Background. In 1990, the Florida Legislature established the Florida Healthy Kids Corporation to implement the concept of school enrollment-based health insurance coverage for children. The county school districts are used as a grouping mechanism to negotiate health insurance policies. The Florida Healthy Kids Corporation negotiates contracts with health maintenance organizations (HMOs) to assume financial risk and to provide health care services at each program site. In 1994, there were five sites with four different participating HMOs. Assessing quality of care is particularly important when contracting with HMOs because of the perception that financial and utilization review arrangements may restrict the enrollees' access to needed health care. One essential component of health care quality is the extent to which health care services are used in a manner consistent with the expected pattern of use for the population of enrolled children. The purpose of this study is to compare children's health care use across five different Florida Healthy Kids Program sites. Specifically, we compare the enrollees' actual health care use across HMO settings and program sites to the expected health care use based on the enrollees' case-mix.
Methods. Each HMO provided child-specific health care use data including Physician's Current Procedural Terminologycodes and International Classification of Diseases,9th Revision codes. We used the Ambulatory Care Groups (ACGs) software to compare the children's actual health care use to the expected health care use at each site adjusted for case-mix. Several steps were then taken to determine if the children were receiving the anticipated number of health care visits based on their diagnoses. First, we divided the average number of encounters at each site by the group average across all of the sites, without adjusting for the case-mix of the enrollees. We then divided the average number of visits at each site by the expected number of visits based on the case-mix adjustment. A value of 1.00 means that the actual use and the expected use are identical. Values below 1 indicate underuse and values over 1 indicate overuse of health care services. Statistical comparisons of the actual versus expected average health care use across the five sites were performed by deriving the appropriate χ2 statistics.
Results. A census of all children (N = 14 688) enrolled in the Florida Healthy Kids Program at each of the sites for 6 months or longer were included in the analysis. The average number of health care encounters across all sites for a 12-month time period was 2.98 ± 4.6 visits. After adjusting for the case-mix of the enrollees in each site using the ACG software, several of the five sites differed from one in a statistically significant way. However, these statistical assessments must be tempered with assessing the practical magnitude of the observed differences.
Conclusions. The number of public and private efforts to insure children who are not eligible for Medicaid and whose parents cannot purchase private insurance has grown dramatically. These programs are vital for ensuring financial access to care for uninsured children. However, it is essential that such programs are not viewed as merely cost containment efforts. Assessing the degree to which children receive the health care services they need across multiple delivery settings is an essential yet challenging component of quality assurance. Generally, our analysis indicates that children in the Florida Healthy Kids Program are receiving the amount of health care expected based on their health care needs; which is one component of a high-quality health care program.
Beginning in the 1980s, many states began developing health insurance programs for uninsured children that were designed to bridge the gap between Medicaid and private insurance. By 1995, 31 states offered special health insurance programs for children funded through public and/or private sources. In an effort to control costs, most of these programs offer care to children through some type of managed care arrangement; predominantly health maintenance organizations (HMOs).1 However, little is known about the health care that children receive within these programs.
Over 49 million persons were enrolled in HMOs in 1994 with approximately 33% of them under the age of 17.2 With increasing numbers of children enrolled in these and other managed care arrangements, it is important to assess the quality of care provided to the enrollees. One essential component of health care quality is the extent to which health care services are used in a manner consistent with the expected pattern of use for the population of enrolled children.3
Assessing quality of care is particularly important when contracting with HMOs because of the perception that financial and utilization review arrangements with providers may restrict the enrollees' access to needed health care.4 For example, HMOs typically require a physician to seek prior authorization before rendering certain types of services in an effort to reduce health care use and control costs. Concern has been raised that some of the reductions in use and costs may be excessive and possibly detrimental to the enrollee.5
In many instances, when contracting with more than one HMO to deliver services, it may be necessary and desirable to compare health care use across multiple HMOs. Such comparisons can be difficult to make unless the health care needs of children within each HMO are taken into consideration. Various risk adjustment methods have been developed that allow for meaningful comparisons of health care use from one setting to the next by controlling for differences in patient severity or case-mix.6 However, to our knowledge, state health care reform initiatives have not yet been evaluated using these methods.
The purpose of this study is to assess children's health care use after enrollment in the Florida Healthy Kids Program, a state initiative designed to provide comprehensive health insurance to uninsured children. Specifically, we compare the enrollees' actual health care use across five different sites to the health care use that is expected based on the enrollees' case-mix.
