier enrollment in Medicaid, earlier
initiation of prenatal care, and improvements in the adequacy of care
for Medicaid patients, although they were not able to identify
consequent improvements in birth outcomes. Differences between states
in their success at getting women onto Medicaid early in the pregnancy
appear to be related to efforts to simplify enrollment procedures.
California, which made few changes in enrollment procedures, enrolled
only 39 percent of its Medicaid expansion population in the first
trimester of pregnancy compared to 54 percent for Michigan, which
adopted a number of enrollment changes (Alpha Center, 1995) .
In addition, the content and scope of prenatal care, not just the
quantity of care, is believed to be an important factor in birth
outcomes (Starfield, 1985). Some research supports this view. Several
studies of Medicaid patients find the incidence of low birthweight and
infant mortality is lower for low-income women who receive prenatal care
from the public health system, which provides coordinated maternity care
and related support services, than for other women (Buescher et al.,
1987, 1991, 1992; Thompson et al., 1993; Clarke et al . , 1993).
Our study addresses the issues raised in the earlier research. We
combine cross-section and time-series analysis methods. To control for
self -selection in the cross-section comparisons, we develop a proxy
measure of income and we distinguish between women whose enrollment in
the public insurance program is related to the pregnancy versus other
enrollees. We also control for the mother's health risk. We study a
state which adopted a number of programs and procedures designed to
maximize the effectiveness of the improved financial access for low-
income women. And finally, we compare outcomes for Medicaid women in
different delivery systems to investigate the effects of content and
scope of services available.
Expansions for Pregnant Women in Florida1
In October 1987, Florida became one of the first 15 states to take
advantage of the option that was authorized by Congress as part of the
Omnibus Budget Reconciliation Act of 1986 (OBRA86) to provide Medicaid
benefits to pregnant women with income below poverty. Two years later,
in July 1989, the' state further expanded eligibility for pregnant women
by increasing the income threshold to 150 percent of poverty. Finally,
in May 1992, Florida raised the income limits to 185 percent of poverty.
As a complement to the expansion of financial access to coverage,
Florida also enacted a broad range of other strategies to help ensure
that the eligibility expansions had their desired impact. These
included changes in the Medicaid eligibility determination process,
outreach and public information campaigns, policies to improve physician
participation in Medicaid, and technical assistance to county health
departments to foster more effective Medicaid enrollment and billing
practices.
In October 1987, the Medicaid program adopted several changes to
simplify and streamline the determination of eligibility for pregnant
women. The processing changes involved: dropping the assets test for
pregnant women; granting continuous eligibility to pregnant women
1 This section is based on a report prepared by Ian Hill, which is
included as Appendix B.
throughout their pregnancies regardless of changes in income; adopting
presumptive eligibility, which permits state-selected providers to grant
immediate, temporary eligibility to low-income pregnant women while the
formal application is being processed; shortening the application form;
and outposting Medicaid eligibility workers at health care provider
sites to facilitate the application for Medicaid benefits. In an effort
to increase the supply of obstetrical providers from which the newly
eligible Medicaid beneficiaries could obtain prenatal care, the state
increased the Medicaid reimbursement rates for prenatal care during the
late 1980s and early 1990s to encourage greater physician participation
in the program. The largest single increase occurred in 1988 when the
global fee for maternity care was increased by more than 250 percent
(from $315 to $800) . The fee was raised again by 25 percent in 1989 and
increased 50 percent (to $1500) in 1992.
However, the strategy that most Florida officials credit with
having the greatest impact on the state's ability to effectively
implement the Medicaid expansions was the creation of Technical
Assistance and Coordination teams (TACTs) between 1988 and 1990. TACTs
were designed to provide assistance to counties in implementing their
indigent care programs. As is the case in many Southern states, public
health departments in each county in Florida have traditionally played a
major role in providing services directly to low-income individuals and
families. A significant portion of each county health department's
service capacity is devoted to the delivery of prenatal care. However,
prior to 1987, county health departments rarely billed Medicaid when
serving Medicaid-eligible women and children. TACTs worked with county
staff to increase the Medicaid revenues that the county health
departments collected when serving Medicaid-eligible clients by helping
them with procedures to improve eligibility determination and enrollment
of clients and increase Medicaid billings. (More detail about the
implementation of the Medicaid expansions and related reforms in Florida
is contained in Appendix B) .
DATA AND METHODS
Data
We studied all births in Florida occurring in the period July 1988
through June 1989 and in calendar year 1991. This study period includes
the twelve month period just prior to the expansion of eligibility to
women with income between 100 percent and 150 percent of poverty and the
second full calendar year after the expansion was implemented. We chose
July 1988 through June 1989 as the "baseline" year so that women with
income below poverty who delivered during the baseline period would have
been eligible for Medicaid throughout their pregnancy under the October
1987 expansion. We chose 1991 as the post-expansion period to allow
time for the new eligibility policy to be implemented. The other
strategies that the state adopted to complement the eligibility
expansion were all in place during both of these periods.
Data come from the Florida birth and death certificates, the
hospital discharge abstracts, Medicaid eligibility files and claims
files, individual encounter records for personal health services
provided through each county health department in the state, the
American Hospital Association (AHA) annual survey files for Florida
hospitals, and the 1990 Census.
The birth and fetal death records define our study universe and
provide information about the mother's demographic characteristics, the
amount and timing of prenatal care, and the birthweight of the newborn.
The birth records matched to infant death certificates also indicate
whether the newborn survived its first year.
The hospital discharge data provide information to identify the
primary payer for the delivery and to measure hospital charges. We
selected discharge records for the study periods that had an ICD-9
indicating a delivery or a hospitalization for prenatal care. A
computer algorithm linked the vital statistics and hospital discharge
data files using common variables including hospital, mother's date of
birth, date of birth or death (vital statistics) and date of first
procedure (hospital discharge file), and the mother's zipcode. We
matched 93 percent of the birth records for which the reported delivery
location was a hospital in the Florida hospital discharge data system
and 87 percent of the hospital discharge records that we identified as
deliveries using this algorithm (for more detail on the matching see
Appendix C) .
The Medicaid eligibility data allow us to distinguish reasons for
entitlement and, in particular, to classify separately pregnant women
covered by Medicaid because they are receiving AFDC cash assistance,
those who qualify under the medically needy program, and those covered
by the Medicaid income expansions. We linked the Medicaid eligibility
data to the matched vital statistics and discharge file using social
security number, which is also present in the vital statistics data. We
successfully matched 80 percent of birth records in which the hospital
discharge record indicated Medicaid as the payer to the eligibility file
(see Appendix C for further information) .
We constructed summary records for each episode of prenatal care
provided through the county health departments over the period 1987
through 1991 (we include care delivered prior to each study year in
order to measure all of the prenatal care associated with the deliveries
occurring in the study year) . These episodes were matched to our other
three data sets using social security number; we were able to match
about 75 percent of the episodes to the other three linked files.
Finally, we used the AHA data to measure the type of hospital and
added this information to our analytic file. Information from the 1990
Census on the income and poverty status of residents in each Florida
zipcode area was also added to the file in order to develop a control
measure for income status as we discuss later.
Methods
Our evaluation of the effects of the Medicaid expansion consists of
two parallel analyses. First, we investigate aggregate changes in the
quantity, delivery provider, and sources of financing for maternal
health care between the baseline period and 1991. Then we estimate
differences in the amount and timing of prenatal care and in the birth
outcomes for a woman covered by the Medicaid program and an uninsured
woman or one covered by some other payer. We also contrast birth
outcomes for Medicaid patients who use the county health departments for
their prenatal care with those using the private delivery system.
8
Aggregate Analvsis-
Overview. Our aggregate analysis examines the changes in the
primary source of financing for deliveries in the two study periods;
changes in the quantity of maternity related services, the type of
provider delivering care, and the financing of the services; and changes
in the flow of payments for maternal health care. We present a series
of matrices that categorize deliveries, services, and payments according
to the financing source--private insurance, Medicaid, and "other payer."
This last category includes those whose care was paid for by some other
third-party payer such as Medicare, CHAMPUS, or state and federal
programs that make payments on behalf of a patient receiving care, and
the uninsured. The hospital discharge data--our primary source for
payer--does not allow us to further classify these other payers. We
also are unable to separately identify a woman's source of insurance, if
any, for prenatal care.
The quantity of prenatal care and the number of prenatal and
delivery admissions are categorized by both payer and site of service.
For ambulatory care we distinguish between care provided in county
health departments and care provided at other sites, including physician
offices, hospital clinics, and hospital outpatient departments.
Hospital admissions are categorized by the type of hospital--public,
voluntary, or proprietary. Our measure of quantity and our measure of
payment covers care received by women who delivered in the study period,
irrespective of whether the care was provided in that period.
Measuring deliveries . The vital statistics records for all births
and fetal deaths registered to Florida residents measure the total
number of deliveries in each year. We distributed these deliveries among
the three payer categories based on the distribution of primary payer at
delivery for deliveries included in the matched hospital discharge file
and vital statistics file for each year.
Measuring use of services. The linked file also provides us with
an estimate of the average number of prenatal care visits received by
women in different payer statuses. We multiply these estimates by the
number of deliveries to measure aggregate prenatal care visits. The
2 This section overviews the aggregate analysis methods. Detail is
contained in Appendix D.
encounter data from the county health department system yielded a count
of the total number of prenatal care visits provided by county health
departments. We distributed this total among the different payers based
on the distribution of county health department visits that we were able
to match to the vital statistics/hospital discharge file.
The total number of hospitalizations for deliveries was counted
from the vital statistics data on location of delivery. We allocated
this total among the types of hospitals and payers based on the
distribution observed in our linked analysis file. We estimated the
number of prenatal admissions for women delivering in each study period
from the number of admissions in the period with ICD-9 codes related to
prenatal or maternity care that did not result in delivery. Because the
hospital discharge data lack individual identifiers, we are not able to
track the prenatal hospital admissions of women who gave birth in the
study period. We therefore approximate these by looking at all prenatal
admissions in a period, regardless of whether the woman actually gave
birth in the period. Because the number of births changes little from
year to year, this method provides a good estimate of the number of
prenatal admissions for women delivering in the study period.
Measuring payments . The third matrix we present shows the flow of
payments for maternal care in the two periods. It measures the direct
payments for care by patients and on account of patients by third-party
payers. That is, it measures what was actually collected by the
provider for the care of a particular patient. It does not include
contributions that are not tied to particular patients, such as federal
block grants to states for Title V programs and general contributions by
local governments to public hospitals for charity care.
To measure the payment flows, we started with an estimate of the
total charges for inpatient hospital services and for physician and
related services (such as laboratory tests and x-rays) categorized by
the delivery payer. The total charge estimates for each payer are based
on estimates of the average charge per quantity of service multiplied by
our estimates of the units of service as described above. For hospital
admissions, estimates of the average charge for women in different payer
statuses and in different hospital types are from the hospital discharge
data. For physician and other services, we use an estimate of charges
10
per prenatal visit from the Florida Medicaid claims file for 1991 and
from the claims data for two large employers in Florida as the basis for
assigning total charges per payer.
In order to distinguish direct patient payments from third-party
payments, we need to separate the charges for visits and admissions by
the uninsured from those for persons with a third-party source of
payment who are included in "other payer" . Because our primary data
sources do not provide this information, we used information about the
distribution of delivery payer for Florida sample persons in the 1988
National Maternal and Infant Health Survey (NMIHS) to estimate the share
of deliveries for uninsured women and the share of their visits that are
included in the "other payer" category to allocate the charges. We
allocated the charges for "other payers" in both of our study periods
using the estimated 1988 ratios. Because the Medicaid expansions would
be expected to decrease the share attributable to the uninsured, our
procedure will somewhat understate the increase in payments over the
period. More detail on the per case charges and on our procedures for
estimating the aggregate matrices are in Appendix D.
The resulting matrix of charges was converted to payments using
estimates of the ratio of payments to charges for different payers.
Payment to charge ratios for hospital care were provided by the Agency
for Health Care Administration in Florida. A Medicaid payment to charge
ratio for physician and related services was derived from the Florida
Medicaid claims files for maternity care. The ratio for private
insurance payers was based on the maternity care claims data from two
large Florida employers. Absent other data sources to provide payment
to charge ratios for physician services in Florida for the uninsured and
those covered by third-party payers other than private insurance and
Medicaid, we used the hospital ratios. Based on the similarity of the
private insurance and Medicaid ratios for hospital and physician
services, this seemed a reasonable assumption. Payments to county
health departments for prenatal care were measured from state Health
Office budget and revenue statistics for the county health department
system.
11
Person Level Analysis of Outcomes
Overview. Our objective is to measure whether expanding Medicaid
coverage to a previously uninsured woman improves her access to prenatal
care and reduces the occurrence of poor birth outcomes. The method is
to compare outcomes for women in different insurance statuses,
controlling for demographic and health factors that are known to affect
service use and birth outcomes. We compare four groups of women: those
enrolled in Medicaid under the eligibility expansions; those enrolled in
Medicaid because of their participation in AFDC; women covered by "other
payers" who reside in low-income areas; and women covered by private
insurance who reside in low-income areas. We compare outcomes among
these groups in each of the two study periods.
Because many believe that birth outcomes for low-income women
depend on the scope and content of care provided rather than the
quantity of care, we also compare outcomes for Medicaid beneficiaries
who use the county health departments for their care with those using
other providers. Through a special grant program in Florida, the former
emphasize comprehensive systems of prenatal care and provide support
services such as health education, nutritional counseling, social work
services, home visits, in addition to clinical prenatal services.
Comparison Groups. Our data do not permit us to precisely identify
uninsured women who would be eligible for the Medicaid expansions for
two reasons. First, our insurance measure from the hospital discharge
file does not distinguish among the uninsured and those who are covered
by other non-private third party payers (except Medicaid) . Second, we
do not have a measure of the woman's income. As a proxy for Medicaid
income eligibility, we measure outcomes for women who live in areas in
which over 30 percent of the population has a family income below 150
percent of poverty (the threshold of the Medicaid expansion in 1988) .
This definition encompassed the poorest quintile of neighborhoods; on
average, over 40 percent of the population in these neighborhoods had
income below 150 percent of poverty. The difference in outcomes between
women whose deliveries were covered by Medicaid and women in low-income
areas with "other payer" is our measure of the effect of providing
Medicaid coverage to the uninsured. This measure may understate any
positive effects of providing public coverage to uninsured women to the
12
extent that not all of those with "other payers" are uninsured women.
This bias, however, should be small because the NMIHS of 1988 shows that
about two-thirds of all women with "other payer" are uninsured, and this
fraction would be expected to increase among low- income women.
We also compare outcomes for those on Medicaid and the uninsured
with those for women who are covered by private insurance. Again, we
limit our comparison to women residing in low- income areas as a proxy
measure to control for economic circumstances of the individual. Other
research (Starfield et al., 1991; Rosezweig and Schultz, 1982) has shown
that there are income effects related to seeking prenatal care and to
birth outcomes .
The difference in outcomes between Medicaid beneficiaries and low-
income uninsured women who do not enroll in Medicaid may overstate the
effect of providing Medicaid coverage if those who enroll in Medicaid to
receive maternity care are healthier, or more likely to use prenatal
care than those who do not enroll. To control for this selection, we
distinguish between those who are enrolled in the expansions for
pregnant women and women whose enrollment in Medicaid is unrelated to
their pregnancy but stems from their participation in AFDC. Differences
between the two Medicaid populations is a measure of selection. We
interpret differences between the AFDC Medicaid population and the low-
income uninsured as a measure of the insurance effect of Medicaid,
controlled for any self -selection in the expansion population.
For Medicaid beneficiaries, we look at differences in outcomes for
patients using the county health departments for their prenatal care and
those using other providers. In this analysis, we categorize a woman
according to where she received the majority of her prenatal care
visits. Users of the county health department system are therefore
women who used the clinics for at least half of their prenatal visits.
Our results were not sensitive to this definition. We obtained similar
results if we included women who received any of their care at the
county health department as using this system. We examine whether the
effect of choice of delivery system on outcomes differs between Medicaid
enrollees in the expansion program and the AFDC Medicaid beneficiaries.
Outcome Measures. The outcomes that we investigate are measures of
prenatal care use and birth outcomes. We examine whether any prenatal
13
care was obtained, the timeliness of initiating care among those who
sought care, and the number of prenatal care visits for those receiving
prenatal care. Timeliness is defined as seeking care prior to the third
trimester. We also measure access using two indices of adequacy of
care: the Kessner Index (1973) and Kotelchuck Index (1994) . Both
indices are based on the timing of initiation and the number of visits,
adjusted for gestational age. Neither index reflects the content of
care, which many analysts cite as an important correlate of birth
outcomes. However, we indirectly investigate the effect of the scope
and content of care on birth outcomes by comparing Medicaid women who
use the county health departments for their care and women using the
private delivery system.
We also investigate differences in the occurrence of adverse
outcomes among women with different insurance status. The adverse
outcomes we examine are low birthweight (less than 2500 grams) , very low
birthweight (less than 1500 grams) , and infant death (death within the
first year after birth) . Because we did not have death certificates for
deaths occurring in 1992, we have missed some infant deaths for babies
born late in 1991. This will cause us to understate the number of
infant deaths in that year, but should not bias our comparison of
different insurance groups. To correct for the underestimate, we
multiplied our estimates of infant deaths in 1991 for all subgroups by a
constant factor (1.20) that reflects the ratio of all infant deaths to
infant deaths occurring in the calendar year of birth based on the 1988-
1990 observations.
Sample. Our person-level analysis includes women with a live birth
for whom we could find a matching hospital discharge record. We exclude
fetal deaths in this analysis because of missing data on important
demographic characteristics that we wish to control for in our
comparisons. Most notably, we do not have zip code to assign an income
status. These fetal deaths, however, account for only about one-half of
one percent of all deliveries in a year. We also exclude births of
less than 500 grams (0.15 percent of records).
Our analysis is restricted to women whose birth record includes
information to calculate the key outcome measures and demographic
characteristics. This restriction eliminated 5.4 percent of records in
14
our matched file in the baseline period and 5.2 percent of records for
1991. Our final analysis sample included 156,453 women in the baseline
period and 164,039 women in 1991. Excluding women with missing
information about any outcome disproportionately excludes babies who die
soon after birth because their birthweight is often not recorded. Thus,
the infant death rate for our analysis sample is lower than the infant
death rate for all births. However, our estimates of differences in the
death rates among payer groups are not affected by this exclusion. We
verified this by producing estimates (not reported) of infant death
rates by payer for all births as well as the analysis sample estimates.
To study differences in the outcomes among women treated in
different delivery systems, we limit our contrast to women enrolled in
Medicaid in 1991, either in the expansion program or because of their
AFDC participation. Our sample for this analysis included 58,751 women.
Analysis. We use regression analysis to control for the effects of
differences between our contrast groups in demographic characteristics.
Indicators for insurance status measure the effect of payer on the
outcomes. The indicators in the model distinguish among women enrolled
in Medicaid according to the reason for their entitlement, women with
private insurance according to the income of the area in which they
reside, and women without private insurance or Medicaid according to the
income of the residence.
The regression models fit to estimate differences in outcomes for
women in different delivery systems use the Medicaid enrollees in 1991.
The models include the indicator for the delivery system used, an
indicator for whether an AFDC or expansion enrollee, and the interaction
between these in order to investigate whether the effects of the choice
of delivery system differ for those in the expansion program and other
Medicaid patients.
In addition to the insurance indicators, the explanatory variables
in the model measure the mother's age, education, race, marital status,
parity, and whether the birth was a singleton birth. For 1991, we also
include measures of ethnicity and whether the mother had any medical
risk factors. We fit ordinary least squares regression for the number
of prenatal care visits. For the other outcome variables, which are
dichotomous, we fit logistic regression.
15
We use the regression model to estimate the outcomes for our
comparison groups standardized to a common set of demographic
characteristics. The reference for our comparisons is a woman whose
characteristics assume the average value of these characteristics for
women in the Medicaid expansion population in 1991.
To illustrate how we use the regressions to produce our results,
let Y be the outcome of interest. We fit a regression model of the
form:
Y = p0 + PjfAFDC) + P2 (Uninsured) + P3( Privately Insured) +
P,*x,
where AFDC, Uninsured, and Privately Insured are indicators for the
source of the financing of the delivery and X denotes the demographic
characteristics of the women. In our tables of results, we present
predicted estimates of the outcome Y for a woman with characteristics
given by Xe, which denotes the average value of the characteristics for
women in the expansion population in 1991, while varying the insurance
group. Table 1 summarizes the predicted values in terms of the
regression coefficients.
Table 1. Predicted Values for Comparing Insurance Groups
Privately
Expansion AFDC Uninsured Insured
Outcome Y p0+p/X. P.+fc+P/X. P„+P,+P4*X. P„+P3+P4*Xe
The difference in the predicted value for any two groups,
therefore, reflects the difference due only to insurance status because
the effects of demographic characteristics are the same across groups.
(See Appendix E for the definitions of our explanatory variables and the
values for the reference population. See Appendix F for the regression
parameter estimates.)
16
RESULTS
Aggregate Analysis
Chancres in Coverage
The Medicaid expansions led to substantial shifts in the source of
payment for deliveries in Florida between the baseline period and 1991.
In just two and one-half years, the number of births covered annually by
Medicaid rose from 47,000 to 70,000, a 47 percent increase (Table 2 and
Figure 1) . Because there was only a 2 percent increase in total births
per year over this period, nearly all of the Medicaid growth represented
a shift among payment sources. The proportion paid for by Medicaid rose
from 25 percent to 36 percent. In contrast, the proportion of births
covered by private insurance remained essentially the same over the
period. Therefore, the public expansions did not substitute for private
insurance. Instead, they either covered the uninsured or replaced other
sources of third party coverage (for example, Medicare, CHAMPUS, and
other federal and state programs that pay for patients' care). Although
our data do not permit us to distinguish between these two groups of
"others," data from the NMIHS in Florida show that in the baseline
period two-thirds of the group without Medicaid or private insurance
(about 33,000 women) were uninsured. Thus it is likely that the
expansion, which covered 23,000 additional deliveries per year, served
largely to cover women who otherwise would have been uninsured.
Table 2. Deliveries by Primary Delivery Payer
Payer
Number of Deliveries
7/88-6/89
1991
Percent
Change
Private Insurance
Medicaid
a
Other Payer
91,948 (48.7%)
47,413 (25.1)
49,432 (26.2)
89,108 (46.1%) -3.1?
69,643 (36.0) 46.9
34,541 (17.9) -30.1
Total
188,793 (100.0)
193,292 (100.0)
2.4
Note: Column percent in parentheses.
Includes other non-private third party payers (e.g., CHAMPUS,
Medicare, other state and federal programs that make payments on behalf
of patients) and the uninsured.
17
100%
90%
80%
CO
70%
LU
rr
>
60%
_l
LU
Q
50%
LL
o
(-
40%
z
LU
()
(T
30%
LU
0.
20%
10%
0%
49%
26%
25°/
□ Private
insurance
□ Other payer
I Medicaid
Baseline (7/88-6/89)
TIME PERIOD
46%
18%
36%
1991
Figure l--Distribution of Deliveries by Primary Payer
Changes in Ambulatory Prenatal Care Visits
Total ambulatory prenatal visits by all women rose by 150,000, or
about 7 percent between the baseline period and 1991 (Table 3). This
was greater than the increase in total deliveries, and so there was
about a 5 percent increase in prenatal visits per person, on average.
The increase in prenatal visits per person occurred among all three
coverage groups. However, the increase in prenatal care visits per
person for those on Medicaid and others without private insurance (the
group that previously included the newly entitled Medicaid
beneficiaries) increased by about 7 percent over the period, whereas the
increase for the privately insured was about 4 percent. So viewed in
the aggregate, the Medicaid expansion appears to have led to a small
increase in access for the non-privately insured relative to what we
Table 3. Ambulatory Prenatal Visits by Primary Delivery Payer
7/88 - 6/89 1991
County Other County Other
Health Delivery Health Delivery
Total Dept . System Total Dept ■ System
Number of Visits
Private Insurance 1,133.9 13.0 1,120.9 1,141.8 1674 l, 125.4
Medicaid 466.6 177.2 289.4 729.7 433.3 296.4
Other Payer' 473.7 59.7 414.0 352.9 48.1 304.8
Total 2,074.2 249.9 1,824.3 2,224.4 497.8 1,726.6
Percent Distribution by Payer (row percent)
Private Insurance 100.0 1.1 98.9 100.0 FTi 98TeT
Medicaid 100.0 38.0 62.0 100.0 59.4 40.6
Other Payer' 100.0 12.6 87.4 100.0 13.6 86.4
Total 100.0 11.9 88.1 100.0 22.4 77.6
Percent Distribution by Payer (column percent)
Private Insurance 54.7 5.2 61.4 51.3 3 . 3 55 2
Medicaid 22.5 70.9 15.9 32.8 87.0 17.2
Other Payer' 22.8 23.9 22.7 15.9 9.7 17.6
Total 100.0 100.0 100.0 100.0 100.0 100.0
'Includes other non-private third party payers (e.g., CHAMPUS, Medicare, other state and federal
programs that make payments on behalf of patients) and the uninsured.
