Objective. We calculated expenditures for children with one of eight selected chronic health conditions who were enrolled in the Washington State Medicaid program and compared them with payments for all Medicaid-enrolled children. We examined variation in mean, median, and total expenditures and identified expenditure sources.
Methods. This study analyzed Medicaid claims data for 310 977 children aged 0 to 18 who were enrolled at any time in fiscal year 1993. Tracer conditions were used to examine expenditure variation within and between diagnostic groupings. A total of 18 233 children (5.9%) had at least one of the conditions. Expenditures were calculated based on payments made by the Medicaid program.
Results. Children with one of the eight selected conditions incurred mean expenditures of $3800, compared with $955 for all Medicaid-enrolled children. Mean payments associated with the selected conditions ranged from 2.5 times to 20 times more than payments to all children. Approximately 10% of children accounted for approximately 70% of the payments in general and in each diagnostic grouping. Variation in mean, median, and total expenditures was extensive among the conditions. For most conditions, inpatient stays accounted for the greatest proportion of expenditures; for some conditions, durable equipment, home nursing, and medication-related services accounted for substantial proportions of total expenditures.
Conclusions. Medical care for children with selected chronic health conditions is 2.5 to 20 times more expensive than children in general, depending on the condition. A relatively few children account for the majority of expenditures. Extensive variation in mean, median, and total expenditures suggests that different conditions will need to be kept distinct for purposes of establishing payment rates. Children with certain conditions are vulnerable to restrictions in specific services, depending on what restrictions are imposed by a financing program. Further analyses are needed to identify risk-adjustment strategies to support delivery of high-quality services to this population of children as they migrate into managed-care environments.
- chronic illness
- cerebral palsy
- chronic respiratory disease
- cystic fibrosis
- muscular dystrophy
- malignant neoplasms
- spina bifida
Approximately 15% of children enrolled in Medicaid have a chronic health condition.1,2 Although these children are commonly believed to represent the high-cost segment of the childhood population, information regarding variation in cost and expenditure patterns among different diagnostic groupings is sharply limited.1,3
A recent study suggests that service expenditures are somewhat elevated for Medicaid-enrolled children with asthma compared with children in general, but no direct comparisons were made between children with other chronic health problems.4 In the mid-1980s, per capita annual health care expenditures in a national sample of youths with severe mental retardation were found to range from <$100 to $43 000.5 These data, however, are now more than a decade old and may have little relevance to children with chronic medical conditions that fluctuate in symptom expression. Variation in life-span expenditures have been examined for selected conditions,6but these estimates shed little light on yearly expenditures. Overall, better estimates of expenditures are needed to assist in developing policy options as states consider enrolling these children into managed-care systems.7,8 This information is also relevant to managed-care plans and individual physicians contemplating their own capitation arrangements.
A major challenge in estimating expenditures of care for this population involves understanding expenditure variation between and within diagnostic categories. In any given year, for example, services for children with asthma are likely to cost on average less than most children with leukemia because of the clinical treatment protocols associated with the two conditions. In the same year, however, some children with asthma will require many services, with correspondingly high expenditures. Similarly, some children who have been treated for leukemia in previous years will be in remission; they may still carry the diagnosis but require relatively few follow-up services. Variation in service needs and expenditures is likely to be substantial both between and within diagnostic categories. Few data are currently available that can be used to compare annual cost and expenditure patterns for children with diverse chronic health conditions, or to compare these children with all children enrolled in a particular financing program.
Knowledge of cost variation for children with chronic illnesses and disabilities is essential for purposes of program planning. Clinicians, administrators of managed-care organizations, and leaders of advocacy organizations require this information to allocate resources in a manner that will assure that these children have access to and receive needed services of high quality.9
In this study we estimate expenditures of care for children with eight selected chronic illnesses who are enrolled in a state Medicaid program, compare these to expenditures for services provided to all children enrolled in the program, and examine the nature of between-group and within-group variation. We focus on Medicaid-enrolled children with chronic conditions because this population presents particular challenges to states as they consider implementing managed-care programs for the Medicaid population.
