Variation in Approval by Insurance Companies of Coverage for Autologous Bone Marrow Transplantation

During the current debate on health care reform, access to state-of-the-art care for serious illness has been stressed as an essential part of universal health care. The American medical-research establishment has been praised as a major strength of the current health care system, and policy experts have frequently cited outcomes research, particularly involving new and expensive forms of technology, as fundamental to any progressive medical health reform1. However, the participation of patients in peer-reviewed, grant-sponsored clinical trials of cancer therapy often depends on prior approval by insurance carriers of the proposed treatment, a process that is termed predetermination.

Over the past decade, the use of high-dose chemotherapy with autologous bone marrow transplantation (ABMT) to treat both metastatic and primary breast cancer has been the subject of clinical research2. Because of the frequency of the disease, the expense and difficulty of the treatment, the requirement for technology-intensive facilities, and the lack of completed randomized trials, this therapeutic approach has been controversial in parts of the medical community3. However, continued clinical research into this approach has been generally advocated3-7. In the past five years, as this therapy for breast cancer has begun to show clinical promise, an increasing number of denials of insurance coverage for high-dose chemotherapy and ABMT for breast cancer have been noted at major academic centers. The use of the clause contained in most health insurance and benefit contracts allowing third-party payers to deny health care coverage for women enrolled in clinical research trials has resulted in concern in the academic world,8,9 outcries by patients and the public,10,11 litigation,12 and interest on the part of the media13,14. The ability to study such coverage has been hampered by the large number of agencies providing third-party payment, the small size of most research studies, and the geographic dispersion of both patients and payers. Funding for the costs of care associated with clinical trials is vitally important to patients' participation in these trials, particularly in oncologic trials, because participation in peer-reviewed research using a protocol is generally considered the best medical therapy available. Although many components of research, such as data collection, statistical analysis, the cost of unapproved medicines, research-related laboratory studies and radiologic evaluations, and support staff, are largely funded by federal, private, and corporate research grants, the costs of clinical care for the underlying medical conditions have generally been covered by third-party insurance15.

We report here our experience over a four-year period with the predetermination process and its relevance to clinical outcomes in 533 patients referred to our center from a very wide geographic area in order to participate in clinical research trials of ABMT for breast cancer.

METHODS

Patient Population

From February 1989 through March 1993, 671 patients with breast cancer were referred to be evaluated for high-dose chemotherapy and ABMT for which coverage from an insurance company was sought. The insurance decisions for 135 patients were still pending when the study group was formed, and 3 patients were paying for their treatment themselves. These patients were excluded from further analysis, leaving 533 patients in the final study group. No other method of selection was used.

Treatment Protocols

The patients had stage II, III, or IV breast cancer as documented by histologic methods. On the basis of an initial history taking and physical examination, patients were selected as candidates for enrollment in protocols approved by the scientific review process at a Comprehensive Cancer Center and by an institutional review board. Two hundred nine of the 378 patients who received bone marrow transplants were enrolled in prospective randomized trials for patients with either high-risk primary breast cancer or early, chemotherapy-sensitive metastatic breast cancer. One hundred thirty-two patients received the same high-dose treatment in phase 2 trials for either high-risk primary breast cancer6 or early metastatic disease involving both viscera and bone. The remaining 37 patients were treated in phase 1 trials evaluating alternative programs of high-dose chemotherapy for metastatic breast cancer. All patients provided written informed consent. Our program did not treat patients who were not enrolled in an approved trial.

Each protocol except the phase 1 studies involved an induction course of two to four cycles of conventional-dose chemotherapy administered over a period of approximately three months, before ABMT. Patients with stage II or IIIA breast cancer received four cycles of cyclophosphamide, doxorubicin, and fluorouracil,16 and patients with stage IIIB or IV disease received doxorubicin, fluorouracil, and methotrexate (AFM)17 for a variable number of courses (from two to four) until the best response was achieved.