To provide background information for the study, we first describe the Florida Healthy Kids Program.7 Second, we describe the methods and the computer software program that were used to control for the enrollees' case-mix when evaluating their health care use across settings.
THE HEALTHY KIDS PROGRAM
In 1990, the Florida Legislature passed the Healthy Kids Act establishing a nonprofit Florida Healthy Kids Corporation to implement the concept of school enrollment-based health insurance coverage for children. This program targets families with school-aged children, which comprise approximately 66% of uninsured families nationally.8 The Florida Healthy Kids Program is an alternative to employment-based health insurance because the county school districts are used as a grouping mechanism to negotiate group health insurance policies.9
A pilot project was implemented in Volusia County, Florida in 1992 and expanded to five other sites by 1994; with more than 16 000 enrollees. The purpose of the program is to provide comprehensive health insurance to uninsured school-age children and their siblings. Children must not be eligible for Medicaid and must be 3 years of age or older to participate. Families are offered subsidized premiums based on a sliding scale so that the financial barrier is reduced (Table1). The National School Lunch Program is used as a method to verify income and eligibility for insurance premium subsidy.
Funding for the subsidies is provided from state general revenue funds and from local community funds at the various program sites. Each year, the Florida Healthy Kids Corporation submits a budget request to the state legislature. This request is based on the number of children the Corporation would like to enroll during the upcoming year in addition to maintaining funding for current enrollees. Following review by House and Senate Health Committee members, the Florida Legislature appropriates funds for the upcoming fiscal year. During the 1995 to 1996 fiscal year, the state appropriation of approximately four million dollars comprised 45% of the Healthy Kids Corporation funding.
During that same time period, local funds made up about 18% of the total funding for the medical premiums. Local leaders at the various program sites work with businesses and philanthropic organizations in the community to secure funding for the premium subsidies. Each community is required to provide a 5% minimum contribution to begin the program based on the number of children they wish to enroll, with increasing responsibility for funding each year. The maximum amount required from a community is 40% by the fifth program year for nonrural sites and 40% by the eighth program year for rural sites. Families contribute about 37% of the funding for the program by paying a portion of the premium based on their incomes.10
A key feature of the program is the provision of care through the private sector. The program is not intended to extend Medicaid coverage or to provide health care as a variation of the current Medicaid system for children in Florida. The Florida Healthy Kids Corporation negotiates contracts with HMOs to assume the financial risk and to provide health care services at each program site. In 1994, there were five sites with four different participating HMOs. Three of the HMOs deliver services at only one site and the fourth HMO provides services at two different sites through two branch offices.
Table 1 contains a description of the HMOs and the provider financial compensation and utilization review approaches used. Different compensation methods are used including salary, discounted fee-for-service, and capitation. None of the HMOs withhold any portion of the primary compensation to cover financial deficits that may arise from hospitalizations or from making specialist referrals. With one exception, the companies require prior authorization for referrals, nonemergent hospitalizations, and procedures.
Care is delivered through private physicians' offices and clinics in the children's communities. Both pediatricians and family practitioners serve as the children's primary care providers. The Healthy Kids benefit package includes well-child visits and immunizations with no copayment required. Other benefits with minimal copayments include acute care visits, inpatient care, maternity benefits, mental health services, prescriptions, eyeglasses, physical therapy, and emergency services and transportation.
A census of all children (N = 20 275) enrolled in the Florida Healthy Kids Program across all sites were considered for inclusion in the analyses. However, the developers of the Ambulatory Care Groups (ACGs) software that was used to assess the children's health care use after adjusting for case-mix, recommend only using children enrolled 6 months or longer.11 In their experience children enrolled less than 6 months do not have an adequate diagnostic profile in the claims databases to yield valid results. Thus the 5587 children enrolled for less than 6 months were excluded, resulting in a final sample size of 14 688. The somewhat high percentage enrolled for less than 6 months was caused by two sites having recent open enrollments into the program.
We chose to limit our analyses to a 1-year time period for the following reason. The amount of health care use is partially related to the time frame considered—the longer the period of observation, the more visits one will observe. Computing an average number of visits for a group of children with widely divergent enrollment lengths would be difficult to interpret, because the average number of visits, in part, depends on the length of time in the program. The ACG software does not provide a mechanism to adjust for the number of months the child is enrolled in the program and so the developers recommend reviewing the enrollees' diagnoses during a fixed time period, typically 12 months.12 These analyses can be repeated at regular intervals to determine health care use trends.
