Unfinished Business: Inadequate Health Coverage for Privately Insured Seriously Ill Children

HERE

About 10% of children account for 60% to 70% of children's total health care expenditures (Hobbs, Perrin, & Ireys, 1985;Neff & Anderson, 1995;Newacheck, Hughes, & Cisternas, 1995;Newacheck & McManus, 1998). High medical care use children come from all socioeconomic groups and so have both public and private health coverage. Many children are ineligible on income grounds for public programs. We report on the adequacy of health insurance coverage for a sample of children with serious and rare illnesses treated at the single tertiary care pediatric hospital in Indiana. Our empirical examination indicates a problem of inadequate coverage for privately insured seriously ill children that has been not addressed by recent coverage expansions or federal and state reform legislation targeted at insurer underwriting practices restricting access to private coverage for seriously ill people.
Although several studies describe health insurance coverage for seriously ill children (Aday, 1992;Aday et al., 1993;Centers for Disease Control, 1995;Fox & Newacheck, 1990;Newacheck, 1990), the problem of inadequate insurance coverage for children with serious illness is still poorly documented. We find that although most seriously ill children have health coverage, many with private insurance have coverage deficiencies according to prevailing definitions of health insurance adequacy. Families with an inadequately insured child also often delay uncovered care for the child or, more likely, delay care for another family member. Our exploratory examination of seriously ill children in Indiana leads us to conclude that there is cause for concern and a need for more research regarding the adequacy of coverage for seriously ill children with private coverage even in view of recent legislative reforms.

HEALTH INSURANCE COVERAGE FOR SERIOUSLY ILL CHILDREN
According to U.S. Census data for 1994-1995, just over 10% of children in both Indiana and the nation were uninsured (Liska, Brennan, & Bruern, 1998). The major source of coverage for all children is private health insurance. Of all children in 1994-1995, about 62% had employer-sponsored health coverage and 4% had other private or public health coverage (Liska et al., 1998). In Indiana, 69% of children had employer-sponsored coverage, and 5% had other public or private coverage (Liska et al., 1998).

Private Insurance Programs
There is great variation in benefits and coverage among private health insurance plans, due chiefly to limited regulation of employer-sponsored plans under federal law (Ginsburg, Gabel, & Hunt, 1998;Jensen & Gabel, 1988). Specifically, the Employee Retirement Income Security Act of 1974 (ERISA), which regulates em-ployer-sponsored health plans, preempts most state insurance regulation, including state law protections for plan enrollees. Congress has not yet amended ERISA to ensure comparable consumer protections for ERISA-regulated plans. In addition, many employer-sponsored plans self-insure rather than purchase state-regulated commercial health insurance to avoid state regulation, including state-mandated benefit laws. The availability, affordability, and adequacy of employer-sponsored health insurance have decreased in recent years (Blumberg & Liska, 1996;Bovbjerg, Griffin, & Carroll, 1993;Ginsburg et al. 1998;Jensen, Morrisey, Gaffney, & Liston, 1997). Cost sharing and benefit limits have increased in recent years in many private health insurance plans (Ginsburg et al., 1998;Sullivan & Rice, 1991). The decline in coverage has been most substantial among low-income children (Fronstin, 1995;Holl, Szilagyi, Rodewald, Byrd, & Weitzman, 1995;U.S. General Accounting Office [GAO], 1996a;Weinick, Weigers, & Cohen, 1998), which inspired the Children's Health Insurance Program (CHIP) in 1997.
Several recent state and federal legislative initiatives have tried to expand private health coverage for the seriously ill using so-called mandated coverage. An example is the creation of state high-risk insurance pools to provide coverage for the seriously ill who cannot otherwise obtain health insurance privately. High-risk insurance pools are now available in most states, including Indiana. Federal law, specifically the Comprehensive Omnibus Budget Reconciliation Act (COBRA), requires employers to extend coverage for extensive periods to family members of terminated employees although at the employee's expense. Coverage through state high-risk pools and the COBRA extension can be expensive and limited.
States and the federal government have also enacted legislation to address problems with scope of health insurance coverage. To limit insurer practices that narrow health coverage, such as preexisting conditions exclusions, most states have enacted so-called small market legislation, which limits medical underwriting and other exclusionary insurer practices (GAO, 1995;Oliver & Fiedler, 1997). In 1996, Congress imposed restrictions similar to the states' on federally regulated employer-sponsored plans as well as on independent state-regulated plans to maximize portability of health insurance and to restrain underwriting limits for seriously ill with health coverage (Nichols & Blumberg, 1998). However, policy measures to recast underwriting and other insurance practices have not addressed all problems related to scope of coverage and have done little to control the cost of private health coverage (GAO, 1998;Kuttner, 1997;Oliver, 1999).