Children with selected tracer conditions were used for this study because there is little consensus in the field regarding the definition of children with disabilities or chronic conditions. In addition, a tracer methodology allows for a focused investigation of variability within and among the selected conditions. The specific conditions were selected using a broad set of criteria, including comparatively high occurrence rates within a Medicaid sample and the likelihood of high expenditures for care. In addition, we included conditions that would illustrate a range of care patterns involving medical and surgical services to varying degrees. We excluded conditions such as mental retardation or attention deficit disorder in which associated expenditures are influenced very heavily by services for comorbidities, rather than services for the condition itself. The tracer conditions selected for in this study are asthma, cerebral palsy (CP), chronic respiratory disease (CRD), cystic fibrosis (CF), diabetes, muscular dystrophy (MD), malignant neoplasms, and spina bifida. Associated ICD-9 codes are listed in Table 1. The specific rationale for selecting each of these conditions is available from the lead author.
We obtained from Washington State the enrollment and claims data for all noninstitutionalized Medicaid patients. For the purposes of this study, we identified children aged 0 through 18 who were enrolled in the Washington State Medicaid program at any time in fiscal year 1993. A child was included if one of the selected diagnoses was found in any of the five diagnosis code fields in the claim record for any inpatient or ambulatory encounter during the year. This approach may lead to including children for whom this condition was listed as a rule-out condition but who do not actually have the condition.
For the purposes of this study, children with two or more of the selected conditions were included in the calculations for each of the conditions. For example, if a child had diabetes and asthma, his or her expenditures were included in both the calculations for diabetes and asthma. When calculations were performed on the entire group of children with the selected chronic conditions, however, this child was counted only once.
Child's age was based on age at last encounter in fiscal year 1993. Child's race was determined with information provided on encounter records.
Expenditure Estimation Procedures
Services rendered to children in the Washington State Medicaid program in fiscal year 1993 were reimbursed on a fee-for-service basis according to payment rates established by the Medicaid program for that year. Total payments per enrolled child were calculated by summing payments made on behalf of the child for any encounter during the year. Thus, for the purposes of this report, expenditures are defined to be what the Medicaid program paid to providers as shown on the claim form; payments do not represent costs to service providers for actually delivering the services. Inpatient services generally were reimbursed at a higher percent of charges than outpatient and some physician services.
Payments were categorized into discrete payment categories reflecting broad types of services (eg, inpatient, outpatient, or physician services). The expenditure categories were based on Washington State Medicaid service codes. For the purposes of this report, eight categories were examined: 1) inpatient expenditures (excluding residential treatment facilities); 2) physician expenditures (excluding expenditures for physician services rendered in hospital, which are included in the inpatient category); 3) private nursing services (ie, home health care services); 4) outpatient services other than physician encounters (including emergency room visits); 5) medications and related medication services; 6) durable medical equipment; 7) other providers (including case managers, midwives, genetic counseling, maternity services, and miscellaneous nonphysicians); and 8) other services. The other services category includes laboratory services, radiology-related services, hospice care, and miscellaneous services. We did not routinely separate services within this category because cell sizes became quite small for individual diagnoses; as a result, expenditure estimates are likely to be unreliable.
We elected not to calculate condition-related expenditures because claims data cannot be used to determine clinically whether a service is related to a child's particular condition or to a more general need for care. Thus, expenditure figures represent all payments made for any reason to children with the selected condition.
We elected to use all children enrolled in the Medicaid program as the comparison group because it allows for a standard comparative framework. Another alternative would have been to delete children with the selected conditions from the group of all enrolled children. This decision would have yielded a nonoverlapping comparison group, but the group would still have included children with numerous other chronic conditions. Furthermore, it would have made it impossible to compare our data with other studies of state Medicaid programs if a different group of chronic illnesses were selected. Our decision was determined by the importance of including a standard comparison group (in this case, all children enrolled in a state Medicaid program).
