The Effect of Lowering Public Insurance Income Limits on Hospitalizations for Low-Income Children

We conducted a retrospective cohort study using the 2014 Agency for Healthcare Research and Quality’s State Inpatient Databases (SID). The SID are a set of state-based all-payer inpatient databases for all hospitalized patients.¹⁴ The SID do not include information about individual family income. Therefore, states were included only if the zip code for patients’ home residence was available (14 states; Table 1). To obtain the percentage of families living at various thresholds of the FPL (<100%, 100%–199%, and 200%–299%), and thus the percentage of children insured by public insurance, we geocoded each patient’s zip code to the US Census Bureau’s 2014 American Community Survey (ACS).¹⁵ This procedure is similar to previous work in which zip codes were used as a proxy for household income to estimate the effects of income on pediatric health services use.¹⁶ The FPL threshold for state-level Medicaid and CHIP eligibility in 2014 was collected from the Kaiser Family Foundation’s Commission on Medicaid and the Uninsured.¹⁷ 

We included all pediatric (age <18 years) hospitalizations in the 14 states from January 1, 2014, to December 31, 2014, with public insurance as the primary payer. Hospitalizations with missing or erroneous zip codes were excluded (0.1%).

The main exposure was lowering public insurance eligibility thresholds to 300%, 200%, and 100% of the FPL for children currently insured through public insurance.

The main outcomes were the number, estimated costs, and demographic characteristics of children whose hospitalizations would become ineligible under each of the 3 different public insurance income eligibility thresholds.

Normal newborns were defined as any hospitalization with a principal diagnosis of birth (International Classification of Diseases, Ninth Revision, Clinical Modification) and a length of stay of 3 days or less. Age groups were then defined as normal newborn, other infants <1 year, children 1 to 4 years, 5 to 9 years, and 10 to 17 years. Race and ethnicity were categorized as non-Hispanic white, non-Hispanic African American, Hispanic, other, and missing. Hospitalizations missing race and/or ethnicity were primarily from Nebraska, which does not report race and/or ethnicity. Hospitalizations for children with complex chronic conditions (CCCs) were determined by using a previously established set of International Classification of Diseases, Ninth Revision, Clinical Modification codes and represent conditions that are expected to last longer than 12 months and are associated with high rates of morbidity and/or mortality.^18,19 Rurality for each patient was determined by linking each patient’s home zip code to the rural-urban commuting area code.²⁰ Hospitalization costs were estimated from billed hospital charges by using hospital-specific cost-to-charge ratio files.²¹ 

For states where the public insurance eligibility limit exceeded 300% of the FPL (n = 6 of 14 states), we lowered the FPL limit to 300%. Next, we calculated the number, percent, estimated costs, and characteristics of hospitalizations in 2014 that would have become ineligible on the basis of that new public insurance income eligibility threshold. We repeated this procedure for public insurance income eligibility limits of 200% of the FPL (lowering eligibility limits for 13 of 14 states) and 100% of the FPL (lowering eligibility limits for all included states).

Categorical variables were summarized with frequencies and percentages, whereas continuous variables were summarized with medians with interquartile ranges (IQRs). Linear relationships between state-level measures were assessed with linear regression. The actual family household income level for each patient was unknown. However, the ACS provided the percentage of households living at different levels of the FPL (<100%, <200%, and <300%) in each patient’s residential zip code. We performed 1000 simulations for each FPL eligibility scenario (100%, 200%, and 300%), and during each simulation, we randomly set each hospitalization to be from a child living above or below the eligibility limit using a Bernoulli trial, with the probability being the proportion of children in the patient’s zip code living at less than the eligibility limit (Supplemental Fig 2). After each simulation, the number and percent of 2014 hospitalizations with public insurance as the primary payer that would have been ineligible for public insurance (ie, over each FPL scenario limit) and the characteristics of these hospitalizations (estimated costs, clinical characteristics, and demographics) were determined as stated above. The results of simulations were summarized with medians and IQRs.

All statistical analyses were performed by using SAS v 9.4 (SAS Institute, Inc, Cary, NC), and P values <.05 were considered statistically significant.

The Office of Research Integrity at Children’s Mercy Hospital deemed this study exempt from institutional board review.

The 14 states examined in this study included 30.6% of family households within the United States. Of included families, 43.1% lived below 300% of the FPL, 27.2% below 200% of the FPL, and 11.2% below 100% of the FPL (Supplemental Fig 3). Six states had public insurance eligibility limits >300% of the FPL, 13 states had limits >200% of the FPL, and all 14 states had limits >100% of the FPL. Public insurance eligibility limits differed considerably across states (Table 1). Public insurance FPL eligibility limits tended to be lower in states with a greater percentage of families below 300% of the FPL (Fig 1; P = .01).

