At the Intersection of Health, Health Care and Policy Cite this article as: Marian Jarlenski, Sara N. Bleich, Wendy L. Bennett, Elizabeth A. Stuart and Colleen L. Barry Medicaid Enrollment Policy Increased Smoking Cessation Among Pregnant Women But Had No Impact On Birth Outcomes Health Affairs, 33, no.6 (2014):997-1005 doi: 10.1377/hlthaff.2013.1167
The online version of this article, along with updated information and services, is available at: http://content.healthaffairs.org/content/33/6/997.full.html
For Reprints, Links & Permissions: http://healthaffairs.org/1340_reprints.php E-mail Alerts : http://content.healthaffairs.org/subscriptions/etoc.dtl To Subscribe: http://content.healthaffairs.org/subscriptions/online.shtml
Health Affairs is published monthly by Project HOPE at 7500 Old Georgetown Road, Suite 600, Bethesda, MD 20814-6133. Copyright © 2014 by Project HOPE - The People-to-People Health Foundation. As provided by United States copyright law (Title 17, U.S. Code), no part of Health Affairs may be reproduced, displayed, or transmitted in any form or by any means, electronic or mechanical, including photocopying or by information storage or retrieval systems, without prior written permission from the Publisher. All rights reserved.
Not for commercial use or unauthorized distribution Downloaded from content.healthaffairs.org by Health Affairs on May 31, 2015 at UNIV OF GEORGIA LIBRARIES
Medicaid By Marian Jarlenski, Sara N. Bleich, Wendy L. Bennett, Elizabeth A. Stuart, and Colleen L. Barry 10.1377/hlthaff.2013.1167 HEALTH AFFAIRS 33, NO. 6 (2014): 997–1005 ©2014 Project HOPE— The People-to-People Health Foundation, Inc.
doi:
Medicaid Enrollment Policy Increased Smoking Cessation Among Pregnant Women But Had No Impact On Birth Outcomes
Marian Jarlenski (jarlenski@ gmail.com) recently completed doctoral studies in health policy and management at the Johns Hopkins Bloomberg School of Public Health, in Baltimore, Maryland. In fall 2014 she will become an assistant professor of health policy and management at the Graduate School of Public Health, University of Pittsburgh, in Pennsylvania.
Cigarette smoking during pregnancy is an important cause of poor maternal and infant health outcomes in the population eligible for Medicaid. These outcomes may be avoided or attenuated by timely, highquality prenatal care. Using data from the Centers for Disease Control and Prevention’s Pregnancy Risk Assessment Monitoring System for the period 2004–10, we examined the effects of two optional state Medicaid enrollment policies on smoking cessation, preterm birth, and having an infant who was small for gestational age. We used a natural experiment to compare outcomes before and after nineteen states adopted either of the two policies. The first policy, presumptive eligibility, permits women to receive prenatal care while their Medicaid application is pending. Its adoption led to a 7.7-percentage-point increase in smoking cessation but did not reduce adverse birth outcomes. The second policy, the unbornchild option, permits states to provide coverage to pregnant women who cannot document their citizenship or residency. Its adoption was not significantly associated with any of the three outcomes. The presumptiveeligibility enrollment policy will continue to be an important tool for promoting timely prenatal care and smoking cessation. ABSTRACT
C
igarette smoking in pregnancy accounts for a substantial portion of poor maternal and infant health outcomes and infant deaths.1–3 The prevalence of prenatal smoking in the United States has declined in recent decades,4 but it is nearly twice as high among lowincome women enrolled in Medicaid than it is in the US population as a whole.5 Since the late 1990s, many state Medicaid programs have begun providing more generous coverage of smoking cessation services for pregnant women.6 However, the process of enrolling in Medicaid may be a barrier to obtaining these services.7 The Medicaid application process is complex—requiring documentation that verifies the applicant’s income, residency, citizenship, and pregnancy—and a determination of eligibil-
Sara N. Bleich is an associate professor of health policy and management at the Johns Hopkins Bloomberg School of Public Health. Wendy L. Bennett is an assistant professor of medicine in the Division of General Internal Medicine, Johns Hopkins University School of Medicine, in Baltimore.
ity may take weeks.8 States have two policy options they can use to reduce barriers to Medicaid enrollment during pregnancy. Under the first, presumptive eligibility, low-income pregnant women are presumed to be eligible for Medicaid when they arrive for care at participating organizations. Thus, they can receive care while their Medicaid applications are still pending.9 A second policy, known as the unborn-child option, allows states to consider a fetus to be a “targeted low-income child” and to provide coverage of prenatal care and delivery to low-income pregnant women even if they cannot provide the documentation of citizenship or residency that is required for eligibility in Medicaid’s pregnancy category.10 These policies can lead to a greater probability of Medicaid enrollment and earlier initiation of June 2014
33:6
Downloaded from content.healthaffairs.org by Health Affairs on May 31, 2015 at UNIV OF GEORGIA LIBRARIES
Elizabeth A. Stuart is an associate professor of mental health and biostatistics at the Johns Hopkins Bloomberg School of Public Health. Colleen L. Barry is an associate professor of health policy and management at the Johns Hopkins Bloomberg School of Public Health.