19
expect based on the trend for the privately insured. But we return to
the question of effects on access and outcomes later.
Striking changes occurred in the quantity of prenatal care
delivered by different parts of the delivery system. The total number
of visits provided by county health departments rose by 250,000 (100
percent) , a number greater than the total increase in visits at all
sites combined. The proportion of visits provided by health departments
rose from 12 percent to 22 percent. In contrast, total visits to all
other ambulatory care sites fell somewhat.
The Medicaid eligibility expansion was almost fully accommodated by
this huge growth of care provided by county health departments. Of the
263,000 additional prenatal visits by Medicaid women in 1991, 256,000
took place there. As a result, a significant shift in where Medicaid
women received their care accompanied the expansions (Figure 2) . By
90% •
80% •
46%
UJ
b
</>
>-
m
(0
70%-
60% •
62%
D Other sites
8
>
u.
o
i
o
2
UJ
K
50%-
40%-
30% ■
^^^H
I 59%
20% ■
38% 1
■ Health
Dept.
i^^H
10% -
i^H
i^H
0% -
i ^ — ^ 1
Baseline (7/88-6/89) 1991
TIME PERIOD
Figure 2--Distribution of Prenatal Care Visits Among Medicaid Women by
Site of Care
20
1991, county health departments provided 59 percent of ambulatory-
prenatal care for this coverage group, up from 38 percent only two and
one-half years earlier. As a result, care delivered to Medicaid
beneficiaries at other sites decreased.
Another effect of increasing Medicaid coverage was to reduce the
number of visits provided to uninsured patients or to those covered by
other non-private third party payers by 121,000, or almost 2 6 percent.
According to the 1988 NMIHS, about 63 percent of these visits were
visits by uninsured patients, and so a large fraction of the decrease is
visits previously provided as charity care by hospitals or physicians.
Visits to county health departments fell by 12,000, or 10 percent of the
total decrease. Most of the decrease was in visits to other sites which
fell by 109,000 visits, or almost 90 percent of the total decrease.
Changes in Hospital Admissions
The Medicaid expansion had a much smaller effect on the types of
hospitals women used, when compared to the substantial shifts in the
sites of ambulatory care. For all payers combined, use of public
hospitals declined about 3 percent, from 29 percent of admissions in the
baseline period to 27 percent in 1991 (Table 4) . The use of voluntary
hospitals declined by a similar percentage. In contrast, the increase
in use came at proprietary hospitals, whose share of maternity care rose
from 14 percent to 17 percent. However, most of the increase in the
admissions to proprietary hospitals is a general trend for all payer
groups and not a consequence of the expansions.
Although there was little change in the number of admissions of
pregnant women to public and voluntary hospitals, there was a sizable
increase in the share of these admissions that are financed by Medicaid
and a sizable decrease in the share that are self-pay or financed by
some other government program (Figure 3). For public hospitals, the
Medicaid share increased from 31 percent to 47 percent whereas for
voluntary hospitals it increased from 25 percent to 36 percent. For
public hospitals, the share attributable to other payers fell from 42
percent to 28 percent, for voluntary hospitals the decrease was from 22
percent to 16 percent. There was also an increase in the share of
admissions to proprietary hospitals that was financed by Medicaid and a
21
100%
80% ..
50% -■
□ Other
Voluntary,
Baseline(7/88-
6/89)
TYPE OF HOSPITAL AND TIME PERIOD
Voluntary,
1991
Figure 3--Distribution of Source of Financing by Type of Hospital and
Time Period
reduction in the share financed by other payers, but the magnitude of
these changes was not large because the vast majority of admissions to
proprietary hospitals are for the privately insured (70 percent in
1991) .
Medicaid eligibles use a very different mix of hospitals than the
privately insured. For example, in 1991, public hospitals provided 36
percent of maternity-related admissions for Medicaid women versus 15
percent for the privately insured. In that same year, voluntary
hospitals accounted for 56 percent of Medicaid admissions and 58 percent
of privately insured admissions. Finally, proprietary hospitals only
accounted for 9 percent of Medicaid admissions versus 27 percent for the
privately insured. The pattern for the uninsured and those covered by
other payers is similar to that of Medicaid, but with greater use of
Table 4. Maternal Hospital Admissions by Primary Delivery Payer
7/88 - 6/89 __ZZZ!Zm 1991
Public Voluntary Proprietary Public Voluntary Proprietary
J__J Hospital Hospital Hospital Total Hospital Hospital Hospital
Number of Admissions
Private Insurance 102.2 15.9 65.4 20.9 9777 TTTs 5679 2 6 0
Medicaid 53.6 19.0 30.2 4.4 77.1 27.4 42.7 7 o
Other Payer' 56.4 25.5 26.1 4.8 38.8 16.3 18.6 3.9
Total 212.2 60.4 121.7 30.1 213.6 58.5 118.2 36.9
Percent Distribution by Payer (row percent)
Private Insurance 48.2 26.3 53.7 69.4 45 .7 25 3 48~_2~
'Includes other non-private third party payers (e.g., CHAMPUS, Medicare, other state and federal
programs that make payments on behalf of patients) and the uninsured.
Private Insurance 100.0 15.5 64.0 20.5 100 . 0 157! 5871 26 6
Medicaid 100.0 35.4 56.3 8.3 100.0 35.5 55.4 g.j
Other Payer' 100.0 45.2 46.3 8.5 100.0 42.0 47.9 10. 1
Total 100.0 28.5 57.3 14.2 100.0 27.4 55.3 17.3
Percent Distribution by Payer (column percent)
70.4
Medicaid 25.2 31.5 24.8 14.6 36.1 46.8 36.1 19.0
Other Payer' 26.6 42.2 21.5 16.0 18.2 27.9 15.7 10.6
Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0
K>
23
public hospitals (42 percent versus 3 6 percent) and lower use of
voluntary hospitals (48 percent versus 55 percent) .
Changes in Payments for Maternity Care
Payments to providers for maternal care in 1991 totaled $928
million dollars, up 3 percent from the baseline period total of $902
million, expressed in constant 1991 dollars (see Table 5) . Medicaid
payments, which are financed with about 55 percent federal funds and 45
percent state funds in Florida, increased by about $52 million, or 38
percent. Medicaid payments for persons in the expansion population
increased by 140 percent; they totaled $105.7 million in 1991, up from
$43.3 million in the baseline period. Medicaid payments for other
beneficiaries declined by 11 percent or $10.3 million over this period,
from $91.9 million in the baseline period to $81.6 million in 1991. In
contrast, payments by other third parties and the uninsured fell by
almost 30 percent.
Hospitals benefited from the expansions. Their maternity-related
revenues grew by 5 percent, whereas admissions of maternity patients
remained fairly constant (Table 4) . This extra revenue largely results
from Medicaid payments for admissions that were previously provided as
bad debts or charity care.
In contrast, payments for physician care and related services rose
by only 0.4 percent in constant 1991 dollars, whereas ambulatory care
prenatal visits increased by 7 percent. Two main reasons account for
this difference. First, total physician payments are a combination of
payments for ambulatory prenatal care and inpatient care, mostly for
deliveries. Deliveries and total physician payments remained relatively
constant. Second, the increase in ambulatory prenatal care was
accommodated almost entirely by county health departments.
Medicaid accounted for 20 percent of total payments for maternity-
related care in 1991, although 36 percent of deliveries in that year
were to women with Medicaid coverage. This difference arises because
Medicaid pays a smaller proportion of what it is charged for maternal
health services than private insurers. Private insurers accounted for
63 percent of payments in 1991, although only 46 percent of deliveries
24
Table 5. Flow of Payments for Maternity Care
(Millions of 1991 dollars)
Paver
Physician
Services
Hospital
Services
7/88 - 6/89
Private Insurance
Medicaid
Other Third Party Payer
Self Pay
Cost Sharing
Uninsured
Total
298.3 i 67.0%) 280.1 ( 61.3%)
51.4 ( 11.5) 83.8 ( 18.3)
35.0 ( 7.9) 56.5 ( 12.4)
44.5 ( 10.0)
15.9 ( 3.6)
445.1 (100.0)
14.7 ( 3.2)
22.0 ( 4.8)
457.1 (100.0)
1991
Private Insurance
Medicaid
Other Third Party Payer
Self Pay
Cost Sharing
Uninsured
Total
299.5 ( 67.0)
65.1 ( 14.6)
25.9 ( 5.8)
44.6 ( 10.0)
11.6 ( 2.6)
446.7 (100.0)
287.5 ( 59.7)
122.2 ( 25.5)
41.0 ( 8.5)
15.1 ( 3.1)
15.5 ( 3.2)
481.3 (100.0)
Percent Change
private Insurance
0
. 4
Medicaid
26
.7
Other Third Party Payer
-26
.0
Self Pay
Cost Sharing
0
.2
Uninsured
-27.
0
Total
0.
,4
2.6
45.8
-27.4
2.7
-29.5
5.3
Total
578.4 ( 64.1%)
135.2 ( 15.0)
91.5 ( 10.1)
59.2 ( 6.6)
37.9 ( 4.2)
902.2 (100.0)
587.0 ( 63.3)
187.3 ( 20.2)
66.9 ( 7.2)
59.7 ( 6.4)
27.1 ( 2.9)
928.0 (100.0)
1 . 5
38.5
-26.9
0.8
-28.5
2.9
Note: Column percents in parentheses.
25
were to women with private insurance coverage. Direct patient payments-
-including cost sharing by insured patients and payments made by
uninsured patients--accounted for 9 percent of funds. This distribution
of payments by source in Florida is similar to national patterns (Long,
Marquis, and Harrison, 1994) .
Person Level Analysis of Outcomes
Our analyses of the individual effects of the Medicaid eligibility
expansions are based on the following causal model. For the expansions
to improve birth outcomes for those who enroll rather than remain
uninsured, several conditions must be met. First, women must enroll for
Medicaid-paid services early in their pregnancies. Then, they must
initiate prenatal care earlier and receive more care than they would if
uninsured. Third, the content of the expanded care must be associated
with an increased likelihood of good birth outcomes. We examine each
stage in this process. Although a finding of better overall birth
outcomes for the expansion population would be suggestive of an effect,
findings that establish all of the steps in this causal chain serve to
increase our confidence that effects we observe can be attributed to the
Medicaid eligibility expansion and other Medicaid program changes that
were introduced along with the eligibility changes.
Timing of Medicaid Eligibility
An increasing proportion of the Medicaid expansion population
enrolled early in their pregnancies over the study period (Figure 4) .
Although only 27 percent of women in the expansion population who
delivered between July 1988 and June 1989 enrolled in the first
trimester, by 1991 the rate increased to 46 percent. Most of the
increase occurred early in our study period; in the first 12 months
following the increase in the income threshold to 150 percent of
poverty, 40 percent of the expansion population enrolled in the first
trimester. This improvement is not explained by earlier enrollment
among the higher income women newly eligible as a result of the increase
in the income threshold. We tested for this possibility by examining
the early enrollment rates over time for the subset of the expansion
population having income below 100 percent of poverty. The change in
26
timing of their enrollment was similar to that for the entire expansion
population. This suggests that the procedures Florida adopted in 1987
to simplify eligibility and enroll women earlier in their pregnancy were
successfully implemented by the end of 1988.
46
45-
40
40-
IT
Ul
1-
Cfl
Ul
s
E
i-
i-
N
35-
30-
27
z
a
z
-i
_i
o
IT
Z
111
25-
20-
■
z
Ul
o
i
E
15-
10
5-
|
Baseline period
12 months after expansion
Calendar
(7/88-6/89)
(7/89-6/90)
TIME PERIOD
1991
Figure 4--Percent of Medicaid Expansion Population Enrolling in First
Trimester, by Time Period
The share of the expansion population enrolled early in their
pregnancy in 1991 compares favorably with rates in other states (cited
earlier) that have implemented a range of strategies to simplify the
eligibility determination process. Nonetheless, half of women in
Florida who ultimately enroll under the expansion rules did not do so
during the first trimester.
Prenatal Care Use
Women enrolled in the Medicaid expansion program use more prenatal
care than women living in low-income areas who are without Medicaid or
private health insurance- -hereafter termed "uninsured" women. In both
study periods, the Medicaid beneficiaries were less likely to forego
27
prenatal care (Table 6) . The proportion not receiving prenatal care
ranged from 1.3 percent to 1.4 percent for the Medicaid expansion group,
significantly less than the rate of 3.4 percent to 3.7 percent for
uninsured women. Among those seeking care in 1991, the percent of
Medicaid beneficiaries initiating care late in the pregnancy (4.7
percent) was below the rate for uninsured women (5.4 percent); however,
there was not a significant difference in timeliness in the baseline
period. Medicaid expansion women obtaining prenatal care had more
visits than the uninsured in each period. Overall, the Medicaid
beneficiaries were less likely to receive inadequate care than the low-
income uninsured women.
This result does not appear to be due to the selection of healthier
women or those who are more disposed to seek prenatal care into the
Medicaid expansion population. We found similar differences between
women who are enrolled in Medicaid because they receive AFDC cash
assistance and uninsured women. Medicaid eligibility for this group is
based only on having low-income and assets, and not on being pregnant.
However, some women applying for Medicaid pregnancy benefits may have
found they were also eligible for AFDC cash assistance, signed up for
the cash benefits, and been designated as an AFDC recipient by Medicaid.
Their participation in Medicaid is related to being pregnant. Including
these women in our "selection control" group would bias the comparison.
To test for this, we also contrasted the expansion population with AFDC
recipients who were enrolled in the program for most of their pregnancy
(enrolled prior to the pregnancy or during the first trimester) and
found similar results. Moreover, our estimates indicate that the number
of pregnant women in eligibility groups other than the expansion group
did not increase over time, and in fact, decreased slightly.
Although providing Medicaid benefits to low-income women appears to
increase their use of prenatal services relative to what they would be
expected to use if uninsured, still the Medicaid women do not receive
the level of prenatal care obtained by privately insured women. On all
of our measures in both periods, Medicaid women have significantly
Table 6. Prenatal Care and Birth Outcomes for Low- Income Women by Insurance Status
Measure
July 8 8 - June 89
1991
Medicai d
Expans ion
Medica id
AFDC
Oilier Low-
Income*3
Private
Low-
Income"
Medicaid
Expansion
Medicaid
AFDC
Other Low-
Income3
Pi l vat e
Low-
Income"
Prenatal Care
No prenatal (%)
User initiates third
trimester (%)
Visits/user
Inadequate care (%)
Kessner
Kotel chuck
1 .4 +
1.9
7.3.
6.4
11.0
10.9
8.9 +
3 5.2*
8.3*
33.9*
3.7*i
0.8**
1.3 +
1 .6
7.0.
2.6*.
4.7
4.6
9.8*.
11.7*+
11.2
11.2
11.2'
4 1
3.4* +
5.4* +
10.6*+
0.7'
1.5'
12.2'
3.5* +
6.4
6.5
9.4* +
2.5
20.8*+
22.5
22.1
24.7*4
H.H
Birth Outcomes
Low birthweight/1000
Very low birthweight/1000
Infant death/100QC
67.2
65.1
72. 1 +
55.4*+
60.6
61.3
68.2
8.9
7.8
9.9 +
9.0
9.2
8.1
9. 1
6.9
6.6
8.5.
8.6
6.5 +
5.0*
6.7
i 4 . 8 '
8.7
7.2
Pregnant women with "other" payer living in areas with more than 30 percent oE population with ncome below 150 percent poverty
b
Pregnant women with private insurance living in areas with more than 30 percent of population with income below 150 percent
poverty,
c
Analysis sample estimate underestimates population rate; 1991 rate adjusted to include deaths occurring in 1992. See l^xt for
detai Is .
•Significantly different from Medicaid Expansion population, p =.05.
.Significantly different from Medicaid AFDC population, p =.05.
29
poorer access to care than the privately insured. While there is
improvement over time in access for the Medicaid women, similar
improvements are also seen for the privately insured and the uninsured.
The changes in access over time thus seem to be the result of an overall
trend in the state, rather than the effect of the Medicaid program
changes. The expansion, however, improved access for low-income women
because it provided coverage to many previously uninsured women--who
have poorer access than those enrolled in Medicaid.
Birth Outcomes
The rate of low and very low birthweight infants and the rate of
infant death is generally lower among women enrolled in the Medicaid
expansion program than it is for the uninsured (Table 6) . However, only
the difference in the incidence of low birthweight infants in 1991 is
statistically significant; in that year, the rate of low birthweight
infants per 1000 was 60.6 for the expansion population versus 68.2 for
the uninsured. Women enrolled in AFDC also have better birth outcomes
than the uninsured; we find statistically better outcomes on all three
measures in the baseline period and on the low birthweight measure in
1991.
Consistent with the utilization findings, women covered by Medicaid
remain significantly more likely to give birth to a low birthweight
infant than women with private insurance.
Role of Different Delivery Systems
The additional prenatal care financed by Medicaid that resulted
from the expansion of eligibility to more low-income women was provided
largely by county health departments. As shown in Table 7, this may
have been an important factor in the better outcomes for the expansion
population compared to those for the uninsured. The rate of low
birthweight infants per 1000 among the Medicaid expansion mothers using
the county health departments was 49.9, versus 70.4 for the other
expansion mothers. The incidence of very low birthweight infants and
infant deaths was also significantly lower among women enrolled in the
Medicaid expansion who used the county heath department for their
prenatal care than among similar women who obtained their care in
Table 7. Prenatal Care and Birth Outcomes for Medicaid Women Using Different Delivery Systems, J 991
Medicaid Expansion Population Medicaid AFDC Populatioi
County Health Other Delivery County Health Other Delivery
Measure Department System Department System
Prenatal Care
User initiates third trimester (%) 6.2 3.7* 6 \ g
Visits/user 10.7 11.8* 10.5
Inadequate care (%)
Ressner 6.7 4.1 7.2
Kotelchuck 25.5 18.8* 27.3
Birth Outcome
a
Infant death/1000 4.3 6.0* 3.
4.2*
11.7*
4.6*
21.3*
Low birthweight/1000 49.9 70.4* 5476 ' 599*
Very low birthweight/1000 6.9 10.9* 6.4 9 5* ej
4.2
Analysis sample estimate underestimates population rate; 1991 rate adjusted to include deaths occurring
in 1992. See text for details.
♦Significantly different from county health department, p=.05.
31
another delivery system. We find better outcomes for women using the
county health department even though they initiated care later and had
fewer visits than women using another delivery system.
Because the county health departments made special efforts to
enroll low-income pregnant women in the Medicaid expansion program,
there may be differences between the expansion population in the two
delivery systems not accounted for by our control variables that might
explain this result. However, women using the county health department
who are enrolled in Medicaid because of their AFDC eligibility also have
better birth outcomes than similar women using another delivery system.
The finding is also not a result of referrals to another delivery
system for women at risk of poor birth outcomes. Our estimates are
adjusted for a measure of medical risk factors. In addition, we obtain
similar results (not shown) when we compare outcomes for women who
received any of their prenatal care in the county health department with
all other Medicaid women.
DISCUSSION
Findings From the Florida Expansion Compared to Other Studies
The Florida Medicaid eligibility expansion from 100 percent of
poverty to 150 percent of poverty led to a large increase in Medicaid
enrollment by pregnant women who would otherwise have lacked insurance
coverage to pay for their prenatal care and delivery. We found strong
evidence that women in the expansion population had better access to
prenatal care than they would have had if they remained uninsured. Our
results also consistently point to improved birth outcomes for the
expansion enrollees.
We come to stronger conclusions about the benefits of the expansion
than most of the earlier literature. Is this because our methods are
more precise, because Florida differs from other states, or both? We
believe the answer is both.
Large sample sizes are needed to precisely measure the birth
outcomes we are trying to study, because they are very rare events.
This suggests that research needs to focus on either the most populous
states or places where policy held constant long enough to permit
32
pooling several years of data. Hence, Florida represented an
opportunity to have large enough samples to find effects if they were
present, and to stratify to more homogeneous subgroups to refine the
comparisons. Moreover, our use of area income to identify the subset
of women who are most likely to be uninsured and eligible under the
expansion is a methodological improvement over other studies. Finally,
earlier cross-section comparisons have been cautious in attributing
differences between the expansion population and uninsured women to
effects of the Medicaid program because of possible selection bias. We
rule out selection as an explanation by finding similar results for
Medicaid beneficiaries who are not enrolled in the program as a result
of their pregnancy.
The Florida experience, however, may differ from other states.
Most of the additional prenatal care financed by Medicaid was
accommodated in the county health departments, and this resulted in a
substantial increase in the quantity of prenatal care provided by public
clinics. This may not have happened in other states that have been
studied. From our finding of better birth outcomes among Medicaid
enrollees using the county health department compared to those using
another delivery system, it appears that the county health department
expansion was an important feature of the Florida intervention. Without
it, the improvements probably would have been more modest. Some
observers of the expansions have pointed to the importance of expanded
services for low-income women. In Florida, Medicaid did not pay for
these services, but the effect of offering Medicaid revenues to the
county health departments was to provide revenue to cover basic medical
services so that non-Medicaid funds could be used to support those
additional services.
Policy Implications
The Congress is currently considering major changes in federal
funding for Medicaid and public health, most of which would control the
growth of federal spending and provide the states with greater
flexibility to use federal funds as they see fit. In this context our
study findings are important. First, they suggest that the expansions
may indeed have had an impact, so as policymakers consider spending
33
reductions, they should be cautious about cutting back on eligibility
for the expansion population.
Second, the results emphasize the inter-relationship of expanding
insurance coverage and providing for a delivery system to accommodate
peoples' needs. Specifically, our findings suggest that of the low-
income women benefiting from the expansion in insurance for their
pregnancies, those who used the county health departments for prenatal
care had the better outcomes. On the other hand, some states have
financed their Medicaid expansions, in part, by a contraction of their
public health systems, assuming that the increased financial access
provided by Medicaid would lead more low-income women to use the private
delivery system (Alpha Center, 1995) . Our findings indicate the
possibility that this could have unintended unfavorable effects on birth
outcomes .
Some states have accompanied eligibility expansions with fee
increases to try to remove barriers to office-based care. Some are
emphasizing enrollment in managed care. Although increasing financial
access would be expected to have beneficial effects, our results suggest
that it is not clear what the ultimate outcome of these trends will be,
especially in states with a strong tradition of direct delivery through
the public health system. Birth outcomes might deteriorate if these
efforts to shift care to the private sector are not complemented by
programs to provide the non-clinical support services to pregnant women
that the public health system now provides.
Several of our findings suggest that despite its contributions, the
Florida intervention may not have achieved the full potential of such
efforts. First, although the improvements in public awareness of the
expanded eligibility limits and in eligibility processing appear to have
had an impact on the percent of women becoming eligible early in their
pregnancies, by 1991 it was still the case that about half of women who
became eligible for Medicaid-paid deliveries did not become eligible
during their first trimester. This suggests room for further
improvement. Second, there remains a significant gap between Medicaid
eligible women and low- income, privately insured women in use of
prenatal care and in birth outcomes.
34
Our study provides new information about the Medicaid expansions in
one state. However, national policy can not be based on one case study
alone. Therefore, it will take study of more states with varied
circumstances to fully evaluate the effects of this major initiative in
Medicaid from the last decade that is still playing out over this one.
Directions for Further Research
The Medicaid expansions were the major policy change in the program
in the 1980s. The program is now undergoing many other new changes and
facing new challenges--including continued income expansions, raising
physician fees to encourage use of office-based care, enrolling Medicaid
patients in managed care, and Congressional consideration of options to
control federal costs. While these changes do not all focus specifically
on pregnant women, they may alter the availability of public programs to
finance care for low- income pregnant women, the amount and kind of
prenatal care that low- income women receive, and birth outcomes.
The findings from our study of the 1989 Florida eligibility
expansion raise a number of important questions about the effect that
the new directions in Medicaid will have on access to care for low-
income pregnant women and on their birth outcomes .
• Will eliminating financial barriers have the same effect on
access to care and outcomes for the near poor--who are the
subject of the more recent expansions in Florida and other
states?
