OBJECTIVE: We examined the association between state Medicaid and State Children's Health Insurance Program (SCHIP) income eligibility and the financial burden reported by low-income families raising children with special health care needs (CSHCN).
SAMPLE AND METHODS: Data on low-income CSHCN and their families were from the National Survey of Children With Special Health Care Needs (N = 17039), with a representative sample from each state. State Medicaid and SCHIP income-eligibility thresholds were from publicly available sources. The 3 outcomes included whether families had any out-of-pocket health care expenditures during the previous 12 months for their CSHCN, amount of expenditure, and expenditures as a percentage of family income. We used multilevel logistic regression to model the association between Medicaid and SCHIP characteristics and families' financial burden, controlling state median income and child- and family-level characteristics.
RESULTS: Overall, 61% of low-income families reported expenditures of >$0. Among these families, 30% had expenses between $250 and $500, and 34% had expenses of more than $500. Twenty-seven percent of the families reporting any expenses had expenditures that exceeded 3% of their total household income. The percentage of low-income families with out-of-pocket expenses that exceeded 3% of their income varied considerably according to state and ranged from 5.6% to 25.8%. Families living in states with higher Medicaid and SCHIP income-eligibility guidelines were less likely to have high absolute burden and high relative burden.
CONCLUSIONS: Beyond child and family characteristics, there is considerable state-level variability in low-income families’ out-of-pocket expenditures for their CSHCN. A portion of this variability is associated with states' Medicaid and SCHIP income-eligibility thresholds. Families living in states with more generous programs report less absolute and relative financial burden than families living in states with less generous benefits.
Our aim was to examine how low-income families' financial burden related to caring for children with special health care needs (CSHCN) is associated with variability in the generosity of state Medicaid and State Children's Health Insurance Program (SCHIP) income-eligibility criteria. Our general hypothesis was that poor and near-poor families (defined here as having a household income at ≤200% of the federal poverty level [FPL]) who lived in states with more generous public benefits would experience lower financial burden, controlling for family demographics, the severity of their child's conditions, and the affluence of their state of residence. We tested this hypothesis by using data from the 2005–2006 National Survey of Children With Special Health Care Needs (NS-CSHCN) and state-level measures of Medicaid and SCHIP income-eligibility standards.
CSHCN use more health care and related services than typically developing children, which increases the likelihood of high financial burden.1,2 Financial burden has been operationally defined for a previous 12-month period in 3 different ways: whether a family had any health expenditures related to the child's special needs; absolute burden (actual expenditures); and relative burden (amount of expenditures relative to income).1–3 High absolute burden is associated with poor child health, being black or Hispanic, being uninsured, and having high family income and socioeconomic status (those who have more, spend more).1–4 High relative burden is associated with poverty. Poor families spend a larger proportion of their household income compared with nonpoor families, although they spend less in absolute dollars.2
Financial burden varies significantly among states as well as among families.3,4 State variability in mean burden could be a result of differences across states' populations. States with healthier populations might have lower mean burden because they require less care. However, Shattuck and Parish3 found that families with similar demographic and child-need characteristics had widely different financial burden depending on their state of residence; state mean annual absolute burden ranged from $562 to $972, and mean relative burden ranged from $14.5 to $32.3 per $1000 of household income after controlling for a range of child- and family-level factors, including child health.
State variability in financial burden could also be a result of differences in the extent of public supports, most notably Medicaid and SCHIP. Families who live in states with more generous benefits might have lower financial burden, especially low-income families targeted by these programs. Indeed, financial burden in families with publicly insured CSHCN is lower among those with public insurance.3,5 However, we are unaware of research that has examined the simultaneous contributions of child-, family-, and state-level policy factors to variability in financial burden.
Understanding whether Medicaid and SCHIP buffer the financial burden of low-income families raising CSHCN is important for several reasons. First, these families spend a disproportionately large share of their limited income on their child's care.2 Second, state Medicaid and SCHIP policy decisions are modifiable. States have considerable leeway in determining program funding and eligibility.6 Third, Medicaid and SCHIP policies affect a substantial number of children. SCHIP enrollment in June 2007 was ∼4.4 million children,7 and ∼1 in 4 children are insured through Medicaid.8 Fourth, advocating for adequate funding of Medicaid and SCHIP is a top priority of both the American Academy of Pediatrics Division of State Government Affairs and American Academy of Pediatrics state chapters.9 Ranking states on malleable policy factors and related family outcomes can help target finite advocacy resources more effectively. This issue is especially salient given that the SCHIP program was reauthorized in February 2009. This reauthorization permits states to cover children in families with an income of up to 300% of the FPL. Fifth, information about the associations between family financial burden and public insurance can help clinicians better understand the challenges that face children's families. Sixth, these findings can help advance our conceptual understanding of the linkages between political context and individual health, which are often discussed but seldom investigated.10–14 Finally, we hope that these findings will stimulate and inform further research into the connections between state policies and child outcomes.