Predetermination Process

At the time of the initial evaluation or shortly thereafter, patients considered by the transplant team to be appropriate candidates for the protocol were given the telephone number of a person in our business office in order to facilitate the predetermination process.

The business officer obtained information from the patient or family about their primary insurance company. The business office then contacted the benefits representative at that company and requested a predetermination of approval for bone marrow transplantation for breast cancer. This telephone request was followed by a letter to the insurer that included the relevant codes from Physicians' Current Procedural Terminology,18 the estimated charges, the initial clinical assessment of the patient, a letter stating the medical necessity, and a packet of information containing reprints from the literature and institutional and protocol-specific data on outcomes when appropriate and available. A copy of the protocol and a consent form were provided if requested. Written confirmation of the approval or denial of insurance coverage was requested. All denials were routinely appealed in writing by the attending physician to the medical director of the insurance company.

During this time, the patient underwent pretransplantation induction chemotherapy and evaluations as prescribed by the protocol. At the end of the induction therapy, a final medical decision about the patient's candidacy for high-dose therapy was made at a multidisciplinary conference, and a bone marrow-harvesting procedure was scheduled (when indicated), generally within the next two weeks. At this point, the decision by the insurance company was archived in the data base. The hospital required that prior approval by the insurance company be given or that other arrangements be made for financial coverage, including charity care, before hospital admission for the harvesting of bone marrow. In order to determine the methods of financial coverage applicable to patients proceeding to transplantation who had received denials of coverage, the records of the business office were reviewed.

Statistical Analysis

The probability of a response to ABMT in the patients receiving insurance approval was compared with that in patients receiving a denial with Fisher's exact test. Heterogeneity in the frequency of predetermination approvals granted by various insurance companies was tested with a chi-square test of homogeneity of proportions. All P values are two-sided.

RESULTS

Effect of Insurance Approval on Receipt of Treatment

Decisions on coverage by an insurance company had been made for 533 patients at the time of the final medical decision about high-dose chemotherapy and marrow transplantation. Of these patients, 412 (77 percent) were approved for insurance coverage, and 121 (23 percent) received denials. In the case of denials, the reason cited in the majority of cases was the investigational or experimental nature of the therapy. In a few cases, a specific exclusion of such treatment in the contract or a preexisting condition was cited, and in other cases no reason was given.

Of the 412 patients whose coverage was approved, only 316 (77 percent) actually received transplants. The remaining 96 patients were not given high-dose therapy either for medical reasons or for reasons involving the protocol. For 22 patients, portions of the pretreatment evaluation resulted in the patient's ineligibility for the trial or in exclusion from treatment for other reasons. Ten patients withdrew from their trials before transplantation, and 23 were at too early a stage of treatment to proceed to transplantation. Twenty-six patients were excluded from transplantation because of a lack of response or a progression of disease during standard-dose therapy, and 15 patients were randomly assigned to the delayed-transplantation arm of the protocol and were not given transplants.

Of the 121 patients who were denied coverage, the majority (62 patients, or 51 percent) received transplants. Fifty-nine patients did not. In general, the reasons for the exclusion of these patients were similar to those for the patients who received insurance approvals; that is, patients were not treated (25), withdrew (3), were at too early a stage of treatment (4), did not respond or had progressive disease (12), or were randomly assigned to delayed transplantation (4). After the exclusions for medical reasons, 11 of the initial 533 patients (2 percent) who appeared to be eligible for ABMT did not go on to transplantation.

For 62 (51 percent) of the 121 patients whose request for coverage was denied, some financial arrangement was secured between the completion of the predetermination process and the initiation of high-dose therapy. Thirty-nine of these patients (63 percent) eventually had the procedure fully or partially approved by the insurer or a secondary insurance carrier. At least 19 of these 39 patients received approval for coverage after the patient hired an attorney. Litigation was undertaken by seven patients and was decided in favor of all but one. In this latter case, the insurer reversed the denial of coverage after publicity appeared in the media. The remaining 23 of the 62 patients proceeded to transplantation without approval for coverage by an insurance company. These patients paid all, none, or a portion of the charge, and some had outstanding bills.