• Does Medicaid eligibility improve access and outcomes for
important subgroups of pregnant women--especially teenagers
and women at high risk for poor birth outcomes?
• What are the likely effects on birth outcomes if increasing
physician fees leads to a shift to more prenatal care
delivered by office-based physicians? What are the likely
effects on birth outcomes for Medicaid beneficiaries of the
new emphasis on managed care?
•
•
Will the effects of the expansions change when the public
health system can not expand further to meet the increased
demand?
Does providing care directly to uninsured women through the
public health system have the same effect on prenatal care use
and birth outcomes as providing public insurance to pay for
care received in the private sector?
35
Answers to these questions are critical to understanding the
effects the recent policy changes in Medicaid will have on maternal
health care and infant health, and therefore to evaluating some of the
costs and benefits of these policy changes. This kind of information
will be of paramount importance to state policymakers if federal
proposals to restructure the Medicaid program and limit the growth in
federal contributions are implemented. Some of these proposals would
provide block grants to states, providing states with new flexibility in
how they structure programs to provide health care services to their
low-income population. If this occurs, state policymakers will need to
make tradeoffs between expenditures on public insurance programs and
direct delivery of services to the low-income population; they will need
to evaluate the costs and benefits of care delivered in different
settings; and they will need information about how different programs
benefit especially vulnerable populations.
Florida remains a good candidate state to study to answer these
questions. A number of these questions can be answered with further
analyses of the database that we constructed for the years 1988-1991.
In addition, we have developed a technology for linking data collected
in the state of Florida that could readily be applied to extend our
database to cover additional years.
Florida has experienced a number of additional changes in Medicaid
eligibility and in delivery systems since the end of our 1991 study
period. The income eligibility limit for pregnant women was increased
from 150 percent of poverty to 185 percent in 1992, allowing one to
observe the effects of expansions over a broader range of the near poor
income distribution.
The state has also seen substantial shifts in the mix of sites at
which prenatal care for low-income women is provided since 1991. Over
the initial years of our proposed study period, most of the prenatal
care services newly financed by Medicaid were accommodated by expansions
in the county health departments. Following the increase in obstetrical
fees in 1991, however, there was a shift in the prenatal services for
Medicaid patients from the county health departments to the private
sector. For example, the number of pregnant Medicaid beneficiaries seen
in county health departments fell from 66,000 to 47,000 between 1991 and
36
1994, despite growth in the number of Medicaid pregnant women over this
time. Within the private sector, the period 1991 through 1994 resulted
in a sizable increase in the number of Medicaid patients enrolled in
managed care plans--from 125,000 to about 600,000. Thus, this period
would provide additional information about the effect of changing sites
of care on birth outcomes.
Expanding our database to cover the 1992-1994 period would provide
an invaluable source for answering the critical questions facing
policymakers as they reform health care policy for low-income women.
37
REFERENCES
Alpha Center, The Medicaid Expansions For Pregnant Women and Children.
Washington, D.C. : Alpha Center; 1995.
Braveman, Paula, Trude Bennett, Charlotte Lewis, Susan Egerter, and
Jonathan Showstack, "Access to Prenatal Care Following Major Medicaid
Eligibility Expansions", JAMA. 1993; 269: 1285-1289.
Buescher, Paul A. and Nancy I Ward, "A Comparison of Low Birth Weight
Among Medicaid Patients of Public Health Departments and Other
Providers of Prenatal Care in North Carolina and Kentucky". Public
Health Reports. 1992; 107: 54-59
Buescher, Paul A., Clinton Smith, Joseph L. Holliday, and Ronald H.
Levine, "Source of Prenatal Care and Infant Birth Weight: The Case
of a North Carolina County". Am J Obstet Gynecol. 1987; 156: 204-210
Buescher, Paul A., Marcia S. Roth, Dennis Williams, and Carolyn M.
Goforth, "An Evaluation of the Impact of Maternity Care Coordination
on Medicaid Birth Outcomes in N. Carolina. AJPH. 1991; 81; 1625-1629.
Clarke Leslie L., Michael K. Miller, W. Bruce Vogel, Karen E. Davis,
Charles S. Mahan, "The Effectiveness of Florida's 'Improved Pregnancy
Outcome' Program". Journal of Health Care for the Poor and
Underserved. 1991; 4, 117-132.
Haas, Jennifer S., Steven Udvarhelyi, Carl N. Morris, and Arnold M.
Epstein, "The Effect of Providing Health Coverage to Poor Uninsured
Pregnant Women in Massachusetts". JAMA. 1993; 269: 87-91.
Kessner D. M. , J. Singer, C.E. Kalk, and E.R. Schlessinger, Infant
Death: An Analysis by Maternal Risk and Health Care. Washington D.C:
Insitutie of Medicine and National Academy of Sciences; 1973.
Kotelchuck, Milton, "An Evaluation of the Kessner Adequacy of Prenatal
Care Index and a Proposed Adequancy of Prenatal Care Utilization
Index". AJPH. 1994; 84: 1414-1420.
Long, Stephen H. , M. Susan Marquis, and Ellen R. Harrison, "The Costs
and Financing of Perinatal Care in the United States". AJPH. 1994;
84: 1473-1478.
Piper, Joyce M. , Edward F. Mitchell, and Wayne A Ray, "Presumptive
Eligibility for Prenant Medicaid Enrollees: Its Effects on Prenatal
Care and Perinatal Outcome". AJPH. 1994; 84: 1626-1630.
Piper, Joyce M. , Wayne A. Ray, and Marie R. Griffin, "Effects of
Medicaid Eligibility Expansion on Prenatal Care and Pregnancy Outcome
in Tennessee". JAMA. 1990; 264: 2219-2223.
38
Rosenzweig, Mark R. and T. Paul Schultz, "The Behavior of Mothers as
Inputs to Child Health: The Determinants of Birth Weight, Gestation,
and Rate of Fetal Growth", In The Economic Aspects of Health, Victor
R. Fuchs (Ed.). Chicago: University of Chicago Press, 1982.
Starfield, Barbara et al . , "Race, Family Income, and Low Birth Weight".
A J Epidemilogy. 1991: 134, 1167-1174.
Starfield, Barbara, "Low Birthweight" . In The Effectiveness of Medical
Care: Validating Clinical Wisdom. Baltimore: Johns Hopkins University
Press, 1985.
Thompson, Daniel, Diane Dimperio, Ronald G. Humphries, C. Meade Girgg,
and Charles S. Mahan, "Low Birth Weight Rates For Florida Medicaid
Recipients Receiving Prenatal Care in Public Health Units Compared to
Those Receiving Care Elsewhere", Talahassee: Florida Department of
Health and Rehabiitative Services; 1993.
A-l
APPENDIX A. SELECTION OF THE STUDY STATE
Ian T. Hill, Stephen H. Long, and M. Susan Marquis
This appendix details the criteria we established to guide our
selection of the study state, describes states along the selection
dimensions, and presents the reasons for selecting Florida as the study
state.
CRITERIA FOR SELECTION
We established the following criteria for selecting candidates for
the study state:
A significant Medicaid eligibility expansion. The state must
have increased Medicaid eligibility thresholds for pregnant
women by a substantial amount, and the increase must be
surrounded by lengthy periods of stable eligibility rules. To
observe the effects of an eligibility change, it made sense to
look in places where the policy change was "substantial" and
where we can take accurate measures of "before" and "after"
conditions. We also preferred a state that simultaneously
adopted policies to encourage program participation—including
information outreach campaigns, expanded services, and
eligibility streamlining--because the effectiveness of the
Medicaid eligibility expansion will depend on these policies as
well as the change in the threshold.
Strong Title V and other direct delivery programs, with a
reputation for good data. To observe the effect of the
Medicaid eligibility expansions on the direct delivery system,
we looked for a state with a strong tradition of direct
delivery through Title V programs and other systems, including
community and migrant health centers.
Availability of Medicaid eligibility and claims, vital
statistics, and uniform hospital discharge abstract data. Our
analysis plan required that we obtain and merge data from
Medicaid, vital records, and hospital discharges at the person
level. The eligibility files also needed to distinguish in the
"post" period those who would have been eligible under the
pervious rules from those eligible under the new rules.
Uniform discharge data were required to identify the insurance
status of women who were not covered by Medicaid.
Favorable prospects for complete cooperation by state and local
officials. We preferred to work in a state where we know
people and expect full cooperation, because without it we would
not be able to gain access to the necessary data.
Minimize overlap with other studies. Other things equal, it
was desirable to work in a state that remained unburdened by
other studies, especially studies of financing issues.
A-2
CHARACTERISTICS OF THE STATES
A descriptive typology of the fifty states and the District of
Columbia along many of our criteria is provided in Tables A.l through
A. 4, which appear at the end of this Appendix.
Tables A.1-A.3 present characteristics of state Medicaid programs
for pregnant women, infants, and children. Specifically, these tables
show Medicaid eligibility thresholds and effective dates for expansions
of coverage, the presence of statewide prenatal care outreach/public
information campaigns, the presence of enhanced prenatal care benefit
programs, and the effective dates of efforts to streamline Medicaid
eligibility processes. These data provide measures for our first
criterion for selecting a study state--the level of increase in Medicaid
eligibility and the degree to which each state engaged in other efforts
to improve access to and the quality of Medicaid-f inanced prenatal care.
Table A. 4 presents characteristics of state and local health care
delivery systems--specif ically , total maternal and child health
spending, and the number and proportion of women served by both local
health departments and community and migrant health centers (C/MHCs).
These data provide information regarding the role of the direct delivery
system in providing prenatal care (selection criterion 2).
STATE SELECTION
We concluded that Florida was the preferred state for our study.
Florida was quick to respond to the new eligibility opportunities. The
eligibility threshold for pregnant women was increased from 47 percent
of poverty to 100 percent in October 1987, then to 150 percent in July
1989 (Table A.l). Florida adopted all six of the eligibility
streamlining methods shown in Table A. 3 before the end of 1987, well in
advance of most states.
The state also has a well established direct delivery system.
Florida ranks third in the nation in total public spending on maternal
and child health (MCH) , third in number of women served by MCH clinics,
and third in number of women served by community and migrant health
centers (Table A. 4) . This large absolute size of the direct delivery
effort is important to our ability to detect changes.
Florida has Medicaid data that distinguish enrollees by their
basis of eligibility (income at or below AFDC threshold, above AFDC but
at or below 100 percent of poverty, and over 100 percent of poverty but
at or below 150 percent) . Florida has a uniform hospital discharge
abstract data system. In addition, the state of Florida maintains a
database of individual encounter records for all personal health
services provided through each county health department. This unique
database greater facilitates our analysis of the substitutions between
programs following the Medicaid expansions. Importantly, the Medicaid
Director and the state Health Department Director offered to cooperate
with us in the study. Finally, there were no other major studies of the
Medicaid eligibility expansions being conducted in Florida at the time
we selected it.
A-3
REFERENCES
Association of Maternal and Child Health Programs, Caring for Mothers
and Children: A Report of a Survey of FY 1987 State MCH Program
Activities, Washington, D.C., March 1989.
Health Resources and Services Administration, U.S. Department of Health
and Human Services, 1990-1991 Grant Applications, Rockville, MD:
Division of Primary Care Services, Bureau of Primary Health Care
1992.
Hill, Ian T., The Medicaid Expansions for Pregnant Women and Children:
A State Program Characteristics Information Base, Washington, D.C.:
Health Systems Research, Inc. for the Health Care Financing
Administration, February 10, 1992.
Hill, I.T., Reaching Women Who Need Prenatal Care: Strategies for
Improving State Perinatal Programs, Washington, D.C. : National
Governors' Association, 1988.
National Association of Community and Migrant Health Centers, Directory
of Community and Migrant Health Centers, Washington, D.C, 1990.
The Public Health Foundation, Public Health Agencies 1991: An Inventory
of Programs and Expenditures, Washington, D.C, 1991.
U.S. General Accounting Office, Prenatal Care: Early Success in
Enrolling Women Made Eligible by Medicaid Expansions, Pub. No.
GAO/PEMD-91-10, Washington, D.C, February 1991.
Table A.l
Medicaid Eligibility Thresholds and Effective Dates for
Expansions of Coverage for Pregnant Women and Infants
January 1987 - December 1991
STATES
Apr
Jul
Jan Apr
1989
Jan Apr
Jul
1990
Apr
1991
Apr
Jul Oct
Alabama
Alaska
Ar i zona
Arkansas
Ca lif ornia
Colorado
Connect icut
Delaware
D.C.
Florida
Georgia
16%
1 1 1 inois
«
64%
45%
i .1 1
67%
61%
75%
4/87
100%
4/87
100%
10/87
100%
1/88
100%
2/88
100%
4/88
100%
1/88
100%
7/88
100%
7/88
50%
7/88
100%
7/88
100%
1/89
185%
1/89
150%
1/89
185%
7/89
b
75%
7/89
150%
7/89
100%
1/89
100%
185%
1/89
1/90
67%
b
75%
1/89
7/89
100%
7/89
185%
7/89
150%
7/89
133%
4/90
b
133%
4/90
b
133%
140%
4/90
10/90
133%"
4/90
133%
4/90
133%
4/90
133%
4/90
133%'
4/90
13 3%"
4/90
185%
7/90
185%
7/91
160%
4/91
150%
7/91
>
Table A.l
Medicaid Eligibility Thresholds and Effective Dates for
Expansions of Coverage for Pregnant Women and Infants
January 1987 - December 1991 (Continued)
Apr
Jul
Jan Apr
Jul
19B9
Jan Apr
Jul
Apr
199 1
Jan Apr
Kentucky
Louisiana
Maine
Maryland
Massachussetts
Michigan
Minnesota
Mississippi
35%
Nevada
New Hampshire
New Jersey
New Mexico
New York
Nor t h Carol ina
71%
53%
38%
71%
34%
100%
7/87
100%
7/87
100%
7/87
100%
10/87
100%
10/87
100%
1/88
100%
7/88
125%
10/88
185%
10/88
185%
7/88
100%
185%
1/88
185%
7/88
10/88
100%
185%
0/87
100%
1/88
7/88
100%
10/88
100%
1/89
185%
7/89
100%
7/89
75%
7/89
75%'
7/89
133%
185%
4/90
7/90
b
133%
4/90
133%
4/90
b
133%
4/90
b
133%
4/90
b
133%
4/90
b
133%
4/90
133%
4/90
185%
1/90
150%
1/90
133%
4/91
185%
7/91
185%
7/91
185%
10/90
>
I
Tabla A.l
Medicaid Eligibility Thresholds and Effective Dates for
Expansions of Coverage for Pregnant Women and Infanta
January 1987 - December 1991 (Continued)
Apr Jul
Jan Apr Jul Oct
1989
Jan Apr Jul
Jan Apr Jul
Jan Apr Jul
North Dakota
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
56%
4 5*
)',*
100%
4/87
85%
11/87
100%
10/87
100%
7/87
Utah
91%
Vermont
81%
Virginia
47%
Washington
73%
90%
7/87
West Virginia
38%
100%
7/87
100%
10/87
100%
1/88
100%
4/88
100%
9/88
100%
7/88
100%
9/88
185%
7/88
100%
7/88
150%
7/88
185%
10/88
100%
1/89
100%
1/89
75%
7/89
85%
7/89
185%
7/89
130%
9/89
185%
7/89
150%
1/90
133%
4/90
b
133%
4/90
b
133%
4/90
b
133%
4/90
b
133%
4/90
133%
4/90
133%
4/90
b
133%
4/90
133%
4/90
185%
7/91
o>
185%
12/91
Table A.l
Medicaid Eligibility Thresholds and Effective Dates for
Expansions of Coverage for Pregnant Women and Infanta
January 1987 - December 1991 (Continued)
STATES
1988
Apr
Jan Apr
1989
Jan Apr
Apr
1991
Apr
Wi scons i
Wyi jin i ng
84*
100%
10/88
133*
1SS%
4/90
7/90
b
1 33%
4/90
Figure represents stares' AFDC or Medically Needy Income threshold (whichever is most generous) as a percent
of the Federal poverty level for a f.imily of three.
states complying with federal mandate.
From 9/88 to 1/89, the state funded a program covering pregnant women and infants up to 120% of poverty.
Between 7/H9 and 4/90, the income limit for this program was raised to 130% of poverty.
lource: Hill. Ian T. Pis. Medicaid Expansions for Preaanant Women and rhldren: A state
Pic-gram Characteristics 'njormation Base. Washington, D.C. : Health Systems Research, Inc.
for Health Care Financing Administration. February 10, 1992.
>
i
A-8
Table A. 2
Presence of Outreach Campaigns and Enhanced Prenatal Care Benefits
Statewide Public Enhanced
States Information/Outreach Campain Prenatal Care
Alabama X X
Alaska X X
Arizona a
Arkansas X . X
California X X
Colorado X
Connecticut X
Delaware X
D.C. X
Florida
Georgia X
Hawaii X X
Idaho X X
Illinois X X
Indiana X
Iowa X X
Kansas X
Kentucky
Louisiana X
Maine
Maryland X X
Massachusetts X X
Michigan X X
Minnesota X X
A-9
Table A. 2
Presence of Outreach Campaigns and Enhanced Prenatal Care Benefits (Continued)
States
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey-
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
Statewide Public
Information/Outreach Campain
Enhanced
Prenatal Care
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
b
X
X
X
X
X
X
X
X
X
Source: Hill, Ian T. The Medicaid Expansions for Pregnant Women and
Children: A State Program Characteristics Information Base.
Washington, D.C. : Health Systems Research, Inc. for the
Health Care Financing Administration. February 10, 1992.
Arizona provides these services, when medically indicated, through
the Arizona Health Care Cost Containment System (the state's
Medicaid program), begun in 10/82.
Starting in January 1988, Rhode Island's state - funded Rite Start
program extended enhanced prenatal services to low- income pregnant women.
A-10
Table A. 3
Effective Dates for State Strategies to Streamline
Medicaid Eligibility
States
Drop
Assets
Continuous
Eligibility
Presumpt
Eligibil.
ive
ity
Outstation
Eligibility
Workers
Shorten
Application
Form
Expedite
Eligibility
Alabama
7/88
7/88
7/88 to 7/91
1/89
4/90
9/91
Alaska
1/89
1/89
7/91*
1/89
Arizona
1/88
1/88
7/91a
Arkansas
10/88
4/87
4/87
7/91°
California
1/91*
5/90
11/91
1/89
Colorado
7/90
7/90
1/90
7/91*
7/91
Connecticut
Delaware
4/88
1/88
4/88
1/88
7/91
7/91*
1/88
7/91
1/88
7/91
1/88
D.C.
Florida
Georgia
4/87
10/87
1/89
4/87
10/87
1/89
7/91
10/87
7/91*
7/86
1/89
7/86
1/89
11/86
7/89
Hawaii
1/89
1/89
1/89
7/91'
10/91
Idaho
1/89
1/89
1/89
7/91*
7/91
Illinois
9/91
10/88
1/89
9/91*
9/91
Indiana
Iowa
7/88
7/88
7/89
7/88 to 7/91
7/89
7/91*
9/90
6/92
Kansas
7/88
1/91*
7/91*
7/88
Kentucky
Louisiana
6/89
1/89
8/88
1/89
1/89
7/91*
1/89
9/89
1/89
Maine
10/88
10/88
10/88
7/91*
Maryland
7/87
7/87
7/87
7/91*
7/87
Massachusetts
Michigan
7/87
10/88
7/87
10/88
7/87
7/91*
10/88
2/90
1/89
A-ll
Table A. 3
Effective Dates for State Strategies to Streamline
Medicaid Eligibility (Continued)
States
Drop
Assets
Continuous
Eligibility
Presumptive
Eligibility
Outstation
Eligibility
Workers
Shorten
Application
Form
Expedite
Eligibility
Minnesota
Mississipi
Missouri
7/88
10/88
7/90
7/88
10/87
7/90
7/90
7/91*
10/87
9/89
7/88
7/91
7/88
Montana
7/89
1/91*
1/91
7/91*
Nevada
7/89
11/91*
7/91*
New Hampshire
7/89
1/91*
New Jersey-
New Mexico
7/87
12/88
7/87
7/89
5/88
4/89
7/91*
1/90
8/88
1/90
New York
North Carolina
1/90
10/87
1/90
10/87
1/90
10/87
7/91*
10/87
1/91
North Dakota
1/91*
7/91*
Ohio
1/89
11/91*
4/91
4/91
4/91
Oklahoma
Oregon
4/88
11/87
4/88
11/87
7/90
7/91*
9/90
7/88
7/88
Pennsylvania
4/88
1/91*
4/88
7/91*
Rhode Island
South Carolina
4/87
10/87
4/87
10/87
7/91*
11/87
12/89
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
7/88
3/88
12/91
1/89
7/89
7/88
7/88
7/87
9/88
1/89
4/87
7/89
2/89
9/89
3/89
7/91*
7/87
1/90
1/89
10/89
7/89
7/88
9/89
10/89
7/88
10/89
7/88
Washington
7/89
7/87
7/91*
2/90
1/90
West Virginia
7/87
7/87
7/91*
4/88
7/88
Wisconsin
7/89
7/89
4/88
7/91*
10/90
1/90
Wyoming
10/88
10/88
7/91*
Source: Hill,
Ian T.
rharactei
The Medicaid Exoansions for
Preanant Women
and Children
: A
State Proaram C
ristics Information
Base. Wa
ishington, D.C.
: Health Sys
terns
Research, Inc. for the Health Care Financing Administration, February 10, 1992
a States complying with federal mandates.
A-12
Table A. 4
State And Local Health Care Delivery System Characteristics, 1989
Total MCH
Total women
Monies
served by
States
(in thousands)
MCH
Alabama
$28, 532
20,172
Alaska
6,318
25
Arizona
21,418
4,553
Arkansas
17,262
8,726
California
206,608
30,197 (1986 data)
Colorado
16,592
4,000
Connecticut
12,176
4,933
Delaware
5,791
1,779
D.C.
16,454
1,032
Florida
109,121
43,081
Georgia
57,416
11,033
Hawaii
14,973
219
Idaho
4,652
553
Illinois
41,460
4, 184
Indiana
24,393
2,809
Iowa
14,312
2,179
Kansas
8,049
2,398
Kentucky
39,494
12,874
Louisiana
37,593
17,539 (1986 data)
Maine
7,950
1,066 (1986 data)
Maryland
150,064
N/A
Massachusetts
26,841
12,000
Michigan
86,159
N/A
Minnesota
15,311
4,660
Mississippi
29,914
17,290
Missouri
$20,390
8,069
Montana
c
0
Nebraska
3,644
946
Nevada
6,378
596
Ratio women Total women Ratio women
served by MCH served by served by C/MHCs
to all women C/MHCs to all women
2.1%
0.0%
0.5%
1.7%
0.4%
0.5%
0.6%
1.1%
0.6%
1.5%
0.7%
0.1%
0.2%
0.2%
0.2%
0.4%
0.4%
1.5%
1.8%
0.4%
N/A
0.8%
N/A
0.5%
2.9%
0.7%
0.0%
0.3%
0.2%
6
,245
342
1
,609
208
20
,592
9
,361
1
,688
249
622"
13
,378
2
,319
584
1
,836
2
,502
404
1
,544
511
2
,511
1008"
8
2
,578
4
,555
2
,792
1
,313
2
,413
4
,754
562
288
509
0.66%
0.25%
0.19%
0.04%
0.29%
1.15%
0.22%
0.15%
0.37%
0.48%
0.14%
0.22%
0.82%
0.09%
0.03%
0.25%
0.09%
0.29%
0.10%
0.00%
0.22%
0.31%
0.13%
0.13%
0.40%
0.41%
0.32%
0.08%
0.18%
A-13
Table A. 4
State And Local Health Care Delivery System Characteristics, 1989 (Continued)
Total MCH
Total
women
Ratio women
Total women
Ratio women
Monies
served by
served by
MCH
served by
served by C/MHCs
States
(in thousands)
MCH
to all women
C/MHCs
to all women
New Hampshire
6,484
811
0.3%
367
0.13%
New Jersey
23,553
9,819
0.5%
6,310
0.35%
New Mexico
8,249
4,646
1.3%
254
0.07%
New York
59,879
53,325
1.2%
18,095
0.42%
North Carolina
53,687
28,741
1.8%
4,877
0.31%
North Dakota
3,278
1466'
1.1%
N/A
N/A
Ohio
48,889
23,096
0.9%
7,504
0.30%
Oklahoma
17,329
5,214
(1983
data)
0.7%
355
0.05%
Oregon
18,712
6675*
1.0%
3,752
0.58%
Pennsylvania
53,557
25,445
0.9%
3,470
0.13%
Rhode Island
7,147
3,553
1.5%
2,513
1.06%
South Carolina
45,446
22,865
2.7%
3,927
0.47%
South Dakota
3,492
54
0.0%
13
0.01%
Tennessee
42,506
14,195
1.2%
1,810
0.16%
Texas
90,204
77,429
1.9%
7,647
0.19%
Utah
10,809
468
0.1%
949
0.24%
Vermont
6,719
4,431
3.3%
41
0.03%
Virginia
56,283
16500*
1.1%
910
0.06%
Washington
22,514
3,669
0.3%
5,444
0.47%
West Virginia
18,219
5,282
1.3%
4,437
1.10%
Wisconsin
15,990
301
0.0%
1,203
0.11%
Wyoming
3,788
0
0.0%
10
0.01%
Source: Public
Health Aaencies
1991: An
Inventorv
of Proarams
and
tion, 1991.