H ea lt h A f fai r s
9 97
Medicaid prenatal care,11,12 which enables women to access smoking cessation services earlier in pregnancy than would otherwise be the case. Smoking cessation early in pregnancy has been shown to reduce adverse birth outcomes.13–15 No published research has examined the effects of these two policies on prenatal smoking cessation or smoking-related adverse birth outcomes. In the context of a new requirement under the Affordable Care Act (ACA) that all state Medicaid programs cover counseling and pharmacotherapies for smoking cessation for pregnant women,16 it is important to understand how these optional state Medicaid enrollment policies affect the use of smoking cessation services and birth outcomes. We addressed this gap in the literature by examining the effects of the policies on prenatal smoking cessation, preterm birth, and having an infant who was small for gestational age. We hypothesized that a state’s adoption of one of the two optional enrollment policies (presumptive eligibility and the unborn child option) would increase the probability that its female residents would stop smoking during pregnancy and decrease the probability of their having a preterm birth and an infant who was small for gestational age. We also hypothesized that the effects of the two optional enrollment policies would be greater in states with more-generous coverage of services for smoking cessation during pregnancy than in states with less-generous coverage.
Study Data And Methods Data Sources The Pregnancy Risk Assessment Monitoring System (PRAMS) of the Centers for Disease Control and Prevention (CDC) is a staterepresentative survey about women’s health, behavior, insurance status, and health care before, during, and shortly after pregnancy.17 States mail women a questionnaire two to four months after delivery, and the women who do not respond to the mailed questionnaire are contacted by telephone. Respondents’ answers to survey items are linked to birth certificate data. PRAMS research data are available for states that achieved a response rate of at least 70 percent before 2007 or a response rate of at least 65 percent from 2007 on. Between 2004 and 2010, nineteen of thirty-five participating states had sufficient response rates in all years and are thus included in our study. Although our study sample is not nationally representative, it is representative of women residing in these nineteen states in each of these years. To assess how similar the populations of the study states were to the US population as a whole in terms of key demographic and 99 8
Health A ffairs
June 2014
33 : 6
smoking-related characteristics, we used the Census Bureau’s Current Population Survey. We used data from the Kaiser Family Foundation to examine Medicaid program characteristics. Demographically, the states included in our study were similar to the country as a whole (Exhibit 1). The proportion of women who reported that they were currently smoking was 17 percent for both the study states and the United States, and the proportion of smokers who reported a quit attempt in the past year was similar in the two populations. A smaller percentage of states had adopted presumptive eligibility among the study states than in the entire country, but the reverse was true for those covered through the unborn-child option. Data on Medicaid presumptive-eligibility and unborn-child policies by state and year were collected from published annual surveys of state Medicaid officials regarding their states’ eligibility and enrollment procedures for pregnant women.18–23 Data on coverage of smoking cessation benefits by state and year were collected from published surveys of state Medicaid officials regarding their states’ smoking cessation benefits for pregnant women.24–29 We also included data on whether states had prohibited smoking in worksites, bars, and restaurants30 and each state’s excise tax on cigarettes in each year.31 To identify relevant state Medicaid and tobacco control policies, we first calculated the year each respondent’s pregnancy began, based on the gestational age of the infant at birth. Then we merged state-specific data on Medicaid and tobacco control policies with PRAMS data based on each respondent’s state of residence and the year when her pregnancy began. We defined Medicaid eligibility for each respondent for the relevant state and year based on household income as a percentage of the federal poverty level. PRAMS asked respondents about annual household income and the number of people in the household who depended on that income. Income data were grouped in categories; we took the midpoint of each income category and counted it as the household income amount.32 This income value was compared with the annual federal poverty guidelines33 to calculate income as a percentage of poverty. Respondents with missing income values (5 percent of the 24,544 women in our sample) were considered to be eligible for Medicaid if they reported that Medicaid paid for their prenatal care or delivery.We found state variation in Medicaid takeup rates that was consistent with the results of previous studies that had used simulation models to estimate Medicaid eligibility and take-up.34 Our study included women ages 19–44 in nineteen states who smoked during the three months
Downloaded from content.healthaffairs.org by Health Affairs on May 31, 2015 at UNIV OF GEORGIA LIBRARIES
Exhibit 1 Characteristics Of The Nineteen States In The Study Sample And Of The United States As A Whole 19 states (mean or %)
Characteristic Demographic characteristics Mean age (years) Race (%) White Black Asian American Indian or Native Alaskan or Hawaiian Other or multiple races Hispanic ethnicity (%) Educational attainment (%)a Less than high school High school diploma Some college or college degree Advanced degree Smoking characteristics of women Smoke cigarettes (%)b Made a quit attempt in previous year (%) Medicaid coverage Mean income eligibility threshold (2004–07) for pregnant women (% of FPL) Enrolled in Medicaid (%) State has adopted presumptive eligibility for pregnant women (%) State has adopted unborn-child option (%)
US (mean or %)
36
36
78 14 5 1 3 11
79 13 5 1 2 16
14 24 34 7
16 23 33 6
17 34
17 33
197 20 51 32
188 21 63 25
SOURCE Authors’ analysis of data from the Current Population Survey Annual Social and Economic Supplement, 2004–10 (demographic characteristics); Current Population Survey Tobacco Use Supplement, 2006–07 (smoking characteristics); and Kaiser Family Foundation State Health Facts and published surveys from the Kaiser Family Foundation and the Centers for Disease Control and Prevention (see Notes 18–31 in text). NOTES Demographic and smoking characteristics employ survey weights. The nineteen states are Alaska, Arkansas, Colorado, Georgia, Hawaii, Maine, Maryland, Minnesota, Nebraska, New Jersey, New York, Ohio, Oklahoma, Oregon, Rhode Island, Utah, Vermont, Washington, and West Virginia. FPL is federal poverty level. aCalculated only for adults. bHaving ever smoked 100 cigarettes and currently smoking every or some days.
