PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of hsresearchLink to Publisher's site
 
Health Serv Res. 2004 June; 39(3): 665–692.
PMCID: PMC1361030

Emergency Department Use among Michigan Children with Special Health Care Needs: An Introductory Study

Abstract

Objective

To describe patterns of emergency department (ED) use among children dual-enrolled in Medicaid and Michigan's Children's Special Health Care Services (CSHCS).

Data Sources

Individual claims and enrollment data from Michigan's Medicaid and CSHCS programs for the period January 1, 1998, to June 30, 1999. Claims data were linked with eligibility data and then used to develop a 100 percent sample of claims for individuals enrolled in both Medicaid and CSHCS.

Study Design

Poisson regression analysis was used to examine the rate of ED use for dual-enrolled children. A time-varying hazard analysis was also used to examine the impact of changes over time. The key variables were gender, age, race, county of residence, Medicaid eligibility category, and qualifying diagnosis.

Principal Findings

Dual-enrolled children under one year of age, and those with qualifying diagnoses of anemia, hemophilia, asthma, epilepsy, and juvenile diabetes displayed especially high rates of ED use. Significant geographic variation in ED use remained after controlling for qualifying diagnoses, race/ethnicity, and other factors. African Americans displayed higher rates of ED utilization than non-Hispanic whites. Supplemental Security Income (SSI) recipients demonstrated higher utilization than other groups.

Conclusions

Children dually enrolled in CSHCS and Medicaid face diverse challenges of both poverty and chronic illness. Differences in patterns of use highlight the importance, but also the difficulty, of developing systems of care to manage complex chronic conditions in low-income populations.

Keywords: Emergency department, Medicaid, chronic illness, children with special health care needs, Title V

In 1998 children under 15 years of age made more than 22 million visits to emergency departments (EDs) (McCaig 2000). Children's patterns of ED use have important implications for public policy, clinical practice, and the well-being of children and their families. Inappropriate ED use may produce needless expenditure or may reflect cost-ineffective patient care. Emergency department utilization for emergent but preventable conditions may also reflect access barriers to primary care. Such patterns may also reflect poor medical management of chronic conditions. In contrast, overly stringent efforts to discourage ED use may lead to improper handling of an emergent condition. Underutilization may produce adverse health outcomes or may ultimately produce needless expenditure due to delayed provision of appropriate care.

This constellation of issues has spawned a large literature on ED use in the general pediatric population. In recent years, much research has focused on the implications of managed care and different types of insurance on ED use (Butler 1998; Chande, Krug, and Warm 1996; Gadomski et al. 1995; Holl et al. 2000; Hodge 1999). Other pediatric research examines characteristics of children who are frequent ED users or who visit the ED as their usual source of care (Bond, Stearns, and Peters 1999; Halfon et al. 1996; Wilson and Klein 2000; Yamamoto et al. 1995; Zimmerman et al. 1996).

However, little research has been published regarding the causes and correlates of ED use among children with chronic or complex health conditions, also referred to as children with special health care needs. This population includes as many as 18 percent of U.S. children younger than 18—depending on the definition used—and may be growing (Newacheck et al. 1998).

Research on ED use within the general pediatric population may not be applicable to the unique circumstances of chronically or severely ill children. Care for children with special health care needs is often expensive and complex, and children with special health care needs often have frequent contact with health care delivery systems (Neff and Anderson 1995; Newacheck et al. 1998). Although some recent studies have described ED use by children with asthma (Halfon and Newacheck 1993; Lozano et al. 1997; Miller 2000; Price et al. 1999; Szilagyi et al. 2000), children with other chronic illnesses may experience quite different utilization and cost trends (Neff and Anderson 1995). A study by Reynolds et al. (1996) did describe variation in ED utilization by children with chronic conditions, but the study was conducted at a single pediatric ED over a brief period and did not examine the possible correlates of the variation observed.

Children with special health care needs are disproportionately poor and socially disadvantaged (Newacheck et al. 1998). On the other hand, children with acute or chronic conditions often have access to public assistance not available to other children. Most notably, child enrollment in Supplemental Security Income (SSI) has sharply increased in recent years (Garrett and Glied 2000; Perrin et al. 1999). Children eligible for SSI are generally provided with Medicaid coverage. Children with disabilities also receive non-Medicaid services based on Title V of the Social Security Act.

Previous research indicates that publicly insured children are more likely than uninsured or privately insured children to visit the ED (Chande, Krug, and Warm 1996; Halfon et al. 1996; McCormick et al. 2000; Shatin et al. 1998). No study has yet examined health services use by chronically ill children across the realm of public assistance programs. This research gap is increasingly important, as a growing number of states seek to improve access and to manage care and costs by moving Medicaid-enrolled children with special health care needs into capitated health plans (U.S. General Accounting Office 2000).

This paper provides a preliminary analysis of one important population of chronically ill children, those dually enrolled in Michigan's Medicaid and Title V programs for children with special health care needs. It provides the first detailed analysis of ED use within a dual-enrolled population of low-income children who have chronic health needs. It also provides a detailed analysis of how qualifying diagnosis, age, gender, race/ethnicity, Medicaid eligibility, and geography jointly influence ED use.

Data and Methodology

Program Description

Michigan's Title V program for children with special health needs, referred to as Children's Special Health Care Services (CSHCS), serves eligible children from birth to age 21. The program also serves some older individuals who have cystic fibrosis or blood disorders such as hemophilia. Because more than 99 percent of our study subjects were younger than 21, we refer to them as children unless otherwise noted.

Eligibility for CSHCS is based on Michigan residency and qualifying diagnosis. Factors considered include the type, severity, and chronicity of medical condition, and the need for pediatric specialty care. Some prevalent disorders, such as autism, dyslexia, emotional and learning disorders, and mental retardation are not qualifying diagnoses. Type 2 diabetes is beyond the present scope of the CSHCS program.