Child- and family-level data were drawn from the 2005–2006 wave of the NS-CSHCN, described elsewhere in this supplemental issue of Pediatrics.15 States' Medicaid and SCHIP income-eligibility guidelines were drawn from the National Academy on State Health Policy.16 State median income values for families with children were drawn from the Annie E. Casey Foundation.17
We focused our study on low-income families, whose incomes were at ≤200% of the FPL. Our final analytic sample included 17039 children from the NS-CSHCN. Table 1 describes these children and their families. We excluded 506 families for whom information on out-of-pocket costs was missing.
Three measures of financial burden were based on NS-CSHCN questions that asked families to report how much they paid during the previous 12 months for their child's medical care: $0, $1 to $249, $250 to $500, $501 to $999, $1000 to $5000, or $5001 or more. The definition used for medical care included out-of-pocket payments for a variety of health-related needs including copayments, medications, special foods, and durable equipment but excluded insurance premiums and reimbursable costs.
The first dependent variable was whether the family had any expenditures. The second dependent variable (computed for families with expenditures greater than $0) was a 3-category indicator of the amount of absolute burden: $1 to $249, $250 to $500, or $501 or more. The third dependent variable was a 3-category measure of relative burden (total expenditures as a percentage of family income): less than 1%, 1% to 3%, or more than 3% of income. This latter measure was created by using a multistep process. First, we transformed the survey's categorical measure of expenditures into dollars by using the midpoints of the first 4 strata. For those who reported $5000 or more, we used the median out-of-pocket health expenditure for CSHCN who had more than $5000 in expenditures from the 2005 Medical Expenditure Panel Survey ($5920).18 We obtained a measure of families' median household income through direct correspondence with the National Center on Health Statistics (S. J. Blumberg, PhD, National Center for Health Statistics, “Median Income From the National Survey of Children With Special Health Care Needs Stratified by State, Household Size, and Federal Poverty Level,” 2007, personal written communication). Then, we calculated a measure of relative burden as the ratio of dollars spent on care to income. Finally, we created our 3-category relative-burden indicator from the burden/income ratio.
A categorical variable, rather than the burden/income ratio variable itself, was used to model relative burden, because neither income nor burden were themselves available, only intervals representing ranges in which each participants' income and burden fell. The thresholds of less than 1%, 1% to 3%, and >3% were selected, because exploratory analyses demonstrated that the models did not run successfully with more than 3 categories, and nonlinear models of categorical data perform better when the dependent variable is balanced.
Covariates included an indicator of household income relative to the FPL (income < 100% or between 100% and 200% of the FPL); binary indicators of the child's race (white or nonwhite, which included children reported as being black, Asian, multiracial, Native American, Aleut, or Pacific Islander) and Hispanic ethnicity (yes or no); child's age, mean centered within each state; parent's high school drop-out status (yes or no); parent ratings of the severity of the child's condition (minor, moderate, or severe); and the stability of the child's needs related to his or her condition (needs are or are not stable). Finally, measures of insurance coverage and service participation included: child participated in early intervention or special education services regulated by the Individuals With Disabilities Education Act; child was ever uninsured in previous 12 months; child had public health insurance only; child had private and public health insurance; and child was currently uninsured.
One state-level covariate was modeled: median income for families with children aged 17 or younger in 2005, measured in thousands of dollars.17
Two state-policy variables were investigated: (1) the Medicaid income-eligibility standard for children aged 6 to 18 years; and (2) the SCHIP income-eligibility standard for children aged 6 to 18 years. The income-eligibility standards were expressed in multiples of the FPL. The Medicaid income-eligibility standard ranged from 1 to 2.25 times the FPL. The SCHIP standard ranged from 1.4 to 3.5 times the FPL, excluding Tennessee, which did not have an SCHIP program.