Public Insurance Programs
The Medicaid program requires states to insure all children under age 6 in families with incomes up to 133% of the federal poverty income level. States also have the UNFINISHED BUSINESS option of covering children in families with incomes up to 185% of poverty income and most, including the state we studied (Indiana), have expanded coverage for children. In 1994 of children nationally and about 16% of children in Indiana had Medicaid coverage (Liska et al., 1998). Benefits under Medicaid are comprehensive and provide what has been deemed adequate health insurance coverage for eligible children. Many parents elect to take Medicaid coverage in lieu of available private insurance coverage because of Medicaid's free comprehensive benefit package (Cutler & Gruber, 1996;Dubay & Kenny, 1995).
The joint federal-state Children's Special Health Care Needs (CSHCN) program targets poor and near-poor children with serious illnesses who are financially ineligible for Medicaid. CSHCN serves approximately 755,000 children nationally and about 5,000 in Indiana (Aron, Loprest, & Steverle, 1995). State programs vary. Indiana's CSHCN program provides health insurance coverage for 23 medical conditions, and Indiana covers all children families with incomes of up to 200% of the federal poverty level through Medicaid CHIP.
In 1997, Congress established the CHIP to expand health insurance coverage for children with family incomes up 200% of poverty (Rosenbaum, Johnson, Sonosky, Markus, & DeGraw, 1998;Bruen & Ullman, 1998;Weinick et al., 1998). Under CHIP, states have great flexibility in program design, although benefit packages must be comparable to benchmark plans such as the standard Blue Cross and Blue Shield plan for federal employees, the state's health plan for state employees, or the commercial HMO having the largest enrollment in the state. States are also prohibited from imposing preexisting condition exclusions or excluding children on the basis of diagnosis in designing CHIP benefit packages. Indiana covers all children up to 150% of the federal poverty income level through Medicaid or CHIP. Estimates of Indiana children with family incomes below 200% of the federal poverty level range from 129,000 to 143,000 (Indiana Children's Health Insurance Program, 1998). Neither CHIP nor Medicaid directly addresses the issue of insurance generosity for privately insured children.

