By Genevieve M. Kenney, Sharon K. Long, and Adela Luque. legislation in April 2006 that has moved the state to nearuniversal

Transcription

1 Update On Massachusetts doi: /hlthaff HEALTH AFFAIRS 29, NO. 6 (2010): Project HOPE The People-to-People Health Foundation, Inc. By Genevieve M. Kenney, Sharon K. Long, and Adela Luque Health Reform In Massachusetts Cut The Uninsurance Rate Among Children In Half Genevieve M. Kenney (JKenney@urban.org) is a senior fellow at the Health Policy Center, Urban Institute, in Washington, D.C. Sharon K. Long is a senior fellow at the Health Policy Center, Urban Institute. Adela Luque is a research associate at the Urban Institute. ABSTRACT Massachusetts 2006 health reform cut the uninsurance rate for children approximately in half in the first two years following implementation. The state now has the lowest rate of uninsurance among children in the nation. More children became enrolled in MassHealth, the state s Medicaid program, and in employer-sponsored insurance. Most of the coverage increases occurred among lower-income children, many of whom were eligible for but not enrolled in MassHealth prior to reform. We derive a major lesson for national health reform: that outreach, enrollment simplifications, and coverage expansions to parents and children can lead to substantial reductions in the number of uninsured children, particularly among children in the lowest-income families. Massachusetts enacted comprehensive health reform legislation in April 2006 that has moved the state to nearuniversal insurance coverage. As of 2008, only 4.1 percent of all state residents and 2.1 percent of children in the state were uninsured, compared to 15.1 percent and 9.9 percent, respectively, in the nation as a whole. 1 A number of studies have documented the impact of Massachusetts health reform initiative on insurance coverage for nonelderly adults. 2,3 However, there has been only limited examination of the impacts of health reform on children. 4 Massachusetts health reform initiative encompassed an extensive set of changes that had the potential to affect the coverage of children, both directly through their eligibility for subsidized coverage and indirectly through their parents insurance coverage. A summary of key components of the law affecting children, and their effects, can be found in Appendix A.1. 5,6 Of direct relevance to children is the expansion of MassHealth, the state s Medicaid program, to children with family incomes of up to 300 percent of the federal poverty level an increase from 200 percent of the poverty level prior to health reform. Beyond that policy change, the remaining elements of health reform are largely targeted at adults in the state. These include a new subsidized health insurance program, Commonwealth Care, for lower-income adults with family incomes below 300 percent of the poverty level; an insurance exchange offering private health insurance for higher-income families called Commonwealth Choice; insurance market reforms, including new requirements for employers; and an individual mandate that requires adults with access to affordable coverage to obtain health insurance or pay a penalty. Prior research found that as a result of these policies, insurance coverage among adults in Massachusetts overall, including parents, has shown strong gains. 3,7 We would expect these gains in parental coverage to have an indirect effect on children. Previous studies have found that expansions in coverage for parents result in gains in coverage for their children. 8 We would also expect the strong outreach and enrollment simplification efforts associated with health reform to affect the coverage of the lowest-income children. Prior research suggests 1242 HEALTH AFFAIRS JUNE :6

2 that such efforts are likely to lead to increased enrollment among those who were already eligible for public coverage. 9 Enrolling this group is key, because such eligible but unenrolled children constitute the majority of uninsured children nationally. 10 In this paper we build on research addressing the effects of health reform on adults in Massachusetts to examine the effects of the reform on children. We estimated the overall impact of health reform on children in the state, as well as the impact on key subgroups of children, including lower- and higher-income children and those newly eligible for MassHealth. Study Data And Methods In assessing how Massachusetts health reform initiative has affected the coverage of children in the state, we compared the change in insurance coverage before and after health reform was implemented for both the treatment group (that is, children in Massachusetts) and the comparison group (that is, children in other states that did not implement health reform). To do this, we used a difference-in-differences 11 