As Table 2 illustrates, of the 310 977 children enrolled in the Washington State Medicaid program in 1993, 18 233 children (5.9%) had at least one of the eight selected chronic conditions. Within this total sample of Medicaid-enrolled children, occurrence rates for the selected conditions ranged from 46 per 1000 enrolled children (asthma) to <1 per 1000 enrolled children (CF, MD, and spina bifida). This database cannot estimate true prevalence rates of children with any particular condition because it includes information only on children for whom a claim was actually made. Most children who have conditions that are complex (eg, spina bifida) will appear in the database because they are likely to receive reimbursable services at least once in a year; consequently, the number of children with spina bifida, MD, and CF, as indicated in Table 2, is close to the Washington State prevalence figures for these conditions as reported elsewhere (J. Neff, personal communication). In contrast, a smaller percentage of children with asthma may appear in the database because many of these children have mild cases and may not require special services that would be coded as asthma-related; consequently, the occurrence rate of asthma in this data set is likely to be less than the true prevalence rate in the state.
Age, gender, and racial distributions across diagnoses generally reflect the expected patterns associated with clinical manifestations of each diagnosis. For example, as expected, the proportion of nonWhites with CF is comparatively low; in addition, the mean age of children with CRD is much lower than the other categories because this diagnosis is generally made in infants and young children. As expected, children with asthma are the single largest subgroup within the sample.
The sample included 525 children who had two or more of the selected conditions. We do not present data on these children separately because this subgroup is small. In addition, many combinations of conditions are not represented in this subgroup because we are including only eight selected conditions; thus, analyses would have pertained only to a sharply limited group of children with multiple conditions. Other reports have investigated the cost implications of multiple conditions.10
The percentile distributions of expenditures for the sample of children with any of the eight conditions and for all Medicaid-enrolled children are shown in Fig 1. Ten percent of both groups account for approximately 70% of total payments. We plotted the same distribution separately for each of the eight selected conditions and found essentially similar curves, as Fig 2 illustrates. Within each diagnostic condition and within the general population of Medicaid-enrolled children, relatively few individuals account for the major share of the expenditures.
Total and Mean Expenditures
As Table 3 illustrates, payments for all Medicaid-enrolled children averaged $955 in fiscal year 1993, compared with $3800 for the group of children with one of the selected conditions. Mean payments ranged from $2359 for children with diabetes to $19 104 for children with CRD. These mean payments vary from 2.5 to 20 times the mean payment for all children enrolled in the state's Medicaid program.
Mean payments tell only part of the story. Table 3 also lists median and total payments for children with the selected condition. Overall, the median payment for the sample of children with one of the selected conditions was $891, compared with a median payment of $290 for all enrolled children. Median payments range from $776 for children with diabetes to $4595 for children with spina bifida. Total payments for the sample of children with any one of the selected conditions is >$69 million, ranging from $2.1 million for children with spina bifida to $37 million for children with asthma.
Table 3 includes condition rankings to illustrate the different ways to compare children with the selected conditions. Children with CRD are generally high on any of the indices (ie, mean, median, and total payments). For other conditions, rankings based on mean, median, and total payments are substantially different because of differences in occurrence rates and service needs across these conditions. Asthma is a clear example of one pattern; it is ranked second to the lowest in mean and median payments but highest in total payments because of high occurrence rates. Children with spina bifida rank fifth highest in mean payments, the highest in median payments, and lowest in total payments because of low occurrence rates. Depending on which payment index is selected, different conditions will be viewed as expensive relative to the others.