Effect of Disease Status or Study Design on Approval

The frequency with which insurance approval was granted did not differ according to either the patients' disease status or the design of the protocol in which the patient was enrolled. Patients with advanced metastatic breast cancer who were enrolled in dose-finding phase 1 trials received approval as frequently as patients enrolled in multicenter randomized trials given a high priority by the National Cancer Institute (86 percent vs. 90 percent). Patients with metastatic breast cancer who were enrolled in phase 2 studies received insurance approval (84 percent) as often as patients enrolled in phase 2 studies who had primary disease (76 percent).

The clinical characteristics before treatment of the patients approved for insurance coverage were similar to those of the patients whose requests were denied. For example, patients with metastatic breast cancer who were enrolled in the randomized trial of AFM therapy had characteristics in common before treatment and did not differ from each other at entry in age or general health status. Table 1TABLE 1Clinical Responses to AFM Therapy in a Randomized Trial, According to the Insurer's Decision to Provide Coverage. shows that for patients treated according to this protocol, the frequency of complete and partial responses to the induction therapy (a measure sometimes cited as a pretransplantation indicator of the response to ABMT) did not differ between patients who received approval of coverage and those who did not. Furthermore, the frequency of responses, and especially of complete responses, after ABMT was the same in both groups. Differences in survival between the groups will be compared after the completion of the randomized trials. The findings were similar in patients with stage II disease.

Effect of Insurer on Approval

The 533 patients had health insurance provided by a total of 187 different companies, covering an average of 2.84 patients each (range, 1 to 47). Table 2TABLE 2Insurance-Company Decisions about Requests to Provide Coverage for ABMT for Breast Cancer. shows the frequency of insurance approvals of ABMT for breast cancer that were granted by the 20 companies making the most such decisions in our study, with the data for all other insurers pooled. The results show significant heterogeneity between insurers with regard to the frequency of approvals (P<0.001) and indicate that many companies approve some requests but not others. Although some companies consistently approved predetermination requests, others varied in their responses. The differences did not appear to result entirely from the language of the contract, since the reason for denial in nearly all cases was the investigational or experimental nature of the therapy. The same heterogeneity between requests and among insurers is seen as well in a single trial of women with metastatic disease who underwent AFM therapy (P = 0.005).

Because we received referrals for these studies from a very wide geographic area, it was possible to compare in a limited manner the frequency of insurance approvals granted by 38 individual Blue Cross-Blue Shield plans for the patients in these studies. The same heterogeneity in the frequency of approvals was seen among the various Blue Cross-Blue Shield plans (P<0.001). Fourteen of the 38 plans approved all of a total of 43 predetermination requests. Five of the plans declined one request each. The remaining 19 plans approved some (112, or 67 percent) of a total of 167 requests but not others, with the frequency of approval ranging from 20 percent to 96 percent. Three referrals from the Canadian health system were all approved for coverage; of requests made to seven health maintenance organizations (HMOs), all those from a given plan were either approved or denied, except for one HMO for which the decisions were heterogeneous. There was no change in the annual frequency of approvals over the period in which the observations were made.

DISCUSSION

Cost-effective development and evaluation of new therapies require that patients be enrolled in clinical research trials designed to test hypotheses and assess outcomes. In this regard, the data presented here have important implications for health policy. Most requests for insurance coverage for ABMT were approved, a finding consistent with our previous observations6. However, the frequency of approval for patients enrolled in these clinical research trials varied among third-party payers, appeared to bear little relation to available medical or scientific information, and often seemed arbitrary and capricious. Of two requests to the same insurer by similar patients in the same protocol, one request might well be approved and the other denied.

