Data Needs for the State Children's Health Insurance Program (2002)

Since SCHIP was established as a state program, each state has developed its own approach in attempting to reduce the number of uninsured children, taking into account its programs that were already in existence through Medicaid and private insurance when they implemented the program. However, SCHIP funds are allocated to the states on the basis of a formula that depends on the proportion of all uninsured children from low-income families who are residents of a given state. Thus, eligibility data for each state are important for at least four purposes: (1) to allocate the total national funding for SCHIP to the states; (2) to measure the success, or lack thereof, of enrollment and reenrollment efforts within a given state; (3) to allocate SCHIP funds across areas (e.g., counties) within a given state; and (4) to present the state legislature with valid estimates of the numbers of children who are eligible for coverage. The latter is critical, since state legislatures must match federal SCHIP funding and be persuaded that the program is effectively meeting its goals. If the program underestimates the number of eligible children, it may be difficult to persuade the legislature to appropriate the matching funds needed to pay for the insurance for the eligible children. Given the way in which the system currently operates, such a deficit cannot be addressed until the following year.

The SCHIP allotments to states for a given year are based on a complex formula that uses the average of the number of low-income children from the March supplement of the Current Population Survey for the previous three years, adjusted to take into account factors accounting for dif

ferences in health care costs among the states. Several other surveys, including the National Health Interview Survey and the State and Local Area Integrated Telephone Survey, have been used to estimate the number of children eligible for both SCHIP and Medicaid. Differences in methodology and in question format make for substantial variations in these estimates. In his presentation at the workshop, Thomas Selden concluded that each of the national surveys has positive and negative attributes and there is no single benchmark against which to compare results. The challenges include having to estimate eligibility from income data that do not map directly to state eligibility criteria, either in terms of the eligibility criteria for the state’s Medicaid program, the definition of insurance, the definition of countable income, or the length of the reference period for measuring income or insurance coverage; respondents’ difficulties in accurately responding to survey questions on income; differences in income thresholds across the states; nonresponse to the survey; and inadequate sample sizes to estimate the number of eligible children in all but the largest states, even when multiple years of survey data are combined.¹

SCHIP is a state program with only general guidelines from the federal government, thus eligibility criteria vary from state to state. In particular, states were required to maintain Medicaid eligibility for all children who were eligible prior to June 1997 and to use SCHIP to provide insurance for children whose family income exceeds the level for Medicaid but whose income falls below the level set by the state as the income ceiling for SCHIP. The assessment of eligibility for SCHIP is complicated by the relationship between family income and the ceiling for Medicaid eligibility and SCHIP eligibility. The income band for eligibility for SCHIP is sufficiently narrow (in some states, for example, between 100 percent and 140 percent of the federal poverty level) that its boundaries are very difficult to identify from survey data, particularly in view of the difficulty in allocating survey responses to income questions and the difficulty that respondents have in supplying the information with the precision needed to accurately model eligibility. In most states, when family income changes only slightly, eligi

bility can shift from SCHIP to Medicaid, or on the other end of the scale, to private insurance or to none at all.

In many states the income eligibility rules for SCHIP are so complicated that survey respondents have difficulty supplying the information needed to determine eligibility. For example, some states attempt to more closely approximate disposable income by allowing applicants to deduct such expenses as child care and child support in calculating income for purposes of determining eligibility. Survey responses are further complicated by respondents giving inconsistent answers to questions about the time period during which they have been uninsured. This creates a problem, because in some states SCHIP requires at least a 6-month period without insurance for eligibility. All of these problems make it difficult to obtain reasonable national data on SCHIP eligibility or enrollment or to make valid comparisons among states.

The difference between what families see as their income and what states count as income for the purpose of eligibility determination, combined with the narrow band between Medicaid and SCHIP eligibility, have resulted in many families who are inquiring about SCHIP finding out that they are actually eligible for Medicaid. The federal law that established SCHIP requires that those applying for SCHIP be screened to determine whether they are eligible for Medicaid. If they are Medicaid eligible, they will be referred to that program for potential enrollment. In her presentation, Kristen Testa described SCHIP as an important motivator for getting individuals enrolled in Medicaid. Pamela Paul-Shaheen cited the example of a family’s calling the 800 number for the Michigan SCHIP thinking that they were eligible, but finding out that instead they were eligible for the Medicaid program. To deal with this problem, the Michigan program located the offices handling Medicaid and SCHIP next to each other so that applications coming to one office that qualified for the insurance provided by the other were simply handed over. This solution resulted in a substantial increase in enrollments by much more quickly identifying proper eligibility.

Furthermore, using national surveys is problematic because states vary in what they consider countable income and assets. Lisa Dubay’s simulation of the eligibility rules in the various states required many different formulations of a large number of variables, including age of child, family

size, work status of parents, how income is counted, whose income is counted, types of income (pension income, Social Security income, wage income, etc.), child care expenses, employment status, reasons for not working, and a host of other factors. The factors included and the way they are defined vary considerably among states. For example, Steve Norton reported that in New Hampshire the guidelines for specification of income to determine eligibility are much less detailed than those described by Dubay.

Another problem in determining eligibility is the volatility in family income among low-income families. Therefore, as Linda Bilheimer indicated, to understand eligibility data, access to Medicaid and frequent changes in family income must be taken into account. She presented data from a period prior to the SCHIP legislation that indicated tremendous movement among Medicaid, employer-sponsored insurance, other insurance, and being without insurance. In some cases, the insurance status for a given child changed several times during a year. Vicki Grant pointed out that one applicant may have several denials of insurance before acceptance. Since the transitions into and out of a particular insurance category tend to balance out during the year, the use of longitudinal data is necessary to properly interpret the data.

Furthermore, in nationally representative data, even surveys with large household samples like the Current Population Survey, sample sizes at the state level are too small to provide reliable estimates, particularly for the smaller states, and they are certainly too small to provide estimates at the regional or county level. One result has been that when states do attempt to make estimates based on national data, the standard errors can be so large as to make point estimates meaningless.

Many states, frustrated by not being able to use national survey data as a basis for their estimates, have developed their own surveys. Thus, at least 27 states were conducting their own household surveys, each of them using their own questions. Cynthia Shirk pointed out that the fact that so many states are conducting their own surveys using their own methodologies makes it very difficult to compare the resulting data among states and to relate outcomes to program characteristics.

Norton expressed the need to move away from nationally based surveys in favor of state-based surveys to be able to better meet the state’s

needs. In New Hampshire, the legislature wanted data on eligibility at the county level in order to understand in which areas problems needed to be addressed. In Connecticut, as Mary Alice Lee showed, data by town revealed a decrease in enrollment in Hartford. A subsequent survey in Hartford of those leaving the program indicated that the major reason was transfer to employer-based insurance. These experiences emphasize the importance of analyzing the distribution within the state of the uninsured population and movements in and out of SCHIP in order to better understand the reasons for changes.

In its discussion of the issues presented at the workshop bearing on the problems in estimating the numbers of children eligible for SCHIP, the panel drew the following conclusions:

• Better ways are needed to estimate eligibility and insurance coverage status from state and national survey data.

• More work is needed to determine what data elements needed for estimating eligibility are included in each of the national surveys, how they are defined, and how differences in their content and definitions can be reconciled.

• States should explore ways to supplement national sample efforts in such surveys as the Behavioral Risk Factor Sample Survey (BRFSS), the State and Local Area Integrated Telephone Survey (SLAITS), and other national surveys in order to expand the size of their state samples.

• Measurement of SCHIP eligibility can be understood only while simultaneously measuring eligibility for Medicaid. It is necessary to measure both in SCHIP eligibility surveys.