Expenditures. Washinaton. D.C.:
The
Public Health Founda'
Carina
for Mothers and
Children:
A ReDort
of a Survev
of 1
Hf 1987
State MCH Proaram Activities. W
ashinc
fton,
D . C . : The .
Association
of Maternal and Child Health Programs, March 1989.
Directory of Community and Migrant Health Centers. Washington, D.C. :
The National Association of Community and Migrant Health Centers, 1990.
1990-191 Grant Application. Rockville, MD. : Division of Primary
Care Services, Bureau of Primary Health Care, Health Resources and
Services Administration, U.S.. Department of Health and Human Services,
1992
a Estimated figure AMCHP, 1987.
b Data from calendar year 1990.
c Montanta did not report to Public Health Foundation in 1989.
KcS
Health Systems Research, inc
2021 l Street nw Suite 400
Washington OC 20036
'202) S28.5100
Fax: (202) 728.9469
Appendix B
IMPLEMENTING THE MEDICAID EXPANSIONS FOR PREGNANT WOMFX-
THE EXPERIENCE IN FLORIDA "WUH.
Prepared for:
Office of Research and Demonstrations
Health Care Financing Administration
Baltimore, MD
Cooperative Agreement 18-C-901 13/9-01
Prepared by:
Ian Hill
Health Systems Research, Inc.
Washington, DC
Under Subcontract with:
The RAND Corporation
Washington, DC
Subcontract #91-18
24 July 1995
Acknowledgments
The author would like to express his sincere gratitude to the many persons who devoted their
time and energy to site visit interviews and were kind enough to submit additional document
and reports describing their efforts. Specifically, thanks are to be extended to the follow?™
Florida Department of Health and Rehabilitative Services officials: Charles Mahan Donna
Barber. Meade Grigg, Fred Roberson. Lynn Bodiford. Molly Melor. Gary Cravton Manlvn
Reaves. Michael Cuppoli. and Phyllis Siderits. The author would also like to thank Garv"
Clarke, tormer Deputy Assistant Secretary for Medicaid, for his critical insights into the'
historical events leading up to Florida's Medicaid expansions.
Many thanks are also extended to Project Officer Marilyn Hirsch at the Office of Research and
Demonstrations. Health Care Financing Administration, for her input on the design of the
evaluation and for her guidance throughout the development of this report.
At RAND, both Stephen Long and Susan Marquis are to be thanked for their energetic
participation in the Florida site visit as well as their critical advice and guidance in carrying out
this qualitative evaluation.
Finally, at Health Systems Research, Inc., the author thanks Renee Schwalberg for her
thorough analysis of the HCFA 2082 data.
Health Systems Research, Inc.
Table of Contents
I. Background „
1
II. The Impetus for Medicaid Expansions in Florida.
A. Passage of the Health Care Access Act of 1984 ..
.j
.j
B. Lack of Impact and the Need for Additional Expansion
III. Florida's Systems for Providing Maternity Care
A. Organization of State Agencies
6
B. The Public Health Service Delivery System 8
C. Federally-Funded Primary Care Centers 8
D. The Role of Private Physicians .Q
IV. Expanding Medicaid and Improving Its Ability to Serve Mothers 1 1
A. Expanding Medicaid Eligibility , -,
B. Simplifying the Medicaid Eligibility Process j 2
C. Promoting Use of Prenatal Care Through Outreach 14
D. Improving Physician Participation in Medicaid 1 5
E. Providing Technical Assistance to Counties to Facilitate Effective Implementation 17
V. Key Findings and Implications for the Evaluation 19
A. Improvements in Enrollment, Expenditures, Revenues, and Outcomes 19
B. Qualitative Impressions of Program Impact and Lessons Learned 25
Works Cited
Appendix A.
Health Systems Research, Inc.
[. Background
In March 1992. the federal Health Care Financing Administration (HCFA) awarded a
Cooperative Agreement to the RAND Corporation (RAND) and its subcontractor Health
Systems Research, Inc. (HSR), to study the impact of the Medicaid eligibility expansions for
pregnant women on state service delivery and financing sv stems for maternity care The
methodology tor the study includes both quantitative and qualitative components The
quantitative analysis, which will be conducted by RAND, involves a detailed studv of how the
Medicaid eligibility expansions affected how pregnant women flow through prenatal care and
delivery systems and how the amount and distribution of funds supporting prenatal and
delivery services changed among payers. The qualitative analvsis. to be completed by HSR
involves a similarly detailed study of how a selected jurisdiction enacted Medicaid expansions
and. in particular, how this jurisdiction implemented strategies in an effort to ensure that access
to. and quality of, prenatal care improved as a result of the expansions.
During the summer of 1992, the project developed a report entitled Impacts of Medicaid
Eligibility Expansions and Innovative Programs for Maternal Health Care: Methodology and
State Selection Report (Hill, Long, and Marquis, 1992). In it, the authors detailed the analytic
framework and estimation methods to be used in conducting the study. Further, the report
presented a 50-state typology describing selected characteristics of state Medicaid and Title
V/Maternal and Child Health (MCH) programs, as well as state and local health care delivery
systems for maternity care. A primary purpose behind the development of the typology was to
assist the RAND/HSR research team with the selection of a state, or states, to include in the
study. The report concluded that the research effort should focus on the State of Florida
because it is a state that significantly expanded Medicaid eligibility for pregnant women in
response to federal statutory changes, aggressively implemented numerous strategies to
streamline access to Medicaid coverage, possesses a strong prenatal care service delivery
system composed of a mix of public and private providers, maintains multiple databases that
would support a sophisticated quantitative analysis, and has no other significant studies of the
Medicaid expansions already underway. In addition and importantly, the director of the state
Medicaid program and the state health officer both offered to cooperate with the research team
in the conduct of the study.
In November 1992, HSR Project Leader Ian Hill and RAND Project Leaders Stephen Long and
Susan Marquis conducted a two-day site visit to Florida. During the visit, team members
conducted interviews with numerous officials representing the Medicaid, MCH, and Children's
Medical Services programs within the state Department of Health and Rehabilitative Services
(HRS), as well as officials from the newly created Agency for Health Care Administration
(AHCA). While RAND team members focused their interviews on issues surrounding the
existence and collection of data for the quantitative analysis, Mr. Hill spoke with state officials
about the impetus for, and objectives behind, the Medicaid expansions, the specific policies
and programs that were enacted, strategies to implement initiatives at the local level, and
impressions regarding the successes and failures of the changes in the program.
This report represents the final qualitative analysis of Florida's experience with implementing
the Medicaid expansions for pregnant women. Based on information gathered during the site
I
Health Systems Research, Inc.
S^tSIT* ob,a,ned from s,a,e 0,T,cials' ,h,s repm «—• >■««— in ,he
following sections:
■ In Section II. the various factors leading up to Honda's decision to expand
Medicaid coverage o low-income pregnant women will be discussed, including
summaries of the health and socio-economic status of child-bearing women
during the mid-1980s and the political environment of the penod that pr" vLd
the impetus for change. ynwaea
■ In Section III. an overview of Florida's prenatal care systems will be provided
including summaries of the organizational structure and roles and
responsibilities of the state agencies involved with the delivery and financing of
care, and the public and private delivery systems in place at the local level.
■ In Section IV. the specific policy and program changes enacted to improve
access to Medical- financed prenatal care will be described, including detailed
descriptions of strategies that were employed to ensure effective implementation
ot these initiatives.
■ In Section V, the key findings of the qualitative evaluation will be presented
including summaries of state officials' impressions of the impacts of their
efforts, and descriptive data illustrating how the Medicaid expansions affected
Medicaid enrollment and expenditures and documenting positive trends related
to birth outcomes and access to prenatal care.
The qualitative knowledge gained through this case study provides the project with critical
insights into the factors that contributed to Florida's successes and/or failures in expanding
Medicaid coverage for pregnant women. In addition, the narrative allows the research team to
more fully understand and accurately interpret the outcomes and implications of the various
quantitative analyses being conducted by RAND.
Health Systems Research, Inc.
on
II. The Impetus for Medicaid Expansions in Florida
In Florida during the early 1980s, the health of the state's newborns was verv poor fn
1983. infants in the state died at a rate of 12.2 per 1.000 live births compared to the national
average or 1 1.2. On this indicator. Florida ranked 41st among the 50 states Further 7 4
percent ot all intents bom that year were low birthweight (less than 2 500 grams) a rate rh.t
did not compare favorably to the national average of 6.8 percent and ranked the state 39th o
this indicator. Not surprisingly, indicators of women's access to, and utilization of nrenata
care were similarly poor. For example, the percentage of women who received prenatal care
during their first trimester of pregnancy was 68.2-compared to 76.2 percent nationally and
ranking the state 45th--while the percentage who received late or no prenatal care was 8 5
HedSgStetis?« 46tH "* C°mpared t0 a nati°nal aVerage °f 56 percent- (National Center for
Poor health status among mothers and children was but one symptom of a much broader
medical indigency problem in Florida during the period. Roughly 15 percent of the state's
total population was uninsured for health coverage, compared to the national average of 1 1
percent (Current Population Survey, 1986-89). And, after excluding the elderly this rate rose
to 22 percent (Clarke, 1995). The state Medicaid program, the principal source 'of health
insurance for low-income families, was limited in its coverage. Specifically, the income
eligibility threshold for a family of three under the state's Aid to Families with Dependent
Children (AFDC) program, to which Medicaid eligibility was linked, was set at just 3 1 percent
of the federal poverty level, a threshold that was more strict than that of all but 1 1 states (Social
Security Administration, 1984). Further, in 1983, Florida was one of just 16 states that did not
cover the optional Medically Needy group under its Medicaid program (HCFA, 1984).
Overall, Medicaid had not succeeded in providing coverage to a significant proportion of the
state's poor; just 24 percent of all persons living below the federal poverty level were Medicaid
recipients in 1984 (Health Care Financing Administration, 1985).
The burden of medical indigency was falling heavily upon the hospitals in the state and
disproportionately on public and not-for-profit institutions. For example, three of the more
than 200 general hospitals in Florida accounted for more than 20 percent of all Medicaid
hospital days and 35 percent of all charity care provided in the state (Clarke, 1986). These
same three facilities provided more Medicaid and charity care than all 85 for-profit hospitals
combined (Clarke, 1986). As health care costs generally and hospital costs specifically
continued to rise faster than national averages, state officials, politicians, and industry officials
all recognized the need for change.
A. Passage of the Health Care Access Act of 1 984
In 1983, a newly-formed commission, headed by State Senator Bob McKnight and called the
Task Force on Competition and Consumer Choice, conducted hearings across the state and
began developing proposals for statewide policy changes. After extensive analysis of a range
of options, including those of both a competitive and a regulatory nature, the commission
Health Systems Research, Inc.
spearheaded the development of a broad-based consensus among state officials hn.«i. ,
;~^Tmunity vvhich resuited in the ^ °f - «^^^
The law contained several provisions intended to contain the rate of growth of hospual costs
reduce the number ot persons who were medically indigent, and finance and more ecu tab v
distribute the bad debt absorbed by hospitals who served the poor. Perhaps the mo Umponant
provision ot the law .mposed a one-percent tax on the net operating revenues of FlondT
hospitals to be raised to 1.5 percent in the second and all subsequent vears, to finance
Medicaid and primary care expansions. The law also created the Public Medical Assistance
Trust Fund into which state appropriations and hospital tax revenues would be deposited and
trom which monies would be drawn to pay the health care costs of newly covered Medicaid
recipients. Importantly revenues placed in the Fund would be used to draw down additional
federal Medicaid matching funds at the rate of 56 percent.
The Health Care Access Act also authorized expansions of Medicaid eligibility to include
several previously uncovered optional groups as well as an appropriation to expand county
health departments capacity to provide primary care services to low-income persons as
described below. K '
■ Effective July 1,1985, Medicaid eligibility was expanded to include married
pregnant women, children under age 21 in intact families, and unemployed
parents and their children under age 1 8 in families meeting the AFDC
program's eligibility thresholds.
■ Effective July 1 , 1 986, Medicaid was expanded to include coverage of the
Medically Needy, that is, persons who met the program's categorical
requirements but whose income was too high to qualify them for either the
AFDC or Supplemental Security Income (SSI) programs. This expansion
effectively raised the income eligibility threshold for families to 47 percent of
the federal poverty level.
■ An appropriation of $ 1 0 million was authorized to support county health
departments' provision of primary care services, with the distribution of funds
based on a county's need and willingness to participate.
These and other provisions of the law reflected the desire among all the various public- and
private-sector partners to develop systems reforms in a comprehensive and consensual manner.
The hospital revenue tax and creation of the Trust Fund, in particular, were seen as strategies
that would "level the playing field" for hospitals by allowing each to recover its costs in direct
proportion to the amount of Medicaid care it provided. At the time, the law seemed quite
likely to succeed; according to one estimate over 300,000 persons would be made eligible for
Medicaid as a result of the eligibility expansions (Lou Harris and Associates, 1985). In a
relatively short time, however, problems arose.
Health Systems Research, Inc.
B- Lack of Impact and the Need for Additional E^aosifln
By 1987. policymakers and providers were becoming increasingly frustrated bv an apparent
lack of impact ot the law. By July of that year, only slightly more than 16.000 persons had
enrolled under the expansions of both categorically and medically needy eligibility eroum
Vorse. the Public Medical Assistance Trust Fund reported a SI 80 million surplus'bv mid
1987 causing hospital officials across the state to cry foul and government officials'to besin
scrambling to identify more effective strategies for expanding low-income families- financial
access to health care and putting the hospital tax revenues to good use. Fortunately this cm <
co.ncided closely with events at the federal level that altered the Medicaid statute to give states
new flexibility to expand their programs for pregnant women and young children.
The remainder of this report will describe how the State of Florida reacted to this opportunity
not only by adopting sweeping expansions of Medicaid eligibility for mothers and children but
also by implementing numerous strategies and initiatives to help ensure that these expansions
actually resulted in improved access to, and use of, Medicaid-financed prenatal care. As will
be described in Section IV of this report, these strategies included efforts to simplify and
streamline the Medicaid eligibility process, to conduct outreach to inform women of the
importance of prenatal care and the availability of Medicaid coverage, and to enhance public
and private providers' willingness and capacity to serve low-income mothers. To provide a
context^ within which to consider these strategies. Section III will first provide an overview of
Florida's systems for providing and financing maternity care.
Health Systems Research, Inc.
III. Florida's Systems for Providing Maternity Care
This section presents an overview of Florida's systems for providing and financing maf
care Specifically, it describes the structure, roles, and responsibilities of the state aSS*
involved in caring tor mothers and children, the organizat.on of the public sector service
delivery; system tor prenatal care, the number and distribution of federallv-runded primary care
centers ,n the state, and the traditional role of private sector physicians in serving low-mcomT
mothers and their families. income
A. Organization of Stare AgenH^
Until 1993, all of the state agencies involved in serving disadvantaged families generally and
low-income pregnant women specifically were housed in the single "umbrella" agency 'the
Department of Health and Rehabilitative Services (HRS).1 As illustrated in Figure 1 the
principal programs serving these groups were organized into three divisions, as described
below.
■ Public Health. Under the direction of the Deputy Secretary for Health, this
division oversaw programs related to environmental health, disease control and
AIDS prevention, technical health services, and personal health and primary
care. Within the office of Personal Health and Primary Care, the MCH program
serves as the designated grantee of the federal Title V Block Grant.
■ Programs. Under the direction of the Deputy Secretary for Programs, this
division housed the programs for regulation of health facilities, developmental
services, aging and adult services, alcohol, drug abuse and mental health, and
children, youth and families. Of particular relevance to this report, the division
also housed the Medicaid program, the office of Economic Services
(responsibility for determining eligibility for Medicaid, AFDC, Food Stamps,
and a variety of other public welfare programs), and the Children's Medical
Services program (responsible for administering the Title V Children with
Special Health Care Needs program as well as the Regional Perinatal Intensive
Care Centers program for high-risk pregnant women and newborns).
■ Operations, Under the direction of the Deputy Secretary for Operations, this
division served a liaison function between the state and local levels by
administering all HRS programs through its network of 1 1 District Offices and
67 local public health units.
The Health Care Reform Act of 1992 created the new Agency for Health Care Administration (AHCA) in an
effort to consolidate under one structure all state agencies involved with health care purchasing and regulation.
The Medicaid program, as well as programs related to provider licensure and certification, certificate of need,
comprehensive health planning, and hospital cost containment were moved out of HRS.
6
Health Systems Research, Inc.
Figure 1
ORGANIZATIONAL STRUCTURE OF THE
DEPARTMENT OF HEALTH AND REHABILITATIVE SERVICES*
Deputy Secretary
for Health
Environmental
Health
Disease Control
- & AIDS Prevention
Technical
Health Services
Personal Health
- & Primary Care
Organizational structure in effect October 1988
HRS
Secretary
Deputy Secretary
for Operations
District
Administrator
HRS Public
Health Units
1
Deputy Secretary
for PrograiiK
Medicaid
Children's
Medical
Services
Aging and
Adult Services
Economic
Services
J
Regulation &
Health
Facilities
Developmental
Services
Children, Youth
and Families
Alcohol, Drug
Abuse & Mental
I lea lib
B. The Public Health Service Delivery Svsjffim
The State of Florida is divided into 67 counties, each of which has its own local health
department (LHD). [n«taL these 67 LHDs operate some 220 clinic service sites As
illustrated in Figure 2. Florida's counties are also organized into 1 1 Health Districts each of
which has a headquarters office that serves a liaison role between the state and local levels
Importantly, all district and local offices are "arms" of state government and are staffed bv
state employees. Therefore, the central HRS office has direct authority over how policies and
programs are implemented at the local level.
As is the case in many Southern states, LHDs in Florida have traditionally plaved a major role
in providing services directly to low-income individuals and families. In most instances LHDs
are staffed by a full complement of public health nurses, health educators, nutritionists and
other ancillary staff. Approximately one-half of all LHDs have physicians on staff, while the
remainder work with community physicians under contract with the state. A typical LHD
provides a broad array of preventive and primary care services, including well-child exams
immunizations, pregnancy testing and family planning services, screening for sexually-
transmitted diseases, and WIC. Importantly, a major portion of LHDs' service capacity is
devoted to the delivery of prenatal care. Beginning in 1982, the state implemented the
Improved Pregnancy Outcomes (IPO) project to channel special grant monies to the counties to
foster the development of more comprehensive systems of prenatal care. Through the IPO
project, Florida was able to raise the quality of county prenatal programs to a relatively equal
plain. By the mid-1980s, most LHDs provided not only clinical prenatal services using
contract obstetricians (OBs) and certified nurse midwives (CNMs), but also offered a wide
range of psychosocial support services, including health education, nutritional counseling,
social work services, and home visiting.
Historically, state general revenues and federal Title V Block Grant moneys comprised more
than 75 percent of LHD budgets; county funds have never supported a majority of LHD
operating costs. Primary care monies authorized by the Health Care Access Act were initially
distributed to 18 counties through a competitive Request for Proposals process and have been
credited with giving LHDs in those counties a revitalized role in providing and brokering
indigent care (HRS, 1986). Notably, LHDs rarely billed Medicaid when serving Medicaid-
eligible women and children prior to 1987. As will be described in the next section of this
report, increasing LHDs' capacity to capture Medicaid revenues to support service delivery
represented a major objective of HRS after the state implemented the Medicaid expansions.
C. Federally-Funded Primary Care Centers
Florida also boasts an extensive network of Federally-Qualified Health Centers (FQHCs), that
is. Community Health Center (CHC), Migrant Health Center (MHC), and Health Care for the
Homeless projects. In all, 46 grantee providers operate some 118 clinics, most often
Health Systems Research, Inc.
Figure 2
HRS District Boundaries
Effective October 1984
Health Systems Research, Inc.
situated ^relatively urban regions of the state. , Twenty-seven providers receive CHC
bection 3^-funds and operate 70 clinics. 15 providers receive MHC-Sect.on 329-fonds and
operate 38 clinics, and tour Homeless Health Care grantees oversee service delivery in 10
sues.) (National Association of Community Health Centers. 1991). Bv definition and «■
requirement of receiving federal funds, these clinics provide a comprehensive arrav of pnmarv
care services to all populations, regardless of ability to pay. L.ke their LHD counterparts
tederally-tunded climes represent a major source of preventive and primarv care for low
income families. '
In an effort to better integrate their service delivery efforts. LHDs and FQHCs have
increasingly negotiated arrangements whereby public health units take responsibility for
providing prenatal care services to Medicaid patients while FQHCs provide well-child and
other primary care services for families. Of note, nearly 40 percent of the $10 million in new
primary care funds appropriated to LHDs in 1984 were channeled to private providers via
contracting arrangements. Of these, the largest proportion went to FQHCs (HRS, 1986).
D. The Role nf Private Phvsirian^
As is also the case in many Southern states, private physicians have not historically played a
significant role in serving poor and disadvantaged families. Rather, they have preferred to
leave this responsibility to LHDs and FQHCs. This situation is borne out by historically low
participation rates among physicians in the Florida Medicaid program. Low fees and
cumbersome administrative rules are usually blamed for this lack of participation, and often for
good reason; in 1986, Medicaid paid obstetricians a global fee of just $315 for providing both
prenatal care and delivery services to a Medicaid-eligible pregnant woman. In addition, the
economic impact of malpractice has dramatically stifled physician participation in Medicaid
Malpractice insurance crises in both the 1970s and 1980s resulted in extremely high premiums
-for OBs, rates were as high as $200,000 per year, and averaged $65,000 in 1986-and limited
availability (Clarke, 1986). As will be discussed in the next section, a major objective of
perinatal reform efforts in the late 1980s was to restructure payment and administrative policies
so that physicians would be more willing to serve Medicaid mothers.
The State of Florida does possess the largest number of CNMs of any state-350 in 1992. In
contrast to obstetricians, these clinicians have traditionally been very willing to serve low-
income women and have worked closely with both LHD and FQHC providers.
10
Health Systems Research, Inc.
[V. Expanding Medicaid and Improving Its Ability to Serve Mothers
As discussed in Section [I. 1987 found Florida policymakers in the midst of a continued
indigent care crisis. Incremental Medicaid expansions enacted in 1985 and 1 986 were Lnn
to draw significant numbers of new recipients into the program and the Public Med cal
Assistance Trust Fund, established to pay for the health care costs of new Medicaid !Lkl
and supported by a new tax on hospital revenues, had built up a surplus of mUonsoT n
that was sitting unused. Meanwhile, the problems of infant mortaJ^Id mc 2K££S
; o^i h ;hstateH 7Rr of infant monaiity *« iow ^^ - '«s
S™ PerCCm' reSpeCtlVely"COntinued t0 Ia* * behind national avenges
During the previous year, however, the nation's infant mortality crisis had gained
unprecedented attention at the federal level. Citing indicators that placed the United States
behind 17 other industrialized nations, advocacy groups like the Children's Defense Fund
succeeded in getting the attention of national policymakers (Children's Defense Fund 1989)
Documenting that an investment in prenatal care is cost-effective-specifically that over S3 00
can be saved in avoided neonatal costs for every S 1 .00 spent on prenatal care-'the Institute of
Medicine lent credibility to the call for perinatal system reform (Institute of Medicine 1 985)
Finally, a strong and well-organized coalition of southern states, the Southern Regional Task
Force on Infant Mortality, drafted a simple but practical proposal to sever the traditional link
between AFDC and Medicaid eligibility and allow states to target large-scale Medicaid
expansions to pregnant women and infants (Hill, 1987).