These peer-reviewed, grant-supported clinical trials conducted at an academic center selected patients for an expensive research technology on the basis of medical and protocol-based criteria. The academic medical center hospital attempted to provide research care to appropriate patients regardless of their financial coverage. This approach resulted in a substantial financial debit to the hospital related to unpaid bills in this study. Only 11 eligible patients of the 533 evaluated did not subsequently undergo transplantation after a denial of coverage. However, for patients who were denied coverage and who eventually underwent ABMT, outstanding medical bills and charity care amounted to more than $2 million.

Many insurance policies and benefit plans contain exclusions for experimental treatment. The origins of this widely used provision are obscure, but it is generally accepted that by requiring that treatments meet minimally acceptable standards in the medical community, insurers will avoid providing coverage for worthless treatments12,19. The lack of any correlation with the phase of the study in which the patient was enrolled suggests that the state of knowledge about the procedure had little effect on the decisions of insurers.

We found that insurers do not generally make decisions related to coverage for complicated research therapies in a consistent or medically relevant manner. Some insurers, such as Blue Cross-Blue Shield, have developed evaluation programs for new forms of technology20 and provided recommendations to their member plans about these therapeutic approaches. However, the practice of the Blue Cross participating plans in our study was as heterogeneous as that of other insurers.

For patients, these data suggest that the selection of a third-party insurer is important when a major medical illness is faced. Depending on the reputation of a given provider may prove unreliable when the recommended therapeutic option is participation in a research study at a major cancer center. The finding that with many insurers the outcome of the predetermination process varies from request to request is particularly disconcerting. We could not quantitate in this study the effect on the quality of life that resulted from this process, but patients commonly reported the anguish that they endured during this time. Nor does prior experience with an insurance company necessarily prove to be helpful, since for most patients, their diagnosis and treatment were the first catastrophic medical illness their families faced. The fact that denials of coverage were frequently reversed when an attorney became involved may offer patients some consolation, but as a policy this approach is to be strongly discouraged. Furthermore, some insurance companies, particularly certain Blue Cross-Blue Shield plans, have introduced into their contracts specific policy exclusions with regard to ABMT for breast cancer. In most employer-sponsored benefit plans, no options are routinely provided that extend coverage to include treatment in clinical research programs.

These data may come as no surprise to most clinical investigators and many patients. For three quarters of Americans under the age of 65, private health insurance is the main method of financing health care21. The data indicate, however, that there is considerable heterogeneity in the process of granting prior approval for privately insured patients enrolled in grant-supported clinical cancer research. The net result of the predetermination process for patients in these scientifically reviewed trials is at once encouraging and disappointing. Most insurers approved coverage for patients in the trials we studied, but nearly a quarter of patients were denied coverage. Although coverage for AMBT for breast cancer outside the context of peer-reviewed research should reasonably be very limited, and indeed discouraged, at this time, most in academic medicine would support a policy of routine coverage for peer-reviewed, grant-supported clinical research trials. Besides being arbitrary, the predetermination process was also less selective. The insurance companies approved 412 of the 533 patients for ABMT (77 percent), whereas when medical and protocol-based criteria were used, 6 percent fewer patients (378) were selected for ABMT during the study period.

In 1988, representatives of eight major oncology organizations, including the National Cancer Institute and the American Society of Clinical Oncology, issued a consensus statement recommending third-party coverage for the costs of patient care in research protocols of cancer treatment22. Most academic and public authorities examining the issue of research-study reimbursement agree that investigational therapy is equivalent or superior to standard care and should be covered by third-party carriers4,7,9,15,22-29. Some representatives of the insurance industry disagree,20,30 arguing that it is not the responsibility of insurers to cover the cost of research. Policyholders are rarely aware of whether their insurance will cover participation in clinical research studies.

Although these data relate only to a single procedure studied in a very structured fashion, the predetermination process is also applied to other clinical situations, and we believe that similar results would be expected for other treatments and procedures,28 such as off-label use of cancer chemotherapy. In most cases, the best available therapeutic option for a patient with cancer involves treatment according to a clinical research protocol. Policy restrictions that limit access to clinical trials are likely to delay the evaluation of therapeutic programs and to result in the relegation of patients to outdated and inferior treatments4.