before conception, had a live single birth between 2004 and 2010, and were eligible for Medicaid during pregnancy in their state in the year their pregnancy began.We excluded women who had twins or other multiples, because having an infant born preterm or small for gestational age are more common in these cases. We were interested in studying the effects of Medicaid enrollment policies, which might differentially enroll women with different smokingrelated risk factors before conception. Therefore, our sample included all women eligible for Medicaid, instead of only those who reported being enrolled in Medicaid during pregnancy. Similarly, we excluded women who were covered by Medicaid just before their pregnancy. Measures The three outcomes of interest were prenatal smoking cessation, preterm birth, and having an infant who was small for gestational age. Prenatal smoking cessation was a binary variable, indicating whether or not a woman reported smoking any amount in the three months before conception and also reported quitting smoking by the third trimester of pregnancy. Preterm birth was another binary vari-
able, indicating whether or not an infant was born before thirty-seven weeks of gestation according to birth certificate data. The PRAMS data contain two measures of small for gestational age: an infant weighing less than the tenth percentile for weight at a given gestational age, and an infant weighing two standard deviations below the mean weight at a given gestational age. We conducted analyses using each of the two measures and observed qualitatively similar results. Clinical practice guidelines define infants as small for gestational age if they weigh less than the tenth percentile for weight at a given gestational age.35 Therefore, we present results using that outcome measure. The primary independent variables of interest were state Medicaid policy variables. For each year, we created indicators of whether or not a state had adopted the presumptive-eligibility option, the unborn-child option, or either of the two. Additionally, for each year we created indicators of whether or not a state Medicaid program provided comprehensive smoking cessation coverage for pregnant women, which we defined as coverage of both pharmacotherapy (any form June 2014
33:6
Downloaded from content.healthaffairs.org by Health Affairs on May 31, 2015 at UNIV OF GEORGIA LIBRARIES
Health Affa irs
99 9
Medicaid of nicotine replacement therapy or bupropion for smoking cessation) and counseling (individual or group) for smoking cessation.36 We controlled for the following individual variables: maternal age, race or ethnicity, education, marital status, number of cigarettes smoked per day before conception, whether or not alcoholic drinks were consumed during pregnancy, parity (that is, whether a woman had had a previous live birth), pregnancy intention (whether the woman wanted a pregnancy then, at another time, or not at all), number of stressors experienced during the twelve months before delivery (for instance, involuntary unemployment or a death in the family), insurance status before conception, and having previously had a preterm birth.We controlled for the following state-level variables: whether or not a state prohibited smoking in worksites, bars, and restaurants; state excise taxes on cigarettes; state Medicaid income eligibility thresholds; and whether a state had a high, medium, or low proportion of Medicaid beneficiaries enrolled in a managed care organization. Our models also included indicator variables for state and year. The former allowed us to control for time-invariant state characteristics, and the latter allowed us to control for national-level secular trends. Data Analysis To estimate the effects of state Medicaid policies on prenatal smoking cessation, preterm birth, and having an infant who was small for gestational age, we took advantage of a natural experiment based on state variations in when the optional Medicaid policies were adopted. In this approach, regression models are run that use pooled cross-sectional data and that include the policy variables of interest, individual control variables, state control variables, and state and year indicators. This allowed us to compare outcomes before and after the policies’ adoption. States without the policies served as the comparison group to control for secular trends in outcomes. This type of analysis can be conceptualized as a comparative interrupted time-series model in which the policy intervention is implemented at different times.37 First, to examine the effects of Medicaid policies on our three outcomes of interest, we employed multivariable logistic regression to estimate the effects of each of the state Medicaid enrollment policies on the odds of prenatal smoking cessation, preterm birth, and having an infant who was small for gestational age, comparing the rates before and after the implementation of the policies and accounting for secular trends. Next, to examine whether the enrollment policies’ effects differed by states’ generosity of coverage of smoking cessation services, we ex1000
H e a lt h A f fai r s
J u n e 201 4
3 3: 6
tended the models to include an interaction term between the state Medicaid enrollment policies (presumptive eligibility and the unborn-child option) and a variable indicating whether or not a state had comprehensive coverage of smoking cessation services. To examine the magnitude of the policies’ effects on the probability of prenatal smoking cessation, preterm birth, and having an infant who was small for gestational age, we derived average marginal effects from the logistic regression models. Average marginal effects represent the percentage-point changes in outcomes due to the policies, and they are helpful in interpreting the results of logistic regression models in a policy context.38,39 All models used PRAMS sampling weights. Robust standard errors were calculated to account for correlation within each state and year. This approach resulted in standard errors that were slightly larger than those obtained by clustering standard errors using the PRAMS sampling strata. Our results provide similar but somewhat more conservative estimates of the policies’ effects than the results achieved by using the survey sampling strata. Limitations This