Embraced by the nation's Governors and lobbied for effectively in Congress, the proposal was
signed into law m December 1986 as part of the Omnibus Budget Reconciliation Act of 1986
(OBRA-86). Specifically, OBRA-86 gave states the option to expand Medicaid income
eligibility thresholds for pregnant women and infants above Af DC levels to 100 percent of the
federal poverty level. In subsequent years, Congress continued to pass legislation allowing
and then requiring, states to further liberalize their Medicaid coverage of these priority
populations, as summarized below.
■ The Omnibus Budget Reconciliation Act of 1 987 (OBRA-87) further expanded
states' flexibility by allowing them to raise Medicaid income thresholds for
pregnant women and infants up to 185 percent of poverty.
■ The Medicare Catastrophic Coverage Act of 1 988 (MCC A) mandated, for those
states that had not already done so voluntarily, minimum coverage of pregnant
women and infants at 100 percent of poverty.
■ The Omnibus Budget Reconciliation Act of 1989 (OBRA-89) required states to
cover, at a minimum, pregnant women and children up to age six at 133 percent
of poverty.
11
Health Systems Research, Inc.
In general, states responded aggressively to the optional authontv presented bv OBR a «a
OBRA-87: by July 1989. the effective date of MCCVs mandate'. Z^ZslsZ^ **
already raised their eligibility thresholds to 100 percent of poverty or higher (Hill. 1990).
For Florida, the optional authority granted by OBRA-86 presented policymakers with an ideal
solut.on to their mdigent care crisis. The pressing and complementary needs for expanding
Medicaid coverage of the uninsured, putting Trust Fund dollars to good use. and addressing the
state s infant mortality problem were equally addressed by the policy of expandine Medicaid
eligibility tor pregnant women and infants up to the federal poverty 'level As will be described
below. Florida joined many states in quickly adopting eligibility expansions not only under
OBRA-86. but in subsequent years as well. Just as important, however, this section will also
describe how the state enacted a broad range of strategies to help ensure that the Medicaid
expansions had their desired impact. Specifically, this section will discuss Florida's efforts to
simplify and streamline the Medicaid eligibility determination process, implement outreach
and public information campaigns, adopt policies to recruit greater numbers of obstetrical
providers into the program, and provide technical assistance to local health departments to
foster effective enrollment and billing practices.
A. Expanding Medicaid Fligihility
In October 1987, Florida became one of the first 15 states to take advantage of OBRA-86
authority by expanding coverage of pregnant women and infants up to 100 percent of poverty.
This expansion effectively doubled the income limit for these groups, as the medically needy
income threshold had stood at just 47 percent of poverty. Nearly two years later, in July 1989.
the state further expanded coverage of these populations up to 150 percent of poverty using
OBRA-87 flexibility. Finally, in May 1992, Florida expanded coverage up to the maximum
allowed by law- 185 percent of poverty (Hill, 1992).
B. Simplifying the Medicaid Eligibility Prnre^
To complement the expansion of financial access to coverage, Florida also adopted several
strategies to simplify the Medicaid eligibility determination process in October 1987.
Specifically, the state implemented three new options also initially permitted by OBRA-86:
■ Dropped Assets Restrictions. Florida eliminated the assets test from its
application so that eligibility for pregnant women and infants could be based on
a simple test of income.
■ Continuous Eligibility. Florida also began granting continuous eligibility to
pregnant women throughout their pregnancies and a 60-day postpartum period,
regardless of changes in income.
12
Health Systems Research, Inc.
■ Presumptive Eligibility. Florida was one of the first four states to adopt
presumptive eligibility, a process by which state-selected providers were
permitted to grant immediate, temporary eligibility to pregnant women based on
a preliminary assessment of income so that they could receive Medicaid-
financed prenatal care while their formal application for Medicaid was beine
reviewed.
Over a year earlier, however. Florida had implemented two additional strategies to simplify the
process and facilitate access to coverage. These strategies are described below:
■ Shortened Application Form. In July 1986. Florida reduced its Medicaid
application from twelve pages to a single page. This form, a copy of which
appears in Appendix A, was adapted and streamlined to gather essential income
and family composition information and was used as an initial screen for not
only Medicaid eligibility but also eligibility for all federal public assistance
programs except Food Stamps.
■ Outposting Eligibility Workers. Also in July 1986, Florida began outposting
Division of Economic Services eligibility workers at health care provider sites,
including hospitals. LHDs, and FQHCs. The strategy was intended to facilitate
families' access to Medicaid coverage by eliminating the need for a separate trip
to a public welfare office to apply for Medicaid.
Florida made the most aggressive use of this strategy of any state in the nation
(Hill, 1992). As illustrated in Table 1, by April 1987, the state had hired at least
one Economic Services eligibility worker to work at each of 1 10 sites. One year
later, nearly 500 eligibility workers were distributed across 226 provider sites.
By April 1990, while the total number of dedicated outposted workers had
slipped to just over 400, these workers were still distributed across nearly 230
provider sites.
Florida officials hoped that these strategies, combined would help translate eligibility
expansions into real enrollment by making Medicaid coverage more accessible, simple, and
continuous.
13
Health Systems Research, Inc.
Tabie 1
Number of Sites, by Type, with Outstationed Eligibility Workers
and Nam ber of Worker, 1987-1990.
Date
April 87
April -88
November 88
April "90
Number of
Sites
226
229
* At least one worker per site
Source: HRS/Medicaid, 1994
Hospitals
66
n/a
n/a
95
Type of Site
LHDs
19
n/a
n/a
102
FQHCs
8
n/a
n/a
Other
n/a
n/a
16
Number of
Workers
a a
495
316
402
C Promoting \he of Prenatal Care Through Qmaafifa
While Florida has never implemented a comprehensive statewide outreach campaign to
promote the importance of prenatal care, the state has engaged in a number of important
outreach efforts and has periodically targeted selected communities for additional casefmding
activities. These efforts are summarized briefly below.
■ Toll-Free MCH Hotline, Following a directive from the federal Title V
Program, Florida instituted a statewide toll-free MCH Hotline in 1987. Women
who call the hotline can receive information about where to obtain prenatal care
in their community, how and where to apply for Medicaid coverage, as well as
general advice regarding healthy behaviors during pregnancy. Medicaid
administrative matching dollars have supported the operation of the hotline
since its inception.
■ Mass Media Efforts. In collaboration with the Florida Association of
Broadcasters, the state has periodically waged prenatal care promotion
campaigns. These efforts, again supported with Medicaid administrative match,
often used media materials developed by the State of New York featuring the
character "Stella the Stork" to promote early and frequent use of prenatal care.
■ Healthy Mothers/Healthy Babies Efforts. Beginning in 1 99 1 , state officials
delegated much responsibility for perinatal outreach to the various not-for-profit
Healthy Mothers/Healthy Babies Coalitions in existence in communities across
Florida. These campaigns varied in their timing and intensity based on the
initiative of community members and local officials.
14
Health Systems Research, Inc.
■ Casefinding in Target Communities. Judging that certain communities were
more disadvantaged than others, and anticipating that women in these
communities might need extra help in understanding the importance of prenatal
care and the availability of Medicaid coverage, Florida hired professional social
workers to perform casefinding activities in target counties.2 These social
workers have engaged in a variety of activities to identify and recruit women
into care, including door-to-door canvassing and networking with community
agencies and providers likely to serve needy pregnant women.
Once again, the various outreach efforts implemented in Florida were designed to bolster the
Medicaid expansions by advertising the availability of enhanced Medicaid coverage and to
promote the broader message of the importance of prenatal care.
D. Improving Physician Participation in Medicaid
Florida officials understood that the impact of their Medicaid expansions could be severely
undermined if there was not an adequate supply of obstetrical providers from which newly-
eligible pregnant women could obtain prenatal care. To address the state's long-standing
problem of insufficient private provider participation, Florida implemented a number of
strategies to make the Medicaid program more attractive to physicians. These efforts, which
are summarized below, involved significant increases in provider fees, initiatives to minimize
malpractice liability exposure, and efforts to create a more user-friendly administrative system.
■ Raising Provider Fees. When Florida first expanded coverage for pregnant
women, the state paid one of the lowest global obstetrical fees in the nation,
reimbursing obstetricians just $315 for all care provided to a Medicaid-eligible
mother, including both prenatal and delivery services. Recognizing that such a
low fee was undermining the program's ability to attract larger numbers of
physician participants, the state has systematically increased reimbursement
rates and brought them to approximate parity with private payer rates.
As illustrated in Table 2, the largest single increase occurred in 1988 when the
global fee was raised more than 250 percent, to $800. That same year, the
program established a new, separate global fee to reimburse OBs who treat
women determined to be high-risk; this fee was set at $1 ,200, a rate nearly four
times higher than the previous $315 fee. Another increase was made in 1989
when the base global rate was raised to $1,000. Finally, in 1992, fees were
raised once again by a significant amount. That year, the base global fee was
increased to $1,500, while the global fee for high-risk women was increased to
$2,000.
2
In 1986 and 1987, social work staff were placed in Gadsden, Columbia, Clay, Pinnellas, Manatee, Oseola,
Martin, Palm Beach, Broward, and Dade Counties. In 1988 and 1989, social workers were placed in Leon and
Putnam Counties.
15
Health Systems Research, Inc.
Table!
Medicaid Obstetrical Global Fees,
1986-1992
Year
Global Fee
Global-High Risk
1986
$315
n/a
1988
S800
$1,200
1989
$1,000
$1,200
1992
$1,500
$2,000
Source:
HRS/Medicaic
L 1994
Limiting Malpractice Liability. The State of Florida established the
Neurological Injury Compensation Association in January 1989. Under the
program, "no-fault" liability coverage for newborn birth-related injuries is
extended to physicians who elect to pay $5,000 per year into a compensation
fund. Participation is completely voluntary and carries with it no obligation to
participate in Medicaid. To help support the fund, a general tax of $250 per
year is levied on all physicians practicing in the state, as well as a $50 per
delivery tax on all of Florida's private hospitals.
Under the system, physicians are indemnified only in cases where very severe
injuries occur. For example, liability coverage only exists for babies born over
2,500 grams who sustain permanent and substantial mental and physical injuries
during labor, delivery, or resuscitation. Awards granted under the compensation
fund cannot exceed $100,000 plus expenses for medical care incurred over the
injured child's lifetime.
By 1992, the compensation fund maintained a balance of approximately $90
million and had 700 participating physicians. Twenty-five claims had been
made against the fund, 14 of which were judged to be compensable. The fund
did not seem to significantly affect malpractice premiums; just two carriers in
the state (out of approximately 15) had reduced by $5,000 their premiums for
physicians participating in the fund to offset the cost of that participation.
■ Creating New Initiatives in the Provider Relations Office. Beginning
in 1 989, the Medicaid program placed new emphasis on the role of its
Provider Relations Office and began several initiatives aimed at making
Medicaid administrative systems more responsive to the needs of
physicians. First, the office designed and distributed a new provider
recruitment brochure. Unfortunately, while it was endorsed by the
Florida Medical Association, the brochure was generally regarded as a
failure. Next, the office expanded and increased publicity for its existing
toll-free hotline through which physicians can receive personal
16
Health Systems Research, Inc.
assistance in answering billing questions and/or resolving billing
disputes. This efforts has been viewed as highly successful. .Another
strategy employed by the office was to send designated staff to
approximately 20 communities per year to recruit physicians and to
provide hands-on technical assistance to providers regarding their
treatment and billing practices. These two-day conferences, conducted
in collaboration with the Medicaid fiscal agent, enjoyed wide
participation by physicians and their business managers.3
The strategies described above were all designed to attract greater numbers of private
obstetrical providers into the Medicaid program and broaden the base of providers from whom
Medicaid recipients could choose to receive their prenatal care. A major objective of state
policymakers was to create a system that more effectively blended public and private resources
in serving low-income families.
E. Providing Technical Assistance to Counties to Facilitate Effective Implementation
The single strategy that is most often credited by Florida officials as having the greatest impact
on the state's ability to implement effective programs for Medicaid pregnant women was the
creation and deployment of Technical Assistance and Coordination Teams (TACTs) between
1988 and 1990. The TACT concept, established by the same law that originally expanded
Medicaid eligibility to 100 percent of poverty, was designed to provide hands-on assistance to
county and district staff with the implementation of their indigent care programs. Composed of
central office staff from Health Services. Economic Services, Children's Medical Services.
Medicaid, and Aging and Adult Services, the TACTs visited each district each year for a three-
vear period. During their week-long site visits, the TACTs worked with district and local staff
to assess their progress, diagnose their problems, recommend solutions, and identify innovative
practices to better serve clients (TACT Report, 1989). At the conclusion of their visits. TACTs
would prepare detailed site visit reports describing their findings and outlining their
recommendations for operational improvements. TACT reports were distributed to all districts
to further foster the sharing of information across the state.
During their tenure, TACTs tended to focus the majority of their efforts on helping counties
resolve issues in two main program areas: Medicaid eligibility and Medicaid billing. In the
area of Medicaid eligibility, TACT members worked with local staff to identify appropriate
sites for outstationed eligibility workers, smooth operational problems between the systems
and staff involved with presumptive eligibility and formal Medicaid eligibility determination,
and build stronger links between eligibility and health care provider staff.
3 It is important to note that the Provider Relations Unit was formed to serve all physician groups participating
in Medicaid and did not focus its efforts solely on obstetrical providers.
17
Health Systems Research, Inc.
In the area of billing, TACT members worked closely with local public health units to ensure
they were maximizing Medicaid revenues when serving Medicaid-eligible clients. The TACTs
were greatly assisted in this effort by the Medicaid program's development of a Medicaid
Reimbursement Guide for HRS County Public Health Units in 1989. The Guide provided
simple, step-by-step instructions on enrolling as a Medicaid provider, determining Medicaid
eligibility, understanding what services Medicaid will cover, preparing claims, and obtaining
assistance with billing problems. Florida officials credit the Guide, coupled with hands-on
assistance from the TACTs, with dramatically improving local public health units' ability to
capture Medicaid revenues to support their operations.4
Despite this improvement, local public health units are still unable to bill Medicaid for a large portion of the
services they provide. For example, unlike most states, Florida never expanded its Medicaid benefit package to
cover non-medical psychosocial support services such as care coordination, nutritional counseling, social work
services, and health education. As a result, LHDs have never had the ability to bill Medicaid for these
traditional public health services and, instead, has financed these services with federal Title V dollars and state
general revenue appropriations.
18
Health Systems Research, Inc.
V. Key Findings and Implications for the Evaluation
At a number of levels, the Medicaid expansions for pregnant women appear to have been a
success in the State of Florida. By coupling income eligibility expansions with a broad set of
innovative strategies to streamline the enrollment process, promote the importance and use of
prenatal care, and recruit physicians into the program. Florida appears to have succeeded in
providing more low-income women with insurance coverage for their maternity care. In
addition, in the period following the Medicaid expansions, statewide rates of women who
received late, no. or inadequate prenatal care declined. Most important, Florida officials can
point to a promising improvement in the health of newborns since the Medicaid expansions
were enacted.
As will be described below, state program and public health vital records data illustrate a
number of desirable trends with regard to Medicaid enrollment and expenditures, prenatal
Medicaid revenues for LFIDs, prenatal care access and utilization, and infant health. Further.
Florida officials from a broad range of agencies and programs indicated during interviews with
the RAND/HSR research team that these positive outcomes can be attributed to the state's
concerted efforts not only to expand financial access to care, but also to facilitate the
implementation of accessible and effective programs at the local level. Many of the lessons
learned by Florida officials in building these systems should prove valuable to other states
working to improve systems of care for pregnant women.
A. Improvements in Enrollment. Fxpenditures. Revenues, and Outcomes
By observing trend data from a variety of secondary sources, it is possible to describe several
preliminary effects of the Medicaid expansions for pregnant women. Specifically, these data
illustrate dramatic increases in the number of pregnant women enrolled in Medicaid, large
increases in Medicaid expenditures related to maternity care, and a steady increase in Medicaid
revenues for prenatal care provided in local health departments. Over the same period that
these trends occurred, vital records data show a steady improvement in several of the key
maternal and child health indicators of concern to policymakers.
It should be noted, however, that several of the trends presented below are drawn from the
HCFA 2082 Database. This source, while valuable in its ability to provide general descriptive
data on state Medicaid program recipients and expenditures, is limited in that it aggregates
information into broad eligibility and covered service categories. Therefore, the 2082 do not
always permit as precise measurement as researchers would prefer. In the following analysis,
for example, data are presented for "Pregnant Women and Caretaker Relatives" made eligible
through the various federal statutory changes of the late 1980s. These data therefore do not
report on the entire universe of Medicaid pregnant women in Florida, as they exclude those
who were eligible under other categories such as "AFDC Adult" or "Medically Needy Adult,"
nor do they permit pregnant women who are AFDC- or Medically Needy-eligible to be
disaggregated from their broader reporting categories. Similarly, the "Pregnant Women and
Caretaker Relatives" category includes certain recipients who were not pregnant, although
19
Health Systems Research, Inc.
according to Florida officials, "caretaker relatives" make up a tiny proportion of the recipients
reported on this line of the 2082. Recognizing these limitations, the 2082 still permits analysts
to draw at least general conclusions regarding how the Medicaid expansions for pregnant
women affected program enrollment and expenditures in Florida. These trends are described in
detail below.
1 . Increases in Medicaid Enrollment of Pregnant Women
An analysis of the HCFA 2082 data from the State of Florida illustrates how the Medicaid
expansions have led to dramatic increases in the number of pregnant women enrolled in the
program. During federal fiscal year (FFY) 1988, when the state first expanded eligibility for
pregnant women to 100 percent of poverty, over 30,000 women eligible under the new
authority received Medicaid-financed services. The following year, the number of such
pregnant women grew to over 48,000, an increase of 58 percent. During FFY 1990, the year in
which the effects of the state's expansion to 150 percent of poverty are reflected, nearly 76,000
pregnant women eligible under the expansions received Medicaid services. In 1991, this figure
grew by another 24 percent to 94,000.
In every year following the expansions, annual growth in pregnant women recipients eligible
under the expansions outstripped growth in Medicaid recipients overall. Still, such Medicaid
pregnant women continue to represent a small proportion of the overall program population;
between FFYs 1988 and 1991, pregnant women grew as a percentage of total Medicaid
recipients from 4.0 to 7.5 percent.
Table 3
Medicaid Pregnant Women Eligible Under the Expansions
a» a Percent of Total Medicaid Recipients
Federal Focal Yean 1987-1991
Year
Medicaid Pregnant
Women Under the
Expansions
1987
1988
1989
1990
1991
14,000<
30,400
48,200
75,900
94,000
Percent
Growth
117%
58%
58%
24%
Total Medicaid
Recipients
639,900
768,200
875,600
1,038,400
1,248,900
Percent
Growth
20%
14%
19%
20%
Ratio of Pregnant
Women/Total
1.5%
4.0%
5.5%
7.3%
7.5%
1987.
HRS Medicaid officials estimate that approximately 14,000 pregnant women were on Medicaid in
Sources: HCFA, 2082 Data, 1987-1991; HRS Medicaid, 1987.
20
Health Systems Research, Inc.
2. Increases in Medicaid Expenditures for Pregnant Women
Data from the HCFA 2082 also illustrate that Medicaid expenditures on behalf of pregnant
women made eligible under the expansions also rose significantly as the program expanded its
coverage. As shown in Table 4, expenditures for such pregnant women in FFY 1988 totaled
just under $24 million. That total rose by 179 percent the next year to nearly $67 million
While the rate of growth slowed somewhat over the next two years, expenditures for pregnant
women eligible under the expansions reached over $157 million by FFY 1991. As was the
case with recipient data, the annual rate of growth in program expenditures for these pregnant
women far exceeded overall growth in Medicaid expenditure during the same period. Yet this
population accounts for an even smaller proportion of total expenditures than it does of total
recipients. Between FFYs 1988 and 1991, expenditures for pregnant women eligible under the
expansions as a proportion of total Medicaid expenditures rose from 1.6 to 5.3 percent.
Year
1987
1988
1989
1990
1991
Table 4
Medicaid Expenditures for Pregnant Women Eligible Under the Expansions
as a Percentage of Tola! Medicaid Expenditures, 1987-1991 (in thousands)
Medicaid
Expenditures for
Expansion Group
Pregnant Women
$11,400'
$23,900
$66,700
$119,100
$157,300
Percent
Growth
110%
179%
79%
32%
Total Medicaid
Expend tames
$1,178,000
$1,493,300
$1,912,000
$2,360,700
$2,944,400
Percent
Growth
27%
28%
23%
25%
Ratio of Pregnant
Women/Total
Expenditures
1.0%
1 .6%
3.5%
5.0%
5.3%
* Represents expenditures for Medically Needy Adults, not all of which are for pregnant women
Source: HCFA, 2082 Data, 1987-1991
21
Health Systems Research, Inc.
3. Increases in Medicaid-Financed Deliveries
Analysis of HC FA 2082 data reveals that slightly less than one-half of all pregnant women
eligible under the Medicaid expansions use inpatient hospital services during any given year
For this population, we assume that inpatient hospital services represent deliveries Therefore
inpatient hospital users and expenditures for this eligibility group can serve as a proxy measure
of Medicaid-financed deliveries. To the extent that pregnant women may be hospitalized for
other reasons, these results will represent an overestimate of expenditures for deliveries.
Table 5 illustrates the growth between 1988 and 1991 in the number of Medicaid pregnant
women made eligible under the expansions who used inpatient hospital services and the
expenditures they accounted for. In 1988, there were 9,700 of these pregnant women who used
inpatient hospital care, for whom Medicaid paid just under $15 million. By 1991, there were
nearly 43,000 of such pregnant women using inpatient services at a total cost of over $82
million. Once again, these figures provide a proxy measure for growth in the number of
Medicaid-financed deliveries during and after the Medicaid expansions.
Table 5
Growth in Pregnant Women Eligible Under the Expansions
Who Used Inpatient Hospital Services,
and Related Expenditures, 1987-1991
Year
Pregnant Women
Under the Expansions
Using Inpatient
Hospital Services ■
Percent
Growth
Expenditures for
Expansion Group
Pregnant Women on
Inpatient Hospital
Services
Percent Growth
1987
5,100*
$9,300,000*
1988
9,700
90%
$14,700,000
58%
1989
23,600
143%
$41,700,000
184%
1990
36,300
54%
$71,100,000
71%
1991
* n
42,900
18%
$82,800,000
16%
women.
Source: HCFA 2082 Data, 1 987- 1 99 1
22
Health Systems Research, Inc.
4. Growth in Medicaid Revenues to Local Health Departments
As described in Section IV, the Medicaid program developed a guide to Medicaid billing for
local public health units in the late 1980s. With the help of the Technical Assistance and
Coordination Teams, local health departments made a high priority of billing for Medicaid
reimbursement and maximizing resources for the delivery of prenatal care services An
analysis of budget and revenue data from HRS indicates that LHDs were indeed successful in
implementing this objective.
As described in Table 6, revenues from the delivery of prenatal care to Medicaid-eligible
women grew from $7 million in state fiscal year (SFY) 1988 to almost $14 million in SFY
1991. Not surprisingly, Medicaid's relative role as a payer of LHD-sponsored prenatal care
also grew over the same period. The proportion of total LHD revenues made up by Medicaid
rose from essentially zero in SFY 1986 to nearly 30 percent in SFY 1991.
::;':; '■ '
Table 6
LHD Prenatal Care Reven oes from Medicaid, as Percent of Total
Prenatal Revenues, 1987.1992 (is millions)
Year
Medicaid Prenatal
Reven oes
Total Prenatal
Revenues
Ratio of Medicaid/Total
1987-88
$n/a*
$27.5
n/a%
1988-89
$7.0
$34.4
20.3%
1989-90
$11.0
$39.3
28.0%
1990-91
$13.0
$46.1
28.2%
1991-92
* iil:i_ Lin; c i /
$13.7
$46.5
29.5%
the accurate reporting of this indicator.
Source: HRS, 1995
23
Health Systems Research, Inc.
5. Improvements in Key Maternal and Child Health Indicators
Over the period during which Florida expanded its Medicaid program for pregnant women, the
state experienced overall improvement in several key indicators of infant health and maternal
utilization of prenatal care. As displayed in Table 7:
■ The rate of infant mortality fell from 11.0 deaths per 1 ,000 live births in 1 986 to
8.8 deaths per 1,000 live births in 1992. Over this period, Florida's national
ranking on this indicator also improved from 35th to 29th among the states.
■ The rate of babies bom at low birthweight, after rising slightly between 1986
and 1988, fell to an all-time low for the state of 7.4 percent in 1992. Once
again, improvement over the period 1986 to 1992 moved Florida up in national
rankings on this indicator from 40th to 34th.