Underinsurance among Children in the United States

Policy discussions on children's health insurance have been devoted largely to reducing the number of uninsured children.1 As compared with children who have health insurance, those lacking insurance are more likely to have needed care that is delayed or forgone; less access to preventive, specialist, or long-term care; lower utilization rates; care that imposes a high financial burden on their families; and poor health outcomes.2,3 Children with intermittent health insurance (i.e., periods without insurance throughout the year) are at greater risk for delayed care, unmet health care needs, lack of a usual source of care, and nonreceipt of well-child care.4-8

Considerably less attention has been devoted to the problem of underinsurance, or insurance that does not sufficiently meet the child's needs. Recently, the American Academy of Pediatrics issued a policy statement highlighting the importance of this issue.9 The major problems cited were cost-sharing requirements that are too high, benefit limitations, and inadequate coverage of needed services. Although some studies have addressed underinsurance among adults10-12 and others have addressed the adequacy of coverage for children with special health care needs,13-16 more information is needed about the adequacy of health insurance for all children.

We conducted a study that incorporated the multiple dimensions of insurance adequacy cited by the American Academy of Pediatrics: adequate coverage for needed services, an adequate choice of providers, and reasonable coverage of costs. We used information from the 2007 National Survey of Children's Health, a nationally representative study of more than 90,000 children, to address the extent of underinsurance among continuously insured children, the sociodemographic and health factors associated with underinsurance and each of its components, and the association between underinsurance and measures of access and quality.

METHODS

Data Collection and Study Design

With funding and direction from the Maternal and Child Health Bureau at the Health Resources and Services Administration, the National Center for Health Statistics (at the Centers for Disease Control and Prevention) conducted the 2007 National Survey of Children's Health using the State and Local Area Integrated Telephone Survey data-collection mechanism. This random-digit-dial survey provides national and state-specific information on the health and well-being of children younger than 18 years of age that is based on interviews with their parents or guardians. Interviews were conducted in English, Spanish, or one of four Asian languages. When households with children were identified, one child from each household was randomly selected to be the subject of the interview. From April 2007 through July 2008, interviews were completed for 91,642 children. The survey was approved by the Research Ethics Review Board at the National Center for Health Statistics. The interview-completion rate (the percentage of completed interviews among known households with children) was 66.0%. Additional details about the survey methods and response rates are available elsewhere.17 All authors contributed to the design and interpretation of the analysis and vouch for the completeness and accuracy of the reported data.

Measurement of Insurance Adequacy

Data on insurance adequacy were based on parents' or guardians' judgments of whether their children's insurance covered needed services and providers and reasonably covered costs. Specifically, adequacy was derived from parents' or guardians' responses to three questions: “Does the child's health insurance offer benefits or cover services that meet his or her needs?” “Does the child's health insurance allow him or her to see the health care providers he or she needs?” “Not including health insurance premiums or costs that are covered by insurance, do you pay any money for the child's health care?” An affirmative answer to the third question prompted the follow-up question, “How often are these costs reasonable?” A summary measure of adequacy was created by combining the responses to these questions. If a parent or guardian answered “always” or “usually” to all three dimensions of adequacy, then the child was considered to have adequate insurance coverage. (If a parent or guardian reported no out-of-pocket costs, the costs were considered to be “always” reasonable.) If a parent or guardian answered “sometimes” or “never” to any of the three questions, the child was considered to be underinsured. Insurance adequacy was examined only for continuously insured children, defined as those children whose parents or guardians reported that they currently had “any kind of health care coverage, including health insurance, prepaid plans such as HMOs [health maintenance organizations], or government plans such as Medicaid,” and that during the past 12 months, there was not “any time when [the child] was not covered by any health insurance.” Parents or guardians of children without current health care coverage were asked whether, during the past 12 months, there was “any time when [the child] had health care coverage.” These adequacy and health care coverage variables were combined to create a four-level variable: no insurance during the past year, periods without insurance in the past year, continuous but inadequate coverage (underinsurance), and continuous and adequate coverage.