Seventy-five percent of enrollees were living in families with incomes below 135% of the Federal Poverty Level (Table2). The children were 53% male and 47% female with an average age of 10.9 ± 3.56 years. The average length of enrollment during the 12 months was 9.8 months. Over 95% of the children (14 031) were enrolled for more than 12 months but only 12 months of their health care use data were used in these analyses.
Each HMO provided child-specific health care use data includingPhysician's Current Procedural Terminology codes andInternational Classification of Diseases, 9th Revision Clinical Modification codes. Health care use data from August 1994 through July 1995 were used in the analysis.
Case-Mix Adjustment Methods
We used the ACGs software to compare the children's actual health care use to the expected health care use at each site adjusted for case-mix.11 The ACG software was developed to examine variations in ambulatory health care resource use and incorporates age, gender, and diagnoses as classified during health care visits usingInternational Classification of Diseases, 9th Revision Clinical Modification codes. It is one of several approaches designed to assess the case-mix of patients in an effort to more adequately address the issue of risk adjustment. A primary strength of this software, however, is the reliance on elements universally recorded in claims data and its focus on ambulatory visits; which is especially important among pediatric populations.
The ACG software initially groups the most common diagnoses recorded during the time period being studied into 34 ambulatory diagnostic groups (ADG) based on similarities in resource consumption. An individual may have multiple combinations of ADGs over time. Examples of ADG categories include: chronic medical conditions, malignancies, psychosocial conditions, and others. Once these determinations are made, the software then assigns individuals into one of 52 mutually exclusive ACGs based on ADG combinations, age, and gender.13,14 Examples of the ACG categories include: acute minor condition(s) with chronic medical-stable condition(s), acute minor conditions ages 2 to 5, preventive/administrative, and others.
The common unadjusted expected health care use for all sites is defined to be a weighted average of the enrollee use in the ACG categories, after pooling the data over sites. Analyses conducted to date using the ACG software have explained between 40% to 50% of the variation in concurrent health care use, an increase of ten times the explanatory power of age and gender alone.13-15
Statistical comparisons of the actual versus expected average health care use across the five sites were performed by deriving an appropriate χ2 statistic, and adjusting for the multiple testing using a Bonferroni correction. This approach assumes independence among the sites and invokes the Central Limit Theorem for distributional assumptions.a
Description of the Healthy Kids Enrollees Case-Mix at Each Site
All children enrolled in the program were first categorized into the ADGs. A child could be assigned to more than one of the 34 ADGs based on his or her diagnoses. After assignment to the ADG category, each child was assigned to one of 52 mutually exclusive ACG categories. Table 3 shows the most common ACG categories for the enrollees. A total of 28.6% (4200) of the children included in the analyses had never used health care services (ACG 52). The calculation of actual versus expected health care use was based on the ACG assignment.
Actual Versus Expected Health Care Use With and Without Case-Mix Adjustment
Table 4 contains a description of the enrollees actual health care use compared to the expected health care use at each site. The average number of health care encounters across all sites for a 12-month time period was 2.98 ± 4.6 (SD) visits with a median of 2.0 visits. Row 1 contains the average number of health care encounters at each site which ranged from a low of 2.08 at Site 5 to a high of 3.40 at Site 2.
Several steps were then taken per the users guide accompanying the ACG software to determine if the children were receiving the anticipated number of health care visits based on their diagnoses.11First, we divided the average number of encounters at each site by the group average across all of the sites (2.98), to obtain the overall expected level of use, without adjusting for the case-mix of the enrollees (Row 2). A value of 1.00 means that the actual use and the expected use are identical. Values substantially below 1 indicate underuse and values over 1 indicate overuse of health care services. For example, at Site 2 the unadjusted actual versus expected number of encounters was 3.40/2.98 = 1.14. The ratio of 1.14 means that without adjusting for case-mix, there was 14% higher use than expected. Additionally, without adjusting for case-mix, it appeared that Sites 4 and Site 5 had underuse and Site 1 and Site 3 had health care use that was much closer to the expected.