study had several important limitations. First, our measure of Medicaid eligibility was imperfect. PRAMS does not ask about certain Medicaid eligibility criteria, such as types of income that states might disregard (for example, child support payments) when determining eligibility, so we were unable to take these criteria into account. Additionally, PRAMS measures household income in categories, which might lead to misclassification in our definition of eligibility. However, our findings were consistent when we used sensitivity analyses with different definitions of Medicaid income eligibility. Second, prenatal smoking cessation was based on self-report, which tends to overestimate reported cessation in pregnancy relative to biochemical validation.40 It is not clear that such overreporting of cessation would differ by state or across time, however. As a result, this limitation would have the practical effect of biasing our results toward the null. Third, we lacked data on whether states required cost sharing or prior authorization for smoking cessation services—requirements that could be a barrier to receiving these services. Combining enrollment simplification policies with reductions in these barriers could lead to greater reductions in prenatal smoking. Finally, our estimates of the effects of presumptive eligibility were driven by policy changes in three states (Colorado, Maine, and Ohio). And although our study sample was representative of women in the states we included,
Downloaded from content.healthaffairs.org by Health Affairs on May 31, 2015 at UNIV OF GEORGIA LIBRARIES
our results might not be generalizable nationally.
Exhibit 2 Weighted Descriptive Characteristics Of Medicaid-Eligible Women In Nineteen States Who Smoked And Were Not Enrolled In Medicaid Prior To Pregnancy, 2004–10
Study Results
Characteristic
Our final analytic sample included 24,544 women in nineteen states who responded to PRAMS in the period 2004–10. Just over half of the women were younger than age twenty-five, about a third were married, and about three-quarters were white (Exhibit 2). Overall, the women had low socioeconomic status, with a mean household income of 129.9 percent of poverty and with two-thirds of respondents having no more than a high school education. Almost two-thirds were uninsured just prior to conception. In the three months before conception, 55.7 percent of the women reported smoking ten or fewer cigarettes per day, on average. The nineteen study states varied considerably in terms of the number of years in the study period that they offered smoking cessation benefits and coverage based on presumptive eligibility and the unborn-child option (Exhibit 3). Eleven states had presumptive eligibility in place, and seven states had the unborn-child option in place, at some point during the study period. Seven states had comprehensive smoking cessation coverage for the entire study time period, and eight states had it for some of the time. Four states did not have it until after the study period, when it was required by the ACA. States’ adoption of presumptive eligibility led to a 7.7-percentage-point increase in the probability of prenatal smoking cessation but did not have a significant effect on the probability of having a preterm birth or an infant small for gestational age (Exhibit 4). Adopting the unborn-child option did not significantly affect any of the three outcomes. Having either enrollment policy in place was associated with a 6.8percentage-point increase in the probability of prenatal smoking cessation. We observed negative relationships between a state’s adopting either policy and adverse birth outcomes. However, these relationships were not significant. To examine whether the effects of state Medicaid enrollment policies differed in states with different coverage of smoking cessation services during pregnancy, we calculated the average marginal effects of presumptive eligibility, the unborn-child option, or either enrollment policy in states with and without comprehensive coverage of smoking cessation services (Exhibit 4). The effects of presumptive eligibility on any of the three outcomes did not differ by states’ generosity of coverage for smoking cessation services: The policy had a significant effect only on smoking cessation, and there was little varia-
Demographic characteristics Mean household income (% FPL) Maternal age (years) 19–24 25–34 35–44 Race or ethnicity White Black Asian American Indian, Native Alaskan or Hawaiian Other or multiple races Hispanic Married Maternal education Less than high school High school diploma Some college College graduate or more
Mean or %
95% CI
129.9
(126.1, 133.5)
50.9% 42.3 6.8
(49.2, 52.5) (40.9, 43.8) (6.0, 7.6)
75.3 9.6 1.2 3.3 2.3 8.3 34.6
(72.9, 77.6) (8.1, 11.1) (1.0, 1.6) (2.2, 4.5) (1.8, 2.8) (6.8, 9.9) (32.6, 36.6)
21.5 45.4 28.0 5.1
(20.0, 23.0) (44.0, 46.7) (26.3, 29.7) (4.4, 5.9)
55.7 33.0 8.0 3.3
(53.5, 57.9) (31.4, 34.5) (7.1, 8.9) (2.8, 3.9)
93.9 5.5 1.0
(93.3, 94.5) (5.0, 6.0) (0.38, 1.0)
11.1 32.1 40.1 16.7
(10.2, (31.0, (38.9, (15.6,
12.1) 33.2) 41.4) 17.7)
28.3 55.7 16.1 53.2 13.1
(27.0, (54.2, (14.9, (51.6, (12.1,
29.5) 57.2) 17.3) 54.8) 14.1)
63.9 36.1
(61.7, 66.1) (33.9, 38.3)
71.9 28.0 74.5
(69.9, 74.0) (26.0, 30.1) (72.7, 76.2)
Health factors Cigarettes smoked per day before conceptiona 1–10 11–20 21–40 41 or more Alcoholic drinks per week during third trimesterb 0
The online version of this article, along with updated information and services, is available at: http://content.healthaffairs.org/content/33/6/997.full.html
For Reprints, Links & Permissions: http://healthaffairs.org/1340_reprints.php E-mail Alerts : http://content.healthaffairs.org/subscriptions/etoc.dtl To Subscribe: http://content.healthaffairs.org/subscriptions/online.shtml
Health Affairs is published monthly by Project HOPE at 7500 Old Georgetown Road, Suite 600, Bethesda, MD 20814-6133. Copyright © 2014 by Project HOPE - The People-to-People Health Foundation. As provided by United States copyright law (Title 17, U.S. Code), no part of Health Affairs may be reproduced, displayed, or transmitted in any form or by any means, electronic or mechanical, including photocopying or by information storage or retrieval systems, without prior written permission from the Publisher. All rights reserved.