■ The percentage of women who received late or no prenatal care during their
pregnancy fell from 8.6 percent in 1986 to 4.6 percent in 1992.
■ The proportion of women who received adequate prenatal care, as measured by
the Kessner index, rose from 62.0 percent in 1986 to 71.6 percent in 1992.
While these improvements cannot, in this analysis, be directly attributed to the Medicaid
expansions, they provide strong support for the conclusions that women in Florida are
receiving earlier and greater amounts of prenatal care and that birth outcomes are improving.
Table 7
Selected Maternal and Child Health Indicators, 1985-1992
Indicator
J986
1987
198S
1989
1990
1991
1992
Infant Mortality
(Deaths/1000 Births)
11.0
10.6
10.6
9.8
9.6
9.0
8.8
Low Birthweight
(Percent)
7.6
7.7
7.7
7.4
7.4
7.4
7.4
Late/No Prenatal Care
(Percent)
8.6
8.2
8.2
8.1
7.1
5.9
4.6
Adequate Prenatal Care
(Percent)
62.0
62.8
62.8
64.5
66.6
69
71.6
Source: NCHS, 1986-1992, 1990, 1991, 1992, 1994
24
Health Systems Research, Inc.
B. Qualitative Impressions of Program Impact and Lessons Lamed
During extensive interviews, officials representing a broad cross-section of Florida programs
summarized their impressions of the effects of the Medicaid expansions for pregnant women
and pointed to a number of lessons they learned while working to implement effective
programs at the local level. The key points raised during these discussions are summarized
below.
■ Florida officials are well aware of the dramatic increases that have occurred in
Medicaid enrollment of pregnant women since the program expanded its income
limits. They note that never before in the program's history has there been such
a significant and rapid response to a change in eligibility policy. In fact, Florida
Medicaid had had a long tradition of being underused by low-income
individuals and families prior to 1986. For example, the Indigent Health Care
Study conducted in 1985 found that, in addition to the roughly 300,000 persons
who would be made eligible for Medicaid through expansions of categorical and
medically needy coverage in '85/'86, there were also another 300,000 persons
who were already eligible for Medicaid by virtue of meeting either AFDC or
SSI criteria but were not enrolled in the program (Lou Harris and Associates,
1 985). Data from the HCFA 2082 demonstrate that large increases in
enrollment directly correspond to expansions of income eligibility thresholds for
pregnant women.
■ In part, this enrollment trend supports the notion that a great need and demand
for insurance coverage existed among low-income pregnant women during the
late 1980s. However, in translating this need into access to care, Florida
officials point to their efforts to simplify and streamline the Medicaid eligibility
process as the key influencing factors. In particular, the state's aggressive
outstationing of eligibility workers at the sites where women sought care is the
policy that is believed to have had the greatest positive impact on the process.
Bolstering this policy, the state's use of Technical Assistance and Coordination
Teams (TACTs) is also cited as the key strategy that helped local agencies and
providers implement simplified eligibility systems at the local level.
■ Florida officials are also well pleased with the success that LHDs have had in
capturing Medicaid reimbursement in support of their delivery of prenatal care.
Over a short period, the public health system evolved from one that did virtually
no billing of Medicaid to one that secured more than 30 percent of its prenatal
care revenue from the program. A key factor in this change was the
development by Medicaid of the user- friendly Medicaid Reimbursement Guide
for HRS County Public Health Units. And, once again, the use of TACTs to
provide hands-on technical assistance in billing practices helped to ensure
effective local-level implementation.
25
Health Systems Research, Inc.
The influx of Medicaid revenues into the public health system has allowed
LHDs to expand their capacity to provide prenatal and related care. In
particular, Florida officials believe that, as Medicaid revenues have increasingly
supported the delivery of medical prenatal care, public health dollars have been
redirected toward financing the delivery of non-medical support services such as
nutritional counseling, health education, and home visiting, as well as medical
prenatal services for populations not covered by Medicaid, such as migrants.
Public health officials did note, however, that certain inefficiencies in the
system have persisted. In particular, the fact that MCH and Medicaid officials
have never successfully negotiated arrangements to permit Medicaid payments
for local public health units' provision of psychosocial support services for
high-risk pregnant women has meant that the state has continued to finance the
delivery of such care with state funds.
It is the belief of Florida officials that the Medicaid expansions were
implemented in a relatively consistent manner across the state and, therefore,
the potential for examining intrastate variations in implementation and impact
under this evaluation is low. While the county public health system in Florida
could be characterized as variable in its capacity to serve low-income families in
the early 1980s (Clarke, 1986), efforts like the IPO Project and the provision of
local-level guidance through the TACTs served to "level the playing field" and
lent an intrastate consistency to the implementation of the Medicaid expansions.
Florida MCH officials believe that the delivery of prenatal care remained
largely the domain of public health until 1992, when Medicaid increased its
obstetrical fees to $1,500 for normal deliveries and $2,000 for high-risk
deliveries. And while the program had raised OB fees in prior years, Florida
officials do not believe that significant changes in physician enrollment rates
occurred until after the 1992 changes. Therefore, while data will show increases
in Medicaid-financed prenatal care throughout the period of expansion, they
will not show large increases in care received from the private sector until
relatively late in the expansion process.
Finally, Florida officials believe that two related factors were probably most
important to the state's overall success in implementing the Medicaid
expansions. First, effective interagency collaboration, under the strong
leadership of the HRS Deputy Secretary for Health and Deputy Assistant
Secretaries for Medicaid and Economic Services, enabled a coordinated process
of development of policies and programs within state government. Second, the
creation of the Public Medical Assistance Trust Fund and, with it, the tax on
hospital revenues, resulted from a carefully orchestrated process of consensus-
building between public and private partners and safeguarded an environment in
26
Health Systems Research, Inc.
which programs for the medically indigent could be developed and financed in a
fair, equitable, and collaborative manner.
Overall. Florida stands-out as a state that has aggressively addressed the problems of infant
mortality and medical indigency by expanding Medicaid coverage of pregnant women in a
highly effective manner. This case study provides detailed information regarding how the state
implemented the Medicaid expansions, information that should be helpful in interpreting the
results of the evaluation's quantitative analyses. It is also hoped that this report will provide
other states with helpful descriptive information on innovative strategies for implementing
improved maternal and child health programs.
27
Health Systems Research, Inc.
Works Cited
Brown. Lawrence D. ••Commissions. Clubs, and Consensus- Florin, o
Care Reform.- Health Affairs. Summer 1993. ^^ F1°nda ^organizes For Health
Children's Defense Fund. The Health of America s Children- LA„ , ,„
Data Book. Washington. DC: 1 989. Maternal and Child Health
Clarke. Gary. Florida s Health Care Access Acf Radiml <r,„„«. <~
Washington, DC: The Intergovernmental He^h £££& ^IST*" ^
sss^^r1 Rehabi,itative Semces- Heaith *•» ™«-
Florida Department of Health and Rehabilitative Services. Medicaid Office. Tallahassee, FL:
Florida Department of Health and Rehabilitative Services. TACT Report. Tallahassee, FL:
Florida Department of Health and Rehabilitative Services. Health Program Office
Tallahassee, FL: Due January 1995. »wum«.
Health Care Financing Administration. State Medicaid Program Characteristics Chart
Baltimore, MD: Health Care Financing Administration. 1984.
Health Care Financing Administration. "Analysis of State Medicaid Program Statistics, 1984 -
Health Care Financing Program Statistics. Baltimore, MD: Health Care Financing
Administration. August 1985.
Health Care Financing Administration. HCFA 2082 Data Baltimore, MD: Health Care
Financing Adininistration. 1987. 1988. 1989. 1990. 1991.
Hill, Ian. Broadening Medicaid Coverage of Pregnant Women and Children: State Policy
Responses. Washington, DC: The National Governors' Association. February 1987.
Hill, Ian. "Improving State Medicaid Programs for Pregnant Women and Children." Health
Care Financing Review. Baltimore, MD: Health Care Financing Administration. 1990
Annual Supplement
Hill, Ian. The Medicaid Expansions for Pregnant Women and Children: A State Program
Characteristics Information Base. Washington, DC: Health Systems Research, Inc. February
1992.
Health Systems Research, Inc.
Hill. Ion. Long. Stephen, and Marquis. Susan, impact of Medicaid Fha h r c
Innovative Programs for Maternal Health Care Uethodolovv Zi*. J ?' Expan"0^ ^
Washington. DC: The RAND Corporation under HC 'a cZ r^T ***""**"
*)113 9-01. August 24. 1992. cooperative Agreement No. 18-C-
Insmutec^ted^ne. ftwrtf£»toH,k Washington. DC: National Academy
Lou Harris and Associates. Medicaid and Indigent Health Care Sunn w r c
Study. New York, NY. March 1985. ' W Cost Estlma"on
National Association of Community Health Centers. Access to Commit*, u ,t/~
DmBootim Washington, DC: 1991. ^^s to Community Health Care. A
National Center for Health Statistics, "'Advanced Report of Final Mortality Statistics " 1990
Month* Vital Statistics Report. Vol. 41, No. 7 Supplement, HyartsviUe, MD. S^'
National Center for Health Statistics, "Advanced Report of Final Mortality Statistics "1991
Monthly Vital Statistics Report. Vol. 42. No. 2 Supplement, HyattsvUle, MD. Public HeTm
National Center for Health Statistics, "Advanced Report of Final Mortality Statistics " 199?
Monthly Vital Statistics Report, Vol. 43, No. 6 Supplement, HyattsvUle, MD, Public Health"
National Center for Health Statistics, Vital Statistics of the United States, 1990 Vol 1 Natalirv
Washington Public Health Service, 1994. ' ' y'
National Center for Health Statistics, Vital Statistics of the United States, 1991 Vol 1 Natality
To be published 7"
National Center for Health Statistics, Unpublished data, 1986-1992.
Social Security Administration. Program Data for the Aid to Families with Dependent
Children Program. 1984.
U.S. Bureau of the Census. March 1986-89 Current Population Suney. Washington, DC.
Ventura, S.J., Martin, J.A., Taffle, S.M., et al., "Advanced Report of Final Natality Statistics,"
1992, Monthly Vital Statistics Report, Vol. 43, #5 Supplement, Hyattsville, MD, National
Center for Health Statistics, 1994.
Htalth Systems Research, Inc.
APPENDIX A:
Florida's Shortened Medicaid Application Form
F l— t-Uck.nrnaU.Ooc "34*1 MiPM
Health Systems Research, Inc.
REFERRAL
SOURCE
RART 1 - HOUSEHOLD INFORMATION
NAME
FIRST
MIOOLE
IASI
STATE OF FLORIDA
SS^STESlTi* MEALTH AND REHABILITATIVE SERVICES
APPLICATION FOR PUBLIC ASSISTANCE
His
SOCIAl SECURITY
NUMBER
IJVING
ADORE SS
STREET
OTY
DATE
OF BIRTH
PHONE AflEACOK
NUMBER ( )
SHELTER
SITUATION
RELATIONSHIP
TO YOU
US HUH 1 1
YE S / NO
riYi.b dm,,
111 INI)
Yl S / N( I
Dyl;, Dn..
DYeb Dm.
DYeb Dn,.
MAILING
ADDRESS
QYeb n»i
Dft». Dn.
fJv.u Hi*,
LJv. . I iu,
dyc. Dm.,
DYeb DNu
DYeb Dnu
Jll'lllhiuil
QYub Dn..
I'l'.AHl I I)
<\ S / No
II.. •:. Dn,,
Ljji b Dnii
Uy.«. iIni
PHEUNANI
YE S / NO
H Yus D N,
Dvb Dn,,
nt.uijN,,
DYeb Hn., HYr-bflN..
I life* llNl,
Uv.b Dno
Dieb Dn,.|t]y,.> pM
ijY.b tjn..
D Rem D MudQdge
D Own □ Oilier
DYeb D No
DrebDNoJ
PART 2 - FINANCIAL INFORMATION
ivpe
INCOME
WHO HAS
l .1111 li AUDI I
(AI* □ Yeb ( J
&oatfS*aj>«y
C«»tt*on«
881
AfOC
GROSS
AMOUNT
HOW OMEN
TOTAL INCOME
ASSETS
Examples
Cm*
CftKhnu
wxuunl
en/Hut*
MMofcyaa
■»«
htm lunas
mauianc*
PART 3 - INTENT TO APPLY
Bom
StoUa'ttufU*
I— HE
TOTAL ASSETS
RART 4 - INTERVIEW INFORMATION
I am applying loi jbbibl..i« i! I ..(hleibljml lh.,1 I «,,«
lidve lt> yiue line uiIuiiiuiioii itn liws loim II ukJU Ik.
a Clinic il I he oi liKie inloiiiialion auool my olKjiUl.ly
loi abbiblan, e
I ayiue kj lull! in papers lh.,1 ale required lu n»,»
Hidl I dill eligible e«Lepl loi |,apeib I canno! .jel
llHOnyll no IjuII ol my own II I Cdllllol Jol papcib
I ayiee to yiv/e Hie names ol any pursuits m uW.os
llidl may be contacted loi Hie reqmieu .iiiu,,„ai„„,
I ayiee Hi.. I Hie Depat lutein ..I Health an. I
Hellauililalivc Semites Hie Division ,.l l',.i,i.
Abbiblaik.e haul. .111,1 autlioii/eil Fmltu.ll Anem « .
may venly Hie iiiIoiiiuIhhi I give on Hub lurm am ,,i
my inlervrew I agree lli.il lliey may . oniat I ,.,,
prcbeill ill |iabl elliployeib ll ll leldlUS I., ,..,
eligibility I aglet- llial lliey , (>.| mloniial,..,! II... I
allet.lb my elKj.liilily la.n. any ret onto
INTERVIEW Tift
OFFICE ADDRESS
AM/PMDATE
/ /,
0 """il
HMtK»_
LJ Hd«*i» Fiaub L»»
HAS E Sfuin ?U«? N,.v»tj|n«,.<i*„. IKIS FSfuni20«l«l>l,.n
">/«/ nil ft
O fooo SAimp NcKk* ol
*»»»„IV^ I,
«HI««II
SCHEtNERS
DATE
STAMP
C-l
APPENDIX C. EVALUATION OF A MEDICAID ELIGIBILITY EXPANSION IN FLORIDA:
DEVELOPING THE DATABASE
Ellen R. Harrison
Stephen H. Long
M. Susan Marquis
INTRODUCTION
This report documents the construction of the analytic database
used in the study of the 1989 Medicaid eligibility expansion for
pregnant women in Florida. The creation of this database was a key
challenge in the research. The resulting procedures may serve as a
model for other evaluations of the Medicaid expansions.
The Medicaid eligibility expansions for pregnant women and children
were the most important policy changes in the program in the 1980s. Yet
there are only a limited number of studies of the effect of these
expansions and it is not clear from these studies whether the expansions
led to an improvement in prenatal care and birth outcomes.1 Moreover,
to understand the full effects of the interventions it is also essential
to understand how the Medicaid program changes affected other government
programs that deliver prenatal care and how the expansions affected
private payers. The effects on prenatal care access and birth outcomes
are likely to be quite different if Medicaid financed care substitutes
for care previously financed and provided under other programs, such as
Title V or private insurance, rather than providing new coverage for
those who previously lacked insurance or access to other public
programs. None of the previous studies, however, addresses these
substitutions .
The objective of our study was to investigate these interactions
between the Medicaid program and other sources of financing and
providing maternal health care, and, with this perspective, to study
whether pregnant women newly entitled to Medicaid coverage received more
or earlier prenatal care, and whether their maternal and birth outcomes
were improved. We studied the experience in Florida. Florida is a good
1 Alpha Center, The Medicaid Expansions For Pregnant Women and
Children. Washington, DC: Alpha Center; 1995.
C-2
site for this study for a number of reasons. It ranks fourth among the
states in total population, and there are about 200,000 births each
year. Florida significantly expanded Medicaid eligibility for pregnant
women and also aggressively implemented other strategies to ensure that
women who were made eligible by the expansions gained coverage under the
program. Florida relies heavily on county public health department
clinics to provide prenatal care to its low-income women, and hence is a
good place to study interactions between the Medicaid financing changes
and the publicly financed direct-delivery system.
To conduct our analysis, we constructed a linked person-level
database for the years 1988 through 1991 from vital statistics records,
hospital discharge abstracts, public health system encounter data,
Medicaid eligibility files, American Hospital Association annual survey
data, and data from the 1990 Census. This report first briefly
describes the content of each database, and then details our file
linkage procedures and results.
THE DATA
Vital Statistics Data (VS)
Source: Florida Department of Health and Rehabilitative Services
Years: 1988-1991
The Vital Statistics data (VS) contain information obtained from
birth certificates, fetal death certificates, and death certificates
from 1988 through 1991. They provide a detailed record of every birth
and fetal death in the state of Florida for those years and define our
study universe. Table CI presents the total number of birth
certificates and fetal death records in each of the study years. To
determine the actual number of women who delivered in Florida, we
subtracted the 1.2 percent of certificates that correspond to additional
certificates for the same delivery (multiple births), and also excluded
the births that took place outside of Florida. For example in 1991, the
195,756 certificates correspond to 193,393 mothers. After subtracting
the out-of-state deliveries, we are left with 193,292 deliveries in the
state of Florida.
C-3
Table CI. Live Births and Fetal Deaths from Vital Statistics
Data, 1988-1991
Live Births Fetal Deaths Total Certificates
1388 185,034 1,772 186,806
1989 193,893 1,833 195,726
1990 200,334 1,847 202,181
1991 194,043 1,713 195,756
The records include information about the pregnancy such as the
initiation and frequency of prenatal care; measures of birth outcomes
such as birthweight and complications of delivery; and demographic
characteristics of the mother and baby. Table C2 presents the key
variables for analysis and for linking the vital statistics data with
other databases. Because the mother's Social Security number is unique,
universally used, and was included on over 95% of the VS records, it is
the primary linkage variable. The mother's date of birth, name, and
zipcode as well as the hospital identification number were also used to
link to other files.
Table C2. Key Patient-Level Variables for Analysis
and Matching from Vital Statistics Files
Mother's Characteristics
• Name
• Social Security number
• Date of birth
• Zipcode of residence
• Race
• Educational attainment
• Marital status
Baby's Characteristics
• Name
• Birthweight
• Gestational age
• Sex
• Congenital anomalies
Pregnancy /Delivery Characteristics
• Mother smoked while pregnant
• Mother consumed alcohol while pregnant
• Number of previous births
• Singleton/multiple birth
• Initiation of prenatal care
• Number of prenatal care visits
• Labor complications
• Delivery complications
C-4
The infant death records enable us to identify those infants who
did not survive the first year of life. An infant death file for each
year includes babies born during the year who died within the following
twelve months. For 1988 and 1989 we were provided with a matched file
that had already linked the death certificate information with the birth
certificates. For 1990 and 1991 we created the linked file which
combined information from the infant death file and the birth
certificate file by matching death certificate numbers, when available,
or by matching mother's and baby's name and baby's date of birth. We
were able to match 97.5 percent of the 1990 infant mortality records and
95.3 percent of the 1991 records. Table C3 presents the number of
infant deaths and the infant mortality rate by year.
Table C3. VS Infant Deaths, 1988-1991
Infant Deaths % of Live Births
1988 1,892 1.02%
1989 1,871 0.97%
1990 1,883 0.94%
1991a 1,441 0.74%
aWe did not have access to 1992 deaths
so the 1991 infant deaths represent just
those infants who died in 1991.
Hospital Discharge Data (HDD)
Source: Florida Agency for Health Care Administration
Years: 1988-1991
The hospital discharge data contain hospital discharge records from
all non-military acute care hospitals in Florida. We matched records of
the mother's delivery hospitalization in this file to records from the
VS file to obtain a number of key variables that were not available
from other sources. Most important to our analysis was the insurance
status variable which was defined as the principal payer for the
hospitalization and had the following values:
C-5
• Medicaid
• Medicare
• Private
• Other
Because women of child-bearing age do not generally qualify for
Medicare, we combined "Medicare" with "Other" which included Champus,
Veteran's Administration, other government third-party payers except
Medicaid, charity, and uninsured. We also collected primary and
secondary diagnostic codes, procedure codes, and the total charges for
the delivery. The only person-level identifiers available to us were
sex and date of birth.
We obtained all discharge records that had a maternity-related
principal or secondary diagnosis. Table C4 presents the number of
discharge records on the resulting HDD file.
Table C4. HDD Maternity Related Discharge Records, 1988-1991
1988 1989 1990 1991
Pregnancy related 24,364 24,764 23,386 21,442
Delivery related 185,493 193,751 199,145 191,482
Public Health System Encounter Data (PH)
Source: Florida Department of Health and Rehabilitative Services
Years: 1987-1991
The Public Health System Encounter Data (PH) includes information
about each public health system client and his/her visits to public
health department clinics from 1987 through 1991. We included the 1987
data to capture all of the prenatal care visits for births occurring in
the first part of 1988. We received over 13 million records for the
five years of data, so our first task was to select prenatal care
records--visits in the Improved Pregnancy Outcome program and maternity-
related Family Planning visits. We identified roughly 4 million
prenatal care records. We linked all of the prenatal care visits for a
woman into a single episode of prenatal care. For each episode, we
C-6
constructed measures of the number of PH visits and the trimester in
which prenatal care in the public health system was initiated.
Individuals are identified in the PH system using Social Security
number, though 6 percent of the cases did not have valid values.
Because we matched the PH data to the VS file, we kept only those
episodes with valid Social Security numbers. Table C5 presents the
number of prenatal care episodes and total number of encounters per
year, where year represents the ending date of the prenatal care
episode. The end date was determined by either the presence of a
postpartum visit or a lapse of more than 90 days between visits. In
both cases, the end date was set to the previous visit.
Table C5. PH Prenatal Care Episodes and
Total Visits, 1987-1991
Year
# Episodes
Total # Visits
1987
35,243
109,418
1988
43,115
185,495
1989
48,817
263,995
1990
62,646
331,932
1991a
83,299
497,076
aThe number of 1991 episodes and visits
are is artificially high because they
include some prenatal care episodes for
births occurring in 1992.
Medicaid Eligibility Data (ME)
Source: Florida Agency for Health Care Administration
Years: 1988-1991
The Medicaid Eligibility (ME) file we received from the Agency for
Health Care Administration was extracted to provide us with a complete
account of periods of eligibility for women of childbearing age (12
through 55) who were enrolled in Medicaid at some time during 1988
through 1991. It was important to match this file to our other data
sources to verify the insurance status of the mother from the HDD file,
to determine the reason for eligibility, and to examine the timing of
the eligibility relative to the pregnancy. After selecting those
records with valid Social Security numbers, in order to match to the
C-7
other databases, and those with eligibility periods that included any
portion of our study years, we had roughly 1.3 million records for the
study period. Table C6 presents the total number of eligibility records
for each of the years.
Table C6. Medicaid Eligibility
Records, 1988-1991
Year # Eligibility Records
1988 240,905
1989 297,347
1990 364,960
1991 387,817
The ME files contain name, Social Security number, Medicaid
identification number, the periods of Medicaid eligibility, and the
basis of Medicaid entitlement for each eligibility period. We
categorized the basis of entitlement into the following groups:
Aid to Families with Dependent Children (AFDC)
Medically needy
Expansion group below 100% of poverty
Expansion group between 100 and 150% of poverty
Other
American Hospital Association Annual Survey of Hospitals (AHA)
Source: American Hospital Association
Year: 1991
The 1991 American Hospital Association Annual Survey of Hospitals
contains hospital characteristics for all participating hospitals in the
United States, including 306 Florida hospitals. We supplemented the
file with information from the 1988 AHA Guide for nine hospitals on our
files that either closed or merged with other hospitals prior to 1991.
C-8
Our primary use of these data was to measure the ownership of each
hospital in which each delivery occurred. The variable has the
following categories:
• Government, nonfederal
• Nongovernment, non-profit
• Nongovernment, proprietary
• Federal
We also used these data to measure delivery specific
characteristics of each hospital such as number of births, number of
newborn days, and presence of a neonatal intensive care unit, as well as
general information concerning the type of hospital, number of beds,
number of Medicaid admissions, teaching intensity, and total expenses.
Census of Population and Housing Summary Data (CEN)
Source: Bureau of the Census
Year: 1990
The Census Bureau produces a summary tape (Summary Tape File 3B)
for the Census of Population and Housing that provides summary
information aggregated to the zipcode level based on a 100% sample of
all persons and housing units in the United States. We created an
extract of the 826 Florida zipcodes that existed in 1990. Because we
did not have a measure of income for the women in the study, we used
incomes in the residence area as a proxy measure. The CEN file was used
to calculate the percentage of zipcode residents falling into different
income categories expressed as a percentage of the federal poverty
level. By merging these data to the VS file, we constructed a proxy
measure of income for each woman giving birth in a year based on the
income distribution in the zipcode of residence.