However, after adjusting for the case-mix of the enrollees in each site using the ACG software, a different picture emerged. Row 3 shows the different expected numbers of encounters at each site after adjusting for case-mix. We then divided the average number of visits at each site by the expected number of visits based on the case-mix adjustment. For example, for Site 2, 3.40/2.98 = 1.07. The use at Site 2, Site 3, and Site 4 was almost 1.00 in each case. Both Site 1 and Site 5 showed apparent underuse.
The developers of the ACG software provide no guidance on how different this ratio of actual versus expected use must be to attribute the observed difference to more than chance. Because the statistical properties of ratios with dependencies between the numerator and the denominator are difficult to derive, a statistical comparison of the difference in the actual versus the expected use was performed instead. The χ2 statistic, with one degree of freedom was calculated for each site and ranged from 4.5 to 128.5 with associatedP values of .03 to <.0001 (Table 4). This implies that the difference between the observed and the site-specific expected average use was more than could be attributed solely to chance. Given the large sample sizes in each of the groups, this finding invokes a nonstatistical consideration; that is, at what point does the observed magnitude of differences have meaningful clinical and policy implications?
The number of public and private efforts to insure children who are not eligible for Medicaid and whose parents cannot purchase private insurance has grown dramatically. These programs are vital for ensuring financial access to care for uninsured children. However, it is essential that such programs are not viewed as merely cost containment efforts.16 Emphasis must be placed on quality assurance activities including both the processes and outcomes of care.17 Assessing the degree to which children receive the health care services they need across multiple delivery settings is an essential yet challenging component of quality assurance.
The Florida Healthy Kids Program represents an important model to evaluate for several reasons. First, with the continued erosion of employment-based health insurance coverage, other states have implemented or are considering the implementation of school-based models of coverage for children.18 In fact, the Robert Wood Johnson Foundation has just announced a major initiative to provide funding to replicate the Healthy Kids Program in other states. Understanding the quality of care that children receive within such a program will allow for careful planning as expansion occurs. Second, this program insures many children who live in working poor families with incomes of $10 000 to $20 000 per year. Children living in these families are more vulnerable to both acute and chronic health care problems than children living in families with higher incomes.16 Assuring that high-risk vulnerable children receive services based on their health care needs is essential given the concern that HMOs' strict utilization and cost control methods may lead to inappropriately low service use. Third, the Florida Healthy Kids Corporation provides coverage for children across multiple settings and has contracts with different HMOs to assume risk and deliver services. Other states also contract with multiple entities to deliver health care. Evaluating care, while controlling for differences in patient severity, is essential for obtaining meaningful information that can be used for ongoing improvements in the health care system.
The ACG software is a useful tool when comparing health care use across sites. As shown in Table 4, without adjusting for case-mix, it would have been easy to misinterpret the children's health care use. For example, Site 1 showed some health care use that was almost on target (1.04) before adjusting for case-mix. After adjusting for case-mix using the ACG software, Site 1 showed health care use that was about 18% below what was expected based on the children's diagnoses. Site 2 showed health care use that was very close to 1 (1.07); whereas the unadjusted value indicated about 14% higher use than was expected. Both the unadjusted and the adjusted values for Site 3 were very close to 1; 0.92 and 1.05 respectively. Before adjusting for case-mix, Site 4 showed health care use that was about 23% below the group average. However, once we adjusted for case-mix, Site 4 showed health care use that was almost ideal use (0.92). Finally, Site 5 showed health care use that was 30% below the group average without case-mix adjustment; and health care use that was 20% below the expected value after case-mix adjustment. Very different conclusions would have been drawn about the health care delivered at each of the sites if the number of visits had been interpreted without case-mix adjustment.
Our analysis indicates that the average use among sites differs by more than could be attributed by chance, after adjusting for case-mix and given the high level of precision of the estimates. The statistical test used was very sensitive to the large sample size used in this analysis (14 688 children). Therefore, it is important to consider the practical significance of the statistical testing in this instance. For example, Site 2 shows a value of actual to expected use of 1.07. The χ2 test indicates that the difference in the mean actual to the expected use is significant. Similarly, Sites 3 and 4 have actual to expected ratios that are very close to the expected (1.05 and 0.92); yet the observed versus the expected means are statistically different. However, in terms of clinical and policy relevance, none of these differences may be meaningful.