Not for commercial use or unauthorized distribution Downloaded from content.healthaffairs.org by Health Affairs on May 31, 2015 at UNIV OF GEORGIA LIBRARIES
Medicaid By Marian Jarlenski, Sara N. Bleich, Wendy L. Bennett, Elizabeth A. Stuart, and Colleen L. Barry 10.1377/hlthaff.2013.1167 HEALTH AFFAIRS 33, NO. 6 (2014): 997–1005 ©2014 Project HOPE— The People-to-People Health Foundation, Inc.
doi:
Medicaid Enrollment Policy Increased Smoking Cessation Among Pregnant Women But Had No Impact On Birth Outcomes
Marian Jarlenski (jarlenski@ gmail.com) recently completed doctoral studies in health policy and management at the Johns Hopkins Bloomberg School of Public Health, in Baltimore, Maryland. In fall 2014 she will become an assistant professor of health policy and management at the Graduate School of Public Health, University of Pittsburgh, in Pennsylvania.
Cigarette smoking during pregnancy is an important cause of poor maternal and infant health outcomes in the population eligible for Medicaid. These outcomes may be avoided or attenuated by timely, highquality prenatal care. Using data from the Centers for Disease Control and Prevention’s Pregnancy Risk Assessment Monitoring System for the period 2004–10, we examined the effects of two optional state Medicaid enrollment policies on smoking cessation, preterm birth, and having an infant who was small for gestational age. We used a natural experiment to compare outcomes before and after nineteen states adopted either of the two policies. The first policy, presumptive eligibility, permits women to receive prenatal care while their Medicaid application is pending. Its adoption led to a 7.7-percentage-point increase in smoking cessation but did not reduce adverse birth outcomes. The second policy, the unbornchild option, permits states to provide coverage to pregnant women who cannot document their citizenship or residency. Its adoption was not significantly associated with any of the three outcomes. The presumptiveeligibility enrollment policy will continue to be an important tool for promoting timely prenatal care and smoking cessation. ABSTRACT
C
igarette smoking in pregnancy accounts for a substantial portion of poor maternal and infant health outcomes and infant deaths.1–3 The prevalence of prenatal smoking in the United States has declined in recent decades,4 but it is nearly twice as high among lowincome women enrolled in Medicaid than it is in the US population as a whole.5 Since the late 1990s, many state Medicaid programs have begun providing more generous coverage of smoking cessation services for pregnant women.6 However, the process of enrolling in Medicaid may be a barrier to obtaining these services.7 The Medicaid application process is complex—requiring documentation that verifies the applicant’s income, residency, citizenship, and pregnancy—and a determination of eligibil-
Sara N. Bleich is an associate professor of health policy and management at the Johns Hopkins Bloomberg School of Public Health. Wendy L. Bennett is an assistant professor of medicine in the Division of General Internal Medicine, Johns Hopkins University School of Medicine, in Baltimore.
ity may take weeks.8 States have two policy options they can use to reduce barriers to Medicaid enrollment during pregnancy. Under the first, presumptive eligibility, low-income pregnant women are presumed to be eligible for Medicaid when they arrive for care at participating organizations. Thus, they can receive care while their Medicaid applications are still pending.9 A second policy, known as the unborn-child option, allows states to consider a fetus to be a “targeted low-income child” and to provide coverage of prenatal care and delivery to low-income pregnant women even if they cannot provide the documentation of citizenship or residency that is required for eligibility in Medicaid’s pregnancy category.10 These policies can lead to a greater probability of Medicaid enrollment and earlier initiation of June 2014
33:6
Downloaded from content.healthaffairs.org by Health Affairs on May 31, 2015 at UNIV OF GEORGIA LIBRARIES
Elizabeth A. Stuart is an associate professor of mental health and biostatistics at the Johns Hopkins Bloomberg School of Public Health. Colleen L. Barry is an associate professor of health policy and management at the Johns Hopkins Bloomberg School of Public Health.