The CSHCS program provides a wide range of specialty care and community support services related to qualifying diagnosis. Enrollees may also have Medicaid, MIChild (Michigan's CHIP program), or private insurance. Children's Special Health Care Services is the payer of last resort, which reimburses important services uncovered by other public and private plans.

Administrative Database

This study is based upon a 100 percent sample of paid claims from the Michigan Medicaid and CSHCS programs during the period January 1, 1998, to June 30, 1999. Based upon historical analyses of Medicaid and CSHCS claims, state program staff estimate a completion factor of more than 98 percent for available claims. Program enrollment and demographic characteristics were obtained from state CSHCS eligibility data, which were linked with claims data by encrypted ID numbers.

Although implementation of the CSHCS managed care alternative began during our study period, the preponderance of study subjects received coverage through the fee-for-service (known as the Basic Health Plan) service delivery option. Voluntary managed care enrollment, which began in September 1998, accounted for less than one percent of person-months over the period of the current study.

Study Population

The focus of our study is an economically disadvantaged subset of children enrolled in the CSHCS program—those also enrolled in Medicaid. To have the most complete claims data possible for dual-enrolled children, we focused on children who had no health insurance coverage other than Medicaid and CSHCS.1 As discussed below, we performed sensitivity analysis to examine the impact of excluding children with other sources of insurance coverage. We constructed two analysis files: one at the person-month level and one that aggregated person-months to the person level. In the person-month sample, we included only person-months of dual Medicaid and CSHCS enrollment and excluded person-months in which the subject had other health insurance coverage in addition to Medicaid and CSHCS. The resulting person-month sample consisted of 136,096 person-months of dual Medicaid/CSHCS enrollment and no other insurance.

To construct the person-level sample, we allowed for the possibility that some characteristics may change for a given person over time. Enrollment in Medicaid or CSHCS, the presence of other health insurance coverage, primary qualifying diagnosis, and some demographic characteristics may change due to changing health status, changing family composition and financial circumstances, administrative churning, and other factors. In addition, program enrollment dynamics are not yet well understood and could be related to ED utilization. To assess whether changes in enrollment affected ED use, we flagged individuals with discontinuous enrollment or intermittent coverage by other health insurance. Inclusion of this term had a small impact on our results.

We included individuals in the person-level sample if they were dually enrolled for at least one month and had no private coverage in any month during the study period. The resulting person-level sample consisted of 10,800 dual-enrolled individuals. Ninety-two percent of these individuals were continuously enrolled in both programs, with the remainder having some intermittent periods of enrollment. Person-months in which the individual was not dually enrolled were not included in our analyses.

Definition of ED Visits

Emergency department services for dual-enrollees were identified based upon procedure and revenue codes reported on administrative claims for professional and technical services during the study period. A claim was considered to be an ED visit based upon unique date of service for professional services coded with Current Procedural Terminology (CPT) codes 99281 through 99285.

In some instances, professional services claims were apparently not submitted, and only facilities charges were found for a given occasion of service. In these cases, an individual was considered to have had an ED visit if a claim with an identified ED revenue code was found for facility charges on a unique date of service and a corresponding professional services claim was not found. All instances where a given patient had multiple professional or technical service claims for the same date of service were identified, and duplicate claims were eliminated from the analysis dataset.

Several study subjects were observed to have obtained ED services from more than one hospital or physician on a single date. These situations were believed to reflect the same episode of care, and were considered to be a single visit for the purposes of this study. In total, 6,277 unique ED visits were identified for our study population in person-months for which dual-enrolled study subjects had no other insurance coverage.

Independent Variables

The independent variables for our analyses are gender, age, race/ethnicity, Medicaid eligibility category, county of residence, and primary qualifying diagnosis. Medicaid eligibility category refers to seven categories of public assistance programs that form the basis of recipients' Medicaid eligibility. Two of these categories (SSI and TANF) refer to cash assistance programs that confer Medicaid eligibility. In addition, the Low Income Family (LIF) program provides Medicaid to families with low income. Healthy Kids is the main Medicaid program for children under 19. Medicaid for the Disabled is for disabled persons who do not otherwise qualify for SSI. Medicaid for Persons Under 21 and Medicaid for Caretaker Relatives both provide coverage to low-income recipients who do not otherwise qualify for cash assistance.

County of residence was included as a separate dummy variable whenever a county's dual-enrolled population accounted for at least two percent of the total dual-enrolled study sample. Remaining counties were grouped in the “Other” category.

Qualifying diagnosis is defined in CSHCS eligibility data by ICD-9-CM code.2 For individuals with multiple diagnoses, the primary qualifying diagnosis is used for analysis. We grouped these codes into 14 diagnostic categories for detailed analysis. These diagnoses were chosen based on historic CSHCS expenditure and enrollment trends, as well as on the clinical judgment of the study team. See the appendix for definitions of the diagnostic categories.

Statistical Methodology

We employ two statistical approaches, using the PC-SAS software package version 6.12 (DiIorio 1991). Our first approach, Poisson regression, provides an aggregate analysis of ED utilization for each individual over the entire period of dual enrollment and seeks to predict the rate of ED use based upon individual characteristics (Cameron and Trevidi 1998; Pollack et al. 2002). In particular, if individual j is observed for Tj periods of available data, we assume that the probability of M visits to an ED during that period is given by

PM,j=(λjTj)MM!exp[λjTj]
(1)

Here λj is the arrival rate for ED visits and is presumed to be a function of some vector Xj of personal characteristics. Xj includes race/ethnicity, gender, primary qualifying diagnosis, and county of residence. Program exposure Tjis the number of months that individual j is observed as being dual-enrolled without other health insurance coverage.

We assume that λj takes the form

λj=exp[Xjβ]
(2)

Here β is a set of coefficients estimated from the available data. Given this specification, estimated coefficients are most interpretable when converted to the incidence rate ratio, IRR=exp[β]. For each explanatory variable, the IRR represents the proportional change in the rate of ED use associated with a one-unit change in the independent variable. For example, we find below that African Americans have an IRR of 1.22 when compared with the default category of non-Hispanic whites. This implies that African Americans have a 22 percent higher rate of ED use within a given person-month.