Operationalizing the Concept of Inadequate Health Coverage
Private health coverage for children has received little attention although many privately insured children have what can be deemed inadequate coverage (Monheit, 1994). About 17% of the privately insured children under age 19 are underinsured based on the actuarial value of their insurance compared to the Blue Cross and Blue Shield standard option plan for federal employees (Short & Banthin, 1995). Approximately 8% are underinsured based on the probable exposure to medical expenses over 10% of family income. Most studies of insurance adequacy focus on the nonelderly population generally (Bashshur, Smith, & Stiles, 1993;Bodenheimer, 1992;Monheit, 1994;Short & Banthin, 1995).
One way to put into practice the concept of inadequate insurance is to consider exposure to high uncovered financial risk. We take into consideration three variables: (a) the probability of incurring an expense for medical care in a year, (b) the size of the expense, and (c) the relation of the expense to income (Farley, 1985). Insurance could also be considered inadequate if it has an actuarial value less than the standard Blue Cross and Blue Shield plan for federal employees or the more generous fee-for-service package proposed in President Clinton's Health Security Act (H.R. 3600, 103d Cong., 1st Sess. [1993]; Short & Banthin, 1995). Finally, one might categorize insurance adequacy based on both the characteristics of the insurance and the insured's perceptions of its adequacy. Insurance could then be said to be inadequate in one or more of the following circumstances: (a)coverage of too few services, (b) excessive out-of-pocket expenditures regardless of family income, and (c) consumer perceived inadequacy of insurance (Bashshur et al., 1993).
We define insurance as inadequate if the characteristics of the health insurance sharply limit coverage compared to the benefit packages of the Medicaid program or conventional private health insurance such as the Blue Cross and Blue Shield standard option plan. In our empirical research insurance is defined as inadequate if it has 1. A permanent preexisting condition exclusion, or 2. No annual out-of-pocket limit on medical expenditures, or 3. At least two of the following limitations: (a) a deductible greater than $1,000; (b) coinsurance rates greater than 20% (or the complete exclusion of coverage) for hospitalization, physician services, or prescription drugs; or (c) a lifetime maximum payout of less than $50,000.
For comparability the indicators of insurance inadequacy we use are prevalent in previous research (Bashshur et al., 1993;Bodenheimer, 1992; U.S. Bipartisan Commission on Comprehensive Health Care [Pepper Commission], 1990; Short & Banthin, 1995). We appreciate the controversy of defining adequacy of coverage based on insurance characteristics rather than on percentage of potential risk of loss to income. Focusing on insurance characteristics is informative for privately insured seriously ill children for several reasons. First, measures based on loss probability estimates are excellent for typifying inadequate insurance in a population but may not be good predictors of inadequate insurance for identified individuals. Second, for most people with moderate incomes, their insurance policy provisions, particularly permanent preexisting condition exclusions or high deductibles, can have a significant impact on net income when use of services is substantial. Third, we can examine whether our definition of inadequacy of coverage correlates with reports in delays in care, in turn illustrating not only the theoretical but also the practical significance of defining adequacy based on insurance characteristics for the families of children with serious illness.

METHOD
Because few seriously ill people appear in samples of the general population, previous researchers using random samples have had trouble detecting health insurance related problems specific to the seriously ill (Monheit, 1994). We focus on already ill children rather than a general population of children to obtain a sufficient number of seriously ill children for statistically informative estimates. Table 1 reports the low incidence in the general population of serious childhood diseases, which emphasizes the need for a targeted sample of ill children such as ours examined here.
Admittedly our targeted sample is not representative of the general population of children. It is important to recognize first the source of nonrandomness in our sample. To be in our data a child must have an illness, which is a background variable, but need not have any particular characteristics of insurance coverage directly, the outcome of interest. By choosing a sample based on an explanatory variable (health status), what we do is similar to research that overselects African Americans to be able to study racial differences in health outcomes with reasonable statistical precision. The overrepresentation of African Americans by design does not create a biased picture of how race affects health because there is no reverse causality-health does not determine race. Our situation is analogous in that there is no obvious reverse causality running from current insurance adequacy to the current presence of a serious health condition, most prevalently cancer and blood disorders. Our results should be informative on how the presence of a severe health condition can limit insurance coverage, without worry that our estimates are contaminated by unnoticed reverse causality.
A second possible complexity in determining what influences health insurance adequacy is the possibility that persons who have relatively good health insurance, an outcome of interest, are systematically more or less likely to participate in a survey of insurance adequacy. Consider an explanatory variable that one believes to 224 SWIGONSKI, KINNEY, FREUND, KNIESNER be positively related to insurance adequacy, say income. Suppose families with above-average incomes are more likely to have adequate insurance and also more willing to discuss it. The estimated effect of income on insurance adequacy will be overstated because it represents two things: the true positive effect of income on insurance adequacy, and the willingness to tell someone about insurance adequacy. On the other hand, one can imagine the situation where a family with bad health insurance is more likely to participate in a survey of health insurance adequacy. Now let families with low incomes, who are more likely to have inadequate insurance, also be more willing than most to tell someone about their health insurance. Now the estimated effect of income on insurance will be understated because it reflects two things: the true negative effect of low income on insurance adequacy, and the willingness of the inadequately insured to tell someone. Our data source would not release the information to permit study of survey nonrespondents. There is no good way, then, to adjust our multivariate analysis for any latent connection between insurance adequacy and the willingness to report it or even to speculate on the direction of any bias in our results from using our admittedly nonrandom sample of children.