framework, wherein the change for the comparison group (before and after implementation of the law) is subtracted from the change for the treatment group, to control for other changes over time beyond health reform. That is, the comparison group provides an estimate of what would have happened to children in Massachusetts in the absence of health reform. Because there is no perfect comparison group in the absence of random assignment, we explored a number of alternative models based on different comparison groups to assess the sensitivity of our findings to the choice of comparison groups. We discuss this issue further below. Given that we expect the effects of health reform to vary for different groups of children in the state by family income, we examined changes in coverage for all children in the state as well as for two broad groups children in families with incomes at or below 300 percent of poverty, who are eligible for MassHealth, and children in families with incomes above 300 percent of poverty. Within the lower-income group, we also examine two subsets of children: those who were eligible for MassHealth prior to health reform (family incomes at or below 200 percent of poverty) and those newly eligible for MassHealth (family incomes of percent of poverty). Data Source We used data for children ages 0 18 from the Current Population Surveys (CPS), which collected information corresponding to the period. The Current Population Survey, a monthly nationally representative household survey of the U.S. civilian, noninstitutionalized population, focuses largely on labor-force information. For this study we relied on the Annual Social and Economic Supplement to the CPS, which provides data on health insurance coverage and family income, among other measures. As in other work, we excluded children from families that had coverage imputed to them by the Census Bureau, because it has been shown that this process tends to overstate the number of uninsured people in states with low uninsurance rates, such as Massachusetts.We reweighted the remaining sample to be representative of the population in each state in each year. 12,13 Defining Health Insurance Status For much of the analysis, we focused on four coverage categories: employer-sponsored; public (Mass- Health); nongroup and other; and uninsured. 14 However, because there is some evidence of misreporting of coverage type in surveys, 15 we also report the results for a category that combines public and nongroup coverage. In our framework, a child reported as having both MassHealth and employer-sponsored insurance coverage would be assigned to employer coverage. 16 Defining The Comparison Groups Because health reform in Massachusetts was aimed at everyone in the state, using a within-state comparison group was not an option. Thus, for this study, we used children in other states as our comparison group. We compared estimates for models that drew from different groups of children (higher- and lower-income) and different states. The comparison group for our core estimation comprised children in higher-income families, ( percent of poverty), in the other states in the New England census division that did not make major changes in their coverage for higherincome children or for parents during the analysis period. 17,18 We varied income groups for children and comparison states in our alternative comparison-group analysis. In particular, we estimated models using Minnesota and Washington State as the comparison states. Although they are not in New England, they have demographic and economic profiles that are very similar to those of Massachusetts, and they did not implement major health care reforms during the study period. Finally, we estimated alternative models using propensity score reweighting, a method that matches the characteristics of the children in each of the comparison groups with those of Massachusetts children. 19,20 Our basic findings are fairly consistent with respect to the choice of comparison groups or weights, which provides support for the reliability of our estimates. JUNE :6 HEALTH AFFAIRS 1243