Two of the sites show some underuse that in addition to being statistically significant, may be practically meaningful. Site 1 and Site 5 show use that is 18% to 20% below what was expected based on the children's diagnoses. As previously indicated, the ACG software can be used to monitor health care use trends across time. The use at Sites 1 and 5 should be monitored carefully. In addition, this finding could be further explored by looking for other obvious possible causes such as provider underreporting on the encounter forms submitted to the HMOs. Finally, both Sites 1 and 5 are rural; whereas Sites 2, 3, and 4 are urban and suburban areas. It is possible that the underuse in Sites 1 and 5 reflect the limited accessibility and availability of health care providers and facilities in rural Florida.
This study focuses on one process measure of health care quality—the expected number of health care visits based on the child's condition, compared with the actual number of health care visits the children made. For three of the five sites, children are receiving the health care that is expected based on their diagnoses. Two sites potentially have some underuse. Outcomes associated with these health care visits for various acute and chronic conditions were not addressed. Future studies will focus on outcomes of care for selected conditions, such as otitis media and asthma. In addition, future studies will focus on comparing the health care visits reported on encounter forms to the HMOs versus the care documented in the medical record to detect possible underreporting. However, ensuring that children have financial access to care and are, in fact, using the health care services, is an essential foundation for future evaluation efforts.
In summary, health care system reform must focus on both the cost and the quality of health care for children and must include strategies to evaluate the quality of care enrollees receive in diverse settings. The Florida Healthy Kids Program is one mechanism for providing health insurance to previously uninsured children through private sector HMOs and their provider networks. The program represents an opportunity to understand the unique health care experiences of children within managed care environments. Using a method for risk adjustment, children in this program generally are receiving the appropriate amount of health care based on their needs. As legislators contemplate future health care reform initiatives for one of the nation's most vulnerable groups, the school enrollment-based concept of coverage for children represents a promising approach.
- Received January 27, 1997.
- Accepted May 12, 1997.
Reprint requests to (E.S.) Department of Pediatrics, University of Florida and the Institute for Child Health Policy, 5700 SW 34th St, Suite 323, Gainesville, FL 32608.
a A copy of the statistical procedure used is available from the authors on request.
- HMOs =
- health maintenance organizations •
- ACGs =
- Ambulatory Care Groups (software) •
- ADGs =
- ambulatory diagnostic groups
- ↵United States General Accounting Office. Health Insurance for Children: State and Private Programs Create New Strategies to Insure Children. Washington, DC: United States General Accounting Office; January 1996. Health Education and Human Services Division publication GAO/HEHS-96-35
- ↵Group Health Association of America. HMO Industry Profile. Washington, DC: Group Health Association of America; 1994
- ↵Durch JS, ed. Protecting and Improving Quality of Care for Children Under Health Care Reform: Workshop Highlights. Washington, DC: National Research Council, Institute of Medicine; 1994
- Newacheck PW,
- Stein REK,
- Walker DK,
- Gortmaker SL,
- Kuhlthau K,
- Perrin JM
- Hughes DC,
- Newacheck PW,
- Stoddard JJ,
- Halfon N
- ↵Fowler EJ. Health Risk Adjustment for Pediatric Populations: Implications for Children's Hospitals. Economics and Care Delivery Report. Alexandria, VA: National Association of Children's Hospitals and Related Institutions; 1995
- ↵Shenkman EA, Freedman S. Schools as insurance grouping mechanisms: the Florida school enrollment-based health insurance program. Gainesville, FL: Institute for Child Health Policy; 1994
- ↵Silvetta MB, Swartz K. The Uninsured and Uncompensated Care: A Chart Book. Washington, DC: National Health Policy Forum; 1986
- ↵The Florida Healthy Kids Corporation. Annual Report, February 1996. Tallahassee, FL: The Florida Healthy Kids Corporation, 1996
- ↵Johns Hopkins University. The Johns Hopkins ACG Case Mix System, Version 3.01. Baltimore, MD: The Johns Hopkins University; June 1995
- Smith NS,
- Weiner JP
- Fowler EJ,
- Anderson GF
- ↵Berman S. Uninsured children: an unintended consequence of health care system reform efforts. JAMA. 1995;274:1472–1473. Editorial
- ↵United States General Accounting Office. Health Insurance for Children: Many Remain Uninsured Despite Medicaid Expansion. Washington, DC: United States General Accounting Office; July 1995. Health Education and Human Services Division publication GAO/HEHS-95-175.
- Copyright © 1997 American Academy of Pediatrics