Analytic Strategy: 2-Part Hierarchical Generalized Linear Models
A 2-part model distinguished families with no financial burden from those who reported a burden of more than $0.19 In the first part, we used logistic regression to model the probability of having any out-of-pocket costs. In the second part, which excludes families with no out-of-pocket costs, we modeled the 3-category absolute financial-burden variable and then the 3-category relative-burden variable. These part-2 models used multinomial logistic regression to estimate the probability of 2 higher categories of burden relative to the lowest: $250 to $500 and $501 or more vs $1 to $249 for absolute financial burden, and 1% to 3% of income and more than 3% vs less than 1% of income for relative financial burden.
Multilevel regression models are appropriate for nested data. In this case, families are nested in states. Nested data can lead to inference problems if not analyzed by using appropriate methods that correctly adjust SEs for the correlation between families who lived in the same state. A multilevel model facilitates examination of the correlates of financial burden at both the family and state levels. A logistic regression modeled in a multilevel data environment is known as a hierarchical generalized linear model.20 As in regular logistic regression, coefficients can be transformed into odds ratios that describe a family's odds of having the specified level of burden.
We used an informed model-fitting process in Mplus 3.1, entering individual correlates of burden first and then entering state median income and the policy variables.21,22 An approximate measure of the amount of state-level variance explained by the state covariate and independent variables, calculated as the percentage change between the “full” model (with all state-level covariates) and null model (having only individual-level covariates) is reported.
Because of missing data on several NS-CSHCN individual-level variables, our analyses were conducted on multiply imputed data we created by using SAS Proc MI.23 A macro written in SAS combined estimates from Mplus.24–28
Weighting and Variance Adjustment
We are unaware of statistical software that simultaneously accommodates multilevel data and the variance adjustment required for stratified random sampling. Simulations we conducted showed that both multilevel analysis and variance-adjusted analysis resulted in properly corrected and similar SEs for individual-level covariates but variance adjusted analyses produced insufficiently corrected SEs for state-level variables. Therefore, we used multilevel data analysis because of the nested nature of the data. All results were weighted to the US Census estimates for the age, gender, race, and ethnicity of the population.
Table 1 summarizes the dependent measures of financial burden. Among low-income families, 61% reported having some financial burden (out-of-pocket costs > $0). Of those reporting any burden, 30% reported absolute expenditures between $250 and $500, whereas 34% reported expenditures that exceeded $500 for the previous 12-month period. Twenty-seven percent of those who reported any out-of-pocket costs had relative burden that exceeded 3% of their total household income.
Table 2 presents the percentage of low-income families within individual states who reported having any burden, absolute burden of more than $500, and high relative burden (expenditures of >3% of total income). Table 2 also presents the state rankings for the percentage that had high relative burden (>3% of total household income). There was considerable variability in the proportion of states' low-income populations with any burden, which ranged from 33.5% in the District of Columbia to 84.4% in Utah. A wide range of families had an absolute annual burden of more than $500, from 7.3% in the District of Columbia to 35.2% in Utah. Finally, the percentage with high relative burden ranged from 5.6% in the District of Columbia to 25.8% in Montana. In 34 states, at least 15% of the state's low-income population with CSHCN had spending that exceeded 3% of income. It is notable that 25% of the families who reported any burden had expenditures that exceeded 5% of their total income.
Table 3 lists the results of the regression models. Although the child- and family-level findings are reported in the table, we focus here on the state-level results. Controlling for state median income for families with children and all child and family covariates, states' SCHIP and Medicaid income-eligibility standards were not significantly associated with the probability of having any out-of-pocket expenditures. The full model, including all state covariates, explained 11% of the state-level variance (column 2).
As compared with families who reported out-of-pocket costs of less than $250, families who lived in states with higher Medicaid and SCHIP income-eligibility guidelines had significantly lower odds of absolute burdens between $250 and $500, by 23% and 11%, respectively (column 3). In other words, the predicted odds of having an absolute burden between $250 and $500 for a family in a state with a Medicaid-eligibility threshold of 200% of the FPL was 77% of the corresponding predicted odds for a family who lived in a state with an eligibility threshold of 100% of the FPL. The full model with the 3 state variables explained 57% of the state-level variance.
Controlling for all model covariates, families who lived in states with more generous Medicaid and SCHIP income-eligibility guidelines had significantly lower odds of having out-of-pocket costs of more than $500, by 30% and 17%, respectively. The state covariates explained 26% of state-level variance (column 4).