3 Update On Massachusetts Defining The Pre- And Post-Reform Periods The main policy change directly targeted at children the expansion of MassHealth was implemented in July Ideally, we would like to match the pre- and post-reform periods with the exact time the reform took place. However, we were unable to do so because the CPS asks about health insurance coverage in the prior calendar year, and Massachusetts phased in health reform over several years. For our core model, we define the pre-reform period as and the post-reform period as Because some elements of reform were implemented in 2006, we also estimated models that excluded 2006 from the analysis. Our basic findings are generally not sensitive to the inclusion or exclusion of data for Specification Of The Model And Estimation Our multivariate models estimated the independent effects of several variables on a child s health insurance status. We estimated separate linear probability models for each insurance category: employer, public, nongroup, and uninsured. The models controlled for the demographic and socioeconomic characteristics of the child and his or her parent(s) and family 21 and for residence in a Metropolitan Statistical Area. This allowed us, among other things, to exclude the effects of differences in children s characteristics over time and across states (Appendix A.2). 5 Children s characteristics included age, sex, citizenship status, race or ethnicity, disability status, and whether the child was in fair or poor health. Family characteristics included family type (one-parent or two-parent family) and size. Parental characteristics include citizenship status, health and disability status, work status, and largest firm either parent works for included because larger firms are more likely than others to offer coverage to their workers. To account for the complex survey design of the Current Population Survey, we used replicate weights in estimating the standard errors and statistical significance for our models. Study Results Simple Differences Over Time The simple differences over time in Massachusetts show a significant drop in uninsurance rates among children: from 4.6 percent in the pre-reform period to 1.8 percent in the post-reform period (Exhibit 1). 22 The greatest gains in coverage were reported for children in lower-income families. This result is expected, given higher uninsurance rates for this group prior to reform, and given that both the expansion in MassHealth eligibility for children and expansion efforts for adults were targeted at those with family incomes below 300 percent of poverty. The uninsurance rate for children in families with incomes at or below 300 percent of poverty fell by 5.2 percentage points (Exhibit 1). Similar declines in uninsurance rates were reported both for children newly eligible for EXHIBIT 1 Changes In Insurance Coverage For Children In Massachusetts, Overall And By Family Income, Before And After State Health Reform Pre-reform period (2004 6) Post-reform period (2007 8) Pre-post difference Insurance coverage for all children N ¼ 2; 623 N ¼ 1; % 1.8% 2.8** Public and nongroup coverage 21.7% 26.7% 5.0* Public coverage 17.8% 24.5% 6.7** Nongroup coverage 3.9% 2.2% 1.7* Employer-sponsored coverage 73.7% 71.5% 2.2 Uninsurance rate for children, by family income Lower-income children N ¼ 1; 112 N ¼ % 3.0% 5.2** Higher-income children N ¼ 1; 511 N ¼ % 0.9% 0.9 Uninsurance rate for lower-income children, by eligibility for MassHealth Eligible for MassHealth prior to reform N ¼ 743 N ¼ % 3.8% 5.4* Newly eligible for MassHealth under reform N ¼ % N ¼ % 5.6** SOURCE Annual Social and Economic Supplement to the March Current Population Survey, NOTE Income cutoff levels are available in the text. *p < 0:10 **p < 0: HEALTH AFFAIRS JUNE :6

4 MassHealth those in families with incomes of percent of poverty (down 5.6 percentage points) and for children who would have been eligible for MassHealth under the eligibility standards in place prior to health reform those with family incomes at or below 200 percent of poverty (down 5.4 percentage points). Although the uninsurance rate also dropped among higher-income children during this period, that change was not statistically significant. Impacts Of Reform Estimates from the difference-in-differences models, which controlled for changes in coverage over time resulting from factors other than health reform, indicate that most of the gains in coverage over this period were attributable to the effects of health reform in the state. Consistent with the simple pre-post analysis, we found a significant decrease in the uninsurance rate among children in Massachusetts, particularly for children in lower-income families. Focusing first on our core model (Exhibit 2), we find that health reform reduced the uninsurance rate by 2.7 percentage points among all children, largely due to a 5.2-percentage-point decline among children in lower-income families. 