For each level of this variable, we examined the association with five measures of access and quality: delayed or forgone care, difficulty obtaining needed care from specialists, receipt of preventive care, receipt of developmental screening at a preventive visit (for children up to 6 years of age), and care in a medical home. We defined a medical home on the basis of the following five components of the framework established by the American Academy of Pediatrics18: having a personal physician or nurse, having a usual place to receive care when sick, having no problems obtaining needed referrals for care, receiving effective care coordination, if needed, and receiving family-centered, compassionate, and appropriate care. (Family-centered care was assessed with the use of five questions. Care was inferred to be family centered if the parent or guardian reported that the child's doctors usually or always spent enough time with the child, listened carefully to the parent, were sensitive to the family's values and customs, provided needed information, and made the parent feel like a partner.) We also examined each of the components of the medical home separately.

Other variables used in the analyses included the child's age, sex, and race or ethnic group; poverty level; geographic region; family structure; primary household language; household employment status; and child's health status as reported by the parent or guardian, health insurance type, presence or absence of a special health care need, and complexity of special needs.19-23 The presence or absence of a special health care need was determined with the use of a tool developed to screen children for special health care needs, which asks if the child has any of five health consequences (e.g., use of prescription medicine, need for special therapy, and limitations on activity) resulting from a chronic condition.21 Complexity was assessed on the basis of the number of consequences.23

Statistical Analysis

The data analysis consisted of five parts. The first part addressed the distribution of uninsured, underinsured, and fully insured U.S. children according to demographic and health characteristics. The second part used four logistic-regression models to examine factors associated with the overall underinsurance measure and each of its three components among continuously insured children. In these models, underinsured children with continuous coverage were compared with children with continuous and adequate coverage. Maternal level of education and poverty level were not included because of their high correlation with insurance type. The third part examined the association between the measures of access and quality and the four-level insurance variable. The fourth part used adjusted logistic-regression models to examine the association between level of health insurance and the five measures of access and quality plus the five components of a medical home. The fifth part examined the three components of underinsurance in relation to measures of access and quality among continuously insured children. The survey questions used to derive these outcome measures and explanatory variables are available at www.cdc.gov/nchs/slaits/nsch.htm.

Estimates were weighted with the use of survey sampling weights available from the National Center for Health Statistics. These weights reflect the inverse of the probability of being selected for the survey and were adjusted to account for nonresponse and noncoverage of households without landline telephones. Weighted estimates are representative of the U.S. noninstitutionalized population of children. The analyses were conducted using SUDAAN software (version 9.0.1) to account for the complex sample design and to obtain appropriate variance estimates. Unless otherwise specified, a P value of less than 0.05 was considered to indicate statistical significance.

RESULTS

We estimated that 19.3% — or 14.1 million — of all U.S. children (and 22.7% of children with continuous insurance coverage) were underinsured in 2007, exceeding the number of children without any insurance at all during the year (3.4 million) and the number who had insurance during only part of the year (7.6 million). Table 1

TABLE 1

U.S. Children Who Were Uninsured, Underinsured, or Fully Insured in 2007, According to Selected Demographic Characteristics.

 shows that older children, children in fair or poor health, and children with special health care needs were more likely to be underinsured. Among those with continuous insurance, children with private insurance were more likely to be underinsured (24.2%) than children with public insurance (14.7%).

Table 2

TABLE 2

Adjusted Odds Ratios for Factors Associated with Underinsurance among Children Who Were Continuously Insured in 2007, According to Type of Inadequacy.

 shows that the most common reason for underinsurance was that costs not covered by insurance were considered by respondents to be sometimes or always unreasonable (accounting for 12.1 million children). Smaller proportions of children had health insurance benefits that sometimes or never met the child's needs (3.9 million) or had coverage that sometimes or never allowed them to see needed providers (2.9 million).