One shortcoming of Poisson regression is that it predicts narrow variance in the rate of ED use. This assumption does not bias the coefficients, but it can produce overly tight standard errors. Because actual ED use reflects unobserved differences in health status, family characteristics, and preferences that vary widely across patients, observed utilization often displays higher variance than the Poisson model allows. We adjust our standard errors empirically for this overdispersion. In particular, if the estimated ED arrival rate is λ, the corresponding variance is posited to be [var phi]λ. These adjustments do not alter our coefficients, but they do increase associated standard errors. We use the SAS GENMOD procedure to calculate all coefficients and standard errors.

Our second analytical approach was to perform a time-varying hazard analysis using a multivariate logistic regression analysis. The dependent variable is the probability of an ED visit in a given person-month, taking into account any changes in an individual's enrollment status and related characteristics. Because the mean number of monthly ED visits is small, logistic regression closely approximates a proportional hazard model (Abbott 1985). In particular, for individual j in the kth month, we create a dummy variable vj,k which is set to 1 if that individual made at least one ED visit in the given month, and is set to zero otherwise. We then estimate the logistic regression model

Pr[υj,k=1]=exp[γXj,k]1+exp[γXj,k]
(3)

Individual preferences, family circumstances, and other factors suggest that there will be correlations in the error term over time for the same person. We therefore estimate equation (3) within a Generalized Equation Estimation (G.E.E.) framework. We again use the GENMOD procedure to compute coefficients and standard errors. We use an exchangeable within-group correlation matrix in our G.E.E. analysis.

Because we seek to examine a single medical encounter that occurs on a specific day rather than an aggregate measure covering the entire period of dual enrollment, we can also provide more precise examination of time-varying factors such as the presence of other insurance and enrollment in public assistance programs. We therefore use person-month eligibility data to compute appropriate values for enrollment and demographic characteristics during the month in which ED services were received.

Additional Analyses

The impact of specific diagnoses may vary by age, race, Medicaid eligibility category, and other factors. We therefore explored several interactive specifications to examine the magnitude of such effects. In general, we had little information a priori concerning which interaction effects might be most important. We therefore explored interaction effects by estimating the main person-year G.E.E. model within different population subgroups. Because this approach produces a very large set of point estimates, we summarize the main results. A full set of parameter estimates is available from the authors.

Results

Descriptive Results

Table 1 shows descriptive statistics for dual-enrolled children without other health insurance. Fifty-five percent were male. Under one percent (n=43) were over 21 years old. Fifty-two percent of children were listed as white, 31 percent were African American, and 4 percent were Hispanic/Latino. About one-third of study children lived in Wayne County, which includes Detroit.

Table 1
Descriptive Statistics for Dual Medicaid/CSHCS Enrollees without Other Insurance

Cerebral palsy (CP) was the most prevalent qualifying diagnosis, accounting for 16 percent of person-months, followed by diseases of the ear and mastoid process, which accounted for 12 percent. The 15 diagnoses selected for in-depth study accounted for approximately 70 percent of the study population.

Twenty-four percent of the study population visited the ED at least once between January 1998 and June 1999, for a total of 6,277 visits. Just under half of these children had one ED visit, while slightly less than two percent had more than 10 ED visits, with a maximum of 39 visits. The mean rate of ED use, shown in Table 1, was adjusted for duration of program enrollment so that it can be interpreted per person per year.

As illustrated in Table 1, rates of ED use varied substantially across the sample. Emergency department utilization differed substantially by age, with infants (those younger than one year) having about three times the rate of ED use as other age categories. African American and Hispanic/Latino children displayed higher rates of ED use than did non-Hispanic whites. Rates also differed by qualifying diagnosis, with anemia, hemophilia, and respiratory distress syndrome associated with sharply elevated use. Receipt of SSI and TANF were associated with higher ED use than were other categories. Emergency department use among the seven children included in the “Other Eligibility” category was also very high. Two urban counties (Wayne and Saginaw) demonstrated the highest rate of ED use. The relative variation in ED use, indicated by the coefficient of variation (CV) in ED use (see far right column) was remarkably similar for most of the subgroups.

Regression Results

Table 2 shows the estimated incidence rate ratios from both Poisson and G.E.E. regressions. Gender was not a statistically or clinically significant predictor of ED use. In contrast, age was a critical factor even after adjusting for qualifying diagnosis and for other factors. In both models, infants had far higher rates of ED utilization than did older age groups.

Table 2
Emergency Department Utilization Regression Results

African Americans and Hispanics displayed higher utilization than did non-Hispanic whites, though this difference was statistically significant only for African Americans. Unadjusted, observed African American utilization was nearly 50 percent higher than was found among non-Hispanic whites. Adjusting for other covariates, African Americans were predicted to have more than 20 percent higher utilization. In part, differences were reduced because African Americans are overrepresented in diagnostic categories associated with high ED use. African Americans accounted for almost 80 percent of anemias, due to the prevalence of sickle-cell disorders within this population.

We also found statistically significant differences in ED use across Medicaid eligibility categories. In both the Poisson and G.E.E. specifications, SSI recipients displayed higher rates of ED use than did other Medicaid eligibility categories. The small “Other Eligibility” category was also associated with increased use.

The large (relatively low-utilization) diagnosis “diseases of the ear and mastoid process” is the reference category in our statistical regressions. The qualifying diagnoses of anemias, asthma, epilepsy, hemophilia, or juvenile diabetes were associated with higher ED use.3 The qualifying diagnoses of cleft lip or palate and malignant neoplasms exhibited low ED use similar to the ear and mastoid category. Patterns were similar using the G.E.E. regression.