The Sample
The sample we use covers children ages 1 to 18 treated for a serious medical condition at the Riley Hospital for Children of the Indiana University School of Medicine, Indianapolis, Indiana. Riley Hospital is the only children's hospital in Indiana and is the state's major pediatric tertiary care center. The children sampled have diseases for which Riley Hospital treats more than 80% of Indiana's afflicted minor residents. Table 2 displays the frequency of diagnoses, which include childhood cancers, inborn errors of metabolism, rheumatological diseases, and chronic renal failure.
Because of confidentiality concerns, Riley Hospital first contacted families and invited them to participate in the study. The hospital then sent out one letter to 1,943 parents of children that met the sampling criteria. The hospital released contact information on 436 consenting participants. The 22% response rate is typical of surveys in which only one mailing is sent to the respondent pool (Fink, 1995;Frey, 1989;Goyder, 1987;Groven, 1988). About 18% of the respondents were additionally excluded because they either did not meet the study's age criteria (6%) or they did not have a selected serious disease (12%).
Telephone interviews of the usable respondents occurred between April and August 1994. The parent or guardian who knew the most about the child's illness and insurance answered the questions. Interviews focused on the child's health, insurance coverage, cost of treatment, and health care choices using a so-called active interviewing technique (Holstein & Gabrium, 1995). About 91% of the eligible respondents completed interviews. After deleting an additional 8% of the ultimate survey respondents because of missing data on one or more independent variables, we were left with 299 children with the necessary information. We studied in detail the insurance adequacy of the 83% with private insurance.

Multivariate Statistical Analysis
The first multivariate statistical model we use is a logit regression intended to reveal the separate influences on insurance adequacy of key demographic, economic, and health status factors. For each of the insurance adequacy criteria, 3% to 8% of the participants did not know what was covered in their policy. Specifically, in the sample of privately insured persons we studied, 8% did not know if there was a preexisting exclusion on their policies, 6% did not know their yearly deductible, and 8% did not know if there was an upper limit on out-of-pocket expenses. When respondents did not know the characteristics of their insurance, we coded their insurance as adequate, which produces a conservative estimate of the number of children with inadequate insurance to explain in our multivariate model.
In supplemental multivariate logit regressions, we examine how demographic, economic, and health status factors plus insurance inadequacy affect the likelihood that someone will delay seeking or taking treatment. Reported delays in obtaining care because of a lack of insurance for the child or the family meter access to health care. Specifically, participants were asked whether they delayed care for the child 226 SWIGONSKI, KINNEY, FREUND, KNIESNER or themselves or another family member because of lack of insurance coverage. Examining delayed care, too, gives us a more complete picture of the impact of inadequate insurance on families of children with serious illness.
Explanatory variables. Our choice of particular explanatory variables follows Aday's conceptualization where access to health care depends on so-called predisposing, enabling, and need factors (Aday et al., 1993). Predisposing factors, which are deemed to underlie a lack of insurance coverage, here include race, marital status, and education. Enabling factors, which should facilitate access to health coverage, here include family income, employment status, and insurance status. Need factors, which reflect need for health care, here include the child's health status and length of time since diagnosis. Table 3. The parents in our data were predominantly privately insured, White, highly educated, and employed with moderate to high incomes. Children had a variety of serious medical conditions; the most frequent serious illness was a childhood cancer (30%). Despite their illnesses, the majority of children were considered to be in a condition of excellent, very good, or good health.