H ea lt h A f fai r s
9 97
Medicaid prenatal care,11,12 which enables women to access smoking cessation services earlier in pregnancy than would otherwise be the case. Smoking cessation early in pregnancy has been shown to reduce adverse birth outcomes.13–15 No published research has examined the effects of these two policies on prenatal smoking cessation or smoking-related adverse birth outcomes. In the context of a new requirement under the Affordable Care Act (ACA) that all state Medicaid programs cover counseling and pharmacotherapies for smoking cessation for pregnant women,16 it is important to understand how these optional state Medicaid enrollment policies affect the use of smoking cessation services and birth outcomes. We addressed this gap in the literature by examining the effects of the policies on prenatal smoking cessation, preterm birth, and having an infant who was small for gestational age. We hypothesized that a state’s adoption of one of the two optional enrollment policies (presumptive eligibility and the unborn child option) would increase the probability that its female residents would stop smoking during pregnancy and decrease the probability of their having a preterm birth and an infant who was small for gestational age. We also hypothesized that the effects of the two optional enrollment policies would be greater in states with more-generous coverage of services for smoking cessation during pregnancy than in states with less-generous coverage.
Study Data And Methods Data Sources The Pregnancy Risk Assessment Monitoring System (PRAMS) of the Centers for Disease Control and Prevention (CDC) is a staterepresentative survey about women’s health, behavior, insurance status, and health care before, during, and shortly after pregnancy.17 States mail women a questionnaire two to four months after delivery, and the women who do not respond to the mailed questionnaire are contacted by telephone. Respondents’ answers to survey items are linked to birth certificate data. PRAMS research data are available for states that achieved a response rate of at least 70 percent before 2007 or a response rate of at least 65 percent from 2007 on. Between 2004 and 2010, nineteen of thirty-five participating states had sufficient response rates in all years and are thus included in our study. Although our study sample is not nationally representative, it is representative of women residing in these nineteen states in each of these years. To assess how similar the populations of the study states were to the US population as a whole in terms of key demographic and 99 8
Health A ffairs
June 2014
33 : 6
smoking-related characteristics, we used the Census Bureau’s Current Population Survey. We used data from the Kaiser Family Foundation to examine Medicaid program characteristics. Demographically, the states included in our study were similar to the country as a whole (Exhibit 1). The proportion of women who reported that they were currently smoking was 17 percent for both the study states and the United States, and the proportion of smokers who reported a quit attempt in the past year was similar in the two populations. A smaller percentage of states had adopted presumptive eligibility among the study states than in the entire country, but the reverse was true for those covered through the unborn-child option. Data on Medicaid presumptive-eligibility and unborn-child policies by state and year were collected from published annual surveys of state Medicaid officials regarding their states’ eligibility and enrollment procedures for pregnant women.18–23 Data on coverage of smoking cessation benefits by state and year were collected from published surveys of state Medicaid officials regarding their states’ smoking cessation benefits for pregnant women.24–29 We also included data on whether states had prohibited smoking in worksites, bars, and restaurants30 and each state’s excise tax on cigarettes in each year.31 To identify relevant state Medicaid and tobacco control policies, we first calculated the year each respondent’s pregnancy began, based on the gestational age of the infant at birth. Then we merged state-specific data on Medicaid and tobacco control policies with PRAMS data based on each respondent’s state of residence and the year when her pregnancy began. We defined Medicaid eligibility for each respondent for the relevant state and year based on household income as a percentage of the federal poverty level. PRAMS asked respondents about annual household income and the number of people in the household who depended on that income. Income data were grouped in categories; we took the midpoint of each income category and counted it as the household income amount.32 This income value was compared with the annual federal poverty guidelines33 to calculate income as a percentage of poverty. Respondents with missing income values (5 percent of the 24,544 women in our sample) were considered to be eligible for Medicaid if they reported that Medicaid paid for their prenatal care or delivery.We found state variation in Medicaid takeup rates that was consistent with the results of previous studies that had used simulation models to estimate Medicaid eligibility and take-up.34 Our study included women ages 19–44 in nineteen states who smoked during the three months
Downloaded from content.healthaffairs.org by Health Affairs on May 31, 2015 at UNIV OF GEORGIA LIBRARIES
Exhibit 1 Characteristics Of The Nineteen States In The Study Sample And Of The United States As A Whole 19 states (mean or %)
Characteristic Demographic characteristics Mean age (years) Race (%) White Black Asian American Indian or Native Alaskan or Hawaiian Other or multiple races Hispanic ethnicity (%) Educational attainment (%)a Less than high school High school diploma Some college or college degree Advanced degree Smoking characteristics of women Smoke cigarettes (%)b Made a quit attempt in previous year (%) Medicaid coverage Mean income eligibility threshold (2004–07) for pregnant women (% of FPL) Enrolled in Medicaid (%) State has adopted presumptive eligibility for pregnant women (%) State has adopted unborn-child option (%)
US (mean or %)
36
36
78 14 5 1 3 11
79 13 5 1 2 16
14 24 34 7
16 23 33 6
17 34
17 33
197 20 51 32
188 21 63 25
SOURCE Authors’ analysis of data from the Current Population Survey Annual Social and Economic Supplement, 2004–10 (demographic characteristics); Current Population Survey Tobacco Use Supplement, 2006–07 (smoking characteristics); and Kaiser Family Foundation State Health Facts and published surveys from the Kaiser Family Foundation and the Centers for Disease Control and Prevention (see Notes 18–31 in text). NOTES Demographic and smoking characteristics employ survey weights. The nineteen states are Alaska, Arkansas, Colorado, Georgia, Hawaii, Maine, Maryland, Minnesota, Nebraska, New Jersey, New York, Ohio, Oklahoma, Oregon, Rhode Island, Utah, Vermont, Washington, and West Virginia. FPL is federal poverty level. aCalculated only for adults. bHaving ever smoked 100 cigarettes and currently smoking every or some days.