After 1000 simulations, we identified 7 994 484 children as eligible for public insurance in the 14 included states, equaling 21.3% of publicly insured US children. For the 795 427 pediatric hospitalizations for this sample of children, total estimated hospitalization costs were calculated at $5.2 billion.

In 7 states, 2014 public insurance eligibility exceeded 300% of the FPL. By reducing public insurance eligibility to 300% of the FPL in those states, the number of hospitalizations currently reimbursed by public insurance that would no longer be eligible was ∼155 000 (45.5%–58.3% of hospitalizations by state; Table 2, Supplemental Table 4). These reductions would exclude at least 1 hospitalization for ∼144 000 children and their families. Median per-hospitalization estimated costs associated with these ineligible hospitalizations ranged from $5991 to $10 634, accumulating $1.2 billion in estimated costs.

The 2014 eligibility limits exceeded 200% of the FPL in 13 of the 14 states. (Only 1 state included in the study, Arizona, has an eligibility level below 200% of the FPL.) If eligibility limits were reduced to 200% of the FPL, ∼440 000 hospitalizations would no longer be covered by public insurance (61.2%–75.1% of hospitalizations by state). Nearly 412 000 children and their families would be responsible for at least 1 hospitalization under these changes. Median estimated costs of these ineligible hospitalizations varied from $2854 to $10 515, totaling $3.1 billion in estimated costs.

The 2014 eligibility limits of all 14 included states exceeded 100% of the FPL. If the FPL eligibility threshold of those states were reduced to 100% of the FPL, the number of hospitalizations currently reimbursed by public insurance that would no longer be eligible would increase to ∼650 000 (79.8%–90.6% of hospitalizations by state). Close to 608 000 children and their families would be affected 1 or more times by these reductions. The median per-hospitalization estimated costs of ineligible hospitalizations remained relatively stable ($2859–$10 434), but the total estimated costs summed $4.4 billion.

The entire study population of hospitalized, publicly insured children was generally young, with a plurality (38.0%) being non-Hispanic white (20.2% were non-Hispanic African American and 24.5% were Hispanic). Most hospitalizations were for public insurance recipients living in urban settings (84.2%; Table 3). The demographics of patients who would lose public insurance under lowered eligibility thresholds differed only slightly across thresholds. For instance, when eligibility levels were reduced to 300% of the FPL, newly ineligible hospitalizations were 40.1% non-Hispanic white (compared with 38.0% under current thresholds).

Hospitalization types also differed only slightly for each reduction in public insurance eligibility limits. For example, when eligibility levels were reduced to 300% of the FPL, the greatest proportion of hospitalizations were for normal newborns (54.8% compared with 53.2% under current eligibility thresholds). The proportion of hospitalizations for children with CCCs was also relatively equal (12.4% compared with 12.9%) to current eligibility thresholds (12.4% compared with 12.9%). These proportions did not change substantially with reductions of public insurance eligibility to 200% of the FPL and 100% of the FPL.

Normal newborns who are currently publicly insured represent the majority of children affected by reductions of income eligibility requirements at all levels (300%, 200%, and 100% of the FPL). For example, the number of normal newborn hospitalizations covered by public insurance would decrease by 96 375 (23%) if income eligibility levels were decreased to 300% of the FPL (7 states), 249 317 (59%) if decreased to 200% of the FPL (13 states), and 355 828 (84%) if reduced to 100% of the FPL (14 states; Supplemental Tables 5 and 6). Estimated costs associated with normal newborn hospitalizations would equal $149 million, $303 million, and $423 million for reductions to 300% of the FPL, 200% of the FPL, and 100% of the FPL, respectively.

In this study of 795 427 hospitalizations of children with public insurance across 14 geographically diverse states, we found that public insurance would no longer reimburse half or more of currently covered hospitalizations and their associated estimated costs under each of 3 scenarios of lowered income eligibility criteria and would result in up to $4.4 billion of estimated hospital costs ineligible for public insurance reimbursement. With these findings, we suggest that reducing public insurance eligibility may potentially result in a large number of children who are currently publicly insured having to either purchase commercial insurance or become uninsured. With these findings, we also predict substantial shifts in costs to lower income families, commercial insurance (if obtainable by families), and/or the health care institutions that serve them.