We also performed sensitivity analysis in which we included person-months when dual-enrollees had other insurance. Other insurance was associated with sharply lower ED use. We also found larger county and race/ethnicity effects when we included these person-months. Excluding observations with other coverage may therefore somewhat understate the role of sociodemographic factors correlated with private coverage.

Interaction Analyses

Although we estimated many interactive models, we found age by diagnosis, age by Medicaid-eligibility category, race by age, and seasonality by diagnosis to be the most important. Detailed results are available from the authors.

Age by Diagnosis

Because age and qualifying diagnosis were such critical factors, we investigated whether age-patterns of ED use differed across diagnostic categories. For example, among dual-enrollees between one and four years of age, leukemia, and respiratory distress syndrome were associated with especially high utilization when compared with the reference ear and mastoid diagnosis. Point estimates for children with leukemia, and respiratory distress syndrome between the ages of five and nine years were much smaller.

In contrast, hemophilia, asthma, and anemias were associated with heightened ED use among children between the ages of ten and fourteen. For several qualifying diagnoses, we observed virtually no ED visits among children older than fourteen. For other diagnoses, including cystic fibrosis, juvenile diabetes, hemophilia, and epilepsy, ED use was greatest among children aged fourteen and older.

Medicaid Eligibility Category and Age

We also observed differences between SSI and TANF recipients that increased with child age. Among one- to four-year-olds, TANF recipients had an incidence rate ratio of 0.76 when compared with SSI recipients. Among children ages five to nine, the IRR associated with TANF declined to 0.46 and remained below 0.50 in older age groups.

Seasonality by Diagnosis

Because ED use may reflect seasonal differences in children's susceptibility to illness, we included season in the regression analysis. We found slightly higher ED use during the spring (March, April, and May) than in other seasons. Analysis of interactions between season and respiratory diagnoses (asthma, cystic fibrosis, and respiratory distress syndrome) yielded statistical significance only for respiratory distress syndrome. Emergency department visits were highest during winter and lowest during spring. The IRR for spring compared to winter is 0.58.

Interactions by Race

Different patterns in utilization by race/ethnicity are of special concern to both policymakers and researchers. We therefore paid special attention to differences between African Americans and whites in our data. We estimated our Poisson regression models separately for the two groups. We also estimated pooled specifications (not shown) in which we examined specific interactions such as different age-patterns of utilization between the two groups.

Table 3 shows the result of our subgroup regressions. Although the coefficients were significantly different between the two groups (p<.001), utilization patterns were similar across county of residence, programmatic categories, qualifying diagnoses, and age. Within our most prevalent diagnoses, the African American sample yielded notably larger point estimates of incidence rate ratio for asthma, anemias, and hemophilia. Differences in ED use between SSI and TANF recipients were slightly higher in the African American than in the white group, though these differences were not statistically significant.

Table 3
Emergency Department Utilization: Subgroup Regression Results

In our pooled, interaction specification, we examined black–white differences in age and qualifying diagnoses. We found that black–white differences in ED utilization increased significantly with age. Among one- to four-year-olds, controlling for other factors, the estimated IRR for African Americans was statistically insignificant. The IRR for African Americans steadily increased with age, and exceeded 1.35 among children aged 15 and above. (Differences were concentrated in the group aged 15–20. Results were not significantly within the small group of individuals age 21 and older.)

We then estimated diagnosis-specific interactions to examine how black–white differences varied across diagnoses. Racial differences were significant in only two of these categories. Compared with non-Hispanic whites with the same qualifying diagnoses. African Americans showed significantly higher ED use for nervous system disorders (IRR=1.73) and endocrine disorders (IRR=2.54). African Americans showed notably lower rates of ED use for neoplasms (IRR=0.57), though the latter finding was not statistically significant.

Summary and Discussion

This paper provides detailed descriptive analyses of ED use within an important population of children—those challenged both by poverty and by complex illness. Our analyses were not designed to explore whether observed use of the ED was medically or economically appropriate. We do not examine the intensity or specific services performed in the ED. Thus, our results should not be interpreted as indicating overuse or underuse of ED services within any population. Rather, we hope that our findings illuminate key sources of variation in ED use within this important population.

Infants exhibited rates of ED use (1.63 visits per person-year) more than twice as high as those of any other age group. No prior study has permitted such detailed comparisons among children of different ages, whether among children with special health needs or among children in general. Studies that have examined children's ED use by age have shown substantial differences between infants and older children. A recent national estimate of ED use among the general population of children under 15 years was 0.37 visits per person-year (McCaig 2000), and a separate analysis of infants alone found an overall rate of 0.94 ED visits per person-year (Sharma et al. 2000).

The high rate of ED use for infants is consistent with clinical practice. In general, infants have more limited physiologic reserve than older children and are susceptible to more rapid deterioration of their health status in the setting of acute illness. Infants under three months of age, in particular, are more susceptible than older children to severe systemic infection because of their immature immune system. Moreover, infants cannot articulate their complaints and may therefore require more extensive effort to diagnose than older children (McCarthy 2000). As a result, the accepted standard of care sets a low threshold for prompt laboratory evaluation (e.g., tests of blood, urine, and cerebrospinal fluid) for infants who may be acutely ill (Baraff et al. 1993). Emergency departments may be better equipped for such evaluations than many primary care settings.

Children with diagnoses of hemophilia, juvenile diabetes, anemias, asthma, and epilepsy, use the ED especially frequently. These diagnoses are often susceptible to acute, potentially life-threatening exacerbations that require emergent evaluation and treatment. Children with hemophilia may develop severe bleeding, while diabetic children are at risk for dangerously uncontrolled blood glucose levels. As expected, children with conditions typically remediable through elective surgery—for example, cleft lip or palate, or diseases of the ear and mastoid process—had lower-than-average rates of ED use. Because some conditions make children especially susceptible to acute illness or urgent medical needs, expectations for appropriate ED use must be sensitive to specific diagnoses.