before conception, had a live single birth between 2004 and 2010, and were eligible for Medicaid during pregnancy in their state in the year their pregnancy began.We excluded women who had twins or other multiples, because having an infant born preterm or small for gestational age are more common in these cases. We were interested in studying the effects of Medicaid enrollment policies, which might differentially enroll women with different smokingrelated risk factors before conception. Therefore, our sample included all women eligible for Medicaid, instead of only those who reported being enrolled in Medicaid during pregnancy. Similarly, we excluded women who were covered by Medicaid just before their pregnancy. Measures The three outcomes of interest were prenatal smoking cessation, preterm birth, and having an infant who was small for gestational age. Prenatal smoking cessation was a binary variable, indicating whether or not a woman reported smoking any amount in the three months before conception and also reported quitting smoking by the third trimester of pregnancy. Preterm birth was another binary vari-
able, indicating whether or not an infant was born before thirty-seven weeks of gestation according to birth certificate data. The PRAMS data contain two measures of small for gestational age: an infant weighing less than the tenth percentile for weight at a given gestational age, and an infant weighing two standard deviations below the mean weight at a given gestational age. We conducted analyses using each of the two measures and observed qualitatively similar results. Clinical practice guidelines define infants as small for gestational age if they weigh less than the tenth percentile for weight at a given gestational age.35 Therefore, we present results using that outcome measure. The primary independent variables of interest were state Medicaid policy variables. For each year, we created indicators of whether or not a state had adopted the presumptive-eligibility option, the unborn-child option, or either of the two. Additionally, for each year we created indicators of whether or not a state Medicaid program provided comprehensive smoking cessation coverage for pregnant women, which we defined as coverage of both pharmacotherapy (any form June 2014
33:6
Downloaded from content.healthaffairs.org by Health Affairs on May 31, 2015 at UNIV OF GEORGIA LIBRARIES
Health Affa irs
99 9
Medicaid of nicotine replacement therapy or bupropion for smoking cessation) and counseling (individual or group) for smoking cessation.36 We controlled for the following individual variables: maternal age, race or ethnicity, education, marital status, number of cigarettes smoked per day before conception, whether or not alcoholic drinks were consumed during pregnancy, parity (that is, whether a woman had had a previous live birth), pregnancy intention (whether the woman wanted a pregnancy then, at another time, or not at all), number of stressors experienced during the twelve months before delivery (for instance, involuntary unemployment or a death in the family), insurance status before conception, and having previously had a preterm birth.We controlled for the following state-level variables: whether or not a state prohibited smoking in worksites, bars, and restaurants; state excise taxes on cigarettes; state Medicaid income eligibility thresholds; and whether a state had a high, medium, or low proportion of Medicaid beneficiaries enrolled in a managed care organization. Our models also included indicator variables for state and year. The former allowed us to control for time-invariant state characteristics, and the latter allowed us to control for national-level secular trends. Data Analysis To estimate the effects of state Medicaid policies on prenatal smoking cessation, preterm birth, and having an infant who was small for gestational age, we took advantage of a natural experiment based on state variations in when the optional Medicaid policies were adopted. In this approach, regression models are run that use pooled cross-sectional data and that include the policy variables of interest, individual control variables, state control variables, and state and year indicators. This allowed us to compare outcomes before and after the policies’ adoption. States without the policies served as the comparison group to control for secular trends in outcomes. This type of analysis can be conceptualized as a comparative interrupted time-series model in which the policy intervention is implemented at different times.37 First, to examine the effects of Medicaid policies on our three outcomes of interest, we employed multivariable logistic regression to estimate the effects of each of the state Medicaid enrollment policies on the odds of prenatal smoking cessation, preterm birth, and having an infant who was small for gestational age, comparing the rates before and after the implementation of the policies and accounting for secular trends. Next, to examine whether the enrollment policies’ effects differed by states’ generosity of coverage of smoking cessation services, we ex1000
H e a lt h A f fai r s
J u n e 201 4
3 3: 6
tended the models to include an interaction term between the state Medicaid enrollment policies (presumptive eligibility and the unborn-child option) and a variable indicating whether or not a state had comprehensive coverage of smoking cessation services. To examine the magnitude of the policies’ effects on the probability of prenatal smoking cessation, preterm birth, and having an infant who was small for gestational age, we derived average marginal effects from the logistic regression models. Average marginal effects represent the percentage-point changes in outcomes due to the policies, and they are helpful in interpreting the results of logistic regression models in a policy context.38,39 All models used PRAMS sampling weights. Robust standard errors were calculated to account for correlation within each state and year. This approach resulted in standard errors that were slightly larger than those obtained by clustering standard errors using the PRAMS sampling strata. Our results provide similar but somewhat more conservative estimates of the policies’ effects than the results achieved by using the survey sampling strata. Limitations This study had several important limitations. First, our measure of Medicaid eligibility was imperfect. PRAMS does not ask about certain Medicaid eligibility criteria, such as types of income that states might disregard (for example, child support payments) when determining eligibility, so we were unable to take these criteria into account. Additionally, PRAMS measures household income in categories, which might lead to misclassification in our definition of eligibility. However, our findings were consistent when we used sensitivity analyses with different definitions of Medicaid income eligibility. Second, prenatal smoking cessation was based on self-report, which tends to overestimate reported cessation in pregnancy relative to biochemical validation.40 It is not clear that such overreporting of cessation would differ by state or across time, however. As a result, this limitation would have the practical effect of biasing our results toward the null. Third, we lacked data on whether states required cost sharing or prior authorization for smoking cessation services—requirements that could be a barrier to receiving these services. Combining enrollment simplification policies with reductions in these barriers could lead to greater reductions in prenatal smoking. Finally, our estimates of the effects of presumptive eligibility were driven by policy changes in three states (Colorado, Maine, and Ohio). And although our study sample was representative of women in the states we included,
Downloaded from content.healthaffairs.org by Health Affairs on May 31, 2015 at UNIV OF GEORGIA LIBRARIES
our results might not be generalizable nationally.