C-9
MATCHING PROCEDURES
Linking Hospital Discharge Data to the Vital Statistics
Matching Algorithm
The primary challenge in the linkage process was to match the vital
statistics record of a birth with the mother's hospital discharge record
in the absence of a unique identifier on the hospital discharge data.
Our algorithm matched by hospital first, and then within hospital, by
mother's date of birth, baby's date of birth, and mother's zipcode of
residence.
Both files had a hospital identifier, but because the two sources
used different coding schemes, we used county and hospital name in our
matching algorithm. After modifying the two lists to have standard
abbreviations and punctuations, we matched by county and the first eight
letters of the hospital name. This task was complicated by hospitals
that changed names, merged with other hospitals, or closed during the
study years. To aid in our matching of hospitals with similar names,
we made separate calculations from the two files of the number of births
occurring at each hospital and then compared the totals from the
potential matches.
For each of the four study years, roughly 94 percent of the births
on the VS file occurred at hospitals that were candidates to match to
the HDD file. Over 40 percent of the non-candidate births took place
at military hospitals, while another 20 percent were performed at
maternity centers. Neither type of facility is included on the HDD
database. Table C7 presents the distribution of location of births by
year. The top row represents the percentage of cases that were eligible
for the next step of the matching algorithm.
C-10
Table C7. Distribution of Location of VS Births and Fetal
Deaths, By Year
Location
1988
1989
1990
1991
HDD hospital
94.4
94.6
94.1
94.7
Military hospital
2.5
2.2
2.4
2.4
Maternity center
0.9
1.1
1.2
1.3
Non-hospital
0.7
0.7
0.6
0.6
Out of Florida
0.6
0.5
0.5
0.1
Enroute to hospital
0.2
0.2
0.2
0.1
Unknown
0.7
0.8
0.9
0.9
Once we made a hospital-level match, we used a patient-level
matching algorithm. We first identified the variables that were common
to both f iles--birthdate of patient and zipcode of residence. The
mother's HDD discharge record also included the date of first surgical
procedure which is almost always the delivery date, and thus could be
used as a proxy for the baby's birthdate. This additional linking
variable increased our ability to find unique matches and increased our
confidence in the validity of the match. The third data element, zipcode
of residence, was available on 99% of both the VS and HDD files. We
used a hierarchy approach in which we first required a match on all
three linking variables, then subsequently loosened the restrictions
requiring either an exact match on mother's birthdate and baby's
birthdate with missing or nonmatching zipcodes, or else a match on
mother's birthdate and zipcode with the baby's birth falling sometime
within the time of hospitalization, although not on the procedure date.
Using this algorithm, we were able to match between 91 percent and
94 percent of the VS events that occurred at an HDD hospital. If we use
the entire universe of VS births and fetal deaths as the denominator,
the match rate ranges from 86 percent for 1988 to 89 percent for 1991.
Table C8 presents the VS records that had matched to the HDD at the
hospital level, and identifies the percentage satisfying each of the
match criteria, as well as those that remain unmatched.
C-ll
Table C8. VS Match Rate to HDD File, 1988-1991
1988 1989 1990 1991
Match on mother's dob, baby's
dob, zipcode 73.5% 74.1% 76.9% 77.3%
Match on mother's dob, baby's
dob (not zipcode) 12.7% 11.1% 10.6% 11.0%
Match on mother's dob &
zipcode, birth w/in
hospitalization 5.1% 7.6% 6.7% 6.2%
Unmatched VS record (w/match
at hospital level) 8.8% 7.2% 5.7% 5.5%
If we had limited our definition of a successful match to only
those instances in which there was agreement on all three linkage
variables, our match rate would have decreased to an average of 75
percent of the eligible VS records. We felt that it was important to
allow for some disagreement in variables that either tend to have
missing values or are coded less reliably such as zipcode. Similarly,
because procedure date is not a perfect proxy for baby's birthdate, it
was reasonable to allow some leeway.
Alternatively, we could have increased our match rate by loosening
our restrictions or by adding a step in which records were compared
manually. The former approach would have decreased the reliability of
the data to a point where we would begin to lose confidence in the data.
The latter approach of taking the time to manually compare lists of
records to pick up a few additional matches was a trade-off that we
rejected given the immense size of the databases and the high match rate
we were able to achieve using a completely automated approach.
Duplicate Matches
Our confidence in the matching algorithm was bolstered because over
97% of the VS records that matched were one-to-one matches. The non-
unique matches resulted from two main factors: (1) records for more
than one mother in one system matching a single record in the other
system, and (2) multiple births. Because these duplicates accounted for
only 1.2 percent of the matched VS records in 1988 and less than 0.2
percent in each of the later years, we decided to maintain all the
records in our files, but 'randomly assign only one of the matches to be
used for analyses so that the analysis sample accurately represented the
C-12
number of deliveries in Florida. In contrast to the 1989-1991 files,
the 1988 VS file recorded mother's age rather than birthdate. Because
age in years is less unique than birthdate, there were more instances of
multiple records with the same match characteristics. In addition, for
the larger volume hospitals, the largest of which had over 14,000
deliveries per year, there were some cases with non-unique matches.
We also needed to adjust for the 2.3 percent of babies that were
part of multiple births. These cases had a birth certificate for each
baby but only one mother's discharge record. Because our unit of
analysis was the delivery rather than the actual birth, we picked one of
the birth certificate records at random to be used in analyses.
Unmatched HDD Records
Although we used the VS file to define our universe, we also
examined the nonmatching HDD records. An average of 13 percent of HDD
records identified as delivery discharges did not match to the VS file.
Table C9 presents the match rates by year. It is possible that some of
the non-matches are non-residents. It is also conceivable that some of
the discharge records that we defined as delivery hospitalizations based
on the ICD-9 code could have been coded incorrectly and were actually
other pregnancy-related hospitalizations.
We examined total charges and insurance status to determine if the
nonmatches had different characteristics than the matches. The median
total charge did not differ greatly, but there were significantly more
nonmatches covered by "other" insurance than matches.
Table C9. Match Rate for HDD Records, 1988-1991
1988 1989 1990 1991
Match to VS 157,620 167,642 173,671 169,069
(85%) (87%) (87%) (88%)
Unmatched 27,873 25,929 25,474 22,413
(15%) (13%) (13%) (12%)
C-13
Linking in Medicaid Eligibility Data
The VS and ME both contain Social Security number and mother's date
of birth. We linked the files by mother's Social Security number and
then verified the match by comparing the date-of-birth values. For each
study year, approximately 9 5 percent of the VS records had valid Social
Security numbers and were candidates for matching.
Because the ME had one record for each period of Medicaid
eligibility, after matching by Social Security number, we needed to
identify which eligibility period contained the date of baby's birth in
order to correctly measure whether the mother was covered at the time of
delivery and the basis of entitlement at that time.
Of the HDD records with valid Social Security numbers and with
Medicaid as the primary payer, approximately 80 percent matched to an
eligibility record on the ME file. Table CIO presents the match rate by
year.
Table CIO. Percentage of Medicaid
HDD Matching to ME File
Year
Percentage
1988
67%
1989
73%
1990
79%
1991
92%
Although the key reason for linking to the ME file was to identify
the basis of Medicaid entitlement, the match to the eligibility file
also enabled us to identify the potential false positives --women
identified as being covered by Medicaid on the HDD but not eligible at
the time of delivery on the ME file.
For 1991, we were also able to examine the false negatives--
Medicaid eligible mothers not identified on the HDD as having a Medicaid
payer--by linking to the Medicaid inpatient hospital claims data (MC) .
Claims data with delivery diagnosis codes were used for this
investigation in order to determine the number of Medicaid mothers who
had deliveries--that is, the denominator for the estimate of false
negatives. These records were matched back to the ME using the Medicaid
C-14
id, and were then linked to the VS file using the method above. Eighty-
five percent of the delivery claims on the MC file successfully matched
to the VS file.
Linking in Public Health System Encounter Data
Because the dates in the PH data are dates for prenatal care
visits, there was no exact date we could use to match to the baby's date
of birth on the VS file. However, we could not ignore date altogether
because a number of women gave birth multiple times within the study
period. In addition, roughly 120 women per year appear on the VS file
as having delivered twice within the same year.
To ensure that the episode was matched to the correct delivery, we
developed a two-stage algorithm. We first matched on mother's Social
Security number and the year of baby's birth where we used the end of
the prenatal care episode to define the PH value for year (though for
episodes ending in 1987 we coded the year as 1988 to match to the
earliest births in our study) . To verify the matches we compared the
birthdate to the prenatal care start and end dates and flagged those
matches in which the birthdate either occurred prior to the first
prenatal care visit or later than six months after the last visit.
Flagged records were then placed back in with the pool of unmatched
records, the birthyear value on the PH record was recoded to one year
later, and the second stage of the match was performed using mother's
Social Security number and the alternate birthyear.
To accurately count the total number of PH visits associated with
deliveries, we needed to estimate what percentage of the records
corresponded to a birth during the study years. By examining the
records that matched to the VS file, we used the relationship between
the last date of prenatal care and the year of delivery to simulate the
delivery year for the records we were unable to match. For example, 14
percent of the prenatal care episodes ending in October 1990 matched to
a delivery in 1991, so we estimated that 14 percent of the unmatched
records ending in October 1990 corresponded to births that occurred in
1991. Because we did not have access to 1992 records, we also estimated
that 14 percent of the October 1991 episodes corresponded to births
outside of the study period. Assigning unmatched records in this way,
C-15
between 73 and 81 percent of the PH prenatal care visits that
corresponded to a delivery in 1988 to 1991 were matched to the VS file.
Table Cll presents the match rate for the PH records by the year of
delivery.
Table Cll. Percentage of PH Prenatal
Care Visits Matched to VS File, By
Year
Year
Percentage
1988
77%
1989
81%
1990
74%
1991
73%
Linking in AHA Data
The AHA file's hospital identification number was different from
both the VS and HDD identifiers. We used the same algorithm that was
used for the hospital level match between the VS and HDD files, although
we could now compare the AHA name with both the VS and HDD version of
the hospital name. After most matches were made using an automated
routine comparing the first letters of the hospital names, the few
remaining cases were examined manually to identify the correct match.
Of the 306 hospitals on the AHA file, all but twelve matched to either a
VS or HDD hospital. An examination of the "number of births" variable
from the AHA file revealed that none of the unmatched hospitals had any
deliveries .
Linking in the CEN Data
Although there were 4,148 different zipcodes of residence for the
women on the VS file who gave birth between 1988 and 1991, 95 percent of
the women lived in the 825 zipcodes that matched to the CEN file. Of
the VS cases that did not match, 86 percent had zipcodes that were in
the range of Florida values but were either mispunches or zipcodes that
did not exist at the time the CEN file was created. The 825 matched
zipcodes accounted for all but one of the zipcodes on the CEN file.
C-16
The Final Linked File
The final linked file contains one VS record per delivery and
includes only those that matched to a HDD record. Table C12 presents
the original number of records and then the number and percentage of
records with successful matches to the HDD, AHA, and CEN files,
cumulatively. Thus the final linked file is composed of records that
represent between 86 and 89 percent of the deliveries to Florida mothers
between 1988 and 1991. 2 In addition, we were able to match between 79
and 85 percent of the VS cases to the HDD, AHA, and CEN files--the three
link files that were applicable to all of the records.
Table C12. Comparison of Original VS File to Matched Files, By Year
1988
1989
1990
1991
Number of deliveries on
VS filea
184,798
193,336
199,663
193,393
Match to HDD (final
159,077
169,548
177,153
173,033
linked file)
(86%)
(88%)
(89%)
(89%)
Match HDD, AHA
157,217
167,587
174,491
170,758
(85%)
(87%)
(87%)
(88%)
Match HDD, AHA, CEN
145,203
158,622
168,498
165,333
(79%)
(82%)
(84%)
(85%)
aTotal number of deliveries includes deliveries to residents that
took place outside of Florida.
The final linked file also includes the additional information
obtained for the subset of Medicaid eligible and low-income women by
matching to the ME and PH files. As a result of linking the VS file to
these five additional files, each of which provides useful data that was
otherwise unavailable, the final linked file represents a database rich
in information.
2 Because four to five percent of the records in the final linked
file have missing values for some of the key outcome measures or
demographic characteristics, an analysis file containing only those
records with complete data would be slightly smaller.
D-l
APPENDIX D. ESTIMATION FOR AGGREGATE ANALYSIS
This appendix provides more detail on our procedures for estimating
the flow of pregnant women through prenatal care and delivery, the
corresponding use of services, and the flow of payments for this care.
ANALYTIC FRAMEWORK
Our method is modeled after the National Health Expenditures
Accounts, a continuing series currently maintained by the Health Care
Financing Administration which estimates the flow of funds financing all
health care in the United States. We develop three matrices to
investigate the effects of the expansions. The first (see Table 2 in
the report) categorizes women according the primary payer for their care
and describes how this changed after the Medicaid expansion. The payer
categories in the matrix include private insurance, Medicaid, and "other
payer", which includes the uninsured and those whose care was paid for
by some other third party-payer such as Medicare, CHAMPUS, or state and
federal programs that make payments on behalf of a patient receiving
care. The hospital discharge data--our primary source for payer--does
not allow us to further classify these other payers.
The second matrix (see Tables 3 and 4 in the report) shows the
effects of the expansions on the quantity and type of health care
services received by pregnant women. The utilization matrix shows the
number of prenatal care visits and the number of pregnancy related
hospitalizations in each of the study years and the change in these
quantities over time. The utilization measures are categorized by both
payer and site of service. For ambulatory care we distinguish between
care provided in county health departments, that received in physician
offices, and care in other settings such as Community Health Centers,
hospital clinics, and hospital outpatient departments. Hospital
admissions are categorized by the type of hospital--public, voluntary,
or proprietary.
The third matrix (see Table 5 in the report) is a flow of funds
matrix. The columns of the matrix categorize the expenditures by type
of service (hospital inpatient, physician and other services) and the
D-2
rows by payer. The unit of measure in this matrix is dollars, and the
dollars that we measure are the direct payments made by the patient or
by a third-party on behalf of a particular patient.
ESTIMATING THE THREE MATRICES
Flow of Women
The first matrix shows the number of women delivering in each year
and the primary payer for their maternity care. The vital statistics
records for all births and fetal deaths registered to Florida residents
measure the total number of deliveries in each year. This serves as a
control total for all of our other estimates. These deliveries were
distributed among the three payer categories based on the distribution
of primary payer at delivery from the matched hospital discharge file
and vital statistics file for each year.
Utilization Matrix
Ambulatory Prenatal Visits
The total number of prenatal visits is based on information
recorded on the birth certificate. We estimated the average number of
prenatal visits for women in each of the three different payer groups
from the linked vital statistics and hospital discharge file (see Table
Dl) . These per case estimates, multiplied by the total number of
deliveries in each of the three payer groups, yielded our estimate of
the total number of prenatal care visits by payer.
Table Dl . Average Number of Prenatal Care
Visits Per Delivery by Payer
Period
Payer
Baseline
1991
Private Insurance
12.3
12.8
Medicaid
9.8
10.5
Other Payer
9.6
10.2
D-3
We distribute these visits by site of care--county health
department versus other sites--based on the public health system
encounter data . Our matrices measure care provided to women who
delivered in the year of study, irrespective of the year in which the
service was actually delivered. Therefore, we needed to link the county
health department visits to the year of delivery of the patient. We did
this in a two step manner. First, for county health department episodes
that we were able to link to the vital statistics and hospital discharge
data, we have a direct measure of the year of delivery. Using the
relationship between the last date seen in the county health department
and the year of delivery for these episodes, we then simulated the
delivery year for episodes that we were unable to match to vital
records. Then, we counted the total number of prenatal visits to county
health departments associated with deliveries in the year of study as
the health department column total in our matrix. This total was
allocated among the three payer groups based on the distribution of
payer for the county health department visits that were matched to the
vital statistics/hospital discharge file from which we can measure the
primary payer.
Hospital Admissions
The total number of hospital admissions for maternity care include
hospital admissions for delivery as well as inpatient prenatal hospital
admissions. The former admissions are determined from the vital
statistics data on place of delivery--i .e. not all deliveries are
associated with a hospital admission. This count of delivery admissions
is a control total. We categorize deliveries by both type of hospital--
public, voluntary, and proprietary- -and by payer. To do so, we first
calculate the distribution of deliveries across hospital types for all
vital statistic records that we are able to link to the AHA data. We
use this distribution to allocate the total delivery admissions to each
hospital type. We next estimate the distribution of payer by hospital
type from the matched vital statistics and hospital discharge file for
each hospital type to allocate the delivery admissions across payers.
We estimated the number of pregnancy-related admissions as the
number of admissions with ICD-9 codes related to prenatal or maternity
D-4
care that did not result in a delivery. Because the hospital discharge
data lacks identifiers, we are not able to track the non-delivery
hospital admissions of women who gave birth in the study year. We
therefore approximate these by looking at all prenatal admissions in a
year, regardless of whether the woman actually gave birth in the year.
If the number of births is constant over time, then this method should
provide an unbiased estimate of the number of prenatal admissions for
women delivering in the study year. We distribute these admissions
among hospital types and payers using the AHA data on hospital type and
the discharge record information on payer.
The Flow of Payments Matrix
The third matrix we present shows the flow of payments for maternal
care in the two years. It measures the direct payments for care by
patients and on account of patients by third-party payers. That is it
measures what was actually collected by the provider for the care of a
particular patient. It does not include contributions that are not tied
to particular patients, such as federal block grants to states for Title
V programs and general contributions by local governments to public
hospitals for charity care.
To measure the payment flows, we start with an estimate of the
total charges for inpatient hospital services and for physician and
related services (such as laboratory tests and x-rays) categorized by
the delivery payer. The total charge estimates for each payer are based
on estimates of the average charge per quantity of service multiplied by
our estimates of the units of service. We convert the result matrix of
charges to payments using estimates of the ratio of payments to charges
for different payers. The steps are detailed below.
Estimating Physician and Related Service Charges
Our calculation of physician and related charges is based on an
estimate of the total charge for these maternal health services per
prenatal visit. This per visit charge measure is multiplied by our
estimate of the total number of prenatal care visits for each payer
group and site of care to provide an aggregate charge for each payer and
site. When we calculate payment, we wish to distinguish payments made
D-5
by third-parties from those made directly by patients. Therefore, for
this purpose, we produce an estimate of the number of visits made by
women in the "other" payer category that were made by uninsured women
and those made by women with coverage from some other third-party. This
allocation was based on the distribution of visits between these groups
as reported in the 1988 National Maternal and Infant Health Survey by
sample person in Florida. This allocation then permitted us to estimate
the charges for uninsured patients and for patients with other third-
party payers. However, because we have only one survey, the same
allocation factor was used for both the baseline period and 1991.
Because the Medicaid expansions would be expected to alter the mix, and
in particular to decrease the share of uninsured patients, our method
will tend to understate the increase in payments resulting from the
expansions .
The per visit charge number is based on data from the 1991 Medicaid
claims files and on data from two large employers in Florida. The
Medicaid estimate of charges per visit is the ratio of charges on all
Medicaid claims for physician and related services (laboratory, x-ray,
outpatient hospital) to the total number of prenatal care visits as
recorded on birth certificates for women covered by Medicaid who
delivered in the last three months of 1991. The estimate is restricted
to those delivering at the end of 1991 to ensure that we have claims for
the entire prenatal period in the 1991 claims data file available to
this study. We include only charges for services provided during the
prenatal and delivery period. We also restricted the estimate to
patients who receive their care in the private sector (i.e. for whom we
do not have a county public health department encounter record) . We do
this because we are able to estimate payments for visits to the public
health system directly and so do not need to make a estimate of charges
for these visits.
The private claims database covers three years for two large
employers. Our estimate of the charge per prenatal visit from the
private payer data is the ratio of charges (in 1991 dollars) for all
claims for maternity care from physicians or related providers divided
by our estimate of the average number of prenatal care visits in 1991 by
women covered by private payers (see Table Dl) . Our estimate of the
D-6
average charge per visit from the Medicaid data was $32 6; our estimate
from the private claims data was $353. Because these estimates were so
similar, we used the average value of $340 per visit charge for all
payers .
Our procedure applies the same estimate of charges relative to
prenatal care visits for all payers. Estimates of the per delivery
charges, however, vary by payer type because of differences in the
average number of prenatal care visits among the payer groups. Our
procedure assumes that variation in the number of prenatal care visits
reflects differences in patient needs and service mix.
Estimating Hospital Charges
For hospital admissions, estimates of the average charge for women
in different payer statuses and in different hospital types are from the
hospital discharge data. For deliveries, these charges are from the
matched vital statistics and hospital discharge file. For prenatal
admissions, the charges are for all of the hospital discharge records
that were identified as prenatal admissions in the year. Appendix Table
D2 presents the average charges that we used in calculating the
aggregate charge numbers; the baseline period charges shown here were
inflated to 1991 dollars for the payment matrix using the hospital
component of the consumer price index.
Table D2 . Average Hospital Charges Per Delivery
By Payer And Type Of Hospital
(in current dollars)
Baseline
1991
Payer
Public
Voluntary
Proprietary
Public
Voluntary
Proprietary
Private
Insurance
2760
2772
2844
3772
4012
4263
Medicaid
2968
2843
2870
3598
3677
3839
Other Payer
2935
2634
2830
3395
3691
3653
To allocate the charges for women in the "other payer" category
between uninsured women and those with a third-party payer, we used the
distribution of deliveries among these groups based on the NMIHS data.
Again, because we use a constant allocation factor over time, we will
D-7
understate the increase in payments to hospitals stemming from the
expansions .
Calculating Payments From Charges.
Using charges to estimate payments would result in an overestimate
of the actual payment flows. Increasingly over the last decade, third-
party payers have either refused to pay full charges, as has Medicaid,
or negotiated discounts, as have Blue Cross and Blue Shield and most
managed care organizations. Moreover, uninsured patients are frequently
unable to pay much, if any, of a hospital's normal charge for delivery,
but generally they are not denied care.
Therefore, we adjusted the estimated aggregate charges for each
payer using payment-to-charge ratios for each payer category. The
payment -to -charge ratio is always less than one reflecting deductions
from gross charges for contractual adjustments and uncompensated care.
Payment -to -charge ratios for hospital care for each payer were
provided by the Agency for Health Care Administration in Florida. The
Medicaid payment-to-charge ratio for physician and related services was
derived from the Medicaid claims files for the sample of women described
above in calculating charge per visit ratios . The ratio for private
payers was based on the maternity claims data from two large Florida
employers. We also used these claims data to estimate the share of the
total payment that was made by the insurer and the share paid directly
by the patient in copayments . Separate estimates were made of the
copayment share for hospital services and for physician and related
services .
Absent other data sources to provide payment-to-charge ratios for
physician services in Florida for the uninsured and those covered by
third-party payers other than private insurance and Medicaid, we used
the hospital ratios. Payments to county health departments were
measured from state budget and revenue statistics provided by the state
health department .
E-l
APPENDIX E. EXPLANATORY VARIABLES AND REFERENCE POPULATION
This appendix tables shows the definitions of the demographic
characteristics included in our regression model and presents the values
of these characteristics for our reference population- -women enrolled in
the Medicaid expansion group in 1991.
Table El. Characteristics of Reference Population
Value for
Reference
Characteristic Population
Age of mother
Under age 18 * 0.08
18-19 0.13
20-24 0.36
25-29 0.24
30-34 0.13
35 and older 0.06
Mother's education
Less than high school * 0.37
High school graduate 0.44
Some college 0.16
College degree 0.03
Race/ethnicity (1991 regressions)
White, non-hispanic * 0.66
Black, non-hispanic 0.15
Other, non-hispanic 0.01
Mexican 0.05
Puero Rican 0.03
Cuban 0.02
Central/South American 0.04
Haitian 0.02
Other hispanic 0.02
Race (1988 regressions)
White * 0.81
Black 0.18
Other 0.01
Marital Status
Married 0 . 64
Not married 0.36
Indicator for no previous live
births 0.44
Indicator for singleton
birth 0.99
Indicator for any medical risk
factors (1991 regression) 0 .21
* Denotes ommitted class in regression model
F-l
Appendix F. REGRESSION RESULTS
This appendix contains the parameter estimates and standard errors
(S.E.) for the regressions described in the text. The explanatory
variables indicate whether or not the mother has the characteristics.