Emergency department use also varies among children with the same qualifying diagnosis. We suspect that such within-diagnosis variation reflects underlying heterogeneity in illness complexity and severity. For instance, some children with hemophilia develop resistance to available treatment modalities (Gill 1999).

Variation in ED use may also reflect other access barriers to primary care, including the absence of a “medical home,” or varying continuity of primary care services (Halfon et al. 1996; Christakis et al. 1999; Jones et al. 1999; Wilson and Klein 2000). Adjusting for diagnoses and other factors, African Americans had roughly 20 percent higher ED use. Further, unadjusted for potential confounders, African Americans have roughly 50 percent higher ED use than non-Hispanic whites. These patterns are consistent with the possibility that diagnoses prevalent among African Americans are undertreated in the primary care setting.

We also examined interaction effects. The effects of qualifying diagnosis on ED use depended in part on child age and seasonality. These patterns matched clinically plausible accounts. Children with respiratory distress syndrome were significantly more likely to visit the ED in winter; this is likely because of the higher incidence of upper respiratory tract infections during fall and winter months (Herendeen and Szilagyi 2000). We did not observe seasonal effects for such respiratory conditions as asthma and cystic fibrosis. In certain conditions, exposure to airborne irritants or the adequacy of maintenance therapy may be more important in precipitating acute illness. These factors may not exhibit the same seasonality as do upper respiratory infections.

We did not examine whether ED visits led to hospital admission, or attempt to identify evidence of an outpatient medical home. We are now analyzing diverse measures of resource consumption to explore these questions in greater detail. Such analyses may suggest principles of clinical management that help develop standards for medical and nonmedical service delivery in the ED, inpatient, and outpatient settings.

Our analyses also revealed significant differences in ED use across public programs. Recipients of SSI used the ED significantly more often than did recipients of TANF or children who were Medicaid-eligible for other reasons. These effects were greatest among older children. Children eligible for SSI may be more easily identified later in childhood. Higher ED use among SSI recipients may also reflect the fact that SSI is the predominant form of eligibility defined by medical, as well as monetary, need. By design, SSI-eligible children meet stringent disability criteria, while other Medicaid-eligible children experience a wider range of illness severity. The SSI eligibility process, therefore, may select children with greater disability—and perhaps greater need for ED resources—than is found within the broader population of children with special needs.

Differences in ED use by SSI status increase with age. These results may suggest that health differences between SSI and TANF recipients increase over childhood. Alternatively, chronically ill children may sort into SSI as their medical situation becomes sufficiently well understood and well documented to justify SSI coverage.

We observed significant geographic variation in children's ED use, although the effect was smaller than for children's age, diagnosis, and Medicaid eligibility category. Controlling for age, gender, race/ethnicity, qualifying diagnosis, and Medicaid eligibility status, children who reside in Michigan's largest urban area, Wayne County, used the ED significantly more often than did children in many other, more rural areas. Other research has found that childhood ED use is higher in rural rather than urban areas (McCaig 2000; Sharma et al. 2000). Our findings may reflect the increased availability of specific pediatric emergency services in metropolitan areas suited to children with complex medical illness. They may also reflect a more expansive role of primary care providers in rural areas to coordinate outpatient care or to otherwise avoid the need for emergency services. Understanding the underlying determinants of geographic variations in ED use among children with special health needs is an important area for future research.

Our results must be interpreted in light of two main limitations. First, because the data used to conduct these analyses were constructed from CSHCS and Medicaid claims, we were forced to exclude dual-enrolled children with other insurance coverage over the study period. The inability to track ED use among these children underscores the importance of obtaining comprehensive claims data from both public and private sources to provide a complete account of ED use in this population. Understanding the dynamics of service use for children with multiple sources of insurance coverage is an especially complex issue for the development of capitation rates and managed care arrangements for children with special needs.

Second, dual-enrolled children are a self-selected population. An unknown population of Michigan children may have qualifying conditions (and may otherwise be eligible for Title V services) yet are not CSHCS-enrolled. Some of the children may be covered by S-CHIP, standard Medicaid, or private insurance. Our paper is pertinent to policy choices within the enrolled population. Further investigation is required before generalizing to nonenrolled populations.

Conclusions

Much existing research examines the role of inadequate coverage, poverty, lack of a medical home, and family and psychosocial factors in explaining children's use of the ED (Apter et al. 1997; Bond, Stearns, and Peters 1999; Butler 1998; Chande, Krug, and Warm 1996; Christakis et al. 1999; Halfon et al. 1996; Jones et al. 1999; Wilson and Klein 2000; Ziv, Boulet, and Slap 1998). These factors may be less applicable to dual-enrolled children, who have insurance coverage and may make frequent contact with providers in many settings.

The present analysis highlights an obvious clinical fact, but one easily overlooked in policy analysis of ED use. Children's age and underlying medical condition are associated with remarkably different levels of ED use. Although race, location, and other demographic variables matter, illness complexity and severity are the most important sources of variation in ED use among children with special health needs. For this reason, clinical and policy interventions designed to promote appropriate ED use must not only address important sociodemographic disparities and adverse incentives, but must also acknowledge the central clinical challenge raised by chronically ill populations: how to appropriately manage a complex and persistent medical condition over time.

The need to understand variations in children's health services use—and the related need to understand how such variation is influenced by health policy and service delivery—is particularly acute as states enroll children with special needs into managed care. Efforts to establish capitated rates that reflect the mix and intensity of required services raise particularly difficult challenges in this population. The diverse circumstances of dual-enrolled children underscore the importance, but also the difficulty, of these public management and research tasks.

Acknowledgments

We thank Karla McCandless, Dan McCandless, Neil Oppenheimer, and others within Michigan's Medical Services Administration for help in obtaining and interpreting claims data, and for manuscript review. Any opinions expressed here are those of the authors, and do not necessarily reflect those of the state Medicaid or CSHCS programs.