ICD-9 codes were examined further to determine potential subgroupings of children that might be associated with particularly high expenditures. For example, children with diabetes include children with diabetes accompanied by ketoacidosis. This subgroup (N = 55) had substantially higher mean expenditures than all children with diabetes ($12 204 vs $2359); the children in this subgroup represented 3.8% of all children with diabetes and accounted for 20% of total expenditures for children with diabetes. Mean expenditures for children with CP accompanied by quadriplegia (N = 305) were substantially higher than children with CP as a whole ($18 830 vs $9887). These specific subgroup analyses illustrate the general pattern in which relatively few children within each diagnostic group account for a disproportionate share of the expenditures. For these two diagnostic categories, ICD-9 coding conventions permit identification of high cost subgroups; for the other diagnostic groupings, this is not now possible.
The distribution of expenditures as illustrated in Fig 2 and the substantial difference between median and mean payments noted in Table3 suggest the presence of a few high-cost outliers—ie, the 10% of cases that account for the majority of payments. To examine this issue further, we calculated the minimum and maximum expenditures associated with the 90th through the 100th percentile for each condition. Table4 presents data for the four conditions (asthma, CP, CRD, and diabetes) in which the total number of cases exceeded 900, assuring that the number of cases in each percentile would equal or exceed nine. We examined the distribution of the cases within each of the top 10 percentiles.
As Table 4 illustrates, dramatic increases in payment occur only in the last several percentiles. This pattern indicates that only an extremely small number of cases have very high payments. For example, the minimum payment in the group of 143 cases in the 100th percentile for children with asthma is $27 318, whereas the maximum payment is $373 846; the minimum payment in the group of nine cases in the 100th percentile for children with diabetes is $23 249, whereas the maximum payment is $221 932.
It is likely that many children in the highest percentiles have one or more comorbidities.10 In children with asthma, for example, these comorbidities may include bronchopulmonary dysplasia or aspiration associated with mental retardation.
Payments by Service Category
Table 5 illustrates the distribution of payments across selected service categories for all Medicaid-enrolled children, for children with at least one of the selected conditions, and for children with each of the selected conditions.
Inpatient and home nursing expenses for children with at least one of the selected conditions are proportionately higher by a substantial margin compared with all enrolled children (45% vs 29% and 9% vs 3%). Physician and outpatient services account for a lower proportion of total payments for children with at least one of the selected conditions compared with all enrolled children.
Within payment categories, the proportion of expenditures varies considerably across diagnostic groupings. For example, within the medication category, expenditures vary from <1% of total expenditures to 12% of total expenditures. Much of this variation can be explained by the different biological processes of the selected conditions. For example, expenditures of infants with CRD are largely related to inpatient services because these infants remain in intensive care units for relatively long periods of time. Many youngsters with MD, especially as they move through adolescence, require increasing amounts of home health services; hence, it is expected that expenditures for this care will be proportionately high, as Table 5 indicates.
Among the diagnostic groupings, inpatient services account for 25% to 71% of total payments. This category accounts for a substantially high percentage of total costs for each diagnostic category except CP (in which durable medical equipment accounts for almost one-quarter of all payments, approximately the same as inpatient expenditures) and MD (in which home health care services account for the 40% of expenditures, compared with the 28% of total payments accounted for by inpatient services).
The data in Table 5 imply differential patterns in outpatient service use across the different diagnoses. For example, claims for physician services, as a proportion of total expenditures, are comparatively higher for children with asthma or diabetes than for children with any of the other selected conditions. This reflects a heavy reliance on outpatient care for children with these diagnoses. For children with CF, medication-related services (ie, drugs and infusion-related services) reflect the second highest category of expenditures relative to the other categories, suggesting that medications and their administration figure heavily in the care of children with this condition.
Among the selected conditions, durable medical equipment accounted for 24% of the payments for children with CP and 12% of payments for children with spina bifida. These data suggest that these two groups of children would be particularly affected by limitations in insurance coverage for durable medical equipment.