With the exception of the insurance class variables, the indicator was
coded as one minus the proportion of the 1991 Medicaid expansion group
with the characteristic if the mother had the characteristic and 0 minus
the proportion of the 1991 Medicaid expansion group otherwise. (The
proportion of the 1991 Medicaid expansion population with each
characteristic, and the omitted subgroup for each characteristic, are
shown in Appendix E) . The insurance variables are coded 0 or 1 to
indicate the mother's status, and Medicaid expansion is the omitted group.
This coding was adopted so that the intercept reflects the value for a
woman in the Medicaid expansion population with characteristics equal to
the average value of characteristics for women in the Medicaid expansion
population in 1991.
F-2
Regression Parameter Estimates and Standard Errors for Baseline Period (1)
July 88 - June 89
Dependent
of users:
of users:
Variable:
no prenatal
care
initiate
in 1st
# of visits
or 2nd trimester
Logit
Logit
OLS
Explanatory-
Parameter
Parameter
Parameter
variable
Estimate
-4.2248
0
S.E.
.0719
Esitmate
2.5390
0
S.E.
.0369
Estimate
10.9651
0
S.E.
Intercept
.0424
mother 18-19 yrs
0.0631
0
.0686
-0.0362
0
.0492
-0.1190
0
.0612
mother 20-24 yrs
0.0190
0
.0644
0.2408
0
.0479
0.1216
0
.0562
mother 25-29 yrs
-0.1253
0
.0703
0.5663
0
.0538
0.3768
0
.0588
mother 3 0-34 yrs
-0.1697
0
.0784
0.7218
0
.0616
0.4689
0
.0621
mother 3 5 and over
-0.4045
0
.1011
0.7730
0
.0778
0.5529
0
.0693
mother is married
-0.9785
0
.0405
0.6272
0
.0308
0.9200
0
.0311
12 yrs educ
-0.5114
0
.0375
0.2924
0
.0296
0.6426
0
.0317
13-15 yrs educ
-1.0440
0
.0664
0.5135
0
.0450
1.0070
0
.0377
16+ years educ
-1.6962
0
.1373
1.0160
0
.0787
1.2695
0
.0437
black
0.2480
0
.0380
0.0566
0
.0302
-0.7411
0
.0302
other race
0.2405
0
.1583
-0.7321
0,
.0958
-0.8517
0.
.0939
singleton birth
0.4479
0.
.1774
-0.7846
0
.1571
-1.7584
0,
,1006
no prev live births
-0.9786
0
,0426
0.6446
0,
,0299
0.7243
0.
0239
medicaid: afdc elig
0.2774
0
,0794
0.1485
0,
,0456
-0.1046
0.
0541
medicaid: medic needy
0.5783
0,
,1751
-0.0272
0,
,1104
0.0407
0.
1249
medicaid: other
0.5145
0,
,1717
0.1799
0.
,1362
0.2272
0.
1613
private, >=30% poor
-0.6411
0.
,1350
1.0960
0.
0774
0.7333
0.
0590
private, < 3 0% poor
-0.9802
0.
,1045
1.4384
0.
,0560
0.6934
0.
0485
private, no zip
-0.4117
0.
1977
1.0345
0.
1159
0.6429
0.
0762
oth ins, >=3 0% poor
0.9757
0.
0794
0.0438
0,
0486
-1.1984
0.
0565
oth ins, < 3 0% poor
0.9980
0.
0765
0.1829
0.
0443
-0.4655
0.
0493
oth ins, no zip
1.3828
0.
0994
-0.1730
0.
0730
-0.6440
0.
0863
F-3
Regression Parameter Estimates and Standard Errors for Baseline Period (2)
July 88 - June 89
Dependent
variable :
Explanatory
variable
Intercept
mother 18-19 yrs
mother 20-24 yrs
mother 25-29 yrs
mother 3 0-34 yrs
mother 3 5 and over
mother is married
12 yrs educ
13-15 yrs educ
16+ years educ
black
other race
singleton birth
no prev live births
medicaid: afdc elig
medicaid: medic needy
other
>=30% poor
< 3 0% poor
no zip
>=30% poor
< 3 0% poor
no zip
medicaid
private,
private,
private,
oth ins,
oth ins,
oth ins.
inadequate care: inadequate care:
kotelchuck kessner
Logit
Parameter
Estimate S.E.
-0.6102
0.0055
-0.1511
-0.4047
-0.4760
-0.4973
-0.5876
-0.3707
-0.5992
-0.9227
0.2790
0.4629
0.1815
-0.4553
-0.0568
0.2354
-0.0727
-0.7287
-0.8405
-0.7633
0.2688
0.0146
0.2210
0.0212
0.0298
0.0277
0.0297
0.0322
0.0377
0.0158
0.0159
0.0208
0.0284
0.0158
0.0530
0.0598
0.0144
0.0267
0.0616
0.0779
0.0335
0.0259
0.0481
0.0278
0.0248
0.0423
Logit
Parameter
Estimate
-2.3266
0.0487
-0.1487
-0.4235
-0.5348
-0.6667
-0.7868
-0.3993
-0.7154
-1.2423
0.0987
0.6349
0.5075
-0.7756
-0.0777
0.1472
-0.0369
-0.9964
-1.3097
-0.8831
0.2591
0.1643
0.5524
S.E.
0.0330
0.0411
0.0393
0.0436
0.0492
0.0626
0.0248
0.0237
0.0370
0.0668
0.0241
0.0838
0.1107
0.0248
0.0396
0.0952
0.1092
0.0663
0.0489
0.0992
0.0410
0.0380
0.0587
F-4
Regression Parameter Estimates and Standard Errors for Baseline Period (3)
July 88 - June 89
Dependent
variable :
Explanatory-
variable
Intercept
mother 18-19 yrs
mother 20-24 yrs
mother 2 5-29 yrs
mother 3 0-34 yrs
mother 3 5 and over
mother is married
12 yrs educ
13-15 yrs educ
16+ years educ
black
other race
singleton birth
no prev live births
medicaid: afdc elig
medicaid: medic needy
other
>=30% poor
< 3 0% poor
no zip
>=30% poor
< 30% poor
no zip
medicaid:
private,
private,
private,
oth ins,
oth ins,
oth ins,
very 1
ow
survive
low birthweight
birthweii
jht
first
year
Logit
Logit
OLS
Parameter
Parameter
Parameter
Estimate
S.E.
Estimate
S.E.
Estimate
S.E.
-2.6303
0
.0384
-4.7118
0
.0951
4.9752
0.1139
-0.0118
0
.0494
-0.1117
0
.1150
-0.0294
0.1401
0.0942
0
.0451
-0.0335
0
.1039
-0.0971
0.1280
0.1914
0
.0481
0.1071
0
.1102
0.0320
0.1384
0.3719
0
.0516
0.2603
0
.1183
-0.0453
0.1491
0.4920
0
.0589
0.3610
0
.1350
0.0113
0.1754
-0.4119
0
.0273
-0.4397
0
.0643
0.4469
0.0781
-0.2181
0
.0275
-0.1248
0
.0658
0.3294
0.0782
-0.3148
0
.0352
-0.1173
0
.0821
0.2542
0.0988
-0.5205
0
.0449
-0.2793
0
.1052
0.4517
0.1309
0.6318
0
.0256
0.9220
0
.0593
-0.3338
0.0741
0.3131
0
.0894
0.5637
0
.1978
-0.6200
0.2185
-2.9266
0
,0505
-2.4450
0
.0865
1.7410
0.1288
0.2875
0
.0235
0.3156
0.
,0557
0.0675
0.0685
-0.0342
0.
.0462
-0.1351
0.
.1116
0.0314
0.1363
0.2739
0.
.1032
0.8264
0,
,2018
-0.6284
0.2632
0.4386
0.
.1088
0.3075
0.
2588
0.2891
0.4269
-0.2064
0.
0562
0.00985
0.
1301
-0.2277
0.1563
-0.2793
0.
.0457
-0.1567
0.
1115
0.1530
0.1353
-0.0862
0.
0765
0.2761
0.
1705
-0.3168
0.2054
0.0750
0.
0484
0.1105
0.
1157
-0.2184
0.1402
-0.0742
0.
0451
-0.00439
0.
1102
-0.0627
0.1319
0.1913
0.
0740
0.5703
0.
1613
-0.2111
0.2156
F-5
Regression Parameter Estimates and Standard Errors for Post Period (1)
1991
Dependent
of users:
of users:
Variable:
no prenatal
care
initiate
in 1st
# of visits
or 2nd trimester
Log it
Logit
OLS
Explanatory-
Parameter
Parameter
Parameter
variable
Estimate
-4.3578
0
S.E.
.0495
Estimate
3.0054
0
S.E.
.0280
Estimate
11.2394
0
S.E.
Intercept
.0246
mother 18-19
-0.0429
0
.0777
0.1057
0
.0560
0.0076
0
.0586
mother 20-24
-0.1548
0
.0731
0.2470
0
.0534
0.2284
0
.0538
mother 2 5-29
-0.2784
0
.0810
0.4812
0
.0602
0.4030
0
.0566
mother 3 0-34
-0.2435
0
.0892
0.7044
0
.0693
0.4447
0
.0596
mother 35 and over
-0.2808
0
.1065
0.7292
0
.0840
0.5547
0
.0651
mother is married
-0.8578
0
.0483
0.4500
0
.0346
0.7420
0
.0288
12 yrs educ
-0.4335
0
.0442
0.2892
0
.0334
0.5002
0
.0304
13-15 yrs educ
-0.7725
0
.0709
0.4842
0
.0507
0.8707
0
.0362
16+ years educ
-1.5275
0
.1389
0.8409
0
.0859
0.9979
0
.0418
black, nonhisp
0.2102
0
.0470
-0.1544
0
.0361
-0.9446
0
.0301
other, nonhisp
0.1033
0
.1840
-0.7165
0
.1101
-0.7516
0
.0875
mexican ethnicity
0.2331
0
.0888
-0.7735
0
.0587
-1.5247
0,
.0678
puerto rican
0.1795
0
.1160
0.0248
0,
.0926
-0.7244
0,
,0682
cuban
-0.8246
0
.1760
0.7906
0
.1284
-0.5809
0.
,0530
central/south amer
-0.0200
0.
,0882
-0.3607
0,
.0628
-1.5370
0.
0504
other hisp ethnicity
-0.0706
0.
.1676
-0.6287
0,
.0984
-0.5811
0.
0913
haitian
-0.6887
0.
.1564
0.0227
0,
,1009
-1.6555
0.
0799
any med hist factors
0.5351
0,
,0409
-0.1057
0,
,0338
0.6980
0.
0255
singleton birth
0.0996
0.
,1662
-0.4930
0,
,1554
-1.5174
0.
0959
no prev live births
-0.9257
0.
0490
0.6016
0.
0338
0.5870
0.
0227
medicaid: afdc elig
0.2164
0.
0600
0.0242
0.
0392
-0.0384
0.
0368
medicaid: medic needy
0.6644
0.
1898
0.3379
0.
1681
-0.1518
0.
1181
medicaid: other
0.3947
0.
1668
-0.1558
0.
1040
-0.9421
0.
1130
private, >=30% poor
-0.6109
0.
1396
1.1764
0.
0982
0.9701
0.
0487
private, < 3 0% poor
-1.0430
0.
0958
1.5520
0.
0620
0.9919
0.
0329
private, no zip
-0.4194
0.
2836
0.9606
0.
1966
0.5768
0.
0947
oth ins, >=30% poor
1.0012
0.
0718
-0.1387
0.
0542
-0.6507
0.
0545
oth ins, < 30% poor
0.9718
0.
0640
0.1474
0.
0486
-0.1600
0.
0395
oth ins, no zip
1.2251
0.
1569
-0.0470
0.
1493
-0.0351
0.
1404
F-6
Regression Parameter Estimates and Standard Errors for Post Period (2)
1991
Dependent
variable :
inadequate care:
kotelchuck
inadequate care:
kessner
Explanatory
variable
Intercept
mother 18-19
mother 20-24
mother 25-29
mother 3 0-34
mother 3 5 and over
mother is married
12 yrs educ
13-15 yrs educ
16+ years educ
black, nonhisp
other, nonhisp
mexican ethnicity
puerto rican
cuban
central /south amer
other hisp ethnicity
haitian
any med hist factors
singleton birth
no prev live births
medicaid: afdc elig
medicaid: medic needy
other
>=30% poor
< 3 0% poor
no zip
>=30% poor
< 30% poor
no zip
medicaid:
private,
private,
private,
oth ins,
oth ins,
oth ins,
Logit
Parameter
Estimate
-1.2364
-0.1095
-0.3071
-0.5715
-0.6663
-0.6721
-0.5655
-0.2819
-0.5282
-0.9711
0.2944
0.5403
0.6539
0.1010
-0.4066
0.2999
0.4281
0.1857
0.0722
0.2491
-0.5742
-0.0226
-0.0688
0.0971
-1.0981
-1.4435
-0.7778
0.1242
-0.1562
0.0610
S.E.
0.0146
0.0313
0.0297
0.0328
0.0363
0.0426
0.0179
0.0177
0.0251
0.0402
0.0187
0.0636
0.0365
0.0461
0.0509
0.0338
0.0589
0.0494
0.0178
0.0707
0.0174
0.0209
0.0754
0.0615
0.0428
0.0276
0.0831
0.0301
0.0249
0.0813
Logit
Parameter
Estimate
-2.6858
-0.0852
-0.2150
-0.4084
-0.5438
-0.5390
-0.6149
-0.3555
-0.6053
-1.0496
0.2363
0.5214
0.6068
0.0372
-0.7728
0.2232
0.4314
-0.2135
0.2920
0.1913
-0.7159
0.0164
-0.0228
0.2095
-0.9839
-1.4373
-0.8428
0.4182
0.2105
0.4488
S.E.
0.0240
0.0455
0.0431
0.0482
0.0544
0.0650
0.0279
0.0265
0.0406
0.0703
0.0283
0.0956
0.0499
0.0727
0.1004
0.0515
0.0854
0.0831
0.0258
0.1059
0.0275
0.0322
0.1245
0.0891
0.0762
0.0508
0.1588
0.0427
0.0376
0.1097
F-7
Regression Parameter Estimates and Standard Errors for Post Period (3)
1991
Dependent
variable :
Explanatory
variable
low birthweight
Logit
Parameter
Estimate S.E.
Intercept
mother 18-19
mother 20-24
mother 25-29
mother 3 0-34
mother 3 5 and over
mother is married
12 yrs educ
13-15 yrs educ
16+ years educ
black, nonhisp
other, nonhisp
mexican ethnicity
puerto rican
cuban
central /south amer
other hisp ethnicity
haitian
any med hist factors
singleton birth
no prev live births
medicaid: afdc elig
medicaid: medic needy
medicaid: other
private, >=3 0% poor
< 3 0% poor
no zip
>=30% poor
< 3 0% poor
no zip
private,
private,
oth ins,
oth ins,
oth ins,
-2.7418
0.0647
0.1155
0.1969
0.3550
0.5372
-0.2966
-0.1788
-0.3649
-0.4853
0.6615
0.1921
-0.1425
0.3722
-0.00827
0.00849
0.0528
0.2784
0.7065
-2.9368
0.2760
0.0139
0.4396
0.0195
-0.1065
-0.2345
0.0149
0.1273
-0.1307
0.5591
0.0247
0.0505
0.0468
0.0502
0.0534
0.0587
0.0275
0.0280
0.0360
0.0445
0.0271
0.0943
0.0737
0.0650
0.0634
0.0560
0.0980
0.0734
0.0222
0.0500
0.0235
0.0338
0.1003
0.1064
0.0498
0.0347
0.0974
0.0487
0.0411
0.1109
very low
birthweight
Logit
Parameter
Estimate S.E.
-4.6821
-0.1244
-0.1587
-0.0346
0.0500
0.4116
-0.3515
-0.0346
-0.1691
-0.1652
0.8089
0.4008
-0.2756
0.5409
0.3148
0.0132
0.1273
0.6374
0.9844
-2.2392
0.2807
-0.1211
0.8892
-0.4392
-0.0572
-0.2999
0.3314
-0.0130
-0.1451
0.9049
0.0582
0.1124
0.1041
0.1114
0.1194
0.1276
0.0638
0.0657
0.0838
0.1019
0.0618
0.2068
0.2017
0.1461
0.1401
0.1424
0.2350
0.1557
0.0488
0.0864
0.0546
0.0772
0.1872
0.2906
0.1104
0.0815
0.1935
0.1139
0.0956
0.2083
survive
first year
OLS
Parameter
Estimate S.E.
5.2077
0.0391
0.2708
0.4145
0.2249
0.1021
0.3088
0.1624
0.4463
0.5763
-0.4870
-0.3094
0.5034
-0.1352
0.1595
0.3561
0.2103
-0.0951
-0.5641
1.4306
0.0945
0.2595
-0.5483
-0.0652
-0.0982
0.2758
0.3636
-0.0261
0.0349
•0.5026
0.0784
0.1494
0.1430
0.1579
0.1684
0.1881
0.0913
0.0908
0.1235
0.1581
0.0894
0.2948
0.2698
0.2307
0.2302
0.2154
0.3589
0.2504
0.0743
0.1622
0.0804
0.1098
0.3010
0.3509
0.1551
0.1161
0.3883
0.1593
0.1304
0.3455
F-8
Regression Parameter Estimates and Standard Errors
for Medicaid Women with Prenatal Care Visits, 1991 (1)
Dependent
Variable:
initiate
in 1st
# of visits
or 2nd trimester
Log it
OLS
Explanatory
Parameter
Parameter
variable
Esitmate
3.2568
0
S.E.
.0439
Estimate
11.7617
0
S.E.
Intercept
.0369
mother 18-19 yrs
0.1137
0
.0679
0.0313
0
.0761
mother 20-24 yrs
0.2592
o.
.0659
0.2631
0
.0733
mother 2 5-29 yrs
0.4342
0
.0750
0.4537
0
.0821
mother 3 0-34 yrs
0.5385
0
.0870
0.4403
0
.0925
mother 3 5 and over
0.5502
0
.1108
0.5646
0
.1140
mother is married
0.2832
0
.0430
0.5589
0
.0441
12 yrs educ
0.1664
0
.0396
0.2363
0
.0429
13-15 yrs educ
0.2602
0,
.0642
0.6319
0
.0624
16+ yrs educ
0.2845
0,
.1535
0.2841
0
.1311
black, nonhisp
-0.1094
0.
.0418
-1.2312
0
.0451
other, nonhisp
-0.7367
0
.1582
-1.3085
0,
.2060
mexican ethnicity
-0.5607
0
.0864
-1.6128
0
.1105
puerto rican
0.1002
0
.1079
-1.1148
0
.1049
cuban
0.5804
0
.1634
-1.7950
0
.1196
central /south amer
-0.0766
0.
,1157
-2.3343
0
.1105
other hisp ethnicity
-0.3133
0.
,1507
-0.6277
0
.1753
haitian
0.1847
0.
,1542
-2.1772
0
.1427
any med hist factors
-0.0705
0.
.0408
0.5472
0
.0442
singleton birth
-0.5009
0.
,1916
-1.0481
0
.1666
no previous live births
0.7139
0.
.0439
0.8257
0
.0434
medicaid: afdc elig
-0.1296
0.
,0590
-0.0957
0
.0552
>=50% visits @ pubhlth
-0.5411
0.
.0559
-1.0924
0
.0535
interact afdc & pubhlth
0.0623
0,
,0711
-0.0492
0
.0733
F-9
Regression Parameter Estimates and Standard Errors
for Medicaid Women with Prenatal Care Visits, 1991 (2;
Dependent
inadequate
care:
inadequate
care:
variable:
kotelchuck
kessner
Logit
Logit
Explanatory
Parameter
Parameter
variable
Estimate
-1.4658
0
S.E.
.0217
Estimate
-3.1402
0
S.E.
Intercept
.0416
mother 18-19 yrs
-0.1619
0
.0380
-0.1362
0
.0644
mother 20-24 yrs
-0.3492
0
.0370
-0.2758
0
.0624
mother 25-29 yrs
-0.5434
0
.0422
-0.4294
0
.0710
mother 30-34 yrs
-0.5599
0
.0480
-0.5277
0
.0822
mother 3 5 and over
-0.5709
0
.0602
-0.4898
0
.1034
mother is married
-0.2913
0,
.0236
-0.3069
0
.0411
12 yrs educ
-0.1267
0
.0222
-0.1752
0
.0376
13-15 yrs educ
-0.2261
0,
.0341
-0.3048
0,
.0616
16+ yrs educ
-0.2796
0,
.0778
-0.2375
0,
.1418
black, nonhisp
0.2302
0.
.0233
0.1889
0.
.0396
other, nonhisp
0.6386
0
.1013
0.6949
0,
.1553
mexican ethnicity
0.5293
0.
.0539
0.5167
0.
.0847
puerto rican
0.0105
0.
.0568
-0.1020
0.
,1040
cuban
-0.1551
0.
.0697
-0.5240
0.
,1529
central/south amer
0.1653
0.
.0604
0.0923
0.
,1105
other hisp ethnicity
0.2473
0,
,0900
0.2895
0.
,1468
haitian
0.1101
0,
.0762
-0.1259
0.
1435
any med hist factors
-0.0500
0.
,0233
0.1123
0.
0384
singleton birth
0.2125
0.
0914
0.2438
0.
1614
no previous live births
-0.5395
0.
0235
-0.7150
0.
0418
medicaid: afdc elig
0.1576
0.
0303
0.1097
0.
0558
>=50% visits @ pubhlth
0.3937
0.
0296
0.5064
0.
0534
interact afdc & pubhlth
-0.0641
0.
0387
-0.0345
0.
0677
F-10
Regression Parameter Estimates and Standard Errors
for Medicaid Women with Prenatal Care Visits, 1991 (3)
Dependent
variable:
Explanatory
variable
Intercept
mother 18-19 yrs
mother 20-24 yrs
mother 25-29 yrs
mother 30-34 yrs
mother 35 and over
mother is married
12 yrs educ
13-15 yrs educ
16+ yrs educ
black, nonhisp
other, nonhisp
mexican ethnicity
puerto rican
cuban
central/south amer
other hisp ethnicity
haitian
any med hist factors
singleton birth
no previous live births
medicaid: afdc elig
>=50% visits 8 pubhlth
interact afdc & pubhlth
very
low
survive
low birthweight
birthweight
first
year
Logit
Logit
OLS
Parameter
Parameter
Paramet
er
Estimate
S.E.
Estimate
S.E.
Estimate S.E
-2.5809
0
.0326
-4.5072
0.0765
5.1042
0.1069
0.0111
0
.0624
-0.0952
0.1466
0.2981
0.1896
0.0183
0
.0604
-0.1600
0.1419
0.4915
0.1862
0.1544
0
.0679
-0.0245
0.1592
0.7061
0.2216
0.3968
0
.0750
0.1707
0.1753
0.4881
0.2462
0.4430
0
.0912
0.4414
0.1993
0.5759
0.3179
-0.2291
0
.0394
-0.3229
0.0952
0.2294
0.1340
-0.1304
0
.0364
-0.0177
0.0882
0.1198
0.1239
-0.3292
0
.0556
-0.1440
0.1306
0.1631
0.1902
-0.2692
0
.1158
0.2606
0.2282
0.3662 •
0.4666
0.5492
0
.0376
0.5531
0.0899
-0.3845
0.1273
-0.0374
0
.1991
0.0614
0.4586
-0.2364
0.5860
-0.2590
0.
.1158
-0.2566
0.2990
0.5028
0.4197
0.2056
0.
.0923
0.1579
0.2300
0.0812
0.3450
-0.1722
0.
,1199
0.0997
0.2693
-0.0056
0.3892
-0.2858
0.
,1181
-0.3837
0.3111
1.0099
0.5852
0.1222
0.
1593
0.5118
0.3269
1.2398
1.0044
0.2566
0.
1170
0.5761
0.2421
-0.5188
0.3528
0.6182
0.
0335
0.9774
0.0754
-0.5643
0.1120
-2.8413
0.
0806
-2.2430
0.1293
0.9708
0.2878
0.1945
0.
0374
0.2291
0.0889
0.2442
0.1281
-0.0069
0.
0462
-0.1357
0.1066
0.3544
0.1550
-0.3659
0.
0505
-0.4666
0.1201
0.3322
0.1616
0.1025
0.
0650
0.0644
0.1569
-0.2203
0.2148
CHS LIBRARY
3 flDT5 DOOD^^ b