Finally, as compared with families with lower relative burden (<1% of total income), those who lived in states with higher Medicaid and SCHIP income-eligibility standards had significantly lower odds of having high relative burden (>3% of total household income), by 23% and 32%, respectively. We infer from this finding that in a comparison between states in which eligibility differed by 1 multiple of the FPL, families in states with the higher Medicaid-eligibility threshold were 77% as likely, with respect to odds, to have a relative burden that exceeded 3% of income, and those who lived in states with the higher SCHIP-eligibility threshold were 68% as likely to have a relative burden of 3% or higher. This full model explained 30% of state-level variance (column 5).
After controlling for child- and family-level characteristics and state median income for families with children, there was persistent and marked state-level variability in the magnitude of financial burden that low-income families faced in raising their CSHCN. These results support previous research that revealed significant state-level variability in financial burden3 and has indicated that a substantial amount of this variability is associated with states' Medicaid and SCHIP program characteristics.
The important contribution of this study is the finding that relative and absolute burden tend to be lower in states with more generous Medicaid and SCHIP income-eligibility standards. That is, low-income families who live in states with higher income-eligibility guidelines for their Medicaid and SCHIP programs tend to have less burden, both in total terms and relative to their household income, as compared with families who live in states with more restrictive income-eligibility guidelines.
The percentage of families in this low-income sample with any out-of-pocket costs (61%) was lower than that found in a general-population sample of CSHCN (82.5%).3 We do not know the reasons for this finding. Previous research has found that CSHCN who live in poverty are at increased risk of unmet needs for both routine and specialty care.29 It is possible that the lower proportion of poor families with any out-of-pocket costs is a result of delayed and foregone care rather than a lower prevalence of need for care, but additional research is required to fully understand this issue.
The elevated rates of high relative burden (out-of-pocket expenditures that exceeded 3% of family income) are particularly troubling given that our analyses were restricted to the population of families with household income at or below twice the FPL. This low-income population is specifically targeted for assistance by Medicaid and SCHIP. Yet, our findings indicate that despite their eligibility for benefits, these families reported significant levels of financial burden, burden that is associated with less generous state Medicaid and SCHIP programs. Given other evidence that families raising children with disabilities face exceptionally high rates of deprivation and material hardship,30 which likely has a deleterious effect on the children's well-being, policy makers should consider ways to strengthen Medicaid and SCHIP to reduce the financial burdens that these families shoulder.
This study's limitations must be considered to fairly interpret the results. First, these analyses are correlational, and we cannot infer causality between state programs and family financial expenditures. Second, the ordinal measures of household income and families' expenditures may not fully capture a level of detail that would ideally inform policy debates. Third, we were unable to model parental employment, because it was not measured in the NS-CSHCN. However, parental employment is strongly associated with insurance status31 and financial burden.32
A number of important strengths offset the study's limitations. First, the sampling design of the NS-CSHCN resulted in a representative sample of CSHCN from each state. Second, the use of multilevel regression enabled us to examine both individual- and state-level public health program characteristics that are correlated with families' out-of-pocket spending for their CSHCN. To the best of our knowledge, this is the first such study of its kind.
This research used an innovative methodologic approach to examine the association between state-policy characteristics and the financial burdens that low-income families face in raising CSHCN. The inverse relationship found between the generosity of state health insurance eligibility criteria and families' financial burdens suggest that these programs buffer the effects of raising children whose health care needs can often be expensive.
As we write this, the state economies are in a recession that is projected to be deep and difficult. Most states are experiencing budget shortfalls and total state budget gaps for fiscal year 2009–2010 are currently projected to exceed $230 billion.33 To lessen these shortfalls, many state governments are looking to cut their Medicaid programs. Indeed, 25 states made cuts in their Medicaid programs after their 2009 state budgets had been passed, and 25 states have also proposed Medicaid cuts to their 2010 budgets.33 Our results indicate that such cuts may have a particularly detrimental effect on the financial well-being of low-income families raising CSHCN.
- Accepted August 3, 2009.