23 Like the pre-post comparisons, the difference-in-differences estimates do not show a significant reduction in rates of uninsurance among children in higher-income families. According to our estimates, the decrease in the number of uninsured children in the lower-income group accounted for approximately 80 percent of the overall decline in the number of uninsured children. A closer look at children in lower-income families reveals that the uninsurance rate among those directly affected by the expansion in MassHealth eligibility decreased by 6.4 percentage points (Exhibit 3). In addition, those children who were eligible for MassHealth prior to reform also experienced a 5.1-percentage-point decline in the uninsurance rate. The gains in coverage in the state reflect gains in both public and employer-sponsored coverage that more than offset the drop in nongroup coverage that occurred. Although the increases in public coverage and employer-sponsored coverage among all children were not significant at conventional levels (p <0:10), together they contributed to reductions in uninsurance rates among children (Exhibit 2). Despite uncertainty as to the reliability of reported coverage types in surveys, patterns across the income groups of children were as expected: Lower-income children showed significant gains in public coverage, while higher-income children showed significant gains in employer-sponsored coverage. Children at all income levels showed EXHIBIT 2 Difference-In-Differences Estimates Of The Impacts Of Health Reform On Children s Insurance Coverage In Massachusetts, Overall And By Family Income All children N ¼ 10; ** Public and nongroup coverage 0.4 Public coverage 3.2 Nongroup coverage 2.8* Employer-sponsored coverage 2.3 Lower-income children N ¼ 8; ** Public and nongroup coverage 4.4 Public coverage 7.7* Nongroup coverage 3.4* Employer-sponsored coverage 0.8 Higher-income children N ¼ 8; Public and nongroup coverage 3.2 Public coverage 0.8 Nongroup coverage 2.3 Employer-sponsored coverage 4.1* SOURCE Annual Social and Economic Supplement to the March Current Population Survey, NOTES These are regression-adjusted estimates based on models that control for child, family, and parental characteristics and geographic location. Income cutoff levels are available in the text. *p < 0:10 **p < 0:05 drops in nongroup coverage. This suggests that reform may have also reduced underinsurance in the state, as nongroup policies often provide limited coverage at high cost. One issue that arises with an expansion in eligibility for public coverage is the possibility that an increase in public coverage substitutes for or crowds out employer-sponsored coverage. As shown in Exhibit 2, we found a significant increase in employer-sponsored coverage for higher-income children under health reform in the state and no change in such coverage for lower-income children. Thus, we found no evidence of the crowding out of employer coverage for children. Rather, it would appear that the expansions in employer coverage for parents that have been reported elsewhere have also benefited children. Sensitivity Analysis A concern about all difference-in-differences models is that the results may be sensitive to the comparison group chosen. When we estimated models using alternative comparison groups based on different groups of children, different states, and propensity-score reweighting methods, we obtained similar findings, particularly in terms of the direction of the effects, on the impacts of reform on insurance coverage (Appendix A.3). 5 We also obtained similar findings when we estimated models that excluded 2006, a transition year for reform, as well as noncitizens, a population that is not eligible for some of the coverage options created under health reform in Massachusetts. JUNE :6 HEALTH AFFAIRS 1245

5 Update On Massachusetts EXHIBIT 3 Difference-In-Differences Estimates Of The Effects Of Health Reform On Lower-Income Children s Insurance Coverage In Massachusetts, By Eligibility Status For MassHealth Eligible for MassHealth before reform N ¼ 7; * Public and nongroup coverage 6.8 Public coverage 10.4* Nongroup coverage 3.6* Employer-sponsored coverage 1.7 Newly eligible for MassHealth under reform N ¼ 6; ** Public and nongroup coverage 1.7 Public coverage 3.9 Nongroup coverage 2.2 Employer-sponsored coverage 4.6 SOURCE Annual Social and Economic Supplement to the March Current Population Survey, NOTES These are regression-adjusted estimates based on models that control for child, family, and parental characteristics and geographic location. Detailed results on the uninsured from our core model for the sample of all children are presented in Appendix A.4, available by clicking on the Appendix link in the box to the right of the article online. *p < 0:10 **p < 0:05 Only about 3 percent of children in Massachusetts are not citizens. In addition to these models, we also estimated a number of other variations, with similar results. The estimates of the impacts of reform on coverage for children were not very sensitive to the particular specification of the model. 24 Discussion Our findings indicate that health reform in Massachusetts cut the uninsurance rate for children approximately in half in the two years following implementation. By 2008, the rate of uninsurance among children had fallen by 2.8 percentage points. At that point, just 1.8 percent of all Massachusetts children and 3.0 percent of those in families at or below 300 percent of poverty were uninsured. These are the lowest rates for children in the nation. 