The “other” category includes payments for a wide range of services, including laboratory tests and radiology fees. A substantial portion of the other category is coded as “miscellaneous” in the Medicaid database; it is not possible to identify what services were provided under this code. For all children enrolled in the program, approximately 50% of all payments in the “other” category are coded as miscellaneous. For children with at least one of the selected conditions, this figure is 36%.
Children with selected chronic conditions enrolled in the Washington State Medicaid program in fiscal year 1993 incurred mean yearly medical expenditures that exceeded those for children without any of the selected conditions. On average, 10% of children with one of the selected conditions accounted for approximately 70% of their costs. This is quite similar to the expenditure distribution for all enrolled children.
Mean, median, and total expenditures varied considerably among the conditions because of differences in associated occurrence rates and service needs. For example, the condition with the second lowest mean payment (asthma) had the highest total expenditures; the condition with the highest median cost (spina bifida) had the lowest total expenditures. Some conditions had more extreme outliers than others.
For some conditions, it was possible to use subordinate ICD-9 codes to identify subgroups within the overall diagnostic category that included children with particularly high expenditures. For example, the mean expenditures for children with diabetes who have also experienced ketoacidosis are higher than for children with diabetes as a whole, largely as a result of higher inpatient expenditures. For most conditions, identifying the high-cost subgroup was not possible because existing subordinate codes are not consistently associated with high-cost treatments.
Our analyses also indicate that the contributions of particular services to total expenditures vary considerably across the diagnostic categories. Inpatient expenditures represent approximately 71% of total expenditures for children with CRD, for example, but only 25% of total payments for children with CP. Home nursing services accounted for 40% of the total expenditures for children with MD, but <15% for children in most of the other condition groupings.
Overall, the eight selected conditions were found to be similar in certain respects and different in others. They are similar in that 1) children with any of the conditions will incur higher expenditures on average than the entire group of Medicaid-enrolled children and 2) a relatively few children account for the majority of expenses.
The eight conditions are different in terms of 1) associated mean, median, and total expenditures in a given fiscal year, 2) proportional amounts accounted for by specific payment categories, and 3) the absolute number and nature of the outliers. Different outlier patterns can substantially influence cost estimates, and the development of predictive statistical models that account for them represents an important goal for future projects.11 Overall, our analyses suggest that different conditions will need to be kept distinct for purposes of estimating costs and developing accurate capitation rates. Methods for adjusting risk for children with asthma, for example, may not be reasonable for other chronic conditions. Developing the information base needed for calculating capitation rates will require longitudinal cost data on a substantial number of children within each diagnostic grouping.
Our analyses also underscore the potential vulnerability of children with certain conditions to insurance policies that exclude or limit access to certain services.12 For example, children with MD will be particularly affected by policies that limit private duty nurses; families of children with CF will be disproportionately affected by policies that require copayments for medications or medication-related services; and children with CP will be affected by limitations on durable medical equipment. In each situation, limiting these services may lead to increased total expenditures because of costs for subsequent inpatient stays or treatments related to preventable secondary health conditions.
To date, states that have established Medicaid managed-care programs typically carve out children with special health care needs or make enrollment voluntary.8 Throughout time, however, these children are likely to be included in managed-care programs.7-9 Pediatricians and other child health professionals will need to assure that policies adopted by managed-care organizations do not affect this population adversely. We believe that part of the problem involves managed-care organizations' relative lack of knowledge and experience with this population. In the absence of little fiscal information on which to base policies, managed-care programs are understandably hesitant to include this group. Additional information on expenditures and service use is needed to demystify this population for managed-care organizations, to build the foundation for reasonable risk adjustment to account for higher expenditures in this population, and to provide child health professionals with the foundation of knowledge needed to argue for sound policies and comprehensive benefit packages.
Some pediatricians or pediatric groups are being asked to accept a capitated payment to care for children. Results from this study may be useful to help negotiate rates if their practices are likely to attract a disproportionately high percentage of children with chronic illnesses. It will also be useful to help them allocate expenditures across different types of services if they accept capitation for all covered services.