Exhibit 2 Weighted Descriptive Characteristics Of Medicaid-Eligible Women In Nineteen States Who Smoked And Were Not Enrolled In Medicaid Prior To Pregnancy, 2004–10
Study Results
Characteristic
Our final analytic sample included 24,544 women in nineteen states who responded to PRAMS in the period 2004–10. Just over half of the women were younger than age twenty-five, about a third were married, and about three-quarters were white (Exhibit 2). Overall, the women had low socioeconomic status, with a mean household income of 129.9 percent of poverty and with two-thirds of respondents having no more than a high school education. Almost two-thirds were uninsured just prior to conception. In the three months before conception, 55.7 percent of the women reported smoking ten or fewer cigarettes per day, on average. The nineteen study states varied considerably in terms of the number of years in the study period that they offered smoking cessation benefits and coverage based on presumptive eligibility and the unborn-child option (Exhibit 3). Eleven states had presumptive eligibility in place, and seven states had the unborn-child option in place, at some point during the study period. Seven states had comprehensive smoking cessation coverage for the entire study time period, and eight states had it for some of the time. Four states did not have it until after the study period, when it was required by the ACA. States’ adoption of presumptive eligibility led to a 7.7-percentage-point increase in the probability of prenatal smoking cessation but did not have a significant effect on the probability of having a preterm birth or an infant small for gestational age (Exhibit 4). Adopting the unborn-child option did not significantly affect any of the three outcomes. Having either enrollment policy in place was associated with a 6.8percentage-point increase in the probability of prenatal smoking cessation. We observed negative relationships between a state’s adopting either policy and adverse birth outcomes. However, these relationships were not significant. To examine whether the effects of state Medicaid enrollment policies differed in states with different coverage of smoking cessation services during pregnancy, we calculated the average marginal effects of presumptive eligibility, the unborn-child option, or either enrollment policy in states with and without comprehensive coverage of smoking cessation services (Exhibit 4). The effects of presumptive eligibility on any of the three outcomes did not differ by states’ generosity of coverage for smoking cessation services: The policy had a significant effect only on smoking cessation, and there was little varia-
Demographic characteristics Mean household income (% FPL) Maternal age (years) 19–24 25–34 35–44 Race or ethnicity White Black Asian American Indian, Native Alaskan or Hawaiian Other or multiple races Hispanic Married Maternal education Less than high school High school diploma Some college College graduate or more
Mean or %
95% CI
129.9
(126.1, 133.5)
50.9% 42.3 6.8
(49.2, 52.5) (40.9, 43.8) (6.0, 7.6)
75.3 9.6 1.2 3.3 2.3 8.3 34.6
(72.9, 77.6) (8.1, 11.1) (1.0, 1.6) (2.2, 4.5) (1.8, 2.8) (6.8, 9.9) (32.6, 36.6)
21.5 45.4 28.0 5.1
(20.0, 23.0) (44.0, 46.7) (26.3, 29.7) (4.4, 5.9)
55.7 33.0 8.0 3.3
(53.5, 57.9) (31.4, 34.5) (7.1, 8.9) (2.8, 3.9)
93.9 5.5 1.0
(93.3, 94.5) (5.0, 6.0) (0.38, 1.0)
11.1 32.1 40.1 16.7
(10.2, (31.0, (38.9, (15.6,
12.1) 33.2) 41.4) 17.7)
28.3 55.7 16.1 53.2 13.1
(27.0, (54.2, (14.9, (51.6, (12.1,
29.5) 57.2) 17.3) 54.8) 14.1)
63.9 36.1
(61.7, 66.1) (33.9, 38.3)
71.9 28.0 74.5
(69.9, 74.0) (26.0, 30.1) (72.7, 76.2)
Health factors Cigarettes smoked per day before conceptiona 1–10 11–20 21–40 41 or more Alcoholic drinks per week during third trimesterb 0