The adjusted logistic-regression analyses showed that certain groups of continuously insured children were more likely to be underinsured. For children who were 6 to 11 years old and those who were 12 to 17 years old, the odds of being underinsured were 32% and 37% greater, respectively, than for children under 6 years of age. Hispanic and black children were significantly more likely to be underinsured than non-Hispanic white children, and children in the Midwest were more likely to be underinsured than children in the Northeast. Children who were not in very good or excellent health and children with special health care needs had elevated odds of being underinsured. In addition, the odds of being underinsured were twice as high for children with private insurance as for those who had public insurance. The multivariate association of the three components of underinsurance with sociodemographic and health characteristics was highly consistent, except that publicly insured children had lower odds of having adequate coverage for needed providers, whereas privately insured children had more than three times the odds of having inadequate coverage of charges.

As compared with continuously and adequately insured children, underinsured children were significantly more likely to be without a medical home, to have delayed or forgone care, and to have difficulty obtaining needed specialist care (Table 3

TABLE 3

Selected Measures of Access to and Quality of Care According to Insurance Coverage and Adequacy.

). With regard to the components of the medical home, underinsured children were less likely to receive needed referrals, care coordination, and family-centered care. Multivariate analyses revealed that children who were underinsured were at increased risk for each of these indicators even after adjustments for sociodemographic and health factors (Table 4

TABLE 4

Adjusted Odds Ratios for Problems with Health Care Access or Quality According to Insurance Coverage and Adequacy.

).

The strength of the associations between being underinsured and uninsured for all or part of the year was somewhat similar across the measures of access and quality. Using the same multivariate analyses reported in Table 4, we compared the associations between these measures and insurance status (being underinsured, never being insured, and sometimes being insured), with underinsured children as the reference group (see the table in the Supplementary Appendix, available with the full text of this article at NEJM.org). The group of children who were underinsured did not differ significantly from the group of children who were never insured with respect to delayed or forgone care, lack of a medical home, difficulty obtaining referrals, care coordination, family-centered care, or specialist care. The odds of having difficulty obtaining referrals, lacking care coordination, having difficulty obtaining specialist care, and not receiving a developmental screening were similar for children who were underinsured and those who were periodically insured. All other comparisons revealed significant differences between these groups.

Except for lack of developmental screenings, at least one type of inadequacy was significantly associated with each measure of access and quality, after multivariate adjustments (Table 5

TABLE 5

Adjusted Odds Ratios for Problems with Health Care Access or Quality among Continuously Insured Children.

). Lack of a medical home, difficulty obtaining referrals, lack of care coordination, lack of family-centered care, delayed or forgone care, and difficulty obtaining specialist care were significantly associated with all three types of inadequacy. Inadequate coverage for needed providers was associated with lack of a usual source of care, lack of a personal doctor, and lack of a preventive visit in the past year. Inadequate benefits or coverage of needed services was associated with lack of a personal doctor.

DISCUSSION

Using the 2007 National Survey of Children's Health to measure the adequacy of insurance coverage, we found that almost one quarter of children with continuous insurance were inadequately covered, or underinsured, as indicated by having insurance that did not provide adequate benefits, provider choices, or coverage of costs from the perspective of parents or guardians. This multidimensional approach to defining underinsurance differs from the approaches commonly used in studies of the adult population, which analyze out-of-pocket expenses that exceed certain — often arbitrary — thresholds (e.g., 10% of family income), deductibles that exceed certain levels, gaps in insurance coverage during the year, or benefits that fall below certain benchmarks. These varied approaches have resulted in a wide range of estimates regarding the underinsurance problem. One review of 24 population-based studies showed that the prevalence of underinsurance among adults ranged from 4 to 53%, depending on the definition, sampling frame, data-collection methods, and year the study was conducted. Most reported estimates ranged between 5 and 25%. Only two of the studies we reviewed incorporated more than one dimension of underinsurance in developing their estimates.24

Our estimate of underinsurance is based on multiple measures of the adequacy of insurance: adequate coverage for needed services and providers, and reasonable coverage of costs. Although our approach to estimating underinsurance is more subjective than approaches based solely on measures such as out-of-pocket expenses, it provides a more comprehensive view of insurance adequacy by capturing multiple domains from the family's perspective. Estimates based on out-of-pocket expenses generally consider income level alone, disregarding other financial commitments or variations in assets and debts. By considering families' perspectives on, for example, the reasonableness of out-of-pocket expenses, our approach complements and expands on existing single-domain approaches to measuring underinsurance.