These data may also be useful to state regulatory agencies concerned that access to needed services is threatened for children with chronic health conditions. For example, a precipitous decrease in mean expenditures for a particular diagnostic group in a managed-care organization or substantial alterations in sources of expenditures throughout a short period of time may indicate changes in the benefit package that would adversely affect quality of care for children with a particular condition. Although this database does not show what services these children need to have access to, the data do show where they have received services in a fee-for-service environment. Regulatory agencies and Medicaid agencies need to assure that managed-care plans offer appropriate access to these services. Cost and expenditure data on this population of children provide a potential means for monitoring quality of care provided by managed-care organizations.
This study provides basic descriptive information that begins to fill a critical gap in knowledge, but it is limited by the inclusion of Medicaid data from only one state and by constraints found in claims databases. Similar analyses using data from other states are needed to determine whether the patterns identified in this report are similar in other Medicaid populations. In addition, it would be of substantial interest to compare percentile distributions across service categories in fee-for-service and managed-care systems for both Medicaid and employer-based pools. However, assembling the requisite databases (and assuring that data from each are actually comparable) presents major technical challenges because of different classification and coding conventions.
Within the group of children with disabilities and chronic illnesses, a relatively small percentage of individuals will account for most of the medical care expenditures. However, few data are available to identify high-cost cases because most available studies have not accounted for critical factors, including age of child, primary diagnosis, comorbidities, access to care, rates of hospitalization or institutionalization, and insurance status. Also, the pattern of medical and health-related expenditures for children with chronic illnesses can vary considerably from one year the next in response to biological, developmental, and family factors. Additional analyses are needed to address these issues.
Nonetheless, the data presented in this study and other reports represent a foundation on which to build informed advocacy efforts. Pediatricians, the leadership of parent organizations, and other child health professionals must work collaboratively with directors of managed-care organizations to develop the necessary databases that will inform ongoing policy debates concerning quality of care for children with special needs and their families.
This project was supported by funds from the David and Lucille Packard Foundation and the Commonwealth Fund awarded to Dr Anderson; and from a School of Hygiene and Public Health Faculty Development Award to Dr Ireys. The Child and Adolescent Health Policy Center of the Department of Maternal and Child Health also provided support for this project.
We would like to thank Roger Gantz, MA, Department of Social and Health Services, Washington State Medicaid agency, for facilitating access to data and Carol Han for assistance in the early phases of this work.
- Received October 16, 1996.
- Accepted February 19, 1997.
Reprint requests to (H.T.I.) Department of Maternal and Child Health, 624 North Broadway, Baltimore, MD 21205.
- CP =
- cerebral palsy •
- CRD =
- chronic respiratory disease •
- CF =
- cystic fibrosis •
- MD =
- muscular dystrophy
- Newacheck PW,
- Hughes DC,
- Stoddard J,
- Halfon N
- ↵Regenstein M, Meyer JA. Low Income Children with Disabilities: How Will They Fare Under Health Care Reform? Portland, ME: National Academy for State Health Policy; 1994
- ↵Birenbaum A, Guyot D, Cohen H. Health Care Financing for Severe Developmental Disabilities. Monographs of the American Association on Mental Retardation (M. Begab, Series Editor), Number 14. Washington, DC: American Association on Mental Retardation; 1990
- Waitzman N,
- Romano P,
- Scheffler R
- ↵McManus P, Fox HB, Newacheck, PW, McPherson M, Roy B. Strengthening Partnerships Between State Programs for Children With Special Health Needs and Managed Care Organizations. Washington, DC: Maternal and Child Health Policy Research Center; March 1996
- ↵United States General Accounting Office. Medicaid Managed Care. Servicing the Disabled Challenges State Programs. Washington, DC: GAO/HEHS-96–136; July 1996
- Ireys H,
- Grason H,
- Guyer B
- Fowler L,
- Anderson G
- Copyright © 1997 American Academy of Pediatrics