Health Insurance Coverage
Insured persons' health insurance coverage can be public or private. Public coverage includes Medicare, Medicaid, CSHCN, and health insurance for military personnel and their dependents. Private coverage here means (a) either private coverage mandated by federal law or Indiana's high-risk pool, and (b) all other private coverage, referred to as market insurance. The primary source of the children's insurance coverage is private health insurance that is a fringe benefit of a parent's employment. About 79% of the children in our sample had conventional private insurance, 5% had private insurance through mandated programs such as the COBRA extension or Indiana's high-risk pool, 13% had insurance from public programs, and 3% had no insurance.
In our data, parents of children with public insurance were less likely than parents of children with either no insurance or private insurance to be employed, white, have more than 12 years of education, have moderate to high incomes, and to be married. Compared to parents with mandated or market private insurance, parents of children with no insurance were more likely to have low incomes. Table 4 shows that according to our definition of adequacy, a third of the privately insured children had inadequate health coverage, some for multiple reasons. About 66% of the inadequately insured were in that situation because their insurance lacked an annual out-of-pocket limit on medical expenditures. About 29% of the inadequately insured were that way because their insurance had a permanent preexisting condition exclusion. About 16% of the inadequately insured were that way because they had a policy containing two of the three characteristics mentioned ear-228 SWIGONSKI, KINNEY, FREUND, KNIESNER Approximately 90% of the children we study did not have a preexisting condition exclusion in their insurance. Almost one half of the children without a permanent preexisting condition exclusion had the same insurance policy as at diagnosis and were not at risk for the exclusion. About one third of the small number of respondents with mandated insurance in Indiana's high-risk pool reported having had inadequate insurance previously. Table 5 presents our multivariate logit model results. For convenience the individual effect of an explanatory variable is expressed as its so-called odds ratio and the associated likely range into which the odds ratio falls with reasonable statistical certainty. No predisposing or enabling variables have a statistically significant relation to insurance adequacy. Our main finding concerning insurance adequacy is a statistically significant negative relation between the number of years since diagnosis and adequacy of insurance, even holding constant the effects of health status, education, race, sex, marital status, income, and employment. On average, a person who was diagnosed with a serious condition a year ago has about a 90% chance of having adequate insurance compared to someone just diagnosed.

Reported Delays in Care
As noted by the frequency data in Table 6, parents were more likely to delay care for themselves or to delay care for other children than to delay care for their seriously ill child. Compared to parents with adequate insurance, parents with inadequate insur-UNFINISHED BUSINESS No out-of-pocket limit 66 Two of three characteristics b 16 ance were about 2.5 times more likely overall to delay care for their child. Similarly, 43% of inadequately insured parents reported delays in treatment for themselves or another family member, compared to the 26% of parents who were adequately insured who reported delays in treatment in Table 4. We now examine whether adequacy of coverage connects to decisions to seek medical advice and treatment both for our focal seriously ill children and other family members when one also considers the intervening factors we have labeled predisposing variables, enabling variables, and need variables. Our data are not rich enough to allow for the possibility of joint causality running back and forth between the decision to delay care and the adequacy of health insurance in the last two columns of results in Table 5. The statistical solution to disentangling the effect of insurance adequacy on delay of care from a possible effect of likely delay of care on insurance purchase (as reflected by adequacy) requires having at least one variable that a researcher believes determines insurance adequacy but is not itself determined by delay of care decisions. The extra variable, known as an instrument or an instrumental variable, could enable one to infer the causality running from insurance adequacy to care delays. We neither have available an obvious instrument to stand in for observed insurance adequacy nor is it clear how any bias in our multivariate results discussed later may be at work, because it is not obvious whether and how persons who report delaying care will have adjusted their insurance adequacy in light of an anticipated need for immediate health care.

229
The results in the third and fourth columns of Table 5 confirm that even after we hold statistically constant the effects of intervening factors, there remains a negative effect of adequacy of coverage on delay in getting medical care. Compared to persons with inadequate insurance, persons with adequate insurance were half as likely to delay care for either themselves or other family members and two thirds less likely to delay care for their child. For emphasis we cite Figure 1, which directly quotes respondents on how decisions concerning an expensive medical procedure led them to consider extreme strategies for obtaining the requisite health coverage.