APPENDIX

ICD-9-CM Codes Associated with Diagnostic Categories

Diagnostic CategoryICD-9-CM Codes*
Anemias2811, 2820–4, 2826–7, 284, 2850
Asthma49311, 49390
Cerebral Palsy3337, 3421, 343, 3440–5, 3448–9
Cleft Palate and Lip749
Congenital Heart Anomalies7450, 7452–6, 746 (×74687)
Cystic Fibrosis2770
Diseases of the Ear and  Mastoid Process38032, 38050–3, 3811–3, 38160, 38181, 3821–3, 3831, 38322, 38330–2, 38389, 3841, 3849, 3850–1, 38521, 38530, 38582, 38589, 3859, 3864, 3868, 38800, 38843–4, 3885, 38861, 3888–9, 38901–3, 3891–2
Epilepsy3451–7
Hemophilia2860–4, 2866–9
Juvenile Diabetes25001, 25011, 25021, 25031, 25041, 25051, 25061, 25071, 25081, 25091
Leukemia204–6, 2072, 2078, 208
Malignant Neoplasms140 (×1408), 141–9, 150–2, 153 (×1535), 154–9, 160–5, 170–5, 179, 180–9, 190–9
Other Congenital Anomalies2579, 7430, 74310, 74312, 7432–5, 74361–3, 74366, 74369, 7438–9, 7440, 7442, 74441, 74481, 74484, 74489, 7449, 7480, 7482–9, 75010, 75013, 75021, 75026, 75029, 7503–4, 7511–3, 75161, 7518, 75210, 75240–1, 7525–9, 7530–2, 7535, 7537–9, 7540–2, 75481, 75489, 7550–1, 7560, 75610–3, 75615–9, 7564–7, 75681, 75689, 7569, 7570, 7572, 75731, 75733, 7578, 7594–7, 7598 (×75983)
Respiratory Distress Syndrome769
Spina Bifida741
All Other Diagnoses
 Infectious and Parasitic   Diseases (001–139)042, 0785, 0903, 0907, 0993, 11509, 11519, 135, 1370–3, 138–9
 Neoplasms (140–239)200–3, 2120–2, 2150, 2156–7, 2159, 2239, 2240–1, 2245–6, 2248–9, 225, 2270–1, 2273–4, 22802–04, 22809, 2281, 2337, 237 (×2375–6), 2385–6, 2388–9, 2397–9
 Endocrine, Nutritional, and   Metabolic Diseases, and   Immunity Disorders (240–279)242–3, 2449, 2461, 2520–1, 2530–5, 2538–9, 2550–4, 2556, 258, 2592–9, 262, 2630, 2632–9, 2681, 2698–9, 270–1, 2720–5, 2727–9, 2751, 2753, 2754 (×27548), 2771–3, 2775–9, 2788, 2790, 2799
 Diseases of the Blood and   Blood-Forming Organs   (280–289)2831, 2873, 2879, 2880–1
 Mental Disorders (290–319)30490, 30723
 Diseases of the Nervous System   and Sense Organs (320–389)3209, 3222, 3229, 3239, 326, 330 (×3309), 3310–1, 3313–4, 3317, 33189, 3319, 3330, 3332, 3334, 3336, 33382, 33389, 33390, 33399, 3340, 3348, 335 (×33523), 336, 340–1, 342 (×3421, 3428), 3480–1, 3484, 3488–9, 3491–2, 34981, 34989, 3499, 3508–9, 351, 352 (×3521), 3530, 3536, 3560–1, 3563, 3570–3, 3580, 3588–9, 3590–3, 36003, 36011, 36013, 36032, 36034, 3609, 3610, 36111v2, 3612, 3618–9, 36212–8, 36221, 36229, 36230–3, 36235–6, 36240, 36242–3, 36252–3, 3626 (×36264), 3627, 36281, 36283–4, 3630–2, 3634–9, 36411, 36422, 36424, 36442, 3645, 36460, 3647–9, 365 (×3560), 3660 (×36604), 3662–3, 36644–6, 3665, 3668–9, 36752, 37002–7, 37050, 3706, 37100, 37102–4, 37110–3, 37115–6, 3713, 3714 (×37143), 3715–7, 3718 (×37182), 3719, 37451, 3771, 37721, 37723, 37733, 3774–7, 3779, 3780–1, 37821–2, 37824, 3783, 37841, 3785, 3786 (×37861), 37871–2, 37883–7, 3789, 37904, 37911, 37916, 37919, 37923, 3793 (×37931), 37951
 Diseases of the Circulatory   System (390–459)390, 391 (×3912), 392–6, 3970–1, 4011, 4021, 4029, 4031, 4039, 40410, 40413, 4049, 4051, 4059, 4109, 4140, 41411, 4148–9, 4150, 4160, 417, 4210, 4219, 42290–1, 42299, 423 (×4230), 4240–3, 42490, 4250–1, 4253–4, 4257–9, 4260–1, 4264–9, 4270, 42731–2, 428, 4290–1, 4293, 4295–6, 42981, 42989, 4299, 430–1, 4320–1, 4340–1, 4359, 436, 4373, 438 (×43853), 4420–8, 4439, 44489, 4449, 4460–1, 4463–7, 4475–6, 4480, 4533, 4571
 Diseases of the Respiratory   System (460–519)47830–4, 4784, 4841, 4928, 494, 510, 5131, 515, 5163, 5168, 5172, 5181–2, 5185
 Diseases of the Digestive   System (520–579)5200, 5204–5, 5218, 5240–2, 5244–6, 5248–9, 5303, 555–6, 5570, 5589, 5714–6, 5723, 5768, 579
 Diseases of the Genitourinary   System (580–629)5810, 5812–3, 58189, 5819, 582, 583 (×5838), 5845–9, 585–6, 5880–8, 5900, 591, 5934–5, 5937, 59381, 5960, 5969
 Diseases of the Skin and   Subcutaneous Tissue (680–709)6944, 6951–2, 6954, 6960–1, 7092
 Diseases of the   Musculoskeletal System and   Connective Tissue (710–739)7100–1, 7103–4, 7108–9, 71103, 71190, 713, 7140–2, 71430, 71432–3, 7144, 7149, 7150, 71598–9, 71610, 71800, 71820, 71823–4, 71828–9, 71830, 71838, 7184–5, 71860, 71880, 71884, 7198–9, 7200, 7209, 7210, 7212–3, 7215–6, 7218, 72190, 7220–7, 7229, 7230, 7239, 7240, 72702, 72781, 72810–1, 7282–3, 7286, 72889, 7301, 7307, 7320–7, 7329, 73320, 7334, 73382, 7360, 7363–4, 7366–9, 7370–1, 73720–2, 73734, 73739, 7374, 7378–9, 7380, 7386, 7388–9
 Congenital Anomalies (740–759)7420, 7422–9, 7451, 7457–9, 7470–3, 74741–9, 7476, 7478, 7543, 75441, 75450–1, 75459, 75460, 75462, 75469, 7547, 7552–5, 75560, 75739, 7599
 Certain Conditions   Originating in the   Perinatal Period (760–779)765, 7674–7, 7685, 7707–8, 7710–2, 7718, 7744, 7752, 7756–7, 7762, 7775
 Symptoms, Signs, and   Ill-Defined Conditions (780–799)7824, 7843, 7845, 7849, 78609, 7958, 7980
 Injury and Poisoning (800–999)806, 85104, 8520, 8522, 8524, 885–6, 8870, 8872, 8874, 8876, 896, 8970, 8972, 8974, 8976, 9050–2, 9054, 9056, 9059, 9060–1, 9063–9, 907, 9080–1, 9083–4, 9086, 9089, 9090–4, 9099, 940, 94109, 9413–5, 94209, 9423–5, 94309, 9433–5, 94408, 9443–5, 94509, 9453–5, 9460, 9463–5, 947, 94811, 94821–2, 94831–3, 94841–4, 94851–5, 94861–6, 94871–7, 94881–8, 94891–9, 9490, 9492–5, 952, 956, 9586, 9661, 9832, 984, 9895, 9948, 9950, 9956, 9966 (×99660), 99668, 99681, 9970, 9975, 99760, 9989
*Most ICD-9 codes included the Michigan CSHCS program are specified at the four- or five-digit level; however, to simplify reporting in this table, three-digit codes are used if all four- and five-digit codes underneath it apply. Ranges do not necessarily imply a sequential list of codes, but rather than all valid codes within that range are included. Codes in parentheses proceeded by an “×” are excluded from the definition of that diagnostic category.