Under 3 scenarios of decreased income eligibility criteria, loss of public insurance would disproportionately affect healthy newborns. Although they vary by state, existing income eligibility criteria for newborns greatly exceeds that for any other age group, and current law (Children’s Health Insurance Program Reauthorization Act of 2009, Public Law 111-3) allows for fast-tracking public insurance coverage for newborn infants until 12 months provided that their mothers received or were eligible for Medicaid or CHIP coverage during pregnancy.^22,–24 Newborn hospitalization is 1 of the fastest-rising costs in pediatric care.²⁵ In our study, hospitalization costs for normal newborns ranged from ∼$700 to $2000. Depending on whether a family can obtain commercial insurance for their child after losing public insurance eligibility, the cost of the newborn hospitalization (or potentially the charges, which generally exceed the costs) may represent a significant financial strain for those families. Across all types of hospitalizations, however, the estimated costs associated with a hospitalization ranged from ∼$3000 to $10 000. Given that 200% of the FPL for a family of 4 is $59 640, the cost of a single hospitalization may represent ∼5% to 15% of a family’s annual income.^26,27 This may represent an untenable financial burden.

Decreases in health care coverage for children and families would also have implications for health care institutions serving lower income children. The specific impact of those changes would depend on whether newly ineligible children would be able to obtain commercial insurance as well as the specific health benefits of those commercial plans. What is clear is that a large amount of reimbursement (albeit reimbursement that is typically below costs) would be at stake. For the 14 states examined in this study, estimated hospitalization costs for children who would lose public insurance eligibility equaled $1.2 billion when income eligibility was reduced to 300% of the FPL, $3.1 billion at 200% of the FPL, and $4.4 billion at 100% of the FPL. Consequently, changes in public insurance income eligibility criteria may place health care institutions at greater financial risk, especially safety-net hospitals, which already operate at narrower financial margins,^28,29 and other hospitals serving large numbers of newborns and low-income families.³⁰ 

Funding of public insurance comprises a large proportion of state budgets.¹² Outside of our analysis of public insurance eligibility changes, we found that state-level public insurance FPL eligibility criteria were indirectly proportional to the percentage of state population near poverty (ie, states with higher proportions of children in poverty had lower public insurance eligibility thresholds). It is likely that higher income eligibility thresholds in states with many near-poor residents would result in higher marginal costs compared with higher thresholds in states with fewer near-poor residents. State government budgets are more restrictive, including legal requirements for a balanced budget and debt limits in many states.³¹ Therefore, any reductions in state-level funding for pediatric public insurance programs may result in lowering of eligibility thresholds.

This study has several limitations. First, our sample was limited to the 14 states in the SID that reported zip codes. However, these states represented approximately one-third of US families and were diverse in their geography, population characteristics, and public insurance FPL eligibility criteria. Second, we did not account for public insurance eligibility for people with disabilities that may have resulted in some children retaining coverage. Although small in number, children with special health care needs represent a large fraction of child health care spending and may have more frequent and complex admissions. The costs specifically attributable to medical complexity are relevant but difficult to validly assess within this data set. Third, CHIP income eligibility limits equaled or exceeded age-specific limits for Medicaid and/or Title XXI CHIP for children of all ages in all states except Iowa (separate CHIP limits 317% of the FPL versus age-specific Title XXI CHIP funding for infants aged 0–1 year 380% of the FPL). As a result, we likely underestimate the effects of a reduction in benefits for infants aged 0 to 1 year in Iowa but likely not in other states. Fourth, although we used cost-to-charge ratio files provided specifically for the SID from the Healthcare Cost and Utilization Project, estimated hospitalization costs varied significantly on the basis of states. Information about hospitalizations within the SID is collected by states but may include different entities within states, which may or may not include observation status.^32,33 It is also likely that differences in cost of living contribute to differences in hospitalization costs (eg, the wage index for Washington is among the highest at 1.13, whereas Iowa is much lower at 0.89).³⁴ Last, given the nature of our data set, individual child-level income characteristics are unknown and were based on zip codes. Zip codes have been shown to be a reasonable proxy for socioeconomic factors but may not fully reflect the characteristics of all individuals within a given area.^35,–38 

Eligibility scenarios in which reductions in public insurance eligibility thresholds for children are simulated resulted in the loss of public insurance coverage for a large number of pediatric hospitalizations. If families of children losing public insurance were unable to obtain commercial insurance, the health of those children and the economic well-being of their families and health care institutions would be greatly impacted.

ACS

American Community Survey

CCC

complex chronic condition

CHIP

Children’s Health Insurance Program

FPL

federal poverty level

IQR

interquartile range

SID

State Inpatient Databases