The prevalence of underinsurance varied across groups. Older children, Hispanic children, those residing in the Midwest, and children with either compromised health or special health care needs were more likely to be underinsured. Adolescents may be more likely than younger children to be underinsured because they are more likely to require services that are not well covered by insurance, such as mental health care.25,26 A possible explanation for the higher estimates of underinsurance for children with special health care needs is that such children use the health care system more often than children without such needs and may therefore be disproportionately affected by deductibles, copayments, benefit limitations, and limits on the range of health care providers available for patient care.27 Children with public coverage were much less likely to be underinsured than those with private coverage, perhaps because public insurance programs provide better protection against out-of-pocket expenses.28,29

Our measure of underinsurance permits assessment of the overall prevalence of underinsurance and its underlying components. We found that inadequate coverage of charges was far and away the most common source of underinsurance. We also found that children enrolled in private plans were more than three times as likely as their counterparts in public plans to have inadequate coverage of charges. This dramatic difference is probably the result of federal rules that permit only very limited cost sharing under Medicaid and modest cost sharing under the Children's Health Insurance Program. The latter program provides subsidized insurance for children in families with incomes above eligibility limits for Medicaid but below the level at which families could afford to purchase private insurance on their own. Most states set the upper limit of eligibility for the program at 200 to 300% of the federal poverty level. In 2009, a total of 29 million children were enrolled in Medicaid, and 7 million in the Children's Health Insurance Program.30

We examined the relationship between underinsurance and several measures of access and quality. As compared with fully insured children, those who were underinsured had substantially greater odds of being without a medical home, having difficulty obtaining referrals and care coordination, lacking family-centered care, and having difficulty obtaining needed specialty care. In fact, on the measures of access to referrals, care coordination, and specialty care, the problems affecting underinsured children were similar to those affecting children who had never been insured or who had discontinuous coverage. In contrast, the utilization measures reveal that underinsured children were only slightly less likely than fully insured children to have a usual source of care, to have a personal doctor or nurse, and to have had a preventive medical visit in the past year. Among younger children, those who were underinsured were as likely as fully insured children to have undergone a developmental screening. When we examined the components of underinsurance individually, we found that each of the three types of insurance inadequacy was associated with an increased likelihood of being without a medical home, having care that was delayed or forgone, and having difficulty obtaining referrals.

Our study has several strengths, including a large, nationally representative sample and the comprehensive set of questions on insurance adequacy and access-to-care indicators. However, the study also has limitations. First, the cross-sectional study design makes it difficult to establish the direction of causality. Second, although the data were largely representative of children in the United States, the National Survey of Children's Health excludes children in institutions. Third, the results are subject to biases associated with telephone surveys, including the exclusion of households without landlines (although adjustments were made for both nonresponse and noncoverage). Fourth, there is no single, universally accepted definition of “underinsurance,”31,32and we could not compare our findings with more objective measures, such as the percentage of family income expended on health care; alternative conceptions of underinsurance might yield different results from those presented here. Finally, we could not examine the health consequences of underinsurance, an important area for future research.

As discussions about the contours of health care reform evolve, it may be worthwhile to consider not only the number of uninsured children in the United States but also the adequacy of coverage for those with current insurance. Policy discussions often focus exclusively on mechanisms to extend insurance coverage to those with no coverage. This study shows that underinsured children have problems with access to health care that are similar to those of uninsured children and that underinsurance among children is actually more prevalent than lack of insurance.