Footnotes

1At any time, approximately 20 percent of dual-enrolled children have private insurance, often through a parent's employer. If insurers offer generous ED coverage, privately insured children may use the ED without generating a Medicaid or CSHCS claim. State claims data may therefore be incomplete.

2International Classification of Diseases, 9th Revision, Clinical Modification disease coding system.

3We found similar effects for the six children listed with Tourette's syndrome. Tourrette's syndrome is no longer a qualifying condition for CSHCS.

This research was funded by the Agency for Healthcare Research and Quality, the Packard Foundation, and the Health Resources and Services Administration, grant no. 1 U01 HS 10441-0.

References

  • Abbott R D. “Logistic Regression in Survival Analysis.” American Journal of Epidemiology. 1985;121(3):465–71. [PubMed]
  • Apter A J, Reisine S T, Kennedy D G, Cromley E K, Keener J, ZuWallack R L. “Demographic Predictors of Asthma Treatment Site: Outpatient, Inpatient, or Emergency Department.” Annals of Allergy, Asthma, and Immunology. 1997;79(4):353–61. [PubMed]
  • Baraff L J, Bass J W, Fleisher G R, Klein J O, McCracken G H, Powell K R, Jr., Schriger D L. “Practice Guideline for the Management of Infants and Children 0 to 36 Months of Age with Fever without Source.” Annals of Emergency Medicine. 1993;22(7):1198–10. [PubMed]
  • Bond T K, Stearns S, Peters M. “Analysis of Chronic Emergency Department Use.” Nursing Economics. 1999;17(4):207–13,237. [PubMed]
  • Butler P A. “Medicaid HMO Enrollees in the Emergency Room: Use of Nonemergency Care.” Medical Care Research and Review. 1998;55(1):78–98. [PubMed]
  • Cameron A C, Trevidi P K. Regression Analysis of Count Data. Cambridge, U.K.: Cambridge University Press; 1998.
  • Chande V T, Krug S E, Warm E F. “Pediatric Emergency Department Utilization Habits: A Consumer Survey.” Pediatric Emergency Care. 1996;12(1):27–30. [PubMed]
  • Christakis D A, Wright J A, Koepsell T D, Emerson S, Connell F A. “Is Greater Continuity of Care Associated with Less Emergency Department Utilization?” Pediatrics. 1999;103(4 part 1):738–42. [PubMed]
  • DiIorio F C. SAS Applications Programming. Pacific Grove, CA: Duxbury Press; 1991.
  • Gadomski A M, Perkis V, Horton L, Cross S, Stanton B. “Diverting Managed Care Medicaid Patients from Pediatric Emergency Department Use.” Pediatrics. 1995;95(2):170–8. [PubMed]
  • Garrett B, Glied S. “Does State AFDC Generosity Affect Child SSI Participation?” Journal of Policy Analysis and Management. 2000;19(2):275–95.
  • Gill J C. “The Role of Genetics in Inhibitor Function.” Thrombosis and Haemostasis. 1999;82(2):500–4. [PubMed]
  • Halfon N, Newacheck P W. “Childhood Asthma and Poverty: Differential Impacts and Utilization of Health Services.” Pediatrics. 1993;91(1):56–61. [PubMed]
  • Halfon N, Newacheck P W, Wood D L, St R F, Peter “Routine Emergency Department Use for Sick Care by Children in the United States.” Pediatrics. 1996;98(1):28–34. [PubMed]
  • Herendeen N E, Szilagyi P G. “Infections of the Upper Respiratory Tract.” In: Behrman R E, Kliegman R B, Jenson H B, editors. Nelson Textbook of Pediatrics. 16th ed. Philadelphia, PA: W.B. Saunders; 2000. pp. 1261–6.
  • Hodge D., III “Managed Care and the Pediatric Emergency Department.” Pediatric Clinics of North America. 1999;46(6):1329–40. [PubMed]
  • Holl J L, Szilagyi P G, Rodewald L E, Shone L P, Zwanziger J, Mukamel D B, Trafton S, Dick A W, Barth R, Raubertas R F. “Evaluation of New York State's Child Health Plus: Access, Utilization, Quality of Health Care, and Health Status.” Pediatrics. 2000;105(3):711–8. [PubMed]
  • Jones D S, McNagny S E, Williams M V, Parker R M, Sawyer M F, Rask K J. “Lack of a Regular Source of Care among Children Using a Public Hospital Emergency Department.” Pediatric Emergency Care. 1999;15(1):13–6. [PubMed]