DISCUSSION
Our sample parents were predominantly White, highly educated, employed, and had middle to high family incomes. Over 96% had health insurance coverage. Nevertheless, over 30% of the privately insured children were inadequately insured. Some children faced selective exclusions because of preexisting conditions, which often means having no coverage for the disease generating most of their medical expenses. Respondents frequently had insurance with no annual limit on out-of-pocket expenditures and in turn faced exposure to high medical bills. Over one third of children insurance mandated through the state of Indiana's high-risk pool reported having inadequate insurance at some point. Our results indicate that inadequate insurance is prevalent among children with serious illnesses and that inadequate insurance has a significant impact on children's access to care. We found that predictors of adequate insurance differ from predictors of noninsurance reported in the literature (Aday et al., 1993;Comer & Mueller, 1992;Holl, Szilagyi, Rodewald, Byrd, & Weitzman, 1995;Weinick et al., 1998). Children whose parents were poorly educated, single, had low incomes, or had children who were in poor health were no more likely to have adequate insurance than children who had parents who were more highly educated, had higher in-comes, were married, or whose children were in better health. Other researchers have found similar results regarding the lack of an effect for marital status (Comer & Mueller, 1992), income (Short & Banthin, 1995), and health status (Aday et al., 1993) on adequacy of health insurance.
Race (Aday et al., 1993;Comer & Mueller, 1992;Holl et al., 1995;Weinick et al., 1998) is a known predictor of noninsurance, and employment status is a predictor of both noninsurance (Cunningham & Monheit, 1990) and inadequate insurance (Comer & Mueller, 1992). We did not find a significant association between race or employment status and adequacy of insurance. Other studies have shown an association between age and both noninsurance and adequacy of insurance (Short & Banthin, 1995). We found that the number of years since diagnosis had a greater effect on adequacy of insurance than the child's age.
Using a definition of inadequate insurance based on characteristics of insurance rather than the potential risk of loss is admittedly controversial. Limited benefits or high deductibles might not result in coverage that is detrimental to long-run health if the insured has sufficient income to pay for the benefit or the deductible. We tested the saliency of our insurance adequacy definition that is based on insurance characteristics by including adequacy as an enabling variable to predict reports in delay of care. Others have reported that noninsurance is correlated with decreased access to care (Butler, Winter, Singer, & Wenger, 1985;Wood, Hayward, Corey, Freeman, & Shapiro, 1990). We found that despite our study's large fraction of families having moderate to high incomes, adequacy as defined by insurance characteristics had a significant impact on reports of delay in care for both the child with serious illness and other family members.
Finally, our data suggest that lack of adequate health insurance coverage leads parents to delay care for their seriously ill children or, more likely, to delay medical care for themselves or other family members. Our findings suggest an important public health problem because delays in medical treatment often mean disease becomes more serious before any treatment.

CONCLUSION AND IMPLICATIONS FOR PRACTICE
Despite coverage expansions for children and state and federal small-market insurance reforms, we find that a problem exists with inadequate insurance for privately insured seriously ill children. Physicians and other practitioners should not assume that seriously ill children with private health insurance-even with moderate incomes-have the requisite coverage to pay for all the health care services they want to treat serious illnesses. It is important to be sensitive to the nature of insurance coverage for privately insured children in designing and implementing treatment regimens. Children with private insurance coverage may actually have less coverage to defray the cost of care than poorer children on CHIP or Medicaid.
Adequacy of insurance health coverage for seriously ill children, particularly children with private insurance, is an important issue both from a public policy and a medical practice perspective. Our strong findings, even though based on a small sample from one state, suggest that a problem with private coverage for seriously ill children exists and that future research could be quite valuable in designing health policy.

ACKNOWLEDGMENTS
Research was funded by The Robert Wood Johnson Foundation (Grant 0201840).
Views expressed are solely our own. We acknowledge the contributions of Karen Jordan, Mary Beth Camp, Esther Gray, Marion Christopher Mayfield, Kunvichet Chea, and Phyllis Bonds. We also thank John Kennedy and the staff of the Center for Survey Research at Indiana University-Bloomington and the physicians and staff at James Whitcomb Riley Hospital for Children at Indiana University Medical Center. Most of all, we thank the parents of seriously ill children who participated in this survey.