25 Our findings on the effects of health reform on children in Massachusetts are consistent with those findings in earlier studies of adults. 3 The estimated impact among adults is larger in magnitude than that found for children. This probably reflects the fact that the uninsurance rate was higher for adults than children prior to reform, and that the majority of the elements of reform (including the individual mandate) are targeted at adults, not children. The reduction in the number of uninsured children in Massachusetts was driven by increases in both MassHealth coverage and employersponsored insurance. Our results do not provide evidence that the reform resulted in the substitution of public for employer coverage among children, at least in the two years following initial implementation. It is likely that this reflects the gains in employer coverage among parents in response to the individual mandate and additional requirements for employers under health reform, among other changes. It appears that the reform initiative led families to substitute employer-sponsored coverage or MassHealth for coverage purchased in the nongroup market. The reductions in nongroup coverage for children in Massachusetts are expected to increase their access to needed health care, since it is likely that they now have more comprehensive benefits and lower out-of pocket cost-sharing burdens for their families. It will be important for future research to examine how access to care and service use have changed for both lower- and higher-income children in Massachusetts as a result of health reform. Most of the coverage increase occurred among lower-income children, many of whom were eligible for MassHealth prior to reform. The success at increasing coverage among these lowincome children in Massachusetts was probably due to a combination of the states communitybased outreach efforts, the simplification of the MassHealth enrollment process, and coverage expansions that were implemented for parents. The gains for children are particularly notable because Massachusetts did not include an individual mandate for children s coverage, and because children had such low uninsurance rates in the state prior to reform. The findings suggest that the reform provisions that were directed at parents had strong positive spillover effects on the coverage of their children. This study has important implications for the recently enacted national health reform law. That legislation contains many of the same elements as the reform initiative in Massachusetts. Because the majority of uninsured children nationwide are eligible for public coverage but not yet enrolled, achieving substantial coverage gains for children depends on raising participation and retention in public programs. 10 The experience in Massachusetts suggests that increasing take-up of public coverage is possible, and that the policy changes in national reform that address both enrollment barriers and coverage gaps for parents may be critical. This research was funded by the Robert Wood Johnson Foundation. Sharon Long s timewassupportedbythestate Health Access Reform Evaluation, a national program of the Robert Wood Johnson Foundation based at the State Health Access Data Assistance Center at the University of Minnesota. Any opinions and conclusions expressed 1246 HEALTH AFFAIRS JUNE :6

6 herein are those of the authors and do not necessarily represent the views of the Robert Wood Johnson Foundation, the Urban Institute, or its sponsors or trustees. The authors thank Allison Cook, Alshadye Yemane, Karen Stockley, and Emily Lawton of the Urban Institute s Health Policy Center for helpful suggestions and assistance. NOTES 1 Turner J, Boudreaux M, Lynch V. A preliminary evaluation of health insurance coverage in the 2008 American Community Survey [Internet]. Washington (DC): U.S. Census Bureau; 2009 Sep 22 [cited 2010 Apr 12]. Available from: hlthins/2008acs_healthins.pdf. This paper used data for children ages 0 17, whereas the Census Bureau s analysis used data for children ages Long SK. On the road to universal coverage: impacts of reform in Massachusetts at one year. Health Aff (Millwood). 2008;27(4): w Long SK, Stockley K, Yemane A. Another look at the impacts of health reform in Massachusetts: evidence using new data and a stronger model. Am Econ Rev: Papers & Proceedings. 2009;99(2): We are aware of only one study that has looked at the impacts of health reform on children. Yelowitz A, Cannon MF. The Massachusetts health plan: much pain, little gain [Internet]. Washington (DC): Cato Institute; 2010 Jan 19 [cited 2010 Apr 12]. Available from: pub_id= The Appendix is available by clicking on the Appendix link in the box to the right of the article online. 