- Address correspondence to Susan L. Parish, PhD, MSW, 325 Pittsboro St, CB 3550, Chapel Hill, NC 27599-3550. E-mail:
Financial Disclosure: The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Newacheck PW, Inkelas M, Kim SE. Health services use and health care expenditures for children with disabilities. Pediatrics.2004;114 (1):79– 85
- ↵Shattuck PT, Parish SL. Financial burden in families of children with special health care needs: variability among states. Pediatrics.2008;122 (1):13– 18
- ↵Ross DC, Cox L. Making It Simple: Medicaid for Children and CHIP Income Eligibility Guidelines and Enrollment Procedures. Menlo Park, CA: Kaiser Family Foundation; 2000
- ↵Smith V, Rousseau D, Marks C, Rudowitz R. SCHIP Enrollment in June 2007: An Update on Current Enrollment and SCHIP Policy Directions. Washington DC: Kaiser Commission on Medicaid and the Uninsured; 2008
- ↵Kaiser Commission on Medicaid and the Uninsured. Health Coverage of Children: The Role of Medicaid and SCHIP. Washington, DC: Kaiser Commission on Medicaid and the Uninsured; 2007
- ↵American Academy of Pediatrics. 2007 state legislation report. Available at: www.aap.org/advocacy/statelegrpt.pdf. Accessed October 21, 2008
- Andersen RM, Davidson PL. Improving access to care in America: individual and contextual indicators. In: Andersen RM, Rice TH, Kominski GF, eds. Changing the U.S. CA: John Wiley & Sons, Inc; 2007:3– 32
- Jeffrey AE, Newacheck PW. Role of insurance for children with special health care needs: a synthesis of the evidence. Pediatrics.2006;118 (4). Available at: www.pediatrics.org/cgi/content/full/118/4/e1027
- ↵Institute of Medicine, Committee on the Consequences of Uninsurance. Coverage Matters: Insurance and Health Care. Washington, DC: National Academies Press; 2001
- ↵Kogan MD, Strickland BB, Newacheck PW. Building systems of care: findings from the National Survey of Children With Special Health Care Needs. Pediatrics.2009;124 (suppl 4):S333– S336
- ↵National Academy for State Health Policy. Income eligibility levels and costs sharing for children in Medicaid and SCHIP and other populations covered with SCHIP funds: July 2005. Available at: www.nashp.org/Files/Elig_and_cost_sharing_Aug_2005.pdf. Accessed October 9, 2009
- ↵Annie E. Casey Foundation. Median family (with child) income: 2005. Available at: www.kidscount.org/datacenter/compare_results.jsp?i=340&dt=3&rt=2&yr=6&s=a&rtype=&x=140&y=6. Accessed October 9, 2009
- ↵Agency for Healthcare Research and Quality. Medical Expenditure Panel Survey, person-level file. Available at: www.meps.ahrq.gov/mepsweb/data_stats/meps_query.jsp. Accessed September 2, 2008
- ↵Duan N, Manning W, Morris C, Newhouse J. A comparison of alternative models for the demand for health care. J Bus Econ Stat.1983;1(2):115– 126
- ↵Raudenbush S, Bryk A. Hierarchical Linear Models: Applications and Data Analysis Methods. 2nd ed. Thousand Oaks, CA: Sage; 2002
- ↵Muthén BO, Muthén L. Mplus User's Guide. Los Angeles, CA: Muthén and Muthén; 2004
- ↵Snijders TAB, Bosker RJ. Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling. London, United Kingdom: Sage; 1999
- ↵Schafer JL. Analysis of Incomplete Multivariate Data. Boca Raton, FL: Chapman Hall/CRC; 1997
- ↵Rose RA, Fraser MW. A simplified framework for using multiple imputation in social work research. Soc Work Res.2008;32 (3):171– 178
- Von Hippel PT. Regression with missing Ys: an improved strategy for analyzing multiply imputed data. Sociol Methodol.2007;37 (1):83– 117
- ↵Rubin DB. Multiple Imputation for Nonresponse in Surveys. New York, NY: Wiley; 1987
- ↵Mayer ML, Skinner AC, Slifkin RT; National Survey of Children With Special Health Care Needs. Unmet need for routine and specialty care: data from the National Survey of Children With Special Health Care Needs. Pediatrics.2004;113 (2). Available at: www.pediatrics.org/cgi/content/full/113/2/e109
- ↵Parish SL, Rose RA, Andrews ME, Grinstein-Weiss M, Richman EL. Material hardship among U.S. families raising children with disabilities. Except Child.2008;75 (1):71– 92
- ↵NeNavas-Walt C, Proctor BD, Lee CH. Income, Poverty, and Health Insurance Coverage in the United States. Washington, DC: US Census Bureau; 2005
- ↵National Governor's Association; National Association of State Budget Officers. The Fiscal Survey of States. National Association of State Budget Officers, ed. Washington, DC: National Governors Association and the National Association of State Budget Officers. 2009
- Copyright © 2009 by the American Academy of Pediatrics