  • Lozano P, Fishman P, VonKorff M, Hecht J. “Health Care Utilization and Cost among Children with Asthma Who Were Enrolled in a Health Maintenance Organization.” Pediatrics. 1997;99(6):757–64. [PubMed]
  • McCaig L F. Vital and Health Statistics Advance Data No. 313. Hyattsville, MD: National Center for Health Statistics; 2000. “National Hospital Ambulatory Medical Care Survey: 1998 Emergency Department Summary.” [PubMed]
  • McCarthy P L. “The Acutely Ill Child.” In: Behrman R E, Kliegman R B, Jenson H B, editors. Nelson Textbook of Pediatrics. 16th ed. Philadelphia, PA: W. B. Saunders; 2000. pp. 228–312.
  • McCormick M C, Kass B, Elixhauser A, Thompson J, Simpson L. “Annual Report on Access to and Utilization of Health Care for Children and Youth in the United States–1999.” Pediatrics. 2000;105(1):219–30. [PubMed]
  • Miller J E. “The Effects of Race/Ethnicity and Income on Early Childhood Asthma Prevalence and Health Care Use.” American Journal of Public Health. 2000;90(3):428–30. [PubMed]
  • Neff J M, Anderson G. “Protecting Children with Chronic Illness in a Competitive Marketplace.” Journal of the American Medical Association. 1995;274(23):1866–9. [PubMed]
  • Newacheck P W, Strickland B, Shonkoff J P, Perrin J M, McPherson M, McManus M, Lauver C, Fox H, Arango P. “An Epidemiologic Profile of Children with Special Health Care Needs.” Pediatrics. 1998;102(1):117–23. [PubMed]
  • Perrin J M, Kuhlthau K, McLaughlin T J, Ettner S L, Gortmaker S L. “Changing Patterns of Conditions among Children Receiving Supplemental Security Income Disability Benefits.” Archives of Pediatrics and Adolescent Medicine. 1999;153(1):80–4. [PubMed]
  • Pollack H A, Khoshnood K, Blankenship K M, Altice F L. “The Impact of Needle Exchange-Based Health Services on Emergency Department Use.” Journal of General Internal Medicine. 2002;17(5):341–8. [PMC free article] [PubMed]
  • Price M R, Norris J M, Bartleson B B, Gavin L A, Klinnert M D. “An Investigation of the Medical Care Utilization of Children with Severe Asthma According to Their Type of Insurance.” Journal of Asthma. 1999;36(3):271–9. [PubMed]
  • Reynolds S, DesGuin B, Uyeda A, Davis A T. “Children with Chronic Conditions in a Pediatric Emergency Department.” Pediatric Emergency Care. 1996;12(3):166–8. [PubMed]
  • Sharma V, Simon S D, Bakewell J M, Ellerbeck E F, Fox M S, Wallace D D. “Factors Influencing Infant Visits to Emergency Departments.” Pediatrics. 2000;106(5):1031–9. [PubMed]
  • Shatin D, Levin R, Ireys H T, Haller V. “Health Care Utilization by Children with Chronic Illnesses: A Comparison of Medicaid and Employer-Insured Managed Care.” Pediatrics. 1998;102(4):e44. [PubMed]
  • Szilagyi P G, Holl J L, Rodewald L E, Yoos L, Zwanziger J, Shone L P, Mukamel D B, Trafton S, Dick A W, Raubertas R F. “Evaluation of New York State's Child Health Plus: Children Who Have Asthma.” Pediatrics. 2000;105(3, supplement E):719–27. [PubMed]
  • U. S. General Accounting Office. Medicaid Managed Care: Challenges in Implementing Safeguards for Children with Special Needs. Washington, DC: U. S. General Accounting Office; 2000. Report no. GAO/HEHS-00-37.
  • Wilson K M, Klein J D. “Adolescents Who Use the Emergency Department as Their Usual Source of Care.” Archives of Pediatrics and Adolescent Medicine. 2000;154(4):361–5. [PubMed]
  • Yamamoto L G, Zimmerman K R, Butts R J, Anaya C, Lee P, Miller N C, Shirai L K, Tanaka T T, Leung Y K. “Characteristics of Frequent Pediatric Emergency Department Users.” Pediatric Emergency Care. 1995;11(6):340–6. [PubMed]
  • Zimmerman D R, McCarten-Gibbs K A, DeNoble D H, Borger C, Fleming J, Hsieh M, Langer J C, Breckenridge M B. “Repeat Pediatric Visits to a General Emergency Department.” Annals of Emergency Medicine. 1996;28(5):467–73. [PubMed]
  • Ziv A, Boulet J R, Slap G B. “Emergency Department Utilization by Adolescents in the United States.” Pediatrics. 1998;101(6):987–94. [PubMed]

Articles from Health Services Research are provided here courtesy of Health Research & Educational Trust