6 For a more detailed summary of the changes introduced under health reform in Massachusetts, see McDonough JE, Rosman B, Phelps F, Shannon M. The third wave of Massachusetts health care access reform. Health Aff (Millwood). 2006;25(6):w Long SK.Who gained the most under health reform in Massachusetts? [Internet]. Washington (DC): Urban Institute; 2008 Oct 15 [cited 2010 Apr 12]. Available from: 8 Dubay L, Kenney G. Expanding public health insurance to parents: effects on children s coverage under Medicaid. Health Serv Res. 2003;38 (5): Hudson JL, Selden TM, Banthin JS. The impact of SCHIP on insurance coverage of children. Inquiry. 2005;42(3): Kenney G, Cook A, Dubay L. Progress enrolling children in Medicaid/CHIP: who is left and what are the prospects for covering more children? [Internet]. Washington (DC): Urban Institute; 2009 Dec 14 [cited 2010 Apr 12]. Available from: policy/url.cfm?id= Wooldridge J. What s new in econometrics? Lecture 10: differencein-differences estimation [Internet]. Cambridge (MA): National Bureau of Economic Research Summer Institute; 2007 [cited 2010 Apr 12]. Available from: diffindiffs.pdf 12 Davern M, Rodin H, Call KT, Blewett LA. Are the Current Population Survey uninsurance estimates too high? An examination of imputation. Health Serv Res. 2007;42(5): Our regression-based impact estimates were virtually the same whether we used this new weight or the original Current Population Survey weight. We note, however, that annual point estimates of the uninsured may vary depending on the weight used. 14 It is not clear where parents would have classified children enrolled in the state-funded Children s Medical Security Plan prior to reform. To the extent that parents reported such coverage as Medicaid, our analysis will not reflect shifts from that plan into MassHealth. 15 Call KT, Davidson G, Sommers AS, Feldman R, Farseth P, Rockwood T. Uncovering the missing Medicaid cases and assessing their bias for estimates of the uninsured. Inquiry ;38(4): In the Current Population Survey, individuals are classified as uninsured only if they report having no coverage at any point over the prior calendar year. However, the uninsurance rate in the Current Population Survey aligns more closely to point-in-time estimates than fullyear estimates. See DeNavas-Walt C, Proctor BD, Smith J. Income, poverty, and health insurance coverage in the United States: 2006 [Internet]. Current Population Reports no. P Washington (DC): U.S. Census Bureau; 2007 [cited 2010 Apr 12]. Available from: pubs/p pdf 17 Heberlein M, Guyer J, Horner D. Weathering the storm: states move forward on child and family health coverage despite tough economic climate [Internet]. Washington (DC): Georgetown University Center for Children and Families; 2009 [cited 2010 Apr 12]. Available from: weathering-the-storm 18 Kaiser Commission on Medicaid and the. A 50 state update on eligibility rules, enrollment and renewal procedures, and cost-sharing practices in Medicaid and SCHIP. Washington (DC): Kaiser Family Foundation; 2004 Jan 2009 Jan. This annual publication is repeated once a year, representing data from Long SK, Stockley K. Health reform in Massachusetts: an update on insurance coverage and support for reform as of fall 2008 [Internet]. Washington (DC): Urban Institute; 2009 Sep 1 [cited 2010 Apr 12]. Available from: 20 Rubin DB. Estimating causal effects from large data sets using propensity scores. Ann Intern Med. 1997;127 (8 pt 2): For this study, we defined family as the child s health insurance unit (HIU), which corresponds to the child, his or her parents or guardians, and his or her siblings under age nineteen. 22 Because these estimates were based on our analytical sample and adjusted survey weights, they are not strictly comparable to those in published tables using Current Population Survey data (such as those in the Kaiser Family Foundation s State- HealthFacts.org site), but the pattern of change is consistent with what has been reported elsewhere. 23 An example of the regression output from the difference-in-differences analysis is provided in Appendix A.4; see Note Our findings are not directly comparable to those reported by the only other study that has looked at the impacts of health reform on children in Massachusetts (Yelowitz and Cannon, Note 4). That study provided little discussion of data and methods, but it appears that there are a number of methodological issues that raise concerns about its findings. For a discussion of those issues, see Long SK. A comment on The Massachusetts health plan: much pain, little gain [Internet]. Washington (DC): Urban Institute; 2010 Feb 2 [cited 2010 Apr 12]. Available from: 25 Urban Institute tabulations of data from the 2009 American Community Survey. JUNE :6 HEALTH AFFAIRS 1247